A Safety Risk Analysis of a Steel-Structure Building Using an Improved Controlled Interval and Memory Model

Abstract

Scholars and engineers have increasingly focused on the safety of steel-structure buildings. An accurate analysis can substantially reduce the collapse probability of these buildings. This paper proposes a novel risk analysis model to assess the safety of steel-structure buildings. The vector entropy method and weight clustering were used to improve the controlled interval and memory (CIM) model. The proposed model has the advantages of a straightforward modeling approach, strong adaptability, and logical relationships. The new strategy improves the reliability and stability of the CIM model results when the maximum membership principle is not applicable. The Jiangxi Exhibition Center in China, which has a steel truss roof, is used as a case study. The results indicate a high safety risk of the project and the need for maintenance and repair. The improved CIM model has higher stability and adaptability for analyzing the safety risks of steel structure than the standard CIM model.

1. Introduction

Steel structures are used extensively in building construction. Policies favoring prefabricated steel structures and the impetus from market-oriented reforms in construction have increased the utilization and development of steel-structure buildings in China [1]. However, due to low construction quality, poor structural design, and insufficient supervision, collapses of steel-structure buildings have occurred frequently, causing public concern about the safety of steel structures [2]. Therefore, identifying, evaluating, and minimizing the safety risks of steel-structure buildings has become a top priority [3].

Steel structures are utilized extensively, owing to their low weight, high strength, and excellent earthquake resistance. However, many operational problems exist. Engineering construction theory and adequate technical methodologies are required for planning, organization, supervision, coordination, and control to ensure high construction quality [4]. Furthermore, it is crucial to implement science-based and efficient structural safety evaluation methods and risk prediction tools during the building’s operational phase.

Risk is the probability of adverse events. The operation safety risk of steel-structure buildings generally refers to the probability of internal and external factors adversely affecting the use of steel-structure buildings. Research on safety risks in steel-structure buildings requires a comprehensive methodology to evaluate the building’s structural safety performance in various phases, including structural design, construction, operation, management, and maintenance [5]. This research encompasses diverse aspects, such as structural design and analysis, material properties, seismic performance, node connection, environmental impact, and other relevant factors [6]. Sanni-Anibire [7] used paired comparisons and weight ranking to determine risk scores and weights for construction accidents and their potential causes. They developed a risk assessment method to improve the safety performance of construction projects. Ning [8] proposed a quantitative safety risk assessment model for factor identification and classification and used an evaluation function to assess site layout scenarios accurately and comprehensively.

General structural safety risk analysis methods are extremely diverse, including the fuzzy comprehensive evaluation method (FCEM) [9], gray relational analysis method (GRAM) [10], and fault tree analysis (FTA) [11]. Safety risk analysis of steel-structure buildings is a relatively novel and time-sensitive topic that assesses structural safety hazards that have increased due to the rise in steel-structure buildings. An accurate definition of the membership functions is required in the FCEM because the functions affect the safety evaluation result [12]. The GRAM requires a linear and independent relationship between the factors affecting structural safety. This relationship may not apply to steel-structure buildings [13]. The FTA method assumes that a dominant factor affects structural safety, but this assumption may not be realistic. In addition, the complexity of the model limits the application of this method.

Unlike the above methods, the controlled interval and memory (CIM) model has excellent performance in handling the ambiguity and correlation between risk factors affecting structural safety, providing a new method to evaluate the safety of complex steel-structure buildings and eliminating multiple factors. The ambiguity of risk factors has been investigated. Tian [14] used the CIM model to determine the risk probabilities at different levels in a highway construction project. They verified the project’s feasibility and provided insights for project-related investment decisions. Zhang [15] used the CIM model to conduct a cost risk assessment of large hydropower projects and determine the probability distribution. However, these methods require evaluation data to obtain the probability distribution of safety risk factors [16], which is not always possible in engineering scenarios. In addition, the CIM model is based on the maximum membership principle to predict the risk probability. Significant differences in the probability of risk levels are required to obtain reliable results for decision-making. Weight clustering is used to aggregate the objects in several components in the gray clustering coefficient vector [17,18]. Therefore, when the maximum membership cannot be determined, weight clustering is adopted to ascertain the ultimate risk level.

