Next Article in Journal
A Multiscale Study on Substrate Size Effect and Energy Density Regulation on Dynamic Response of Dilution Rate in Laser Cladding Iron-Based Coatings
Previous Article in Journal
Thermochromic Behaviour and Comfort Properties of Printed Woven Fabric
Previous Article in Special Issue
The Properties of Self-Compacting Ultra-High Performance Concrete with Different Types of Mineral Admixtures
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

Review of Research on Prediction Models for Residual Life of Concrete Structures

1
School of Civil and Transportation Engineering, Ningbo University of Technology, Ningbo 315211, China
2
Engineering Research Center of Industrial Construction in Civil Engineering of Zhejiang, Ningbo University of Technology, Ningbo 315211, China
3
Ningbo Roaby Technology Industrial Group Co., Ltd., Ningbo 315800, China
4
Key Laboratory of New Technology for Construction of Cities in Mountain Area, School of Civil Engineering, Chongqing University, Chongqing 400045, China
*
Authors to whom correspondence should be addressed.
Coatings 2025, 15(6), 693; https://doi.org/10.3390/coatings15060693 (registering DOI)
Submission received: 7 May 2025 / Revised: 3 June 2025 / Accepted: 5 June 2025 / Published: 7 June 2025

Abstract

:
The performance degradation of concrete structures directly impacts their safety. As such, accurately predicting their remaining service life is critical for effective operation and maintenance management. This paper reviews the key factors influencing the degradation of concrete structures, providing a comprehensive summary of current research on deterioration mechanisms, steel corrosion, crack propagation, and chloride ion penetration. It also compares the advantages and limitations of physical, empirical, statistical, and machine learning models used for life prediction. A critical aspect highlighted in this paper is the importance of model validation based on real-world field data, which can more effectively determine the applicability of prediction models in actual engineering practice. Model validation incorporates evaluation metrics like sensitivity analysis to gauge how fluctuations in input parameters, such as temperature, influence life prediction models and thereby reveal the uncertainties inherent in complex engineering environments. Currently, life prediction models are widely applied to infrastructure projects like bridges and tunnels. By incorporating environmental factors such as chloride ion concentration, temperature, and humidity, as well as real-time monitoring data, these models effectively predict the remaining service life, aiding engineers in developing optimized maintenance strategies. However, current models still face challenges in terms of data requirements and accuracy. Future research should focus on the integration of hybrid models and intelligent technologies. By leveraging the combined strengths of physical and data-driven approaches, hybrid models can enhance prediction accuracy. Additionally, intelligent technologies and real-time monitoring will help dynamically update and optimize models, resulting in more precise and efficient life predictions.

1. Introduction

Concrete structures find widespread application in the field of civil engineering, especially in the construction of buildings, bridges, tunnels, and other forms of infrastructure development, where they serve a fundamental function. The lifetime of such structures needs to be predicted to maintain public safety, minimize costs of maintenance, and improve the efficiency of resource allocation. Concrete is the reigning material in infrastructure construction as the primary building material. Research shows that nearly half of the global production of ordinary Portland cement goes towards the production of about 11 billion tons of concrete [1]. With such a vast demand, the assessment of the durability of concrete structures has become a more pertinent subject of research with far-reaching implications.
The durability of concrete structures over the long term is especially critical under severe environments, such as those characterized by marine and cold climates. Research has established that, in these specific environments, the rate of corrosion affecting concrete structures significantly exceeds the rate under normal conditions, leading to increased deterioration. For instance, the high level of chloride ions characteristic of marine environments promotes the corrosion of reinforcing bars, thereby compromising the structural integrity of the concrete [2,3]. In cold environments, low temperatures cause the internal moisture in concrete to freeze. As the temperature fluctuates, the expansion and contraction of the ice generate stress within the concrete, particularly under prolonged exposure to low temperatures. While cold environments do not directly cause freeze–thaw cycles, extended periods of low temperatures can significantly reduce the physical properties of concrete, thereby affecting its structural stability [4]. In contrast, freeze–thaw cycles refer to the repeated freezing and melting of moisture within concrete due to temperature fluctuations. During freezing, the moisture expands, and during melting, it contracts. These cycles generate internal expansion stresses within the concrete, which gradually enlarge microcracks, ultimately leading to surface spalling and crack propagation. The effect of freeze–thaw cycles on concrete’s durability is especially pronounced [5,6].
The degradation of concrete structures is a complicated multivariable phenomenon [7,8] involving numerous factors such as corrosion of crack propagation, chloride ion infiltration, and material fatigue. In particular, chloride-induced corrosion causes rust on the surface of the rebar, initiating crack propagation that accelerates the reduction in the structure’s strength and service life. Thus, the precise prediction of the residual service life of concrete structures is crucial to avert structural failure and effectively extend their service life.
The development of machine learning and artificial intelligence in recent years has enabled the creation of lifespan prediction models that involve real-time monitoring data and multi-factor integration, thereby rendering it a forefront research area. As opposed to the conventional physical models, the novel approaches utilize big datasets and machine learning algorithms to optimally address complicated nonlinear relationships, thereby greatly improving the accuracy of predictions [9,10].
This paper offers an extensive overview of current lifespan prediction models for concrete structures and their merits and drawbacks, and proposes ways for enhancement. First, we review the fundamental principles and application domains of physical and statistical models. Next, we examine how machine learning-based models, by processing real-time monitoring data, enable dynamic updates that enhance prediction accuracy. Finally, we discuss future research directions, particularly the integration of multi-factor coupling with intelligent models, and how real-time data feedback can further optimize prediction outcomes. As technology advances, future lifespan prediction models for concrete structures are expected to become increasingly accurate and efficient, offering enhanced support for structural health management in civil engineering.

2. Aging and Failure Mechanisms of Concrete Structures

The aging and failure mechanisms of concrete structures are crucial aspects of concrete durability research. Over time, chloride ions penetrate into concrete, eroding the passive film on the steel reinforcement, initiating electrochemical corrosion and leading to the expansion of corrosion products. This results in concrete cracking, accelerates the ongoing corrosion of the steel reinforcement, significantly reduces the mechanical properties of the concrete, and severely compromises its service life. This process involves the coupling of multiple factors, including rebar corrosion, crack growth, and environmental effects. As corrosion and cracking advance, the concrete’s strength and stability diminish, potentially resulting in catastrophic structural failure.

2.1. Rebar Corrosion and the Decline of Concrete’s Mechanical Properties

Rebar corrosion is a primary factor contributing to the degradation of concrete structures, particularly in environments with chlorides, moisture, and temperature variations, where the corrosion rate accelerates significantly. The expansion of corroded rebar generates internal stress, causing surface cracking in the concrete, which, in turn, accelerates the decline of its mechanical properties. A thorough understanding of rebar corrosion and its effects on the mechanical behavior of concrete is essential for providing scientific guidance in the design, repair, and maintenance of concrete structures.
For instance, Tian et al. [11] reviewed the key characteristics of rebar corrosion in concrete under natural environmental conditions. The study found that, while the corrosion rate decreases over time, crack propagation accelerates, highlighting the importance of establishing a correlation between laboratory accelerated tests and natural corrosion, as well as developing a durability assessment system for structures after repair. Furthermore, to evaluate the corrosion risk of rebar in marine environments, Ju et al. [12] proposed a time-dependent model for the surface chloride ion concentration and diffusion coefficient. This model can predict the variation in chloride ion concentration in concrete under various heights and exposure durations. Zhang et al. [13] developed a probabilistic model for rebar corrosion based on the Markov chain model and Faraday’s law. Their experimental data showed that, as current density increased from 50 µA/cm2 to 100 µA/cm2, the corrosion rate rose from 6.645 × 10−5 mm/h to 13.290 × 10−5 mm/h. The increase in current density accelerated the corrosion process, providing valuable insights for the evaluation and durability prediction of rebar corrosion in concrete structures. As corrosion advances, the concrete cover deteriorates, exposing the rebar to external environmental conditions. Sun et al. [14] proposed an enhanced model for concrete carbonation and rebar corrosion to evaluate crack vulnerability induced by corrosion. This model accurately predicts crack evolution in concrete bridges under corrosion and offers guidance for bridge maintenance. Zhang and Su [15] examined non-uniform crack propagation due to rebar corrosion and developed a corner crack model to describe the degradation of concrete cover. This model provides theoretical support for the durability design, service life prediction, and crack width estimation of concrete structures. To assess the durability of concrete structures exposed to marine environments for extended periods, Wang et al. [16] introduced a chloride ion diffusion model that accounts for the coupling effects of coarse aggregate and rebar. The study demonstrated that, as the volume fraction of coarse aggregate increased from 0 to 0.3, the chloride ion diffusion coefficient decreased from 9.37 × 10−12 m2/s to 5.98 × 10−12 m2/s, a reduction of 20%. This suggests that coarse aggregate improves the concrete’s resistance to chloride ion penetration by modifying its pore structure, thus enhancing durability. Hajkova et al. [17] proposed a chemical–mechanical model for chloride-induced rebar corrosion and examined the impact of corrosion on the serviceability of concrete structures. Their study provides valuable theoretical support for assessing the serviceability of structures affected by rebar corrosion, underscoring the importance of early diagnosis and repair. Wang and Lu [18] investigated the corrosion process of rebar in ordinary concrete and concrete with varying fly ash contents under constant climatic conditions. The results showed that increasing the fly ash substitution ratio from 0% to 45% significantly reduced the corrosion rate of rebar, thereby improving concrete durability. Finally, Ann et al. [19] analyzed chloride ion penetration in concrete using four different surface chloride ion accumulation models. The study found that the constant surface chloride ion model overestimated corrosion risk, whereas the optimized accumulation model more accurately predicted chloride ion distribution and corrosion life. This approach provides a reliable foundation for structural maintenance and durability assessment.

