A Structured Data Model for Asset Health Index Integration in Digital Twins of Energy Converters

Abstract

A persistent challenge in digital asset management is the lack of standardized models for integrating health assessment—such as the Asset Health Index (AHI)—into Digital Twins, limiting their extended implementation beyond individual projects. Asset managers in the energy sector face challenges of digitalization such as digital environment selection, employed digital modules (absence of an architecture guide) and their interconnection, sources of data, and how to automate the assessment and provide the results in a friendly decision support system. Thus, for energy systems, the integration of Asset Assessment in virtual replicas by Digital Twins is a complete way of asset management by enabling real-time monitoring, predictive maintenance, and lifecycle optimization. Another challenge in this context is how to compound in a structured assessment of asset condition, where the Asset Health Index (AHI) plays a critical role by consolidating heterogeneous data into a single, actionable indicator easy to interpret as a level of risk. This paper tries to serve as a guide against these digital and structured assessments to integrate AHI methodologies into Digital Twins for energy converters. First, the proposed AHI methodology is introduced, and after a structured data model specifically designed, orientated to a basic and economic cloud implementation architecture. This model has been developed fulfilling standardized practices of asset digitalization as the Reference Architecture Model for Industry 4.0 (RAMI 4.0), organizing asset-related information into interoperable domains including physical hierarchy, operational monitoring, reliability assessment, and risk-based decision-making. A Unified Modeling Language (UML) class diagram formalizes the data model for cloud Digital Twin implementation, which is deployed on Microsoft Azure Architecture using native Internet of Things (IoT) and analytics services to enable automated and real-time AHI calculation. This design and development has been realized from a scalable point of view and for future integration of Machine-Learning improvements. The proposed approach is validated through a case study involving three high-capacity converters in distinct operating environments, showing the model’s effective assistance in anticipating failures, optimizing maintenance strategies, and improving asset resilience. In the case study, AHI-based monitoring reduced unplanned failures by 43% and improved maintenance planning accuracy by over 30%.

1. Introduction

The energy sector is undergoing a profound transformation driven by the urgent need to decarbonize operations, improve infrastructure efficiency, and enhance system resilience. These objectives are being pursued in an environment shaped by increasingly stringent regulatory frameworks, the integration of renewable energy sources, and growing public expectations for sustainability [1]. Within this evolving landscape, the ability to manage critical energy assets effectively throughout their lifecycle has become a strategic priority for utilities, manufacturers, and operators [2,3].

Digitalization has emerged as a foundational enabler of modern asset management. The convergence of the Internet of Things (IoT), cloud computing, artificial intelligence (AI), and big data analytics has enabled new paradigms in operations and maintenance [4,5,6,7,8,9]. One of the most transformative developments in this space is the concept of the Digital Twin—a dynamic, virtual representation of a physical asset that continuously synchronizes with real-world data. These digital replicas integrate historical records, real-time telemetry, and predictive models, allowing for continuous assessment of performance, degradation, and operational risk.

However, unlocking the full potential of digital twins requires more than just data acquisition. It demands a robust, structured data infrastructure that can organize and contextualize information and link it to actionable decision-support mechanisms [10,11,12,13]. Without this, digital twins remain limited in their ability to support proactive maintenance, long-term planning, and operational optimization [14,15,16,17].

A critical requirement in this context is the capability to assess the health condition of assets in a way that is both comprehensive and operationally meaningful. The Asset Health Index (AHI) has emerged as a powerful methodology for this purpose, integrating diverse variables—including operational conditions, reliability indicators, usage intensity, and degradation patterns—into a single, interpretable score that reflects the current asset condition [18]. By synthesizing both real-time and historical data, the AHI supports predictive maintenance, risk prioritization, and strategic investment decisions [19].

As industries adopt digital transformation initiatives, the integration of Asset Health Index (AHI) methodologies into digital twin ecosystems is becoming increasingly relevant. Digital twins offer a virtual environment in which asset behavior can be simulated, monitored, and optimized in near real-time. In this context, the AHI serves not only as a monitoring tool but also as a predictive feature that enables early detection of failure risks and supports what-if analysis of maintenance strategies [5,20].

A comprehensive framework for the development of the Asset Health Index (AHI) and its integration with IoT platforms has been thoroughly discussed in the literature. For instance, ref. [21] provides detailed methodologies for calculating and applying the AHI in maintenance scenarios. This work outlines the foundational principles for deriving the AHI and demonstrates its utility in predictive maintenance through IoT-driven approaches. The insights from this study significantly inform the integration of AHI into Digital Twin systems, particularly in understanding its practical challenges and scalability considerations.

While the theoretical foundations of the AHI are well-established, its practical application within digital twin environments remains a developing field. For example, health indices have been used in the energy sector for asset prioritization, particularly in power transformers and rotating equipment, using historical and real-time data such as temperature, load, and dissolved gas analysis [22]. In railways and aviation, similar indices—while not always formally referred to as “AHI”—are used to assess asset degradation and plan maintenance interventions based on usage patterns, fault history, and environmental factors [23].

However, many of these implementations remain siloed or tailored to specific asset types. Integration into full-fledged digital twin architectures—enabling real-time updates and predictive simulation based on health indices—remains limited. In many industrial settings, AHI computations are performed periodically and offline, based on export data, which undermines the potential of digital twins as living models for continuous decision-making [24]. Most existing approaches rely on fragmented or ad hoc solutions, such as isolated spreadsheets, SCADA logic, or proprietary implementations, which lack transparency, standardization, and interoperability. These limitations severely constrain the potential of AHI-based decision-making within digital environments [21,25]. In many cases, AHI is calculated offline or periodically, preventing its use in automated workflows and real-time simulations. Furthermore, few solutions align with internationally accepted data standards, making them difficult to scale across industrial systems.