Accurately assessing the operational safety risks of steel-structure buildings is required to ensure the structure’s health and stability. This study proposes a safety risk analysis method for steel-structure buildings using an improved CIM model to address safety challenges in prefabricated steel-structure buildings. The main contributions of this article are as follows: (1) We propose a novel improvement of the CIM model that uses weight clustering to determine the risk when the maximum membership cannot be determined. This strategy improves the adaptability and robustness of the CIM model in assessing the safety of steel-structure buildings. (2) Current research has focused on structural performance evaluations during the construction of steel structures but has not assessed risk management during the building’s operation. In contrast, this study conducts a structural safety evaluation during the operation of steel structures to improve their safety and integrity.

The remainder of this paper is organized as follows. Section 2 describes the proposed method. Section 3 presents the case study. Section 4 discusses the results, and Section 5 concludes the paper.

2. Methods

2.1. The Proposed Index System

Structural failure risks are more common than functional risks. They include a decline in the structural bearing capacity, the failure of components, steel corrosion, or fatigue cracks.

The proposed index system for evaluating the operational safety risk of steel-structure buildings is listed in

Table 1. Index system for evaluating the operational safety risk of steel-structure buildings.

Index	References
$R_1$: Load redundancy	[19,20,21]
$R_2$: Construction error of steel members	[22,23,24]
$R_3$: Steel corrosion degree	[25,26]
$R_4$: Integrity of connectors	[27,28,29]
$R_5$: Fatigue crack of connector	[30,31]
$R_6$: Local buckling of steel members	[32,33,34]
$R_7$: Regular inspection and maintenance	[35,36,37]
$R_8$: Unfavorable Operating Environment	[38,39,40]

.

This index system covers three stages: the design, construction, and operation of steel-structure buildings. In Table 1, $R_1$ reflects the influence of design on the operation safety of steel-structure buildings. $R_2$ refers to the influence of construction errors on the operation safety of steel-structure buildings. The other indicators reflect the influence of structural failures, inadequate maintenance, and environmental factors on the operation safety of steel-structure buildings.

$R_1$ refers to the structural bearing capacity of steel-structure buildings, i.e., the ability of the structure to bear a load exceeding the design standard. The higher the load redundancy, the greater the safety margin of the steel-structure building during its operation.

$R_2$ refers to deviations from the design requirements, codes, or standards during the construction of steel structures. Common construction errors include deviations in the size of steel members, installation position, and prestress load, affecting the structure’s geometry, stability, and bearing capacity.

$R_3$ refers to steel corrosion caused by environmental factors during the building’s operation. Steel corrosion reduces the effective size of the structure, significantly lowers its bearing capacity, and may cause the destruction of steel-structure buildings.

$R_4$ refers to the integrity of the connectors of the steel-structure building. They may break, become loose, or fall off during operation and use. Common connectors are bolts, welds, and rivets. The connector integrity affects the structure’s stability and safety.

$R_5$ refers to fatigue crack of connectors, i.e., the generation and propagation of small cracks due to repeated loading and unloading. Fatigue cracks are a common problem in steel-structure buildings, substantially reducing the structure’s bearing capacity.

$R_6$ refers to the local buckling of steel members, a common failure type of steel-structures. When local buckling occurs, the steel structure’s bearing capacity is reduced.

$R_7$ refers to the periodic inspection and maintenance of steel-structure buildings during their operation. Regular inspection and maintenance can significantly reduce potential safety hazards caused by material fatigue, corrosion, local damage, and other factors.

$R_8$ refers to an unfavorable operating environment, including extreme weather conditions, chemical corrosion, and high-frequency vibration or shock. These factors accelerate the degradation of steel materials and cause corrosion, fatigue, and local damage to components.

2.2. The Proposed Safety Risk Analysis Model

2.2.1. The Standard CIM Model

The following steps were performed:

(1) Establish a risk evaluation set ($V$ = extremely high risk, high risk, medium risk, low risk, and extremely low risk) and invite experts to score the indices to obtain preliminary data.

(2) The probability of element $e_i$ at risk level $h$ is [41]:

$$P_{ih}=n_{ih}/n,$$

(1)

where $h$ is the risk level, $n_{ih}$ is the number of experts scoring risk level $h$, and $n$ is the total number of experts.

(3) Calculate the risk of each element using a parallel response model. For example, the probability of risk level H when $e_1$ and $e_2$ are connected in parallel is:

$$P_h\left(e_{12}\right)=P_h\left(e_1\right)\times {\sum }_{f=1}^h{P}_f\left(e_2\right)+P_h\left(e_2\right)\times {\sum }_{f=1}^{h-1}{P}_f\left(e_1\right),$$

(2)

where $P_h\left(e_1\right)$ is the probability of element $e_1$ at risk level $h$, and $P_h\left(e_2\right)$ is the probability of element $e_2$ at risk level $h$.