2.2. Crack Propagation and Structural Damage

Rebar corrosion not only threatens the strength and durability of concrete, but also significantly compromises the overall safety of the structure. Common corrosion products include carbide, chloride, and sulfate compounds. Table 1 presents a comparison of the effects of these three typical corrosion products on the corrosion process. As observed from Table 1, among the three corrosion products, chloride ions have the most significant impact on both the concrete protection layer and the corrosion process.
As corrosion progresses, the expansion of the rebar generates internal cracks in the concrete. The study in reference [23] primarily investigates the influence of different types of cement and supplementary cementitious materials (SCMs) on the corrosion rate of reinforcement steel, particularly under conditions with cover depths of 20 mm and 40 mm, and crack widths of 0.2 mm and 0.7 mm. The experimental data in Table 2 are derived from the research presented in reference [23], with the corrosion rates obtained under accelerated corrosion conditions in a laboratory setting. These conditions involve a wet–dry cycling process and exposure to a 5% sodium chloride solution. The corrosion rate is calculated by applying a small electrical current and monitoring the changes in the reinforcement steel potential, which allows for determining the polarization resistance of the steel. The corrosion rate is then calculated using the Stern–Geary equation. According to Table 2, significant differences in corrosion rates are observed for different types of cement; ordinary Portland cement (PC) shows higher corrosion rates under both 20 mm and 40 mm cover depths at 2.65 µA/cm2 and 1.20 µA/cm2, respectively. This is primarily due to its lower resistivity, which allows chloride ions to more easily penetrate the steel surface. In contrast, cements containing fly ash (FA) and ground granulated blast furnace slag (GGBS) significantly reduce the corrosion rate. In the case of fly ash concrete, the corrosion rates are 0.64 µA/cm2 for 20 mm cover depth and 0.50 µA/cm2 for 40 mm cover depth, indicating that fly ash increases the concrete resistivity and inhibits chloride ion ingress. GGBS exhibits the lowest corrosion rates at 0.35 µA/cm2 and 0.24 µA/cm2, respectively, due to improved concrete density and resistivity, which reduces chloride ion penetration. Silica fume shows considerable fluctuation in corrosion rates, especially at 40 mm cover depth and larger crack widths, where the corrosion rate reaches 1.03 µA/cm2. This may be attributed to the rapid hydration characteristics of silica fume, leading to an unstable microstructure, which exposes the steel to corrosive environments at the crack sites and accelerates the corrosion process.
Therefore, supplementary materials such as fly ash and ground granulated blast furnace slag significantly improve the corrosion resistance of concrete by enhancing its resistivity and density. In contrast, ordinary Portland cement and silica fume exhibit different corrosion rate variations due to their distinct microstructural characteristics.
Several studies have investigated crack propagation and structural damage. For instance, Guo et al. [24] proposed a new fatigue life prediction model for ordinary Portland cement (PC) beams that takes into account the spatial variability of pitting corrosion. The model uses a probabilistic approach to predict fatigue damage resulting from rebar corrosion. Wu et al. [25], based on Fick’s second law and Paris’ law, developed a model to predict the lifespan of reinforced concrete (RC) beams under chloride ion corrosion and fatigue. Their study revealed that crack initiation significantly affects fatigue life, which decreases with lower stress ratios, longer service times, and higher environmental temperatures, offering valuable insights for structural safety assessments. As corrosion progresses, cracks in the concrete propagate, which usually leads to the deterioration of the structure and, in extreme cases, the premature failure of the reinforced concrete elements. Ge and Kim [26] introduced a probabilistic update method based on Bayesian inference, integrating multiple inspection results using Markov chain Monte Carlo sampling. This approach enhances the prediction accuracy of remaining life under corrosion–fatigue interactions in reinforced concrete structures, providing critical support for preventive maintenance and life prediction. Bhargava et al. [27] developed a model to evaluate concrete cover cracking caused by rebar corrosion. This model considers the effect of corrosion on the concrete cover and is capable of predicting the probability of crack propagation and structural deterioration. Choe et al. [28] also introduced a probabilistic model integrating chloride ion corrosion, time-dependent corrosion rates, and shear capacity analysis for examining the vulnerability of reinforced concrete columns. Sensitivity analysis was employed to identify the most dominant parameters, giving a theoretical basis for reinforced concrete structure life prediction and lifecycle cost estimation. These studies highlight the fact that corrosion is not just a parameter influencing concrete strength, but, by inducing crack propagation and structural damage accumulation, eventually compromises the safety of the whole structure. As the corrosion process continues, the structural integrity is lost, highlighting the need for early detection and evaluation to ensure the long-term safety of concrete structures.

2.3. Chloride Ion Penetration and Concrete Corrosion

Concrete structures, particularly those exposed to marine and cold climates, are prone to corrosion from chloride ion intrusion. Chloride ions are one of the primary causes of rebar corrosion initiation, and their diffusion and penetration significantly affect concrete durability. It is important to distinguish between corrosion efficiency and degree of reinforcement corrosion. Corrosion efficiency refers to the effectiveness of the electrochemical reaction in causing the dissolution of steel, while the degree of reinforcement corrosion describes the overall extent of material loss and damage to the rebar due to corrosion. Figure 1, derived from reference [29], illustrates that the corrosion efficiency of the 5% NaCl solution is highest at a current density of 3 mA/cm2, suggesting that the corrosion rate of the rebar increases at lower current densities. This also highlights the variation in the corrosion efficiency of rebar due to chloride ions under different environmental conditions.
In recent years, extensive research has been directed towards chloride ion diffusion models in concrete, corrosion initiation mechanisms, and the corrosion prediction of reinforcing bars. For example, Yu et al. [30] performed long-term experimental investigations to examine the correlation between the chloride ion diffusion coefficient, exposure time, and curing age. The results obtained showed that the diffusion coefficient is a power function of curing age, but its relation to exposure time was not found to be significant. This observation has provided a scientific basis for estimating the durability of concrete structures. Chen et al. [31] have proposed a probabilistic multi-factor predictive model in which various aggregate sizes (i.e., dmax = 10 mm, 20 mm, 25 mm) are taken into consideration when estimating the chloride ion diffusion coefficient. The results proved a significant increase in the coefficient of diffusion for chloride ions relating to the increased sizes of aggregates. Furthermore, an analysis of a case study of Shenzhen-Zhongshan Bridge with a concrete cover thickness of 78 mm provided evidence that raising the cover thickness reduced the penetration depth of the chloride ions and, hence, enhanced the durability of the structure. Their study highlights the relevance of both the cover thickness of the concrete and the diffusion coefficient of the chloride ions as key factors deciding durability. In practical applications, the concentration of chloride ions plays a role in assessing the durability of concrete structures. To solve this issue, Guo et al. [32] proposed a random forest–fuzzy logic hybrid model, in which the random forest algorithm was utilized to enhance training samples and forecast surface chloride concentrations in concrete. The model effectively deals with complicated environmental conditions. In the same way, Taffese and Espinosa-Leal [33] used five various machine learning models with the goal of predicting chloride ions’ permeability in concrete, thereby developing novel approaches for assessing the lifespan and durability of construction materials. Liu et al. [34] analyzed how precipitation affects chloride ions’ transport dynamics in concrete and established that higher rainfall intensity and duration decreased chloride ion concentration at the surface considerably. For chloride-induced corrosion in concrete, Shao et al. [35] proposed a combined evaluation approach with both deterministic and probabilistic methodologies for evaluating the corrosion initiation life of reinforced concrete hollow piles in chloride environments. Safehian and Ramezanianpour [36] also analyzed chloride ion penetration in a dock exposed to harsh marine environments, taking into account the influence of construction methods and exposure conditions. Their findings underscored the significant influence that microenvironments have on the penetration of chloride ions. The research shows that, with the progress in investigating corrosion mechanisms triggered by chloride, increasingly predictive models and assessment methods are being created, thus giving valuable support for concrete maintenance.
In conclusion, concrete structure ageing and deterioration are complex and multi-factorial processes involving rebar corrosion, cracking, chloride ion ingress, and fatigue. As artificial intelligence and machine learning have advanced in recent times, real-time monitoring-data-driven lifespan prediction models and multi-factor coupling models have gained popularity in research on concrete structure ageing. Future research must further investigate the relationship between these variables, with a particular focus on how real-time monitoring and feedback systems can enhance predictive modeling. This would, therefore, allow for more precise and effective assistance in the maintenance and management of concrete structures.

3. Classification of Lifespan Prediction Models

This section outlines different types of life prediction models and categorizes them into four types: physical models, empirical models, statistical models, and machine learning-based models. It also provides a detailed discussion on each model type, analyzing their principles, application areas, advantages, and limitations.

3.1. Physical Lifespan Prediction Models

Physical lifespan prediction models of structures assess their durability by simulating the interaction between concrete and the environment and physical–chemical processes. Founded on sound theoretical principles, these models yield credible lifespan predictions for varying environmental conditions. Carbonation models and chloride ion penetration models are representative models in this group.

3.1.1. Carbonation Depth Model

Carbonation is a major cause of concrete degradation. Carbon dioxide migrates through the surface of the concrete, reducing its pH and harming the protective coating of the rebar, leading to the corrosion of the rebar. The Fick model, used routinely for the estimation of carbonation depth, assumes that carbonation depth varies with the square root of time [37]. The equation is given as follows.
h = x 0 t
In this Equation (1), h denotes the carbonation depth, or the extent to which carbon dioxide penetrates into the concrete, t represents time, and x 0 , known as the carbonation coefficient, indicates the carbonation rate of the concrete.
The Fick model’s linear nature and strong physical basis make it amenable to particular uses, namely the initial estimation of carbonation depth. Its drawbacks are, however, exposed under complicated environmental circumstances, for instance, where the carbonation rate changes with high humidity and, hence, departs from the Fick model’s assumptions [38]. As a solution to this issue, Cao et al. [39] suggested an improved variant of the Fick model by adding a correction term to the basic equation.
h = x 0 t + x 1 e t e t e t + e t
In this Equation (2), x 1 denotes an unknown fitting parameter in the correction term introduced.
The data in Table 3 come from reference [39]. As demonstrated in Table 3, the improved model not only reduces prediction errors, but also minimizes the variability in the results, thereby enhancing the applicability and reliability of the Fick model for carbonation depth prediction.

3.1.2. Chloride Ion Penetration Model

Chloride ion penetration is a significant cause of corrosion in reinforced concrete structures, particularly in marine and industrial environments. Chloride ions diffuse through the concrete’s surface and penetrate the structure, compromising the rebar’s passive film, which leads to corrosion. The prediction of chloride ion penetration is generally based on Fick’s second law. Yu et al. [40] applied Fick’s second law model to describe the diffusion process of chloride ions in concrete. The basic Equation (3) of the model is as follows:
C x , t = C 0 + C s C 0 1 e r f x 2 t D c l
  • C x , t represents the chloride ion concentration at time t and depth x ,
  • C 0 is the initial concentration, typically the chloride ion concentration in the concrete,
  • C s is the surface concentration, corresponding to the concentration at the concrete surface exposed to water or air,
  • D c l is the chloride diffusion coefficient,
  • erf denotes the error function.
In practical applications, Fick’s law assumes uniform diffusion. However, in concrete structures, the diffusion coefficient is typically not constant due to variations in environmental factors such as temperature and humidity. As a result, numerous studies have modified Fick’s law to account for the time dependence of the diffusion coefficient and other influencing factors. The improved Equation (4) [41] is as follows:
C ( x , t ) = C s 1 e r f x 2 D 0 1 n t 0 t n t
  • D 0 represents the initial chloride ion diffusion coefficient,
  • t 0 is the reference time,
  • n is the coefficient for time-dependent decay.
Equation (4) has been evaluated in both laboratory and field environments. Under laboratory conditions, the model tends to show a larger and more stable diffusion coefficient, near the ideal time exponent, and comparably low computational complexity. For field applications, the diffusion coefficient is significantly affected by external conditions, so there are larger fluctuations. The time exponent can be different from the ideal, and other factors must be taken into account. Although this adds computational complexity, the model provides a more realistic description of actual nonlinear diffusion processes and exhibits great benefits in practical applications. Chen et al. [42] also presented a variable-order fractional diffusion model that has the ability to describe the chloride ion diffusion–absorption process more accurately by introducing a concentration influence factor k. The study shows that the new model describes experimental data more precisely than the classic Fick’s second law and can forecast the distribution of chloride ion concentration for 200 days.
Physical models predict the lifespan of concrete structures by simulating the interaction between environmental factors and physical–chemical processes. Examples are the Fick model and the chloride ion diffusion model. The Fick model estimates carbonation depth as a function of the square root of time and is, thus, suitable for application during the early age, but accuracy declines under complicated environments. The improved model incorporates a time-dependent diffusion coefficient for increased accuracy. The diffusion model of chloride ions using Fick’s second law simulates the rebar corrosion caused by chloride ion intrusion. With the influence of environmental variations on the diffusion coefficient, researchers have proposed more advanced models, such as those with fractional derivatives, to better fit actual conditions. In conclusion, although physical models offer a sound theoretical basis for lifespan predictions, optimization is needed under complex environments.

3.2. Empirical Models

The estimation of the lifespan of concrete structures has the highest priority in civil engineering, especially their durability for long periods under changing environmental conditions and loading. By combining historical data with information about environmental effects and changes in loading, researchers have developed a series of empirical formulas and data fitting models to precisely estimate the lifespan of concrete structures. This section presents an exhaustive overview of empirical models, such as their applicability, limitations, and future directions.