From the literature, two main approaches can be observed. One is standards-based, aligning models with frameworks such as ISO 14224 [26] for failure data or ISO 55000 [27] for asset management. This ensures consistency but often involves significant effort in data modeling and integration. The other is more pragmatic or ad hoc, relying on simplified spreadsheets or proprietary logic within SCADA or CMMS platforms. While faster to deploy, these approaches tend to lack scalability and transparency.

Common implementation barriers include poor data quality, unclear model design, and the need to define update frequency and input selection. Without a robust asset data structure linking condition monitoring, reliability history, and operational metadata, the AHI may lose effectiveness as a tool for comprehensive digital asset management. The problem statement is to propose a structured asset data model and sustainable cloud architecture for energy equipment over automatic and standards-aligned integration of variables in AHI, and in an economic and sustainable way in order to facilitate its application not only in large companies but also in small companies. This paper enables the seamless integration of the AHI into Digital Twins, with a particular case of energy converters.

For this purpose, the model is based on internationally recognized frameworks of asset reliability data including ISO 14224 [26], and asset models digitalization as the Reference Architecture Model for Industry 4.0 (RAMI 4.0) [28] and the Industrial Internet Reference Architecture (IIRA) [29]. Thanks to these foundations, the proposed development tries to ensure compatibility with existing industrial practices, interconnection, and exchanging data via API (Application Programming Interface is a set of rules and protocols that allow different software applications to interact and communicate with each other), while supporting modern digital capabilities common in the market. For this, a crucial capability of the proposed system is the synchronization of real-time telemetry with historical asset records, allowing the continuous and automatic computation of the Asset Health Index (AHI) within a cloud-based architecture.

To demonstrate the practical applicability of this approach, the paper presents a case study involving three high-capacity DC/AC energy converters operating under different environmental and load conditions.

Energy converters are essential components of modern power systems, enabling the transformation of energy from one form to another to support efficient, reliable, and sustainable distribution. Their function is particularly relevant in applications such as renewable energy infrastructures, battery energy storage systems (BESS), and industrial power networks, where they facilitate the integration of intermittent sources and ensure proper load balancing and energy flow optimization. The performance of these units directly impacts energy efficiency, system reliability, and operational costs [30].

However, these devices are subject to a variety of operational stresses that can lead to critical failures. Among the most common issues are thermal degradation, insulation breakdown, semiconductor malfunction, and electromagnetic interference. Such problems often result in unplanned downtime, reduced efficiency, and increased maintenance needs. In critical infrastructures—such as utility grids or industrial manufacturing plants—malfunctions may trigger widespread disruptions and safety hazards [31]. With the support of IoT-enabled sensors, machine learning, and advanced analytics, operators can deploy real-time diagnostics and predictive models that enhance resilience and reduce operational risk.

The analysis shows how variations in stress, degradation, and maintenance history are reflected in the AHI, and how this information supports the planning and execution of differentiated maintenance strategies. The study also illustrates how cloud-based integration allows for continuous monitoring and rapid feedback, enhancing operational efficiency and resilience.

In conclusion, these different and remote sources of data and their logical and standardized computation are essential to the functionality and reliability of today’s energy infrastructure. The complexity of their operation and prediction/simulation of the risks associated with failure demand has to be facilitated in a comprehensive and sustainable approach that integrates lifecycle thinking, predictive analytics, and digital twin technologies, to ensure long-term asset performance and reliability.

The remainder of this paper is structured as follows: Section 2 describes the methodology for the AHI asset data model, including the application context and problem definition, the data model design, and the digital twin implementation in the cloud Environment. Section 3 describes its implementation in a real-world scenario, including results, benefits, and guidelines for its implementation. Finally, Section 4 presents conclusions and future research directions.

2. Methodology for the AHI Asset Data Model

This study follows structured phases to design, implement, and validate a data model that enables the integration of Asset Health Index (AHI) methodologies into Digital Twins (DTs) for energy converters.

For the development of the AHI asset data model, our methodology is based on the application of DevOps and Agile Scrum, which enable the continuous integration of design, implementation, validation, and deployment processes. These methodologies allow iterative cycles, workflow automation, cross-functional collaboration, and improved product quality [32,33]. Then, it searches the continuous refinement of asset ontologies and telemetry ingestion layers and monitors the model accuracy, with the following benefits [34,35]:

• Continuous improvement of data model.
• Promote a culture of collaboration and sharing of knowledge.
• Seamless integration between development and implementation in Digital Twin solutions.
• Automatization of testing and deployment.
• Quick adaptation to new asset types or sensor configurations.
• According to DevOps and Scrum, the core steps of the methodology are:
• Planning (Scrum Planning): User stories are defined for asset components and their properties.
• Development (Coding): Model design using Digital Twins and normalization of asset schemas.
• Continuous Integration and Delivery (CID): Automated tests to ensure semantic and structural integrity of the model, and deployment of intermediate versions in staging environments for validation.
• Monitoring and Feedback for Continuous Improvement (MCI): Use of cloud services to evaluate model performance and data consistency, with retrospectives and refined planning sessions.