(4) The risk $P_h$ of all elements for parallel connection of the elements is calculated step by step.

(5) The maximum membership is used to obtain the final calculation result [41]:

$$P_g=\max \left\{P_h|h=\mathrm{1,2},\cdots ,5\right\},$$

(3)

where the risk level of the object is $g$.

2.2.2. The Proposed Improvement Strategy

The CIM has the advantages of straightforward modeling and strong adaptability [41]. However, it uses the maximum membership, which may not provide reliable results when the risk probability is similar for different risk levels.

For example, if $\left\{0.0147,0.0324,0.0310,0.0187,0.0032\right\}$ is the risk probability, it is difficult to determine if the ultimate risk probability is 2 or 3.

Therefore, this study proposes the following CIM improvement strategy.

The clustering coefficient vector entropy $I\left(P\right)$ is typically used [42,43] when the maximum membership cannot be determined. It is defined as follows:

$$I\left(P\right)=-{\sum }_{h=1}^5{P}_h\ln P_h.$$

(4)

This study uses five risk levels ($h=1,2,3,4,5$), but the improved model is applicable to more risk levels, as discussed in Section 4.2.

Property 1 [42]: the value of $I\left(P\right)$ satisfies $0\le I\left(P\right)\le \ln 5.$ The following conditions apply:

(1) When $I\left(P\right)=0$, $P_h$ is 1, and the probability of the other risk levels is 0, i.e., the result is clearly defined.

(2) When $I\left(P\right)\to 0$, $P_h$ tends to 1, and the probability of the other risk levels tends to 0. A reliable conclusion can be drawn.

(3) When $I\left(P\right)=\ln 5$, the probability of all risk levels is equal. In this case, no conclusion can be drawn.

(4) When $I\left(P\right)\to \ln 5$, the probability of all risk levels tends to be equal. In this case, no reliable conclusion can be drawn.

(5) When $I\left(P\right)>0.7\ln 5$, the maximum degree of membership cannot be determined [42], and weight clustering is used to determine the final risk level:

(a) Let ${\eta }_k\left(k=1,2,\cdots ,h\right)$ be a clustering weight vector group, and ${\eta }_k$ is defined as follows [43]:

$${\eta }_1=\frac{2}{s\left(s+1\right)}\left(s,s-1,s-2,\cdots ,1\right),$$

(5)

$${\eta }_2=\frac{1}{\frac{s\left(s+1\right)}{2}+\left(s-2\right)}\left(s-1,s,s-1,s-2,\cdots ,2\right),$$

(6)

$${\eta }_3=\frac{1}{\frac{s\left(s+1\right)}{2}+\left(2s-6\right)}\left(s-2,s-1,s,s-1,\cdots ,3\right).$$

(7)

Using the analogy, we can obtain ${\eta }_k$:

$${\eta }_k=\left\{\frac{1}{\frac{s\left(s+1\right)}{2}+\left[\left(k-1\right)s-\frac{k\left(k-1\right)}{2}\right]}\right\}\left(s-k+1,s-k+2,\cdots ,s-1,s,s-1\cdots ,k\right),$$

(8)

$${\eta }_{s-1}=\frac{1}{\frac{s\left(s+1\right)}{2}+\left(s-2\right)}\left(2,3,\cdots ,s-1,s,s-1\right),$$

(9)

$${\eta }_s=\frac{2}{s\left(s+1\right)}\left(1,2,3,\cdots ,s-1,s\right).$$

(10)

We can obtain ${\eta }_1=\frac{1}{15}\left(5,4,3,2,1\right)$, ${\eta }_2=\frac{1}{18}\left(4,5,4,3,2\right)$, ${\eta }_3=\frac{1}{19}\left(3,4,5,4,3\right)$, ${\eta }_4=\frac{1}{18}\left(2,3,4,5,4\right)$, and ${\eta }_5=\frac{1}{15}\left(1,2,3,4,5\right)$.