3.2.1. Empirical Formulas

Empirical formulas are predictive models that have been created by analyzing large amounts of historical data, as well as environmental factors such as temperature, humidity, and corrosive substances, along with loading conditions. Empirical formulas are a scientifically based platform for the construction, building, and maintenance of concrete structures.
For example, Lu et al. [43] compared seven empirical models on the basis of 156 experimental datasets and introduced a new model that comprises variables such as concrete resistivity, temperature, humidity, corrosion time, and chloride ion content. This model effectively predicts corrosion rates and, through error probability distribution analysis, provides theoretical support for the reliability-based lifespan prediction of reinforced concrete structures. Chen et al. [44] developed a simplified empirical model using regression analysis, enabling the rapid evaluation of reinforced concrete structure service life and offering a practical computational tool for engineering applications. While empirical models are convenient and user-friendly, they generally rely on historical data and lack the detailed consideration of structural specifics and complex environmental changes.

3.2.2. Applicability and Limitations of Empirical Models

While empirical models are simple to use and computationally efficient, their application presents several challenges. First, these models often fail to account for the impact of specific environmental factors, such as climate change and various corrosive media, which can lead to biases in predictions. Second, empirical models rely heavily on historical data, which limits their predictive accuracy in situations with scarce data or unique environmental conditions. Table 4 summarizes several chloride diffusion models, which are examples of empirical models used to predict chloride penetration in concrete. These models, although computationally efficient, also face challenges such as simplifying assumptions and reliance on specific conditions. The Table 4 provides a comparative overview of these models, highlighting their advantages and disadvantages.
It is noteworthy that, while the empirical models in Table 4 provide practical methods for predicting chloride ion ingress into concrete, these models must be used with great caution, particularly in situations where the data are limited or the environmental conditions are complex.

3.2.3. Integrating Empirical Models with Real-Time Data

Advancements in sensor technology and real-time monitoring systems offer significant potential to enhance the predictive accuracy of empirical models. By integrating real-time data, such as temperature, humidity, and chloride ion concentration, into these models, parameters can be dynamically updated, improving prediction precision. For instance, Ashrafian et al. [49] used 642 field data points and identified the water-to-cement ratio, annual average temperature, and exposure time as key factors in predicting chloride ion concentration on the surface of concrete in marine environments. Wang et al. [50] applied extensive experimental data from marine tidal zone field studies, reported in the literature, to predict long-term chloride ion diffusion in concrete and developed a structural lifespan assessment method suitable for marine environments.
Empirical models, known for their simplicity and efficiency, are widely applicable in predicting the lifespan of concrete structures. However, their accuracy and applicability heavily depend on the quality of historical data and variations in environmental conditions. In practical application, it is recommended that environmental conditions, real-time monitoring, and the strengths and weaknesses of other modeling techniques be integrated to enable a more in-depth analysis that leads to the enhanced accuracy and reliability of lifespan estimation.

3.3. Statistical Lifespan Prediction Models

The prediction of the lifespan of concrete structures has emerged as a key component of civil engineering, specifically in the context of sustainability and maintaining infrastructure over extended durations. Specifically, regression analysis and Bayesian networks, in particular, are widely used to quantify the relationship between various factors and the expected lifespan of concrete structures.

3.3.1. Regression Analysis Models

Linear and nonlinear regression analyses are commonly used to predict the life of concrete structures by defining the interrelations among contributing parameters and durability responses. For instance, Prieto et al. [51] used multiple linear regression analysis with over 650 data points of rebar bond tests, demonstrating that their suggested model is effective and reliable compared to other models for bond strength assessment, and it is suitable for assessing the structural integrity of reinforced concrete elements subjected to corrosion.
Additionally, Jin et al. [52] used nonlinear and hybrid regression analyses, showing that these models offer higher accuracy and better R2 values compared to linear methods. Similarly, Hanif et al. [53] reviewed the latest developments in nonlinear-model-based approaches for structural system damage assessment, identified limitations in field applications, and proposed new methods for damage assessment. While regression analysis is a powerful and widely used tool, it has limitations, particularly in handling nonlinear relationships, poor data quality, and multicollinearity. In practice, additional validation and optimization methods are required to ensure the model’s accuracy and stability.

3.3.2. Bayesian Networks for Lifespan Prediction

Bayesian networks offer a powerful statistical framework to integrate uncertainty and complex dependencies between influencing factors. These models are particularly useful in predicting the lifespan of concrete structures, as various uncertainty factors interact over time. HSPBN is a type of Bayesian network that models both discrete and continuous variables. Figure 2 [54] illustrates an example of an HSPBN. Discrete nodes can only have other discrete nodes as parents, while continuous nodes can have any combination of parent types. The conditional distribution of discrete variables is usually represented as a categorical distribution (CPT), while continuous variables are represented by parameterized or non-parameterized conditional probability distributions (CPDs). When the CPDs for all continuous variables are parameterized, the HSPBN becomes a special case of CLGBN. However, the model is more flexible when non-parameterized CPDs are used.
Bayesian networks use probability distributions to model relationships between factors and predict the probability distribution of structural lifespan. This approach integrates expert knowledge with empirical data to address uncertainties in material properties, environmental exposure, and operational conditions. Guo and Dong [55] developed a dynamic Bayesian network-based framework for assessing reinforced concrete structures, utilizing the finite difference method to focus on the impact of chloride ion transport on damage evaluation. In addition, Guo et al. [56] presented an MBN-based reliability analysis framework for reinforced concrete structures with modules to assess tension, to compute the load-carrying capacity, and to conduct time-dependent reliability calculations. In accordance with their study, it has been concluded that early inspection may overestimate failure probabilities because of the high occurrence of cracks, while neglecting temporal distinctions, environmental effects, and chloride ion transport could result in the underestimation of impending failures.
Statistical models, specifically regression analysis and Bayesian networks, are the most important methods for the prediction of concrete structure lifespan. Regression analysis provides an easy model for the relationship between different factors that affect the life of concrete. Bayesian networks can handle uncertainty in addition to the interaction among numerous variables and, thus, provide more accurate and reliable predictions of lifespan.

3.4. Machine Learning-Based Lifespan Prediction Models

Due to the limitations that are inherent in classic models used for predicting lifespan, especially with respect to complicated environmental factors and large datasets, machine learning algorithms have become effective tools in the area of the longevity prediction of concrete structures. With the computational capabilities of artificial intelligence, these algorithms process nonlinear relationships and large datasets efficiently, thereby enabling more accurate lifespan predictions. In this section, we will go through various popular machine learning models and their uses in predicting concrete lifespan.

3.4.1. Artificial Neural Networks (ANNs)

ANNs are a powerful technique for lifespan prediction, and the reason for this lies in their ability to readily identify nonlinear and complex patterns. The schematic diagram in Figure 3 [57] of a computational neuron in an ANN system is provided with three inputs and a single output. The neuron computes the weighted sum of the inputs as a product of input features I and weights w and compares it to a threshold term called the “bias” to produce the output o. When the sum is greater than or equal to the bias, it fires the output signal; otherwise, it does not.
With the growth of research on neural networks, scientists have increasingly utilized them to estimate the lifespan of concrete structures. Kellouche et al. [58] developed an ANN-based model for the prediction of carbonation depth in fly ash concrete, which accurately predicted it. Chen [59] began an ANN-based new approach to estimate the compressive strength of high-performance concrete and obtained high accuracy. To improve the neural networks’ predictive accuracy, BKA et al. [60] compared various architectures. While the first model with one hidden layer comprising 50 neurons yielded a mean squared error (MSE) of 34.81, the second model with two hidden layers with 50 and 10 neurons each lowered the MSE to 28.33, indicating that the prediction accuracy is improved with an increase in the hidden layers and neurons. Other studies, such as Zhang et al. [61], have expanded ANN applications, proposing an interpretable machine learning framework that combines chloride ion erosion and mechanical aging characteristics to predict the lifespan of medium- and short-span bridges.

3.4.2. Support Vector Machines (SVMs)

SVMs are a key tool for concrete lifespan prediction, particularly due to their ability to handle small sample sizes and high-dimensional data. The basic structure of an SVM is illustrated in Figure 4 [62].
In concrete lifespan prediction, an SVM effectively handles complex nonlinear relationships. Hariri-Ardebili and Pourkamali-Anaraki [63] applied an SVM to assess the structural response of concrete dams. When comparing an SVM with traditional reliability analysis methods, the three commonly used methods are first-order reliability method (FORM), second-order reliability method (SORM), and Monte Carlo simulation (MCS). The failure probability estimated by the SVM closely matched other methods but with higher computational efficiency. Fan [64] proposed a Hybrid Machine Learning (HML) approach, combining critical analysis and an SVM, which is valuable for rapidly evaluating concrete structure damage. Additionally, for predicting concrete strength degradation in complex marine environments, Ling et al. [65] developed an SVM model optimized using K-Fold cross-validation. The results showed that, by optimizing input parameters, the prediction accuracy of the SVM model was significantly improved. Compared to the original model, the average relative error decreased from 34.8% to 27.6%, and the median relative error reduced from 24.7% to 20.8%, further proving the superiority of the optimized SVM model.

3.4.3. Random Forests and Decision Trees

Random forest and decision tree models are pivotal in lifespan prediction due to their ability to perform multivariable analysis and their interpretability. Golafshani et al. [66] applied random forest models to analyze the effects of chloride ion diffusion and environmental exposure on lifespan, thereby enhancing the durability assessment of concrete structures. The model demonstrated strong predictive performance. Random forests are also applied in concrete material design. For example, Liu et al. [67] showcased the potential of random forests in predicting concrete wear resistance and optimizing mix proportions. Concha and Oreta [68] combined random forests with decision trees to assess the impact of rebar corrosion on bond strength, improving the understanding of concrete structure deterioration. Additionally, random forests have been integrated into hybrid models. Ahmad et al. [69] demonstrated how to combine random forests with other machine learning techniques to predict chloride ion concentration accurately, providing a scientific basis for concrete lifespan management. Table 5 summarizes the advantages, disadvantages, and applicability of ANNs, SVMs, random forests, and decision trees.
The machine learning models discussed above (ANN, SVM, random forest, etc.) are capable of handling complex multi-factor data in concrete lifespan prediction. In practice, an ANN is well suited for large-scale, complex, and nonlinear data, while an SVM excels in small sample sizes and high-dimensional datasets. Random forest offers strong multivariable analysis capabilities and high interpretability. However, these models still face challenges, including the need for large datasets, high computational resource requirements, and the risk of overfitting.

4. Model Validation and Prediction

In civil engineering, model validation is required to ensure both scientific accuracy and practical applicability. This section discusses four significant aspects: validation methods, typical evaluation metrics, sensitivity analysis, and uncertainty analysis.