Therefore, this particularized approach for the AHI asset data model consists of four main phases:

• Application Context and Problem Definition (Planning): The first phase involves the review of all the data requirements from standards and best practices for asset health assessment and orientated to their integration with AHI methods, but considering the diverse sources that could be on fragmented architectures and with lack of interoperability among systems. This phase defined the need for a structured and standardized data model fulfilling the mentioned standards and for their implementation in a cloud environment and Industry 4.0 principles.
• Data Model Design (Coding): Once the requirements are compiled, the diverse data needs have to be organized. The proposed model is organized into a set of interconnected domains that support integration and implementation in commercial Digital Twin solutions, and align with descriptions of RAMI 4.0 layers and the Industrial Internet Reference Architecture (IIRA). Based on the identified requirements, we defined a multi-domain data model encompassing physical, logical, property, and decision-making domains. A UML class diagram is developed to formalize the relationships among the model components, with special emphasis on how different data layers contribute to the AHI computation. A Unified Modeling Language (UML) class diagram formalizes the relationships among these domains, providing a transparent and extensible architecture.
• Digital Implementation on cloud environments (CID): Cloud-based and on-premise digital twin solutions present key differences. Cloud solutions offer scalability, remote access, and advanced analytics powered by AI and machine learning. On the other hand, on-premise solutions provide greater data security and lower latency for critical applications, making them suitable for sensitive industrial environments. Major cloud platforms for digital twin implementations include Microsoft Azure Digital Twins, Google Cloud IoT, Amazon AWS IoT TwinMaker, and FIWARE. These platforms provide robust infrastructure and tools for real-time monitoring, simulation, and analytics of energy assets. These cloud platforms permit Digital Twin Implementation, but they are orientated as a catalog of different modules that users can combine to develop the final solution. Therefore, they need customization according to the systems environment of each company. This makes the standardization, integration, and sustainable implementation of digital twins for asset assessment difficult, especially for small companies and without huge amount of budget. In our case, we have selected the most employed cloud environment using Microsoft Azure cloud services for deploying the architecture over our designed data model, searching to support real-time data ingestion, structured processing of operational and environmental parameters, automated AHI calculation, and friendly representation of risk levels according to this assessment. Each AHI component is mapped to cloud-based processing routines, which feed the digital twin system. Throughout the Azure deployment, real-time data from IoT sensors is continuously merged with historical failure and maintenance logs, enabling a hybrid dataset for AHI computation.
• Case Study Validation (MCI): To validate the model, we applied it to three high-capacity DC/AC converters operating under different environmental and load conditions. Real-world operational and maintenance data were used to evaluate the behavior of the AHI in each case. The results confirmed the model’s ability to differentiate asset condition trajectories and support proactive maintenance strategies.

2.1. Application Context and Problem Definition

The model objective constitutes a synchronization of real-time and historical data with a virtual representation of the asset’s condition, enabling AHI calculation and automated maintenance responses through cloud services. Therefore, data of asset identification, criticality assessment, real-time monitoring, health evaluation, and risk-based decision-making have to be managed.

The digitalization of assets is a multifaceted process that requires comprehensive frameworks to manage and optimize assets across their lifecycle. To fully harness the potential of digitalization, a structured approach is essential—one that incorporates different aspects of asset management. Drawing from established standards such as ISO/IEC/IEEE 42010:2011, EN IEC 81346:2022, ISO 14224:2016, etc. [26,36,37].

It is crucial to consider a comprehensive life cycle cost (LCC) approach, evaluating all phases of asset ownership, including design, operation, maintenance, and decommissioning, helping decision-makers evaluate long-term sustainability and cost-effectiveness. These assets are acquired by different companies and from remote sites, so digital monitoring and predictive maintenance strategies are necessary everywhere, to provide analysis that supports extended equipment lifespan and the prevention of unexpected failures [38].

Reliable operation depends on monitoring a range of key parameters, as in the energy converters including voltage, current, temperature, power output, efficiency, and harmonic distortion. These indicators allow operators to assess device conditions in real time and detect early signs of degradation. As highlighted in [39], the convergence of digitalization and asset intelligence enables energy systems to evolve toward smarter, more adaptive infrastructures aligned with Industry 4.0 principles.

In addition, the engineering and operational management of the systems are increasingly being enhanced through digitalization and Model-Based Systems Engineering (MBSE). These methodologies enable robust modeling, simulation, and verification from early design to full deployment. MBSE when coupled with digital twins facilitates the early detection of design inefficiencies, improves traceability of requirements, and shortens integration times [40].

Based on this, a suitable recommendation would be to collect data to populate the following asset model [41]:

• Asset Definition Model (ADM). The Asset Definition Model provides a comprehensive framework for defining the structure, properties, and relationships of an asset within a digital system. This model aligns with the standards of ISO/IEC/IEEE 42010:2011, which emphasizes the need for a well-structured architecture description that defines the fundamental components and relationships of a system. The ADM is essential for creating a digital twin of the asset, ensuring that the digital representation is complete, and can serve as the foundation for further processes, such as monitoring and maintenance. Additionally, the EN IEC 81346:2022 standard on object classification further supports this model by offering a systematic way to categorize and manage asset data within complex systems, ensuring consistency and interoperability across platforms.
• Asset Criticality Model (ACM). The Asset Criticality Model addresses the prioritization of assets based on their impact on operations and their associated risks. This model is critical in ensuring that digitalization efforts focus on the most essential assets, maximizing value and minimizing downtime. RAMI 4.0, with its emphasis on asset lifecycle and hierarchical structures, supports this model by integrating the importance of criticality within the broader context of Industry 4.0. The ACM helps organizations focus on the assets that are most likely to influence operational efficiency and safety, ensuring that resources are allocated effectively in both short-term and long-term management.
• Asset Monitoring Model (AMM). The Asset Monitoring Model facilitates real-time condition monitoring of assets using IoT networks and other digital technologies. This model draws from both RAMI 4.0 and IIRA, which emphasize the need for interconnected systems and continuous data flow between physical assets and digital systems. By integrating IoT sensors and signal processing, the AMM enables real-time visibility into asset conditions, supporting predictive maintenance and minimizing unplanned downtime. This model also resonates with the Asset Administration Shell (AAS) in RAMI 4.0, which acts as the digital representation of the physical asset, ensuring seamless data exchange and analysis.
• Intelligent Asset Management Models (IAMMs). The Intelligent Asset Management Models provide a higher-level view of asset health and performance, utilizing advanced analytics to deliver actionable insights. These models are closely tied to Asset Performance Management (APM) and Asset Investment Planning (AIP) systems and support the goals of ISO 55000 on asset management, which emphasizes the need for integrating asset management practices into broader organizational goals. The IAMMs enable organizations to predict future asset conditions, optimize maintenance schedules, and align asset management with business objectives. Asset’s health index model provides critical insights for long-term asset management planning, ensuring that digitalization efforts lead to sustainable value creation.