(b) We create a new decision vector $w_k$:

$$\left\{\begin{array}{l}{w}_k={\eta }_k{P}^{\mathrm{T}}\\ P^{\mathrm{T}}={\left[P_1,P_2,P_3,P_4,P_5\right]}^{\mathrm{T}}\end{array}\right.,$$

(11)

where $P^{\mathrm{T}}$ is the vector to be evaluated.

(c) We determine the final risk level $g$:

$$w_g=\max \left\{w_k|k=1,2,\cdots ,5\right\}.$$

(12)

2.3. The Proposed Model

The main steps of this model are as follows:

Step 1. Conduct the risk analysis.

Step 2. Establish the evaluation index system.

Differences exist in the steel members, connectors, design, construction, and operation of different steel structures. The indicators in Table 1 are adjusted based on expert opinion.

Step 3. Determine the risk levels.

The risk levels are determined based on the characteristics of the steel structure. We used five risk levels.

Step 4. Data collection and preprocessing.

Experts are invited to review the indices, score them, and obtain preliminary data. Equation (1) is used to determine the probability of the indices for different risk levels $P_{ih}$.

Step 5. Calculate the risk probability of the indices.

Equation (2) is used to determine the probability of risk $P_h$ for each index using the parallel response model.

Step 6. Determine whether the maximum membership can be used.

Calculate the cluster coefficient vector entropy $I\left(P\right)$ using Equation (4). If $I\left(P\right)>0.7\ln 5$, the maximum membership cannot be used; go to Step 8. Otherwise, the maximum membership can be used; go to Step 7. It should be noted that $0.7\ln 5$ corresponds to five risk levels.

Step 7. Obtain the risk analysis results using the maximum membership.

Use Equation (3) to obtain the final calculation result; go to Step 9.

Step 8. Obtain the risk analysis results using the proposed improved strategy.

Equations (5)–(10) are used to establish the clustering weight vector group ${\eta }_k$. Equation (11) is used to establish a new decision vector $w_k$, and the final risk level $g$ is determined by Equation (12).

Step 9. Implement targeted measures based on the results. Apply the risk management strategy to future projects.

The flowchart of the proposed model is shown in Figure 1.

3. Case Study

3.1. Project Overview

The Jiangxi Exhibition Center in China, which has steel roof trusses, was used as a case study. The project was completed on 29 March 2011, and the steel structure was repaired in 2023. It was necessary to evaluate the operation safety risk of the building to determine whether the building required maintenance or repair. The building is shown in Figure 2.

The steel structure was designed according to a seismic fortification intensity of 6 and a seismic level of 4. This is determined according to the national code for seismic design of buildings in China (GB50011-2016, https://www.mohurd.gov.cn/gongkai/zhengce/zhengcefilelib/201608/20160801_228378.html, accessed on 23 May 2024) [44], which also shows that the project is in a moderate or small earthquake activity area. The steel structure consists of Q235B steel. The steel structure has a height of 57.9 m, and the height of the reinforced concrete roof is 43.100 m. The building is covered by steel keel and aluminum veneer.

After interviewing 11 experts in this field, the proposed evaluation index system was adopted. The details of the 11 experts and their evaluation of the index system are listed in

Table 2. Details of the 11 experts and their evaluation of the index system.

No.	Working Years	Professional Title	Number of Similar Projects in Which the Experts	Evaluation of the Index System
(1)	12	Senior engineer	17	No need to modify
(2)	15	Senior engineer	23	No need to modify
(3)	8	Associate professor	11	No need to modify
(4)	14	Professor	24	No need to modify
(5)	15	Senior engineer	20	$\mathrm{Delete} R_5$
(6)	32	Senior engineer	35	No need to modify
(7)	25	Professor	21	No need to modify
(8)	27	Senior engineer	13	No need to modify
(9)	16	Senior engineer	9	No need to modify
(10)	29	Professor	36	No need to modify
(11)	31	Senior engineer	23	No need to modify

. The same experts were used to evaluate the risk factors.

The average working life of the 11 experts was 20.36 years; 90.9% of experts had senior titles, and the average number of projects in which the experts participated was 21.09. These results indicate that the experts had sufficient engineering experience and solid industry knowledge.

We used five risk levels. Extremely high risk (Level I) indicates serious safety problems in steel buildings that may cause immediate structural failure. High risk (II) means a large safety hazard. The structure is not likely to fail immediately, but the risk of structural failure is high in the future, requiring timely intervention and repair. Medium risk (III) indicates some safety problems, with a potential for higher risk in the future, although the building’s integrity and safety are not compromised. Low risk (IV) means good safety conditions with few risks, but regular inspection and maintenance are recommended. Very low risk (V) means that the steel building has excellent safety with relatively low risk.