4.1. Validation Methods

Concrete structure durability models are typically examined using two main approaches: experimental validation and field-data-based verification. Experimental validation provides early observations using simulation procedures, including accelerated corrosion testing and freeze–thaw cycling tests. In contrast, field-data-based verification provides a more effective determination of the suitability of the model in real practice in the engineering field.
As an example, Ann et al. [73] reported that concrete that was naturally exposed for a period of 18 years had a model-predicted carbonation depth of 8.3 mm, while the experimental measurement was 11.62 mm with a standard deviation of 2.45 mm, demonstrating a strong correlation between model predictions and empirical observation, which shows near conformity between calculated model values and experimental values. Papadakis and Demis [74] performed an in-depth examination of carbonation depths using various water-to-cement ratios (w/c) and various cement types with very small experimental measurement errors, which provided a mean calculated value of 7.6%. This observation indicates the high accuracy of the model in experimental use. Xu et al. [75] studied Aramid Fiber Reinforced Polymer-reinforced concrete exposed to multiple freeze–thaw cycles in an alkaline environment, where the measured compressive strength was 38.100 Mpa and the back propagation (BP) model predicted 42.263 Mpa. The relative error of 10.927% suggests room for optimization in the model’s prediction stability. While experimental validation can assess model performance under controlled conditions, it cannot fully capture the complexity of actual service environments; thus, further validation with field monitoring data is necessary. For example, Wang et al. [76] proposed a validation framework using both experimental and long-term field monitoring data to assess the durability of concrete structures in marine environments. Similarly, Fan et al. [77] validated the applicability of concrete durability models in complex corrosive environments by integrating experimental and numerical simulation results. To further illustrate the validation based on field data, Table 6 summarizes several cases of concrete durability models validated by field data. These cases demonstrate the comparison between model predictions and actual field data, highlighting the importance of field data in model validation.
The cases in Table 5 clearly demonstrate the differences between model predictions and actual data, proving the importance of field data in validating the accuracy of a model. While laboratory simulations provide initial validation, field data can more accurately assess the applicability of the model in real-world environments. Future research should continue to strengthen the integration of experimental data and field monitoring to improve the models’ predictive capabilities.

4.2. Expansion and Application of Evaluation Metrics

In concrete structure lifespan prediction modeling, the traditional performance measures such as MSE, Coefficient of Determination (R2), Standard Error (SE), and Prediction Error (PE) have been widely utilized. However, existing studies indicate that the sole utilization of a single measure of assessment is not normally adequate to generate an extensive judgment of model performance, especially in the context of complex engineering problems [82,83,84]. To resolve this, researchers have augmented conventional measures by combining various assessment techniques, incorporating precision, stability, and generalizability into models. Although MSE and R2 remain cornerstone measures, Yu et al. [85] found that, in addressing complicated nonlinear relationships, MSE alone may not account for extreme error impacts. MSE is, thus, sometimes combined with other error metrics, such as Absolute Error or Maximum Error. This approach skillfully recognizes outlier values and unusual cases in prediction analyses, particularly in the estimation of concrete structure service life and explaining abrupt changes due to factors like corrosion and cracking. SE is widely used to assess model reliability, mostly in longitudinal research or repeated experiments. For instance, Kostić et al. [86] combined SE with experimental design methods in the analysis of slope stability. The SE ranged from 0.06 to 0.12 through the comparison of the model predictions and experimental values, showing a very small error of prediction and enhancing the stability and reliability of the model.
In addition, Kefei and Torrent [87] presented a Hong Kong-Zhuhai-Macao Bridge concrete structure lifespan prediction model using electric flux and accelerated chloride ion penetration tests. A confidence interval for the lifespan estimation with a concrete cover thickness of SE = 3.9 mm was employed, which indicated the variability range of model inaccuracy under standard exposure conditions. While SE alone can measure deviations of model error, researchers now combine SE with sensitivity analysis to examine the effects of environmental variables (e.g., chloride ion concentration, temperature, and humidity changes) on predictions concerning concrete durability. The hybrid analytical strategy enhances the understanding of the model’s robustness and stability under variable conditions. Sarmiento et al. [88] carried out a reliability analysis of the lifecycle of prestressed concrete bridges that had reached a high age and observed that the Safety Evaluation Index increased from 1.8 × 10−2 to 3.7 × 10−2 in the course of a 50-year service life, illustrating the significant effect of failures in safety-critical elements, such as external steel strands, on the safety of the infrastructure as a whole. Moreover, the structural robustness index fell from 0.74 to 0.60, indicating a decrease of approximately 18.92% in the structure’s ability to withstand cascading failures when damaged. In contrast, relying solely on the reliability index β (from 1.95 to 1.78) reveals changes in overall risk levels but does not capture the effects on local components or system redundancy. Therefore, combined evaluation methods offer greater diagnostic power for identifying structural vulnerabilities and assessing lifecycle safety. Moreover, Carevic and Ignjatovic [89] incorporated various evaluation metrics into their green concrete carbonation prediction model, predicting the service life by establishing a relationship between compressive strength and carbonation resistance.

4.3. Sensitivity Analysis and Uncertainty Analysis

Both sensitivity analysis and uncertainty analysis are crucial in life prediction models of concrete structures for determining model performance. Sensitivity analysis entails the identification of critical input variables by examining how changes in input parameters affect the output and, hence, allowing for the identification of parameters that have a notable effect on accuracy in predictions of lifespan. Conversely, uncertainty analysis displays the risk and stability range of model predictions, thereby offering insights into model reliability and associated risks. For example, Yu et al. [90] discovered that cover depth is a critical parameter influencing durability. An increase in cover depth from 40 mm to 70 mm increased the 100-year reliability index β from 0.3696 to 2.1042, which implies that a higher cover depth considerably enhances structural durability and safety. This, again, underscores the significance of cover thickness in lifespan-related predictions. In addition, Guo et al. [56] utilized a Hybrid Bayesian Network model to investigate the sensitivity of key parameters in relation to the reliability of reinforced concrete structures in different environments. Strauss et al. [91] proved that sensitivity analysis based on monitoring data together with probabilistic prediction models more precisely identifies the critical variables affecting the durability of concrete.
Prediction uncertainty is a critical issue in engineering practice. Monte Carlo simulation is a commonly used method to address this uncertainty. For instance, Franceschini et al. [92] proposed a method to evaluate the remaining service life of prestressed concrete beams under corrosion conditions. The study focused on the failure mechanisms caused by corrosion and quantitatively assessed the structural degradation process using statistical methods. Additionally, Vishwanath and Banerjee [93] enhanced the reliability of bridge pier lifespan prediction by introducing the uncertainty analysis of chloride corrosion when evaluating the seismic vulnerability of aging bridge piers. Cho et al. [94] combined random variable simulations in the study of a fly ash concrete carbonation, systematically quantifying the confidence intervals of model predictions. Furthermore, Taffese and Espinosa-Leal [33] used machine learning techniques to assess the confidence intervals of concrete durability prediction models, demonstrating that, in large-scale data-driven contexts, the uncertainty in model predictions can be significantly reduced.

5. Application Cases of Models

With the increasing complexity and variability of engineering environments, traditional single models are becoming insufficient. In modern engineering practice, many models combine various factors such as environmental influences, load variations, and material degradation. This is particularly crucial for lifespan prediction in key infrastructure like bridges, tunnels, and high-rise buildings, where model accuracy and adaptability are essential. Below, we review the practical applications of representative lifespan prediction models, categorized by typical civil engineering structures such as bridges, underground facilities, and high-rise buildings. As a vital part of transportation infrastructure, bridges bear substantial loads and long-term usage pressure, making the accuracy of their lifespan prediction models critical. Ultra-high-performance concrete (UHPC) is widely applied in bridge construction. For example, Shafikhani and Chidiac [95] measured the chloride ion diffusion coefficient through concrete materials. The diffusion coefficient for UHPC ranged from 6 × 10−15 m2/s to 5.9 × 10−13 m2/s, while for ordinary concrete, it ranged from 2 × 10−13 m2/s to 7.97 × 10−12 m2/s, demonstrating UHPC’s superior corrosion resistance. Additionally, Liang et al. [96] analyzed the chloride-induced lifespan of the Linbian and Dajin bridges in Keelung, comparing the lifespan predictions of three models: the Langmuir isotherm model, the linear binding model, and the Fick diffusion model [97,98]. In low-chloride conditions, the Langmuir isotherm model predicted a lifespan of 33.12 years for the bridge deck, which was higher than the linear binding model (31.55 years) and the Fick diffusion model (31.54 years).
The findings indicate that the Langmuir isotherm model predicts lifespan and longevity more precisely than other models, particularly in low-chloride-concentration environments. Moreover, Xue and Mao [99] modified the chloride ion diffusion model by introducing variables such as the time-dependent decay of the diffusion coefficient, chloride ion binding capacity, type of concrete, and service environment. The researchers confirmed the model’s accuracy with an application study on the Huangpu Bridge, proving that environmental parameters have a great impact on chloride ion diffusion rates and the corrosion process. The revised model provides a more precise method for predicting bridge durability and service life. Chen et al. [100] performed a lifecycle cost analysis on reinforced concrete bridge edge beams, effectively controlling crack width and improving resistance to chloride ion corrosion, resulting in a 58% extension of edge beam lifespan. Yamamoto et al. [101] found that the relationship between the apparent chloride ion diffusion coefficient and the air permeability coefficient showed different diffusion trends in road bridges exposed to deicing salts and in marine environments or laboratory tests. Their study highlighted the critical role of environmental factors in concrete evaluation. Furthermore, Li and Feng [102] studied the fatigue life of bridge cables, validating the model’s reliability using real-world monitoring data. Fathalla et al. [103] proposed a fatigue life prediction model for concrete bridge decks, further advancing bridge lifespan evaluation methods. Underground facilities are subjected to complex environmental conditions, which increase the demands on lifespan prediction models. Balestra et al. [104] introduced a nonlinear lifespan model that incorporates environmental factors such as chloride concentration, temperature, and humidity to predict the corrosion rate of underground structures. For example, the corrosion of buried cast iron pipes is a complex process [105]. Li and Mahmoodian [106], using corrosion data from pipes in the UK that have been in service for up to 150 years, combined a corrosion pit growth model to calculate the failure probability of pipes with varying diameters and wall thicknesses under both internal and external corrosion effects. Dasar et al. [107] experimentally investigated the natural corrosion of beams with different cover layers and rebar diameters, validating the long-term corrosion performance prediction accuracy of their model. Chen et al. [108] applied a nonlinear chloride ion transport model to simulate chloride ion intrusion in underwater concrete tunnels, and the simulation results closely matched experimental data, providing a useful tool for assessing chloride ion corrosion in cylindrical tunnels. Additionally, Keßler [109] proposed a probabilistic corrosion state assessment method for tunnel structures based on half-cell potential measurements, significantly improving lifespan prediction accuracy. High-rise buildings, airport runways, and similar infrastructure are subjected to heavy loads and complex environmental conditions. Meng et al. [110] evaluated the performance of nano-silica-modified recycled concrete in marine environments and proposed a lifespan prediction model based on material modification for environmental adaptation. Canestro et al. [111] developed a multiscale modeling method to accurately assess the long-term performance of prestressed concrete structures. For airport runway lifespan prediction, Zheng and Pornsing [112] introduced a novel empirical optimization light-gradient-boosting machine model, enhancing prediction accuracy through detailed parameter optimization, feature selection, and model integration. Mortagi and Ghosh [113] presented a vulnerability assessment method for highway bridges under climate change scenarios, offering a new perspective for building lifespan prediction.

6. Analysis of Advantages and Disadvantages of Existing Models

In predicting the lifespan of concrete structures, different models offer distinct advantages and limitations. The selection of an appropriate model depends on the specific engineering environment, data availability, and prediction needs.
Table 7 provides an analysis of the advantages and disadvantages of several commonly used models.
In summary, each model has its own strengths and limitations in practical applications. To achieve more accurate predictions, it is essential to choose the most appropriate model based on the specific needs of the project. With technological advancements, hybrid models and machine learning models have proven especially advantageous in handling complex problems, particularly for real-time monitoring and dynamic updates. The continuous improvement and integration of these models will help better address the diverse needs of engineering projects.