The Intelligent Asset Management Model (IAMM) represents the final layer in asset digitalization, focusing on integrating real-time and historical asset data for advanced decision-making processes. Once an asset’s data has been captured, processed, and structured (via the Asset Definition and Monitoring Models), and its criticality is addressed (via the Asset Criticality Model), the need arises to connect this data with human reasoning and decision-making tools.

Intelligent asset management relies heavily on the development and integration of robust data models that allow for the digital execution of key methodologies, such as Root Cause Failure Analysis (RCFA), Reliability Centered Maintenance (RCM), Condition Based Maintenance (CBM), Reliability, Availability, Maintainability, and Safety (RAMS), Lyfe Cycle Cost (LCC), Asset Health Index (AHI), etc. These data models provide the essential framework to capture, organize, and analyze the wealth of information generated by industrial assets, supporting decision-making processes that drive asset optimization.

The organization of asset data into structured models—covering identification, criticality, monitoring, and intelligent management—lays the foundation for the effective deployment of advanced maintenance techniques in a digital environment.

The integration of these data models ensures that decision-making processes are streamlined, and that predictive analytics and simulation tools are leveraged to generate actionable insights. This approach enables organizations to move beyond manual methods, optimizing maintenance strategies and aligning them with broader business goals through data-driven insights.

For the present article, the methodology proposed to be implemented as part of the Intelligent Asset Management Models (IAMM) is the Asset Health Index (AHI). The AHI provides a key tool for evaluating and monitoring the health status of assets comprehensively, integrating both historical and real-time data. Its ability to synthesize information into a single indicator facilitates informed decision-making and the prioritization of maintenance strategies, aligning asset management with organizational objectives.

The proposed health index data model takes this a step further by interpreting the processed variables through a structured system of well-defined indicators that integrate operational, environmental, and maintenance-related factors. These indicators serve as robust evaluative tools for determining the asset’s condition and performance over time, particularly in cases where degradation mechanisms are not well understood. By incorporating dynamic adjustment factors—known as health and reliability modifiers—the model ensures that the index reflects both expected aging and the actual impact of operational and contextual deviations. These modifiers are normalized and weighted based on their influence, enabling the model to adjust the health index in real time according to current asset conditions and historical data trends. The outcome is a dimensionless health indicator that not only enables accurate diagnostics but also informs risk-based prioritization and predictive maintenance decisions. This facilitates more efficient asset replacement strategies, optimized life cycle management, and long-term sustainability of power generation infrastructure. The underlying mathematical framework that supports this model, including definitions for the aging rate, initial and adjusted health indices, and modifier integration, is presented in detail in

Table 1. Mathematical formulation for Asset Health Index definition [42].

ID	Indicator	Mathematical Formulation
$t_{VNE}$	Estimated normal life	$t_{VNE}=f(asset classification, business)$
$F_{Ld}$	Load factor	$F_{Ld}={\displaystyle \frac{Load under normal operation condicions}{Maximum permissible load}}$
$F_{Lc}$	Location factor	$F_{Lc}=f(functional location)$
$t_{VE}$	Estimated life	$t_{VE}={\displaystyle \frac{t_{VNE}}{F_{Ld} \times F_{Lc}}}$
$HI\left(t_{VE}\right)$	Health index at estimated life	$HI\left(t_{VE}\right)=5.5$
$HI\left(t_0\right)$	Health index at the beginning of life	$HI\left(t_0\right)=1$
$\beta$	Aging rate	$\beta ={\displaystyle \frac{\ln \frac{HI\left(t_{VE}\right)}{HI\left(t_0\right)}}{t_{VE}}}$
$HIi\left(t\right)$	Initial health index	${HI}_i\left(t\right)=HI\left(t_0\right)·e^{\beta · t}$
$F_{CR}$	Real load factor	$F_{LdR}\left(t\right)={\displaystyle \frac{Load consumed \left(t\right)}{Maximum permissible load}}$
$M_{Ld}\left(t\right)$	Load modifier	$M_{Ld}\left(t\right)={\displaystyle \frac{F_{LdR}\left(t\right)}{F_{Ld}}}$
$HIiReal\left(t\right)$	Real Initial health index	${HI}_{i}Real\left(t\right)=HI\left(t_0\right)·e^{M_{Ld}\left(t\right) · \beta · t}$
$MH\left(t\right)$	Health modifier	$MH\left(t\right)=f(monitored variables)$
$MR\left(t\right)$	Reliability modifier	$MR\left(t\right)=f(Operations\&Maintenance indicators)$
$HI\left(t\right)$	Health index	$HI\left(t\right)={{[HI}_{i}Real\left(t\right)]}^{e^{\left(MH\left(t\right)+MR\left(t\right)\right)}}$

.

The proposed AHI method is presented in Figure 1, describing in a flow graph how the different sources of data are incorporated to contribute to the final score of the Asset Health Assessment.