3.2. Risk Level Calculation

(1) Determine the risk probability of the indices.

The average risk scores obtained from the 11 experts are listed in

Table 3. Probability of operational safety risk of steel-structure buildings for the indices.

Factor	Extremely High	High	Medium	Low	Extremely Low
$R_1$	$7/11$	$3/11$	$1/11$	$0$	$0$
$R_2$	$8/11$	$1/11$	$1/11$	$1/11$	$0$
$R_3$	$9/11$	$1/11$	$0$	$1/11$	$0$
$R_4$	$8/11$	$2/11$	$1/11$	$0$	$0$
$R_5$	$9/11$	$1/11$	$0$	$1/11$	$0$
$R_6$	$10/11$	$1/11$	$0$	$0$	$0$
$R_7$	$9/11$	$1/11$	$1/11$	$0$	$0$
$R_8$	$8/11$	$2/11$	$1/11$	$0$	$0$

.

(2) Determine the overall risk probability.

Due to the space limitations, we use $R_1$ and $R_2$ as examples to calculate the overall risk.

The risk probability of $R_1$ and $R_2$ is as follows:

The probability of extremely high risk is $7/11\times 8/11=0.463$.

The probability of high risk is $3/11\times \left(8/11+1/11\right)+1/11\times 7/11=0.281$.

The probability of medium risk is $1/11\times \left(8/11+1/11+1/11\right)+1/11\times \left(7/11+3/11\right)=0.165$.

The probability of low risk is $0\times \left(8/11+1/11+1/11+1/11\right)+1/11\times \left(7/11+3/11+1/11\right)=0.091$.

The probability of extremely low risk is: $0\times \left(8/11+1/11+1/11+1/11+0\right)+0\times \left(7/11+3/11+1/11+0\right)=0$.

The probability of the operational safety risk of the steel-structure building is listed in

Table 4. Probability of operational safety risk of the steel-structure building.

Level	Probability
Extremely high	0.122
High	0.340
Medium	0.289
Low	0.249
Extremely low	0

.

Equation (4), $I\left(P\right)=0.132>0.7\ln 5=0.127$, indicates that the maximum membership criterion is invalid. We use $P^{\mathrm{T}}={\left[0.122,0.34,0.289,0.249,0\right]}^{\mathrm{T}}$ in Equation (11) to obtain:

$$w_1={\eta }_1{P}^{\mathrm{T}}=\frac{1}{15}\left(5,4,3,2,1\right)\cdot {\left[0.122,0.34,0.289,0.249,0\right]}^{\mathrm{T}}=0.222,$$

(13)

$$w_2={\eta }_2{P}^{\mathrm{T}}=\frac{1}{18}\left(4,5,4,3,2\right)\cdot {\left[0.122,0.34,0.289,0.249,0\right]}^{\mathrm{T}}=0.227,$$

(14)

$$w_3={\eta }_3{P}^{\mathrm{T}}=\frac{1}{19}\left(3,4,5,4,3\right)\cdot {\left[0.122,0.34,0.289,0.249,0\right]}^{\mathrm{T}}=0.219,$$

(15)

$$w_4={\eta }_4{P}^{\mathrm{T}}=\frac{1}{18}\left(2,3,4,5,4\right)\cdot {\left[0.122,0.34,0.289,0.249,0\right]}^{\mathrm{T}}=0.204,$$

(16)

$$w_5={\eta }_5{P}^{\mathrm{T}}=\frac{1}{15}\left(1,2,3,4,5\right)\cdot {\left[0.122,0.34,0.289,0.249,0\right]}^{\mathrm{T}}=0.178.$$

(17)

According to Equation (12), $w_g=\max \left\{w_k|k=1,2,\cdots ,5\right\}=w_2$, indicating that the operation safety risk of the steel-structure building is high. Thus, maintenance and repair are required.

4. Discussion

4.1. Comparison of the Standard and Improved CIMs

The standard and improved CIMs are compared to demonstrate the advantage of the latter. The results are listed in

Table 5. Comparison of the results obtained from the standard and improved CIMs.