7. Research Progress and Development Trends

As civil engineering technology continues to advance, concrete structure lifespan prediction models are progressing toward greater intelligence and precision. Recent developments, particularly in the application of high-performance materials and the rapid development of intelligent technologies, have significantly enhanced the accuracy and reliability of predictions of the life of concrete. Jeon et al. [114] presented a maintenance strategy based on a health index for concrete bridges, which combines real-time monitoring data with changes in environmental conditions, thus greatly increasing prediction accuracy.
Multiscale modeling has become a necessity for lifespan prediction. Ying et al. [115], for example, employed multiscale modeling to examine the behavior of concrete under high-temperature conditions and revealed the influences of environmental conditions on the aging process. Through integrated multiscale analysis approaches, researchers are able to analyze concrete material aging and its influence on the overall structure performance more accurately. Kaveh and Alhajj [116] proposed a multiscale model based on the development of microcracks, emphasizing the central role of crack evolution in the life of concrete. Wang and Maekawa [117] developed a multiscale modeling framework that predicts the life of concrete based on moisture transport, temperature gradients, and electrochemical coupling under different environmental conditions.
The fast evolution of data-driven technologies has led to the increased application of intelligent approaches in predicting the life of concrete. The combination of machine learning and big data analytics not only improves the accuracy of predictions, but also facilitates real-time monitoring. Azimi et al. [118] conducted a review of the latest advances of deep learning for structural health monitoring, pointing out that the coupling of real-time data with machine learning improves diagnostic capability, thereby rendering life predictions more accurate.
In the future, concrete structure life prediction models will continue to evolve with the help of even more advanced technologies. Incorporating smart technologies, real-time feedback, and multivariable modeling approaches is particularly of interest, as this will significantly enhance model precision and ease of use. With the deeper integration of interdisciplinary technologies, future research will focus on enhancing prediction accuracy through real-time data feedback and dynamic updates, providing more accurate and efficient support for the maintenance of civil engineering projects.

8. Conclusions

Concrete structures find extensive applications in civil engineering, and it is essential to guarantee their long-term durability and safety for public welfare and the effective utilization of resources. With changing environmental conditions and the degradation of materials under the influence of a multitude of factors, the durability assessment of concrete infrastructures is a critical problem. This review presents a summary of lifespan prediction models such as physical, empirical, statistical, and machine learning models and evaluates their applicability and limitations in accordance with different environmental conditions.
With the development of technology, models based on real-time monitoring data and integrating multiple factors are given more prominence in research forums. While physical models offer a good theoretical model, they are computationally intensive and restrictive when used in complex environments. Empirical models are extremely simple and interactive but are highly reliant on historical data and could overlook certain elements of environmental complexity. Statistical modeling based on regression analysis and Bayesian networks is aptly suited for addressing complex environments and diverse uncertainties but requires large amounts of data. Machine learning models such as artificial neural networks (ANNs) and support vector machines (SVMs) exhibit an enhanced capability of addressing nonlinear complexities and can offer enhanced predictive accuracy when supplemented with large datasets.
Despite these advances, current models still face many challenges, particularly with respect to updating data in real time, dynamic model optimization, and the integration of various influencing factors. Future studies can focus on the following areas:
  • Combining real-time monitoring data with advanced machine learning algorithms to improve prediction accuracy and robustness;
  • Exploring hybrid models that integrate the strengths of both physical and data-driven models to enhance adaptability in complex environments;
  • Developing more precise nonlinear and multiscale modeling techniques to accommodate environmental and material variations;
  • Optimizing model validation and evaluation methods to improve their reliability and applicability in practical engineering projects;
  • By combining experimental data with field monitoring data, future research should continue to explore how to effectively integrate these two sources in order to further optimize the predictive capability of models and improve the accuracy of concrete structure durability assessments.
By continuously advancing technological innovations and improving model capabilities, future concrete structure lifespan prediction models will be more accurate and efficient, providing stronger support for structural health management in civil engineering.

Funding

This work was supported by the Key Projects of Yongjiang Science and Technology Innovation 2035 in Ningbo (2024Z090) and the Major Science and Technology Project of Ningbo High-tech Zone in 2023 (2023CX050001).

Conflicts of Interest

Authors Xi Peng and Bin Xu were employed by the company Ningbo Roaby Technology Industrial Group Co., Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