Assuming a high level of criticality for the asset and without delving into the criticality model itself, Figure 1 illustrates the proposed required data to achieve an advanced level of maturity in asset digitalization, taking into account the information system in which each activity is developed. This data model establishes a comprehensive framework for managing and analyzing asset information, beginning with the precise definition of the physical asset within its designated functional location.

A critical component of this framework is the monitoring model, which plays a pivotal role in ensuring the continuous and accurate assessment of asset conditions. This model systematically collects and organizes the information generated within the Internet of Things (IoT) network, enabling the transformation of raw sensor data into actionable insights. The process follows a structured data pipeline that encompasses four key stages: sensing, extraction, transformation, and loading (ETL).

In the sensing phase, embedded IoT devices capture high-resolution data across multiple operational variables—such as temperature, pressure, vibration, and speed—relevant to the asset’s performance and degradation patterns. During extraction, this data is pulled from diverse sources and platforms, often including supervisory control systems, condition monitoring software, and maintenance logs. The transformation stage involves cleaning, filtering, and standardizing the data to ensure consistency and reliability. Techniques such as correlation analysis, dimensionality reduction, and threshold normalization are applied to refine the dataset, ensuring that only the most informative variables contribute to the model. Specialized analytical tools assist in identifying redundant or low-impact data, thereby enhancing model efficiency.

Once transformed, the data is loaded into the analytical engine where it serves as the foundation for the real-time computation of health and reliability modifiers. These modifiers are integral to updating the health index, as they adjust for deviations in operational behavior and maintenance conditions. The result is a dynamic and scalable monitoring system that supports data integrity, traceability, and responsiveness, ultimately enabling proactive maintenance planning and asset lifecycle optimization. The mathematical formulations supporting each step of this process are detailed in the accompanying formula table.

This integrated approach underscores the critical importance of leveraging structured data models in achieving high levels of digitalization and operational excellence in asset management.

2.2. Data Model Design

The proposed model is organized into a set of interconnected domains based on the identified requirements, and represented by a UML class (see Figure 2), which represents the Asset Health Index (AHI) Data Model for Digital Implementation, emphasizing the sources of information and the relations among them. For this purpose, the diagram is structured into different information domains that interconnect, facilitating comprehensive asset health assessment inside a structured model for complete asset management.

The UML diagram shows an explanation of the primary classes and their interrelations, focusing on the calculation of an AHI and its provided Risks serving a specific purpose in asset health assessment and maintenance management, deploying it in an optimized and economic cloud architecture with the idea to provide support to operational and strategic decisions.

Each multiple domain of the UML diagram is divided into the main classes according to most employed asset management systems.

2.2.1. Physical Domain

This domain represents the physical assets’ attributes and describes the relations of the parts, that is, owing to a hierarchical breakdown of assets according to standards such as ISO 14224.

• System: The highest-level entity that includes subsystems.
• Subsystem: Defines a part of the system (e.g., bogies, wheels).
• Maintainable Item: The specific components subject to maintenance (e.g., wheel bearings, brakes).

These classes are interconnected, in a hierarchical tree with relations of 0 to many upwards.

2.2.2. Logical Domain

This section links assets to their functional locations within the entire organization.

• Organization, Segment as sections of the plant or network, and Functional Location define where the assets are placed.
• Topology and asset relationships between physical and logical domains must be ensured for traceability within the organization.

In this domain, the relationship among classes is also in a hierarchical tree with relations of 0 to many upwards.

2.2.3. Property Domain

This domain handles variables affecting asset condition and operation, some in real time and others offline.

• Operation and Maintenance Variable: Captures operation and monitored parameters, and some reliable calculations, that can affect asset health.
• Property: Based on previous variables defines measurable asset and processed characteristics (e.g., temperature, vibration), and any combination of them (e.g., power efficiency).

Any maintainable item has several O&M variables, that have to be particularized for employment in the asset health assessment, and stored for future and batch analysis.

2.2.4. Data Domain

This domain includes historical data and calculations based on manufacturer and company specifications to be used in asset health assessment.

• Locational Conditions have to be considered to adjust estimations based on environmental factors that can contribute to reducing or improving the health of the same assets in different functional locations.
• Estimated Normal Life and Aging Rate therefore are developed as a previous prediction of asset deterioration considering the specifications and locational conditions in general.

2.2.5. AHI Domain (Asset Health Index)

This domain is the development that integrates the necessary information in real time as a virtual entity of the asset producing the health score of assets using this composition from bottom to top:

• Initial Health Index (HIi)
• Real Initial Health Index (HIiReal), that includes the affections due to load operating circumstances.
• Load Modifier, recommended based on real-time data.
• Health Modifier, mainly based on real-time data.
• Reliability Modifier, based on reliability calculations such as MTBF, MTTR, Availability, number of preventives, frequency of preventives, overhauls, etc.
• Final Health Index (HI(t)) which incorporates the affections to asset life by O&M properties and the reliability history.
• Each modifier adjusts the health score based on the real circumstances that affect the assets.

2.2.6. Decision-Making Domain

This domain supports risk-based decision-making by integrating:

• Strategy Plan: Defines maintenance and asset strategies.
• Proactive Work: Represents predefined maintenance tasks to minimize or eliminate the risks.
• Model Risk: The most critical class, which assesses the risk level of the assets according to the Final Health Index, producing warnings at different prioritization levels and for operative, tactical, and strategic decisions.

2.3. Digital Twin Implementation in Cloud Environment

The aim of this model is to be implemented using a cloud platform of IT services, leveraging its suite of IoT, data processing, and visualization tools to automate AHI calculation and enable scalable deployment. The adoption of digital twin technology in energy equipment provides significant advantages, including real-time performance monitoring, parameter deterioration, fault diagnosis, enhanced predictive maintenance, and lifecycle optimization [43,44,45,46]. By creating a virtual replica of physical assets, digital twins enable improved diagnostics and operational efficiency, reducing downtime and enhancing decision-making processes.