Level	Standard CIM		Improved CIM
	11 Experts in Case Study	10 New Experts	11 Experts in Case Study	10 New Experts
Extremely high	0.122	0.133	0.222	0.216
High	0.340	0.282	0.227	0.232
Medium	0.289	0.323	0.219	0.206
Low	0.249	0.257	0.204	0.197
Extremely low	0	0.005	0.178	0.149

. The results of the standard CIM are listed in Table 4.

It is more difficult to determine whether the project has medium or high risk using the standard CIM. The vector entropy method for clustering the coefficient used in the improved strategy incorporates information entropy into the standard CIM, enabling the analysis of the risk characteristics and improving the interpretation of the risk assessment results.

Therefore, the improved CIM is more suitable for risk assessment in a dynamic environment. When new risk factors are added, or the relationship between factors changes, the vector entropy of the clustering coefficient changes significantly, reflecting the system’s dynamic characteristics. Although the maximum membership is a fuzzy classification approach, it cannot quantify information entropy or uncertainty of risk factors.

We invited 10 other experts to repeat the case analysis and verify the model’s validity. The results of the two strategies are listed in Table 5.

After changing the expert group, the standard CIM provides substantially different results, unlike the improved CIM, whose results change negligibly. Therefore, changing the experts does not affect the results obtained from the improved CIM.

4.2. Model Adaptability

A multi-attribute evaluation model with strong adaptability can cope with different research environments. We analyzed the adaptability of the standard and improved CIM when the index system was changed. The results are listed in

Table 6. Adaptability of the two algorithms.

Research Environments	Standard CIM		Improved CIM
	11 Experts in Case Study	10 New Experts	11 Experts in Case Study	10 New Experts
Original index system	II	III	II	II
Delete $R_1$	II	II	II	II
Delete $R_2$	II	II	II	II
Delete $R_3$	II	II	II	III
Delete $R_4$	III	III	III	III
Delete $R_5$	II	II	II	II
Delete $R_6$	II	II	II	II
Delete $R_7$	III	III	III	III
Delete $R_8$	II	II	II	II
Add: $R_9$	III	II	II	II
Add: $R_{10}$	II	II	II	II
Add $R_9$ and $R_{10}$	III	II	II	II

. $R_9$ denotes fatigue cracks of the main components of the steel structure, and $R_{10}$ refers to the integrity of the main components of the steel structure.

Four deviations occurred in the results of the standard CIM, and two deviations were observed for the improved CIM. The risk level changed in two instances for the standard CIM and in one for the improved CIM. The results indicate that the proposed strategy has higher adaptability to different evaluation indices.

Other cases should be used to compare the standard and improved CIMs to confirm our results.

5. Conclusions

In order to provide more accurate and reliable theoretical and practical support for the risk management of steel-structure buildings, an improved CIM model was established and verified to analyze the safety risks of steel-structure buildings during their operation. Firstly, this paper constructs a comprehensive safety risk evaluation index system during the operation of steel-structure buildings and reveals in detail how these risk factors affect the operation safety of steel-structure buildings. This provides a basis for research in the field of the risk management of steel-structure building operation safety. Then, aiming at the problem that traditional CIM methods often have contradictory evaluation results, the vector entropy method and weight clustering were used to optimize the standard CIM model, significantly improving the reliability and stability of safety risk analysis of steel-structure buildings. Finally, a case study of the Jiangxi Exhibition Center’s steel truss roof demonstrated the applicability of the improved CIM model for safety risk analysis and its potential to reduce the safety risk of steel structures. The operation safety risk of this steel-structure building is high. By comparing the calculation results of classical CIM, this paper finds that the improved CIM model has good adaptability to different expert groups and different evaluation indexes. This proves that the improved CIM model has better adaptability and flexibility in different environments and using conditions. The proposed model is a powerful decision support tool in the steel-structure buildings’ design, construction, maintenance, and operation, and substantially improves the safety of buildings.

The limitations of this paper include the following. (1) The improved CIM model proposed in this paper relies heavily on expert opinions. This is also the biggest limitation of CIM model. The CIM model can be further improved to make it compatible with quantitative data and qualitative data. (2) The improved CIM model proposed in this paper can be used to evaluate the safety performance of other similar buildings. The premise of this is to build an index system with pertinence and building characteristics. (3) The proposed model can only roughly judge the safety risk of steel-structure buildings and provide managers with an efficient tool for daily management. In the future, safety risk assessment based on mechanical performance analysis should be carried out in order to accurately obtain and judge the safety risk of steel-structure buildings.