References

  1. Imbabi, M.S.; Carrigan, C.; McKenna, S. Trends and developments in green cement and concrete technology. Int. J. Sustain. Built Environ. 2012, 1, 194–216. [Google Scholar] [CrossRef]
  2. Shi, X.; Xie, N.; Fortune, K.; Gong, J. Durability of steel reinforced concrete in chloride environments: An overview. Constr. Build. Mater. 2012, 30, 125–138. [Google Scholar] [CrossRef]
  3. Athibaranan, S.; Karthikeyan, J.; Rawat, S. Investigation on service life prediction models of reinforced concrete structures exposed to chloride laden environment. J. Build. Pathol. Rehabil. 2022, 7, 1–15. [Google Scholar] [CrossRef]
  4. Peng, R.; Qiu, W.; Jiang, M. Frost resistance performance assessment of concrete structures under multi-factor coupling in cold offshore environment. Build. Environ. 2022, 226, 109733. [Google Scholar] [CrossRef]
  5. Zhang, K.; Guo, A.; Yu, Y.; Yang, B.; Yu, B.; Xie, C. Freeze–thaw damage degradation model and life prediction of air-entrained concrete in multi-year permafrost zone. Materials 2023, 16, 7703. [Google Scholar] [CrossRef]
  6. Chen, J.; Li, Y.; Li, Y.; Wen, L.; Guo, H. Effects of curing conditions with different temperature and humidity on damage evolution of concrete during freeze–thaw cycling. Mater. Struct. 2022, 55, 80. [Google Scholar] [CrossRef]
  7. Zhang, X.G.; Wang, J. Probabilistic Model of Corrosion in Reinforced Concrete Structures Considering Coupling Effects of Influence Factors. Adv. Mater. Res. 2013, 671, 740–745. [Google Scholar] [CrossRef]
  8. Zhang, P.; Cong, Y.; Vogel, M.; Liu, Z.; Müller, H.S.; Zhu, Y.; Zhao, T. Steel reinforcement corrosion in concrete under combined actions: The role of freeze-thaw cycles, chloride ingress, and surface impregnation. Constr. Build. Mater. 2017, 148, 113–121. [Google Scholar] [CrossRef]
  9. Zhou, L.; Pan, S.; Wang, J.; Vasilakos, A.V. Machine learning on big data: Opportunities and challenges. Neurocomputing 2017, 237, 350–361. [Google Scholar] [CrossRef]
  10. Dinesh, A.; Prasad, B.R. Predictive models in machine learning for strength and life cycle assessment of concrete structures. Autom. Constr. 2024, 162, 105412. [Google Scholar] [CrossRef]
  11. Tian, Y.; Zhang, G.; Ye, H.; Zeng, Q.; Zhang, Z.; Tian, Z.; Jin, X.; Jin, N.; Chen, Z.; Wang, J. Corrosion of steel rebar in concrete induced by chloride ions under natural environments. Constr. Build. Mater. 2023, 369, 130504. [Google Scholar] [CrossRef]
  12. Ju, X.; Wu, L.; Lin, C.; Yang, X.; Yang, C. Prediction of chloride concentration with elevation in concrete exposed to cyclic drying-wetting conditions in marine environments. Constr. Build. Mater. 2021, 278, 122370. [Google Scholar] [CrossRef]
  13. Zhang, W.; Chen, J.; Yu, Q.; Gu, X. Corrosion evolution of steel bars in RC structures based on Markov chain modeling. Struct. Saf. 2021, 88, 102037. [Google Scholar] [CrossRef]
  14. Sun, B.; Xiao, R.; Ruan, W.; Wang, P. Corrosion-induced cracking fragility of RC bridge with improved concrete carbonation and steel reinforcement corrosion models. Eng. Struct. 2020, 208, 110313. [Google Scholar] [CrossRef]
  15. Zhang, Y.; Su, R.K.L. Corner cracking model for non-uniform corrosion-caused deterioration of concrete covers. Constr. Build. Mater. 2020, 234, 117410. [Google Scholar] [CrossRef]
  16. Wang, Y.; Gong, X.; Wu, L. Prediction model of chloride diffusion in concrete considering the coupling effects of coarse aggregate and steel reinforcement exposed to marine tidal environment. Constr. Build. Mater. 2019, 216, 40–57. [Google Scholar] [CrossRef]
  17. Hájková, K.; Šmilauer, V.; Jendele, L.; Červenka, J. Prediction of reinforcement corrosion due to chloride ingress and its effects on serviceability. Eng. Struct. 2018, 174, 768–777. [Google Scholar] [CrossRef]
  18. Wang, W.; Lu, C. Time-varying law of rebar corrosion rate in fly ash concrete. J. Hazard. Mater. 2018, 360, 520–528. [Google Scholar] [CrossRef]
  19. Ann, K.Y.; Ahn, J.H.; Ryou, J.S. The importance of chloride content at the concrete surface in assessing the time to corrosion of steel in concrete structures. Constr. Build. Mater. 2009, 23, 239–245. [Google Scholar] [CrossRef]
  20. Zhou, Y.; Gencturk, B.; Willam, K.; Attar, A. Carbonation-induced and chloride-induced corrosion in reinforced concrete structures. J. Mater. Civ. Eng. 2015, 27, 04014245. [Google Scholar] [CrossRef]
  21. Saeki, T. Effect of carbonation on chloride penetration in concrete. In Proceedings of the Third International RILEM Workshop on Testing and Modelling Chloride Ingress into Concrete, Madrid, Spain, 9–10 September 2002; pp. 1–14. [Google Scholar]
  22. Zhou, Y.; Tian, H.; Sui, L.; Xing, F.; Han, N. Strength deterioration of concrete in sulfate environment: An experimental study and theoretical modeling. Adv. Mater. Sci. Eng. 2015, 2015, 951209. [Google Scholar] [CrossRef]
  23. Scott, A.; Alexander, M.G. The influence of binder type, cracking and cover on corrosion rates of steel in chloride-contaminated concrete. Mag. Concr. Res. 2007, 59, 495–505. [Google Scholar] [CrossRef]
  24. Guo, Z.; Wang, L.; Ma, Y.; Wu, B.; Ren, L.; Gui, S.; Su, X. Probabilistic fatigue life prediction of corroded PC beams considering spatial variability of pitting corrosion. Eng. Struct. 2024, 306, 117721. [Google Scholar] [CrossRef]
  25. Wu, J.; Xu, J.; Diao, B.; Jin, L.; Du, X. Fatigue life prediction for the reinforced concrete (RC) beams under the actions of chloride attack and fatigue. Eng. Struct. 2021, 242, 112543. [Google Scholar] [CrossRef]
  26. Ge, B.; Kim, S. Probabilistic service life prediction updating with inspection information for RC structures subjected to coupled corrosion and fatigue. Eng. Struct. 2021, 238, 112260. [Google Scholar] [CrossRef]
  27. Bhargava, K.; Ghosh, A.K.; Mori, Y.; Ramanujam, S. Model for cover cracking due to rebar corrosion in RC structures. Eng. Struct. 2006, 28, 1093–1109. [Google Scholar] [CrossRef]
  28. Choe, D.E.; Gardoni, P.; Rosowsky, D.; Haukaas, T. Probabilistic capacity models and seismic fragility estimates for RC columns subject to corrosion. Reliab. Eng. Syst. Saf. 2008, 93, 383–393. [Google Scholar] [CrossRef]
  29. Jianyu, Y.; Xin, Y.; Cong, J.; Weijun, Y. Effect of NaCl solution and simulated concrete pore solution environment on the efficiency of steel bar energized corrosion. Materials 2022, 15, 7040. [Google Scholar] [CrossRef] [PubMed]
  30. Yu, H.; Li, L.; Ma, H.; Lu, Y.; Hu, D. Study on time range of definite integral in derivation process of chloride ion diffusion theoretical model of concrete. Constr. Build. Mater. 2024, 428, 135876. [Google Scholar] [CrossRef]
  31. Chen, D.; Guo, W.; Quan, X.; Duan, B.; Guo, L. Mesoscopic statistics-based probability characteristics of chloride transport and reliability-based corrosion initiation life of bridge tower. Thin-Walled Struct. 2024, 198, 111680. [Google Scholar] [CrossRef]
  32. Guo, Z.; Guo, R.; Lin, S. Multi-factor fuzzy prediction model of concrete surface chloride concentration with trained samples expanded by random forest algorithm. Mar. Struct. 2022, 86, 103311. [Google Scholar] [CrossRef]
  33. Taffese, W.Z.; Espinosa-Leal, L. Prediction of chloride resistance level of concrete using machine learning for durability and service life assessment of building structures. J. Build. Eng. 2022, 60, 105146. [Google Scholar] [CrossRef]
  34. Liu, J.; Liao, C.; Jin, H.; Jiang, Z.; Zhou, H. Numerical and experimental research on the effect of rainfall on the transporting behavior of chloride ions in concrete. Constr. Build. Mater. 2021, 302, 124160. [Google Scholar] [CrossRef]
  35. Shao, W.; Nie, Y.; Liang, F.; Shi, D. A novel comprehensive evaluation method for the corrosion initiation life of RC hollow piles in chloride environments. Constr. Build. Mater. 2020, 249, 118801. [Google Scholar] [CrossRef]
  36. Safehian, M.; Ramezanianpour, A.A. Assessment of service life models for determination of chloride penetration into silica fume concrete in the severe marine environmental condition. Constr. Build. Mater. 2013, 48, 287–294. [Google Scholar] [CrossRef]
  37. Papadakis, V.G.; Vayenas, C.G.; Fardis, M.N. Fundamental modeling and experimental investigation of concrete carbonation. Mater. J. 1991, 88, 363–373. [Google Scholar]
  38. Wang, D.; Tan, Q.; Wang, Y.; Liu, G.; Lu, Z.; Zhu, C.; Sun, B. Carbonation depth prediction and parameter influential analysis of recycled concrete buildings. J. CO2 Util. 2024, 85, 102877. [Google Scholar] [CrossRef]
  39. Cao, H.; Xu, Z.; Peng, X. An Improved Fick Model for Predicting Carbonation Depth of Concrete. Coatings 2024, 14, 1345. [Google Scholar] [CrossRef]
  40. Yu, H.; Sun, W.; Ma, H. Research on Theoretical Models of Chloride Ion Diffusion in Concrete I—Non-steady State Homogeneous and Non-homogeneous Diffusion Problems Based on Infinite Bodies. Nanjing Univ. Aeronaut. Astronaut. 2009, 41, 276–280. [Google Scholar]
  41. Yu, H.; Sun, W.; Ma, H. Research on Theoretical Models of Chloride Ion Diffusion in Concrete II—Non-steady State Homogeneous and Non-homogeneous Diffusion Problems Based on Finite Bodies. Nanjing Univ. Aeronaut. Astronaut. 2009, 41, 408–413. [Google Scholar]
  42. Chen, W.; Zhang, J.; Zhang, J. A variable-order time-fractional derivative model for chloride ions sub-diffusion in concrete structures. Fract. Calc. Appl. Anal. 2013, 16, 76–92. [Google Scholar] [CrossRef]
  43. Lu, Z.H.; Lun, P.Y.; Li, W.; Luo, Z.; Li, Y.; Liu, P. Empirical model of corrosion rate for steel reinforced concrete structures in chloride-laden environments. Adv. Struct. Eng. 2019, 22, 223–239. [Google Scholar] [CrossRef]
  44. Chen, X.; Ming, Y.; Fu, F.; Chen, P. Numerical and empirical modeling for the assessment of service life of RC structures in marine environment. Int. J. Concr. Struct. Mater. 2022, 16, 11. [Google Scholar] [CrossRef]
  45. Mangat, P.S.; Molloy, B.T. Prediction of free chloride concentration in concrete using routine inspection data. Mag. Concr. Res. 1994, 46, 279–287. [Google Scholar] [CrossRef]
  46. Mangat, P.S.; Molloy, B.T. Prediction of long term chloride concentration in concrete. Mater. Struct. 1994, 27, 338–346. [Google Scholar] [CrossRef]
  47. Luping, T.; Gulikers, J. On the mathematics of time-dependent apparent chloride diffusion coefficient in concrete. Cem. Concr. Res. 2007, 37, 589–595. [Google Scholar] [CrossRef]
  48. Ma, J.; Yang, Q.; Wang, X.; Peng, X.; Qin, F. Review of Prediction Models for Chloride Ion Concentration in Concrete Structures. Buildings 2025, 15, 149. [Google Scholar] [CrossRef]
  49. Ashrafian, A.; Panahi, E.; Salehi, S.; Amiri, M.J.T. On the implementation of the interpretable data-intelligence model for designing service life of structural concrete in a marine environment. Ocean. Eng. 2022, 256, 111523. [Google Scholar] [CrossRef]
  50. Wang, Y.; Wu, L.; Wang, Y.; Li, Q.; Xiao, Z. Prediction model of long-term chloride diffusion into plain concrete considering the effect of the heterogeneity of materials exposed to marine tidal zone. Constr. Build. Mater. 2018, 159, 297–315. [Google Scholar] [CrossRef]
  51. Prieto, M.; Tanner, P.; Andrade, C. Multiple linear regression model for the assessment of bond strength in corroded and non-corroded steel bars in structural concrete. Mater. Struct. 2016, 49, 4749–4763. [Google Scholar] [CrossRef]
  52. Jin, R.; Chen, Q.; Soboyejo, A.B.O. Non-linear and mixed regression models in predicting sustainable concrete strength. Constr. Build. Mater. 2018, 170, 142–152. [Google Scholar] [CrossRef]
  53. Hanif, M.U.; Ibrahim, Z.; Ghaedi, K.; Hashim, H.; Javanmardi, A. Damage assessment of reinforced concrete structures using a model-based nonlinear approach–A comprehensive review. Constr. Build. Mater. 2018, 192, 846–865. [Google Scholar] [CrossRef]
  54. Atienza, D.; Larrañaga, P.; Bielza, C. Hybrid semiparametric Bayesian networks. TEST 2022, 31, 299–327. [Google Scholar] [CrossRef]
  55. Guo, H.; Dong, Y. Dynamic Bayesian network for durability of reinforced concrete structures in long-term environmental exposures. Eng. Fail. Anal. 2022, 142, 106821. [Google Scholar] [CrossRef]
  56. Guo, H.; Dong, Y.; Bastidas-Arteaga, E. Mixed Bayesian Network for reliability assessment of RC structures subjected to environmental actions. Struct. Saf. 2024, 106, 102392. [Google Scholar] [CrossRef]
  57. Aneja, S.; Sharma, A.; Gupta, R.; Yoo, D.Y. Bayesian regularized artificial neural network model to predict strength characteristics of fly-ash and bottom-ash based geopolymer concrete. Materials 2021, 14, 1729. [Google Scholar] [CrossRef]
  58. Kellouche, Y.; Ghrici, M.; Boukhatem, B. Service life prediction of fly ash concrete using an artificial neural network. Front. Struct. Civ. Eng. 2021, 15, 793–805. [Google Scholar] [CrossRef]
  59. Chen, L. Hybrid structured artificial network for compressive strength prediction of HPC concrete. J. Appl. Sci. Eng. 2022, 26, 989–999. [Google Scholar]
  60. BKA, M.A.R.; Ngamkhanong, C.; Wu, Y.; Kaewunruen, S. Recycled aggregates concrete compressive strength prediction using artificial neural networks (ANNs). Infrastructures 2021, 6, 17. [Google Scholar] [CrossRef]