Azure is one of the most employed platforms for digital twins on industrial assets, and for the energy converters, it has developed an economic and simple architecture for any size of the company. Azure Digital Twins stands out for energy converters due to its seamless integration with Azure IoT Hub, AI-driven analytics, and scalable data processing. It allows energy companies to optimize operations by enabling predictive maintenance, reducing failures, and enhancing energy efficiency.

The UML model can be deployed using Microsoft Azure Cloud Services (Figure 3), where a robust digital twin architecture for energy assets integrates several Azure modules to ensure efficient data acquisition, processing, and visualization. The basic components form the backbone of this architecture for the AHI data model implementation [47]:

• Step 1: Data Acquisition and Ingestion:

• Azure IoT Hub: Focus on Physical Domain exchanging variables from the assets to the cloud and vice versa, this module serves as the main communication gateway between energy converters and the cloud. It enables secure, bi-directional communication, allowing real-time data collection and remote control. This module is essential for managing many connected devices while ensuring data integrity.

  a.

  IoT sensors provide real-time data e.g., on temperature, vibration, and load.

  b.

  Data ingested through Azure IoT Hub collects real-time sensor data and Azure Event Hubs stream high-frequency telemetry data.

• Azure IoT Central: For Physical Domain implementation, this module provides a scalable, low-code interface for managing IoT devices.

  a.

  It simplifies device provisioning, monitoring, and data visualization, making it easier to deploy and manage energy assets networks at scale.

  b.

  Real-time monitoring is realized for the selected Operation and Maintenance Variable class, and for the developed properties from them.

• Step 2: Data Store and Processing:

• Azure Functions: Realizing the Property Domain from raw data, processing them, to the properties needed to manage the digital twin. This automation ensures that only high-quality, structured data is sent to analytics platforms, reducing computational overhead. The combination of Azure Functions and IoT Central automates data processing and device management, ensuring timely insights into asset health.

  a.

  It triggers data cleaning, normalization, and aggregation tasks when new telemetry is received, filtering out incomplete or noisy telemetry data.

  b.

  Transforms O&M variables in properties to be used in the digital assets.

• Cosmos DB. Responsible for the Data Domain, and retrieves Historical Data such as previous maintenance records, failure rates, and environmental conditions. It is a multi-model database service designed for low-latency, high-availability, and massively scalable applications. Cosmos DB supports multiple APIs to match different data models, compliance standards, low-latency queries with JSON-based schema, and security with encryption and private endpoints.

  a.

  Store Manufacturer and organization specifications and statistics parameters for reliability modifiers in a Cosmos DB.

  b.

  Supports custom automatic indexes for performance optimization.

• Azure Data Lake Storage, archives historical asset variables, implemented by the Property Domain.

  a.

  It stores the raw data in a time series.

  b.

  Store the properties that have to be related to the asset for future analysis in the asset health assessment.

• Step 3: Health Index Calculation

• Azure Data Explorer: It generates the AHI Domain and offers advanced time-series analysis and anomaly detection. This module enables real-time analytics on large datasets, supporting predictive maintenance and fault detection in energy converters. It is a powerful analytics module optimized for managing time series data efficiently with built-in compression and partitioning. The mathematical structure of the AHI follows a weighted combination approach, where each modifier reflects a different aspect of asset degradation, by integrating several modifiers—Health, Reliability, and Load—according to a rule-based aggregation model inspired by the framework described in [31].

  a.

  Calculates AHI dynamically according to the defined formulae, using the properties and joining with maintenance history tables, applying the different modifiers: load, health, and reliability. Each modifier represents specific risk adjustments.

  i.

  Health Modifier: Accounts for environmental conditions.

  ii.

  Reliability Modifier: Based on past failures.

  iii.

  Load Modifier: Adjusts for operational stress.

  b.

  Executes time-alignment of asynchronous data streams, and aggregation of rolling statistics to capture degradation trends.

  c.

  Implements the risk model of AHI determining the probability of failure based on asset health, operational conditions, and historical trends.

  d.

  Triggers maintenance workflows based on defined AHI thresholds.

• Azure Machine Learning, although no AI or machine learning models are used in this project, the calculation leverages domain expert knowledge and operational thresholds to adjust the health score in real time. This module can be used for future developments to simulate or train models for asset health prediction and for defining the level of risks archives historical asset variables.

  a.

  Simulation scenarios.

  b.

  Prediction and prescription scenarios.

• Step 4: Digital Twin Integration

• Azure Digital Twins creates virtual representations of assets and predicts maintenance scheduling. This module defines the Logical Domain, integrating the rest of the domains with the logical structure and relations among variables, models, and other assets. This module provides the basis for the visualization of the Decision-Making Domain.

  a.

  Maps health attributes with the virtual assets using the DTDL (Digital Twins Definition Language).

  b.

  Visualizes asset conditions and automated predictive maintenance alerts based on risk level.

• Azure Power BI generates real-time dashboards for the Decision-Making Domain. Provides intuitive dashboards and visualizations, allowing operators to monitor key performance indicators (KPIs) for:

  a.

  Detecting anomalies, and generating predictive insights based on historical data.

  b.

  Representing levels of risk according to different filters dynamically and effortlessly.

The UML diagram represents a comprehensive Asset Health Index model that evaluates energy asset risks using real-time data, historical records, and predictive models. By deploying this framework in Azure Cloud, manufacturers and their clients can optimize asset maintenance, minimize downtime, and enhance safety standards.

The proposed model is implemented as an active Digital Twin component, where each data entity exists as a digital counterpart synchronized with physical telemetry. Data streams trigger real-time updates to the health index, which in turn initiates predictive alerts or maintenance actions through Azure Logic Apps, confirming the model’s digital operability.

In summary, using Azure Cloud Services to implement this UML model provides several advantages:

• Simple: According to the different sources of data and their processing, storing, and visualization.
• Scalability: Azure Digital Twins can model complex systems and replicate scenarios.
• Real-Time Analysis: Azure Event Hubs and Data Explorer process sensor data instantly.
• Automated Decision-Making: Based on Azure Data Explorer and Azure Functions, maintenance actions can be triggered.
• Cost Optimization: Serverless architecture minimizes operational costs.

3. Case Study Validation

This section presents the practical validation of the proposed Asset Health Index (AHI) data model through its application to three high-capacity DC/AC converters operating under diverse conditions. The aim is to demonstrate the model’s ability to detect asset degradation, support maintenance decisions, and enable real-time health monitoring within a digital twin architecture.

Energy converters, particularly DC/AC inverters, are critical in photovoltaic (PV) and battery energy storage systems (BESS), ensuring compatibility with grid infrastructure by transforming DC into AC. In this case study, each converter operates with an input of 1525 V DC and 1600 A, and delivers an output of 690 V AC and 3200 A.

The converters were selected based on three key criteria: (i) high power capacity (above 1.5 MW), (ii) diverse operating environments (ranging from moderate to severe thermal conditions), and (iii) availability of complete operational and maintenance records over a multi-year period.

An overview of the digital twin-based framework applied in the case study is presented in Figure 4. The figure illustrates the integration between the physical energy converters and the digital infrastructure that enables real-time monitoring, data aggregation, and health assessment. Each converter is equipped with sensors that stream operational data (electrical, thermal, and mechanical) to the cloud-based data model. This model feeds a digital twin instance where the Asset Health Index (AHI) is computed using rule-based algorithms. The computed AHI, together with other reliability indicators, is visualized through a dashboard that supports maintenance planning and intervention strategies. This architecture enables continuous fault diagnosis and condition-based decision-making throughout the asset lifecycle.

An analysis of historical failure records revealed that power devices and capacitors accounted for more than 50% of total failures. These components were found to be highly sensitive to thermal stress and aging mechanisms, motivating the focus on these failure modes in AHI computation.

The AHI model integrated data from multiple domains:

• Electrical variables: Voltage, current, harmonics, power factor.
• Thermal indicators: Real-time temperature, overheating events, thermal gradients.
• Mechanical parameters: Vibration levels, mechanical stresses.
• Operational events: Preventive and corrective maintenance history and downtime records.

3.1. Degradation Mechanisms

Two mechanisms were captured and monitored through a digital infrastructure based on Azure IoT services, which processed operational and environmental data to dynamically adjust the AHI over time:

• Power devices were observed to fail due to thermal fatigue in solder joints, leading to progressive resistance increase, overheating, and material degradation.
• Capacitor failures were linked to environmental and operational stress factors—particularly temperature and electrical overload—which contributed to increased internal impedance and eventual performance loss.

The data were ingested through the Azure IoT pipeline, structured via the defined UML data model, and processed using Azure Functions and Azure Data Explorer services (for modifier estimation). All health scores and modifiers were stored in Azure Data Lake and visualized through Power BI dashboards.

3.2. Comparative Health Index Results

Figure 5 shows the AHI evolution for the three converters. The analysis of the results is as follows:

• Converter 1 was the first to be commissioned. Operating under moderate environmental conditions and with a load close to expected values, its degradation was gradual. However, a significant rise in AHI was observed before its first overhaul (values > 7), coinciding with recurring failures. This suggested that even under seemingly stable conditions, long-term accumulation of stress led to performance decline. The overhaul helped to restore operation, but the asset showed signs of reduced lifespan.
• Converter 2 operated under more aggressive conditions, including frequent load peaks and thermal cycling. It experienced early degradation and failures, requiring an overhaul at around 12,000 h—earlier than originally planned. After the intervention, failures ceased and operational stability improved, although the degradation trajectory remained steeper than in the other units. This case underlined the necessity of dynamic maintenance strategies based on real operating profiles.
• Converter 3, despite working in an environment with irregular load and thermal fluctuations, showed low failure rates and a stable AHI close to initial values. This indicates that either better component resilience or more adaptive control strategies were at play. The second overhaul was performed proactively and confirmed that the asset retained higher performance levels than its counterparts.

Table 2. Comparative Analysis of AHI and Maintenance for the Three Converters.

Parameter	Converter 1	Converter 2	Converter 3
Total operating hours before 1st overhaul	17,585.81 h	11,720.26 h	15,919.33 h
Number of failures before 1st overhaul	5	6	0
Average AHI before 1st overhaul	8.23	5.37	7.17
Environmental severity (qualitative)	Moderate	High (thermal cycling)	Variable
Load intensity (qualitative)	Normal	Frequent peaks	Irregular
Number of overhauls	2	3	2
AHI after last overhaul	5.03	2.66	5.00
Current trend in AHI	Slightly increasing	Stable	Stable
Failure rate after overhaul	Reduced	Stabilized	Minimal

summarizes key quantitative indicators derived from the case study. The differences in operational conditions, failure history, and AHI behavior across the three converters highlight the model’s capability to differentiate degradation trajectories and support maintenance planning. Notably, Converter 2 experienced earlier degradation and required more frequent intervention, while Converter 3 maintained better health despite environmental variability.