  61. Zhang, B.; Wang, K.; Lu, G.; Guo, W.; Liu, J.; Zhang, N.; Yang, C. Life-cycle seismic performance analysis of an offshore small-to-medium span bridge based on interpretable machine learning. In Structures; Elsevier: Amsterdam, The Netherlands, 2024; Volume 70, p. 107511. [Google Scholar]
  62. Wang, X.; Yang, Q.; Peng, X.; Qin, F. A review of concrete carbonation depth evaluation models. Coatings 2024, 14, 386. [Google Scholar] [CrossRef]
  63. Hariri-Ardebili, M.A.; Pourkamali-Anaraki, F. Support vector machine based reliability analysis of concrete dams. Soil Dyn. Earthq. Eng. 2018, 104, 276–295. [Google Scholar] [CrossRef]
  64. Fan, C.L. Detection of multidamage to reinforced concrete using support vector machine-based clustering from digital images. Struct. Control. Health Monit. 2021, 28, e2841. [Google Scholar] [CrossRef]
  65. Ling, H.; Qian, C.; Kang, W.; Liang, C.; Chen, H. Combination of Support Vector Machine and K-Fold cross validation to predict compressive strength of concrete in marine environment. Constr. Build. Mater. 2019, 206, 355–363. [Google Scholar] [CrossRef]
  66. Golafshani, E.M.; Kashani, A.; Arashpour, M. Chloride diffusion modeling of concrete using tree-based forest models. Struct. Concr. 2023, 24, 5614–5634. [Google Scholar] [CrossRef]
  67. Liu, Q.; Andersen, L.V.; Wu, M. Prediction of concrete abrasion depth and computational design optimization of concrete mixtures. Cem. Concr. Compos. 2024, 148, 105431. [Google Scholar] [CrossRef]
  68. Concha, N.C.; Oreta, A.W.C. Investigation of the effects of corrosion on bond strength of steel in concrete using neural network. Comput. Concr. 2021, 28, 77–91. [Google Scholar]
  69. Ahmad, A.; Farooq, F.; Ostrowski, K.A.; Śliwa-Wieczorek, K.; Czarnecki, S. Application of novel machine learning techniques for predicting the surface chloride concentration in concrete containing waste material. Materials 2021, 14, 2297. [Google Scholar] [CrossRef]
  70. Ding, S.; Li, H.; Su, C.; Yu, J.; Jin, F. Evolutionary artificial neural networks: A review. Artif. Intell. Rev. 2013, 39, 251–260. [Google Scholar] [CrossRef]
  71. Rothacher, Y.; Strobl, C. Identifying informative predictor variables with random forests. J. Educ. Behav. Stat. 2024, 49, 595–629. [Google Scholar] [CrossRef]
  72. Karbassi, A.; Mohebi, B.; Rezaee, S.; Lestuzzi, P. Damage prediction for regular reinforced concrete buildings using the decision tree algorithm. Comput. Struct. 2014, 130, 46–56. [Google Scholar] [CrossRef]
  73. Ann, K.Y.; Pack, S.W.; Hwang, J.P.; Song, H.W.; Kim, S. Service life prediction of a concrete bridge structure subjected to carbonation. Constr. Build. Mater. 2010, 24, 1494–1501. [Google Scholar] [CrossRef]
  74. Papadakis, V.G.; Demis, S. Estimation and validation of concrete strength and service life using software packages based on predictive models. In Proceedings of the 12th International Conference on Durability of Building Materials and Components, Porto, Portugal, 12–15 April 2011. [Google Scholar]
  75. Xu, W.; Li, W.; Wang, D.; Ji, Y. Mechanical performance degradation investigation on FRP reinforced concrete based on neural network design method. Sci. Rep. 2024, 14, 11807. [Google Scholar] [CrossRef]
  76. Wang, S.; Wu, R.; Gong, F.; Xia, J.; Zhao, Y.; Zeng, B. Service life evaluation of marine concrete structures considering spatial and temporal characteristics: A framework based on multi training-MCS-NLS. Eng. Struct. 2025, 322, 119193. [Google Scholar] [CrossRef]
  77. Fan, J.; Zhu, H.; Yi, C.; Li, Z. Research on durability evaluation of concrete in complex attack environment. AIP Adv. 2019, 9, 115019. [Google Scholar] [CrossRef]
  78. Nanukuttan, S.V.; Basheer, P.A.M.; McCarter, W.J.; Tang, L.; Holmes, N.; Chrisp, T.M.; Starrs, G.; Magee, B. The performance of concrete exposed to marine environments: Predictive modelling and use of laboratory/on site test methods. Constr. Build. Mater. 2015, 93, 831–840. [Google Scholar] [CrossRef]
  79. Pang, L.; Li, Q. Service life prediction of RC structures in marine environment using long term chloride ingress data: Comparison between exposure trials and real structure surveys. Constr. Build. Mater. 2016, 113, 979–987. [Google Scholar] [CrossRef]
  80. Al-Obaidi, S.; Bamonte, P.; Luchini, M.; Mazzantini, I.; Ferrara, L. Durability-based design of structures made with ultra-high-performance/ultra-high-durability concrete in extremely aggressive scenarios: Application to a geothermal water basin case study. Infrastructures 2020, 5, 102. [Google Scholar] [CrossRef]
  81. Chen, S.; Duffield, C.; Miramini, S.; Raja, B.N.K.; Zhang, L. Life-cycle modelling of concrete cracking and reinforcement corrosion in concrete bridges: A case study. Eng. Struct. 2021, 237, 112143. [Google Scholar] [CrossRef]
  82. Sharp, M.E. Simple metrics for evaluating and conveying prognostic model performance to users with varied backgrounds. In Proceedings of the Annual Conference of the PHM Society, San Diego, CA, USA, 27 September–1 October 2013; Volume 5. [Google Scholar]
  83. Bennett, N.D.; Croke, B.F.W.; Guariso, G.; Guillaume, J.H.A.; Hamilton, S.H.; Jakeman, A.J.; Marsili-Libelli, S.; Newham, L.T.H.; Norton, J.P.; Perrin, C.; et al. Characterising performance of environmental models. Environ. Model. Softw. 2013, 40, 1–20. [Google Scholar] [CrossRef]
  84. Gupta, H.V.; Clark, M.P.; Vrugt, J.A.; Abramowitz, G.; Ye, M. Towards a comprehensive assessment of model structural adequacy. Water Resour. Res. 2012, 48, W08301. [Google Scholar] [CrossRef]
  85. Yu, J.; Wang, N.; Wang, M.; Zhang, J. Hou, D. Recyclable rubber-cement composites produced by interfacial strengthened strategy from polyvinyl alcohol. Constr. Build. Mater. 2020, 264, 120541. [Google Scholar] [CrossRef]
  86. Kostić, S.; Vasović, N.; Sunarić, D. Slope stability analysis based on experimental design. Int. J. Geomech. 2016, 16, 04016009. [Google Scholar] [CrossRef]
  87. Kefei, L.; Torrent, R. Analytical and Experimental Service Life Assessment of Hong Kong-Zhuhai-Macau Link; Civil Engng. Dept., Tsinghua University: Beijing, China, 2016. [Google Scholar]
  88. Sarmiento, S.; Thöns, S.; Gonzalez, J.; Björnsson, I.; Sas, G. Comparison and Joint Evaluation of Importance, Redundancy and Robustness Indicators Applied to Aging Prestressed Concrete Bridges. Nord. Concr. Res. 2024, 70, 147–171. [Google Scholar] [CrossRef]
  89. Carević, V.; Ignjatović, I. Evaluation of concrete cover depth for green concretes exposed to carbonation. Struct. Concr. 2021, 22, E1009–E1021. [Google Scholar] [CrossRef]
  90. Yu, B.; Ning, C.; Li, B. Probabilistic durability assessment of concrete structures in marine environments: Reliability and sensitivity analysis. China Ocean. Eng. 2017, 31, 63–73. [Google Scholar] [CrossRef]
  91. Strauss, A.; Vidovic, A.; Zambon, I.; Grossberger, H.; Bergmeister, K. Monitoring information and probabilistic-based prediction models for the performance assessment of concrete structures. J. Perform. Constr. Facil. 2016, 30, 04015081. [Google Scholar] [CrossRef]
  92. Franceschini, L.; Vecchi, F.; Belletti, B.; Andrade, C.; Peiretti, H.C. Analytical method for the evaluation of the residual service life of prestressed concrete beams subjected to corrosion deterioration. Struct. Concr. 2022, 23, 121–137. [Google Scholar] [CrossRef]
  93. Vishwanath, B.S.; Banerjee, S. Considering uncertainty in corrosion process to estimate life-cycle seismic vulnerability and risk of aging bridge piers. Reliab. Eng. Syst. Saf. 2023, 232, 109050. [Google Scholar] [CrossRef]
  94. Cho, H.K.; Lee, H.S.; Wang, X.Y.; Ismail, M.; Park, W.J. Evaluation of CO2 emission–absorption of fly-ash-blended concrete structures using cement-hydration-based carbonation model. Mater. Struct. 2015, 48, 3949–3963. [Google Scholar] [CrossRef]
  95. Shafikhani, M.; Chidiac, S.E. Quantification of concrete chloride diffusion coefficient—A critical review. Cem. Concr. Compos. 2019, 99, 225–250. [Google Scholar] [CrossRef]
  96. Liang, M.T.; Chang, J.J.; Yu, T.W. Studies of the effect of diffusion-induced chloride binding on chlorination life predictions for existing reinforced concrete bridges. J. Mar. Sci. Technol. 2012, 20, 10. [Google Scholar]
  97. Martın-Pérez, B.; Zibara, H.; Hooton, R.D.; Thomas, M.D.A. A study of the effect of chloride binding on service life predictions. Cem. Concr. Res. 2000, 30, 1215–1223. [Google Scholar] [CrossRef]
  98. Sergi, G.; Yu, S.W.; Page, C.L. Diffusion of chloride and hydroxyl ions in cementitious materials exposed to a saline environment. Mag. Concr. Res. 1992, 44, 63–69. [Google Scholar] [CrossRef]
  99. Xue, P.F.; Mao, D.L. Corrected diffusion model of chloride in concrete and its engineering application. Adv. Mater. Res. 2011, 189, 1886–1891. [Google Scholar] [CrossRef]
  100. Chen, E.; Berrocal, C.G.; Löfgren, I.; Lundgren, K. Comparison of the service life, life-cycle costs and assessment of hybrid and traditional reinforced concrete through a case study of bridge edge beams in Sweden. Struct. Infrastruct. Eng. 2022, 19, 39–57. [Google Scholar] [CrossRef]
  101. Yamamoto, M.; Nakarai, K.; Yoshizumi, Y.; Torrent, R. Relationship between chloride ingress and concrete cover quality of inland structures exposed to deicing salts. Case Stud. Constr. Mater. 2024, 20, e03075. [Google Scholar] [CrossRef]
  102. Li, J.A.; Feng, D. Fatigue life evaluation of bridge stay cables subject to monitoring traffic and considering road roughness. Eng. Struct. 2023, 293, 116572. [Google Scholar] [CrossRef]
  103. Fathalla, E.; Tanaka, Y.; Maekawa, K. Fatigue lifetime prediction of newly constructed RC road bridge decks. J. Adv. Concr. Technol. 2019, 17, 715–727. [Google Scholar] [CrossRef]
  104. Balestra, C.E.T.; Reichert, T.A.; Vizzotto, A.L.P.; Pansera, W.A.; Savaris, G. General model for service life analysis of reinforced concrete structures subject to chloride penetration. Constr. Build. Mater. 2021, 305, 124727. [Google Scholar] [CrossRef]
  105. Shreir, L.L. (Ed.) Corrosion: Metal/Environment Reactions; Newnes-Butterworths: London, UK, 1976. [Google Scholar]
  106. Li, C.Q.; Mahmoodian, M. Risk based service life prediction of underground cast iron pipes subjected to corrosion. Reliab. Eng. Syst. Saf. 2013, 119, 102–108. [Google Scholar] [CrossRef]
  107. Dasar, A.; Patah, D.; Hamada, H.; Yamamoto, D.; Sagawa, Y. Life performance of 40-year-old RC beams with different concrete covers and bar diameters in natural corrosion environments. In Structures; Elsevier: Amsterdam, The Netherlands, 2022; Volume 46, pp. 2031–2046. [Google Scholar]
  108. Chen, D.; Mei, G.; Xiao, L. Analysis of chloride invasion process in undersea RC circular lined tunnel based on nonlinear ADE theory: Analytical solution, numerical verification and experimental prediction. Constr. Build. Mater. 2022, 343, 128140. [Google Scholar] [CrossRef]
  109. Keßler, S. Probabilistic corrosion condition assessment of a tunnel structure. Struct. Concr. 2020, 21, 1345–1355. [Google Scholar] [CrossRef]
  110. Meng, T.; Yu, H.; Zhao, Y.; Ruan, S. Strength/Service life-oriented environmental assessment of normal and recycled concrete with nano-SiO2 in a natural and a marine environment: A pioneer case study in Zhejiang Province. J. Build. Eng. 2023, 64, 105623. [Google Scholar] [CrossRef]
  111. Canestro, E.; Strauss, A.; Sousa, H. Multiscale modelling of the long-term performance of prestressed concrete structures–Case studies on T-Girder beams. Eng. Struct. 2021, 231, 111761. [Google Scholar] [CrossRef]
  112. Zheng, L.; Pornsing, C. Prediction of airport road service life based on concrete structure characteristics. Case Stud. Constr. Mater. 2024, 20, e02814. [Google Scholar] [CrossRef]
  113. Mortagi, M.; Ghosh, J. Climate change considerations for seismic vulnerability assessment of aging highway bridges. ASCE-ASME J. Risk Uncertain. Eng. Syst. Part A Civ. Eng. 2020, 6, 04020005. [Google Scholar] [CrossRef]
  114. Jeon, C.H.; Kwon, T.H.; Kim, J.; Jung, K.S.; Park, K.T. Health Index-Based Maintenance of Prestressed Concrete Bridges Considering Building Information Modeling Application. Buildings 2024, 14, 4032. [Google Scholar] [CrossRef]
  115. Ying, J.; Chen, W.; Wang, W.; Tian, Z.; Liang, L.; Chen, B. Multiscale study on high-temperature mechanical behaviors and durability of recycled concrete reinforced by polycarboxylate-modified 3D porous graphene. Constr. Build. Mater. 2024, 450, 138676. [Google Scholar] [CrossRef]
  116. Kaveh, H.; Alhajj, R. Recent advances in crack detection technologies for structures: A survey of 2022-2023 literature. Front. Built Environ. 2024, 10, 1321634. [Google Scholar] [CrossRef]
  117. Wang, Z.; Maekawa, K. Lifetime assessment of structural concrete–multi-scale integrated hygro-thermal-chemo-electrical-mechanistic approach and statistical evaluation. Struct. Infrastruct. Eng. 2022, 18, 933–949. [Google Scholar] [CrossRef]
  118. Azimi, M.; Eslamlou, A.D.; Pekcan, G. Data-driven structural health monitoring and damage detection through deep learning: State-of-the-art review. Sensors 2020, 20, 2778. [Google Scholar] [CrossRef] [PubMed]
Figure 1. The corrosion efficiency of reinforcement under 5% NaCl solution.
Figure 1. The corrosion efficiency of reinforcement under 5% NaCl solution.
Coatings 15 00693 g001