3.3. Overhaul Planning and Maintenance Optimization

The AHI analysis enabled condition-based decisions on overhaul scheduling:

• Converter 1 and Converter 3 underwent their first overhauls at similar operating hours, though Converter 1 had accumulated ~2500 more hours and exhibited more degradation.
• Converter 2’s early overhaul prevented further degradation and improved its lifecycle trajectory.
• A third overhaul was eventually necessary for Converter 2 due to accelerated wear.

These insights revealed that preventive strategies must be tailored not only to operational hours, but to real health data as captured through the AHI model.

3.4. Benefits of AHI-Based Monitoring in Digital Twin Architecture

The deployment of the proposed data model and AHI integration within this basic Azure-based digital twin provided:

• Granular visibility into degradation patterns and performance variation across identical assets.
• Early warning of asset deterioration, enabling predictive maintenance planning.
• Dynamic maintenance scheduling, adapted to operational stress and environmental conditions.
• Operational savings, by preventing unnecessary or untimely overhauls.
• Improved diagnostics and resilience, enabling real-time anomaly detection, identifying potential issues before they escalate into failures, extending asset life, and reducing failure frequency.
• Scalability and Cost Efficiency: The use of serverless computing reduces infrastructure costs while maintaining high performance.
• Comprehensive Visualization: Offering user-friendly dashboards that enhance decision-making, providing stakeholders with actionable intelligence for optimizing energy conversion processes.

Quantitatively, the AHI-based monitoring system achieved a 43% average reduction in unexpected failures across the three converters and improved maintenance scheduling accuracy by over 30%, as measured by alignment between predicted and actual overhaul needs.

This architecture represents a cost-effective, scalable, and high-performance solution for implementing digital twins in energy converters, leveraging Azure’s powerful suite of cloud-based tools to drive operational efficiency and reliability.

3.5. Implementation Guidelines for Practitioners

Based on the case study and implementation experience, we suggest the following steps for practitioners aiming to adopt the proposed AHI-integrated digital twin framework in their own organizations:

• Select a Target Asset Class: Begin with critical or high-maintenance assets (e.g., power converters, pumps, transformers) that already have sensor infrastructure or can feasibly be instrumented.
• Map Asset Structure and Data Sources in a data model for Digitalization: Define the physical hierarchy and identify available data sources (e.g., SCADA, PLCs, IoT sensors) and the relations among them. Ensure coverage of electrical, thermal, mechanical, and maintenance-related variables.
• Deploy a Scalable Data Pipeline: Use a cloud platform (e.g., Azure, AWS, GCP) to centralize and structure data flows. Implement ingestion, transformation, and storage layers compatible with real-time and historical processing.
• Integrate the AHI Model: Implement the AHI computation logic using rule-based engines or cloud functions. Calibrate the model using historical failure data and expert knowledge to define health modifiers and thresholds.
• Visualize Results for Decision Support: Create dashboards (e.g., with Power BI) to display AHI scores, trends, and alerts. Integrate these outputs into maintenance workflows or ERP/CMMS systems if possible.
• Train Stakeholders and Monitor Performance: Educate maintenance teams on interpreting AHI results. Monitor system outputs over time to refine health rules, improve model robustness, and support continuous improvement.

These steps ensure a structured, standards-aligned implementation that can be scaled across asset types and organizational contexts.

4. Conclusions

This paper has introduced a structured and standards-aligned data model designed to enable the integration of Asset Health Index (AHI) methodologies within Digital Twin architectures for industrial energy converters. The model addresses the growing need for scalable and interoperable solutions to support predictive maintenance, real-time condition monitoring, and risk-informed decision-making.

The proposed framework organizes asset-related data into interconnected domains covering asset definition, criticality assessment, operational monitoring, intelligent management, and maintenance planning. This structure provides the necessary infrastructure to compute health indices dynamically and facilitates integration with digital platforms.

The model was formalized through a Unified Modeling Language (UML) class diagram and implemented using a cloud-native architecture based on Microsoft Azure. This implementation demonstrates the model’s applicability in real industrial environments, enabling seamless data ingestion, automated AHI calculation, and health visualization via business intelligence tools.

A case study involving high-capacity DC/AC converters operating under different environmental and usage conditions validated the model’s practical value. The results confirmed the AHI’s ability to differentiate asset behavior over time, supporting early degradation detection, maintenance optimization, and extension of asset life. In particular, the study highlighted how stressors such as temperature, load variation, and maintenance history shape the health trajectories of otherwise identical assets—emphasizing the relevance of contextualized digital representations.

The key contributions of this study are threefold:

• the development of a structured and standard-compliant data model for AHI integration;
• the demonstration of its deployment within a scalable, cloud-based digital twin architecture; and
• the validation of its practical utility through an empirical industrial case study.

To support adoption in real-world settings, the paper provides implementation recommendations for practitioners. These include: aligning with standards such as ISO 14224 and RAMI 4.0, selecting pilot assets for initial deployment, ensuring high-quality data for AHI modifiers, and using business intelligence tools to support transparent maintenance decisions.

Looking ahead, future research will explore the integration of machine learning techniques to dynamically refine AHI modifiers and enhance predictive accuracy. Additional efforts are also needed to generalize the framework to other asset types and industrial domains and to quantify the economic and operational benefits of AHI-driven asset management. While the current case confirms the model’s utility in capturing degradation patterns and enabling condition-based interventions, further validation is required—particularly in large-scale deployments and under uncertain or noisy data conditions.

Ultimately, the convergence of digital twin technologies, cloud infrastructure, and advanced analytics is opening new avenues for autonomous and intelligent asset management. The proposed framework offers a replicable and future-ready foundation for organizations pursuing digital transformation in the energy sector and beyond.