Figure 2. Example representing the structure of an HSPBN. Discrete variable nodes are represented with rectangles and continuous variable nodes with ellipses. Parametric CPD nodes are represented with white nodes and non-parametric CPDs nodes are represented with gray-shaded nodes.
Figure 2. Example representing the structure of an HSPBN. Discrete variable nodes are represented with rectangles and continuous variable nodes with ellipses. Parametric CPD nodes are represented with white nodes and non-parametric CPDs nodes are represented with gray-shaded nodes.
Coatings 15 00693 g002
Figure 3. A computational neuron in the artificial neural network (ANN) model [52].
Figure 3. A computational neuron in the artificial neural network (ANN) model [52].
Coatings 15 00693 g003
Figure 4. The basic structure of an SVM.
Figure 4. The basic structure of an SVM.
Coatings 15 00693 g004
Table 1. Impact of three common corrosion products on concrete protection and the corrosion process.
Table 1. Impact of three common corrosion products on concrete protection and the corrosion process.
Corrosion ProductConditions for Corrosion OccurrenceEffect on Concrete CoverEffect on Corrosion ProcessRelative Effect During Continuous Corrosion Process
Carbonate Products [20]Reaction of carbon dioxide with calcium hydroxide, forming calcium carbonateReduces the alkalinity of the concrete, compromising the concrete cover of the reinforcementLowers the pH of concrete; the steel protection layer gradually fails, promoting steel corrosionModerate
Chloride Products [21]Chloride ions penetrate, particularly in marine or high-salinity environmentsIncreases the permeability of concrete, diminishing its protective capacityDirectly breaks down the steel passivation layer, initiates corrosion, causes pitting (localized corrosion)High
Sulfate Products [22]Sulfates react with cement hydration products to form expansive compoundsCauses expansion and cracking, damaging the concrete structureExpanding reactions create cracks, increasing permeability and accelerating corrosionLow
Table 2. Influence of crack width on corrosion rate (µA /cm2).
Table 2. Influence of crack width on corrosion rate (µA /cm2).
Corrosion rate at 40 mm cover (µA/cm2)
Cement TypeCrack widthPercentage change
0.2 mm0.7 mm
Ordinary Portland Cement 1.20 1.4823%
7% Silica Fume Blended Cement 0.59 1.0375%
30% Fly Ash Blended Cement 0.390.5028%
50% Ground Granulated Blast Furnace Slag Blended Cement 0.350.5351%
Corrosion rate at 20 mm cover (µA/cm2)
Cement TypeCrack widthPercentage change
0.2 mm0.7 mm
Ordinary Portland Cement 2.653.2322%
7% Silica Fume Blended Cement0.671.1267%
30% Fly Ash Blended Cement 0.640.7111%
50% Ground Granulated Blast Furnace Slag Blended Cement0.390. 5131%
Table 3. Average relative error and standard deviation of the fitted carbonation test results at 20 °C and 30 °C.
Table 3. Average relative error and standard deviation of the fitted carbonation test results at 20 °C and 30 °C.
Carbonation ConditionsStatistical IndicatorsFick ModelNew Model
20 °C CarbonationMean error value e (%)1.80%0.78%
Standard deviation σ (%)1.91%0.35%
30 °C CarbonationMean error value e (%)6.67%2.67%
Standard deviation σ (%)7.70%3.39%
Table 4. Comparison of some empirical models.
Table 4. Comparison of some empirical models.
ModelDisadvantagesAdvantages
The model in reference [45]:
C x , t f = C 0 f 1 erf x 2 D c f t
C x , t f : chloride concentration in the pore fluid at depth x from
the concrete surface after time t of exposure to a chloride
environment;
C 0 f : free chloride concentration at the concrete surface;
erf: error function;
D c f : the coefficient of Cl diffusion.
1. Assumes chloride diffusion is uniform, ignoring the pores and cracks in concrete, which affects accuracy.
2. Assumes the diffusion coefficient remains constant, whereas it actually changes with time and environmental conditions.
3. In extreme environments such as high or low temperatures, the model’s applicability and accuracy decrease.
1. The formula is based on Fick’s second law, with a simple structure that is easy to apply.
2. It is consistent with experimental data for short-term and medium-term exposures (90 to 520 days).
3. Suitable for different types of concrete and exposure environments.
The model in reference [46]:
C = C 0 1 erf x 2 D 1 1 m t ( 1 m )
C : chloride concentration at distance x and time t ;
C 0 : chloride concentration at the concrete surface;
erf: error function;
D 1 : the initial chloride diffusion coefficient;
m: the empirical coefficient that represents the time-dependent
variation of the diffusion coefficient.
1. The m coefficient depends on the mix ratio and environment, requiring calibration.
2. Extensive experimental data are needed for validation and adjustment.
3. Long-term predictions (e.g., decades) face uncertainties regarding diffusion coefficient behavior.
1. Incorporating time-dependent diffusion coefficients enhances long-term chloride concentration predictions.
2. Predictions align well with actual data, especially in marine environments.
3. Suitable for various concrete mixes and environmental conditions.
The model in reference [47]:
C C s = 1 erf x 2 D 0 1 1 n 1 + t ex t 1 n t ex t 1 n t 0 t n t
C : chloride concentration inside the concrete;
C s : chloride concentration at the surface of the concrete;
x : depth of chloride penetration;
t : exposure time;
D 0 : initial chloride diffusion coefficient;
n : age factor representing how diffusion changes with concrete age;
t ex : initial exposure time (when the concrete starts being exposed);
t 0 : reference time for the initial measurements.
1. Assumes chloride binding is time-independent and linear, which may not reflect actual behavior.
2. When the age factor is high, it may underestimate chloride penetration depth, leading to overly conservative predictions.
3. For short-term exposure, the model may overestimate the diffusion coefficient, leading to inaccurate predictions.
1. Introduces a time-dependent diffusion coefficient, more accurately reflecting the change in chloride diffusion over time.
2. Provides a more accurate prediction of chloride penetration, especially under long-term exposure conditions like marine environments.
3. Aligns better with measured chloride penetration profiles in real-world applications.
The model in reference [48]:
C ( x , t ) = A 1 D max k 1 ( 1 erf x 4 K D 0 t e x 2 4 K D 0 t R e e g 2 erfc i g )
C ( x , t ) : the chloride ion concentration at position x and time t ;
D max : the stabilized surface chloride ion concentration;
k 1 : the coefficient representing the influence of water–cement ratio on the stabilized surface chloride ion concentration;
A 1 : the impact of sulfate ions on the chloride ion concentration on the concrete surface;
erf: error function;
D 0 : the chloride diffusion coefficient of the concrete measured at the hydration age.
1. The model features complex mathematical expressions, requires large computational resources, and is highly sensitive to input parameters, requiring precise measurements.
2. In real marine environments, the diffusion coefficient undergoes dynamic changes due to the inhomogeneity and variability of the medium, and the boundary conditions are uncertain, which adds challenges to its field application.
1. The model is particularly suitable for describing diffusion processes involving dynamic changes in both time and space variables.
2. It can be applied to diffusion processes under complex boundary conditions, taking into account the changes in substances over time.
Table 5. Comparison of ANN, SVM, random forest, and decision tree models for lifespan prediction.
Table 5. Comparison of ANN, SVM, random forest, and decision tree models for lifespan prediction.
ReferenceTypeAdvantagesDisadvantagesApplicable Conditions
Reference [70]Neural NetworksStrong adaptability, no gradient information required, optimization potentialSensitive to network architecture, difficult to optimize, risk of overfittingComplex and nonlinear problems with large datasets
Reference [65]SVMsHigh prediction accuracy, adaptable to complex problems, strong optimization capabilityRequires many parameters, computationally complexSmall high-dimensional datasets
Reference [71]Random ForestsHandles high-dimensional data, easy to applyLack of interpretability, long computation timeExploratory analysis when relationships are unclear
Reference [72]Decision TreesCan handle multiple data types, requires little data preprocessingInstability, prone to biasData need clear classification, applicable to structured data
Table 6. Studies on the applicability of concrete durability models validated by field data.
Table 6. Studies on the applicability of concrete durability models validated by field data.
ContextModel Prediction AccuracyImportance of Field Data
Reference [78]Field validation of chloride ion diffusion in UK harborsAt 1.17 and 6.17 years, model predictions closely matched the monitoring data. At year 18, the model slightly underestimated chloride ion diffusion at deeper layers, with about 10% error.Highlights the role of field data in validating the model’s performance over time, showing discrepancies at deeper layers at later years.
Reference [79]Study of chloride ion diffusion coefficients in aging structures along the South China SeaFor the HPW6 structure, the measured chloride ion surface concentration was 1.78%, while the model predicted 4.13%.Demonstrates how field data from actual aging structures provides critical validation for model predictions, emphasizing real-world applications.
Reference [80]Geothermal power plant concrete exposed to chloride- and sulfate-rich geothermal waterField data show that, after 18 years, chloride ions had penetrated into deeper layers in normal concrete and began corroding the reinforcement; ultra-high-durability concrete performed better at the same time, although the diffusion occurred slightly earlier, but the overall trend was consistent with the model.Field data confirmed that the actual chloride diffusion rate was faster than predicted in normal concrete, validating the importance of monitoring for accurate durability modeling.
Reference [81]Lifecycle modeling of concrete cracking and reinforcement corrosion in Changfeng Creek Bridge, ChinaThe study predicts, through lifecycle modeling, that reducing the maintenance cycle (e.g., shortening it from 12 years to 4 years) can significantly extend the service life of the bridge. Field data from regular monitoring revealed an extension of approximately 15 years.This case highlights the importance of field monitoring in ensuring the long-term durability of infrastructure and optimizing models.
Table 7. Analysis of advantages and disadvantages of common concrete structure lifespan.
Table 7. Analysis of advantages and disadvantages of common concrete structure lifespan.
Model TypeAdvantagesDisadvantagesApplicable Scenarios
Physical ModelsStrong theoretical foundation, capable of simulating long-term degradation processes, incorporates environmental factorsComputationally complex, requires accurate data, limited applicabilitySuitable for simpler controlled environments, such as laboratory settings or less complex engineering projects
Empirical ModelsSimple, fast, cost-effective, highly adaptableLimited precision, reliant on past data, not applicable to all situationsIdeal for scenarios with available data, especially for quick preliminary assessments
Statistical ModelsBased on data-driven predictions, adaptable to complex environments, capable of handling uncertaintyRequires high-quality data, difficult to interpretBest for data-rich complex environments with numerous variables and uncertainties
Machine Learning ModelsHigh accuracy, capable of handling complex data and problems, real-time prediction updatesRequires large amounts of data and computational resources, difficult to interpret model mechanismsBest suited for environments that require high-precision predictions, real-time monitoring, and substantial data support
Hybrid ModelsStrong theoretical foundation, capable of simulating long-term degradation processes, incorporates environmental factorsComplex models that require multidisciplinary knowledge for integrationIdeal for engineering projects that need to consider multiple factors, complex structures, or dynamic environments
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Qi, L.; Peng, X.; Yang, Q.; Xia, K.; Xu, B. Review of Research on Prediction Models for Residual Life of Concrete Structures. Coatings 2025, 15, 693. https://doi.org/10.3390/coatings15060693

AMA Style

Qi L, Peng X, Yang Q, Xia K, Xu B. Review of Research on Prediction Models for Residual Life of Concrete Structures. Coatings. 2025; 15(6):693. https://doi.org/10.3390/coatings15060693

Chicago/Turabian Style

Qi, Linyuan, Xi Peng, Qiuwei Yang, Kangshuo Xia, and Bin Xu. 2025. "Review of Research on Prediction Models for Residual Life of Concrete Structures" Coatings 15, no. 6: 693. https://doi.org/10.3390/coatings15060693

APA Style

Qi, L., Peng, X., Yang, Q., Xia, K., & Xu, B. (2025). Review of Research on Prediction Models for Residual Life of Concrete Structures. Coatings, 15(6), 693. https://doi.org/10.3390/coatings15060693

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop