Fault Detection And Diagnostics Research Articles

Transformer encoder based self-supervised learning for HVAC fault detection with unlabeled data

Data driven methods are the most studied fault detection and diagnostics (FDD) type in buildings HVAC systems. However, most studies rely on labeled data for specific faults which are hard to find and collect for real systems. While the fault-free data is easier to collect, it is still time consuming to label for large systems operation. Moreover, most of the studies rely on the usage of supervised learning algorithms which do not generalize well beyond the training data making unseen faults hard to detect. In this paper, we define a methodology to use a self-supervised learning method for HVAC systems' FDD using a Transformer encoder, moreover, we tested it on a real case study. By strategically masking portions of the multivariate time-series data using Markov chain approach with two states. The model is trained by predicting these concealed segments. This approach, independent of labeled data, offers a scalable solution for practical HVAC applications. Anomalies are labeled using the Peak Over Threshold (POT) method, which dynamically determines thresholds by fitting reconstruction errors to a generalized Pareto distribution. Subsequent fault diagnostics emphasize features with pronounced reconstruction errors, pinpointing potential HVAC malfunctions. This methodology reduces dependence on labeled datasets and augments the model's generalization, facilitating detection of unobserved faults. This approach was applied to data from a real building. As a results multiple faults were detected mainly due to the malfunctioning of the monitoring system. The model demonstrates the ability to detect both sequential and individual faults. The period from October 19th to December 23rd was detected as a fault period due to the change in the trend of the data because of the monitoring system.

Implementing Industry 4.0: An In-Depth Case Study Integrating Digitalisation and Modelling for Decision Support System Applications

The scientific community has shown considerable interest in Industry 4.0 due to its capacity to revolutionise the manufacturing sector through digitalisation and data-driven decision-making. However, the actual implementation of Industry 4.0 within complex industrial settings presents obstacles that are typically beyond the scope of mainstream research articles. In this paper, a comprehensive case-study detailing our collaborative partnership with a leading medical device manufacturer is presented. The study traces its evolution from a state of limited digitalisation to the development of a digital intelligence platform that leverages data and machine learning models to enhance operations across a wide range of critical machines and assets. The main business objective was to enhance the energy efficiency of the manufacturing process, thereby improving its sustainability measures while also saving costs. The project encompasses energy modelling and analytics, Fault Detection and Diagnostics (FDD), renewable energy integration and advanced visualisation tools. Together, these components enable informed decision making in the context of energy efficiency.

Deep Learning Approach to Within-Bank Fault Detection and Diagnostics of Fine Motion Control Rod Drives

Control rod motion is one of the primary means of regulating the rate of fission in a nuclear reactor core to ensure safe and stable operation. Reactor power distribution and thermal power output can be fine-tuned by adjusting the control rod position. For high-precision control of rod movements, Fine Motion Control Rod Drives (FMCRDs) are often used. The operation of FMCRDs provides a unique opportunity to implement condition monitoring related to the intermittency of motion and the use of control rod banks. This research sets out to detect three types of faults in an electrically driven FMCRD. In addition to detecting faults, this work will attempt to determine both the type of fault and the source of each fault, completing the fault detection and diagnostics (FDD) pipeline on a scarcely researched system. The three types of faults to be investigated are short-circuit faults, ball screw wear faults, and ball screw jam faults. This is a potential advancement to the within-bank FDD of this specific drive system intended for deployment in an advanced nuclear reactor plant. Using encoder-decoder structured convolutional neural networks and autoencoders, the three tested faults were confidently detected and isolated as well as reasonably diagnosed by monitoring the FMCRD servomotor torque.

Machine Learning-Based Automated Fault Detection and Diagnostics in Building Systems

Automated fault detection and diagnostics analysis in commercial building systems using machine learning (ML) can improve the building’s efficiency and conserve energy costs from inefficient equipment operation. However, ML can be challenging to implement in existing systems due to a lack of common data standards and because of a lack of building operators trained in ML techniques. Additionally, results from ML procedures can be complicated for untrained users to interpret. Boolean rule-based analysis is standard in current automated fault detection and diagnostics (AFDD) solutions but limits analysis to the rules defined and calibrated by energy engineers. Boolean rule-based analysis and ML can be combined to create an effective fault detection and diagnostics (FDD) tool. Three examples of ML’s advantages over rule-based analysis are explored by analyzing functional building equipment. ML can detect long-term faults in the system caused by a lack of system maintenance. It can also detect faults in system components with incomplete sets of sensors by modeling expected system operations and by making comparisons to actual system operations. An example of ML detecting a failure in a building is shown along with a demonstration of the soft decision boundaries of ML-based FDD compared to Boolean rule-based FDD analysis. The results from the three examples are used to demonstrate the strengths and weaknesses of using ML for AFDD analysis.

Implementing a Digital Twin-based fault detection and diagnosis approach for optimal operation and maintenance of urban distributed solar photovoltaics

Over the next decades, solar energy power generation is anticipated to gain popularity because of the current energy and climate problems and ultimately become a crucial part of urban infrastructure. Furthermore, in order to facilitate the rapid deployment of solar production in sustainable cities, solar technologies must be digitalized and smart. This will increase their reliability and ensure their optimal performance. This work fits into this perspective and focuses on the effectiveness of processing data using the Digital Twin (DT) framework as a state-of-art information technology for urban distributed solar PV systems. Accordingly, the present paper deals with the DT of five Building-added Photovoltaic (BAPV) systems installed at the solar village in Benguerir city, Morocco. The 3D models were generated in the Rhinoceros platform, the solar AC electrical production of each photovoltaic system is calculated by the ladybug-plugin-integrated statistical model, within the grasshopper environment. Additionally, data assessment and visualization are processed through Python. Then, a hybrid model-based and data-driven fault detection and diagnosis (FDD) approach is proposed to identify and isolate anomalies for decentralized solar PV systems at the urban scale using monitoring and inspection techniques, namely Remote Sensors (RS) and real-time solar production monitoring system. The application of the DT concept for complex dynamic systems has shown its effectiveness in ensuring optimal operating conditions for the energy systems by measuring the spatiotemporal energy performance of distributed solar PV systems at an urban scale, evolving with the changing environment, and identifying degradation based on anomalies occurrence; Only then when the physical models need to be amended. Therefore, the main goal behind this work is to advance knowledge regarding research on condition monitoring of solar systems in cities while considering issues related to anomalies like the degradation phenomenon in PV modules over time. The findings of this study have proved that the state of health and monitoring of PV systems could be successfully detected and efficiently monitored using the combination of real time monitoring and remote aerial sensing.

Deployment of real-time building automation system-integrated inverse-model-based fault detection and diagnostics algorithms

The complex operation of HVAC systems in large commercial buildings warrants regular implementation of advanced analytical approaches to operations and maintenance, and subsequent corrective measures to improve and maintain optimal energy performance. Despite the established capabilities of data-driven fault detection and diagnostics (FDD) to identify suboptimal controls policies and mechanical faults resulting in poor energy performance, few attempts have been made to deploy scalable solutions around these approaches. Furthermore, real-time BAS-integrated FDD methods are predominantly rule-based, offering limited insights to faults with gradual negative impacts to energy performance. This paper demonstrates the application of various established data-driven, inverse-model-based FDD methodologies in a BAS-integrated environment. Traditionally implemented sparingly, the novelty of recursive and automatic execution of advanced FDD methodologies, facilitated through a direct data pipeline to an existing BAS, capitalizes on the BAS’s real-time monitoring capabilities to enable continuously refreshed inverse model generation that can capture the gradual degradation of building performance, and provide up-to-date actionable visualizations and key performance indicators (KPI) to building operators. Since deployment, the application has successfully identified a scheduling fault on two separate occasions in a case study building in Ottawa, Canada, and the visualizations were presented to the building operators who resolved the issues.

A comparative investigation between rule- and inverse model-based fault detection and diagnostics for HVAC control systems

Fault detection and diagnostics (FDD) tools provide valuable information regarding system faults and deviation from expected operation. Most existing FDD tools apply rule-based fault detection algorithms that generate an alarm when a rule is met; however, these tools cannot evaluate the overall performance of a system. Inverse-model-based FDD algorithms can be deployed to complement the fault alarms triggered by rule-based building energy management systems (BEMS). This paper examines the faults detected by rule- and inverse model-based algorithms used to detect faults in multiple zone variable air volume air handling unit systems. The capability of the rule- and inverse model-based algorithms in detecting and diagnosing faults is demonstrated through illustrative examples using data from three commercial buildings in New Brunswick, Canada. The results show that inverse model-based algorithms could diagnose faults that were not detected by the rule-based FDD algorithms implemented in a commercially available BEMS tool.

Fault detection and diagnosis in AHU system using deep learning approach

Energy consumption in buildings increases with the failures of equipment involved in the energy exchange, and control networks in buildings. One of the ways to remedy this issue is to offer high-performance fault detection systems. This article proposes a Fault Detection and Diagnostics (FDD) system based on Convolutional Neural Network (CNN) and Long Term Short Memory (LSTM) neural networks, applied to an AHU and using an hybrid database containing data from simulation and real-world on an actual physical building. The proposed system is designed to effectively identify and categorize faults, whether they occur in the sensors or in the mechanical equipment responsible for critical functions such as heat exchanges, air transfer, and system control. The FDD system provides results with an overall accuracy of around 96.88 %.

Dynamic model-based feature extraction for fault detection and diagnosis of a supermarket refrigeration system†

With the increasing concerns over climate change and carbon emissions, fault detection and diagnostics (FDD) of low–global warming potential (GWP) refrigerant supermarket refrigeration systems has gained great attention from academic and industrial sectors. Various FDD approaches have been developed to detect, identify, and diagnose faults to save energy, improve food quality, and protect the environment. To mitigate the difficulty of collecting high-quality steady-state operational data in field operations faced by most model-based FDD methods, this study developed dynamic models of a low–GWP refrigerant (CO2) supermarket refrigeration system. The model accuracy was validated using manufacturer data and experimental data. Simulations were conducted to predict the system dynamic response under two common operational faults—evaporator air path blockage fault and the display case door open fault—to identify fault patterns and define key dynamic behavior indexes for supporting FDD algorithm development.

Fault detection and diagnostics in the context of sparse multimodal data and expert knowledge assistance: Application to hydrogenerators

Deep learning-based methods for industrial fault detection and diagnostics (FDD) depend strictly on good quality and sufficient quantity of condition monitoring data. However, in real-world industrial settings, data collection is usually limited, leading to sparse and insufficient data to train a data-driven model. Therefore, this work proposes a new methodology to address this issue by leveraging multimodal data anddomain knowledge to develop a data-driven solution. Particularly for large, complex machinery, unimodal sensors may not fully capture the health state information. In such cases, multimodal data may provide complementary insights into the machine degradation. However, challenges mentioned above need to be addressed before these data can be useful. The multimodal learning method presented within the methodology can benefit from useful information from different data modalities and from domain expert knowledge, even when these data are of low volume. The performance of the proposed methodology is investigated through a real industrial case study involving energy production systems. The obtained results demonstrate the potential of the proposed methodology in augmenting the FDD accuracy and tackling the sparse data challenge.

Deployment of crowdsourced occupant data to support fault detection and diagnosis in buildings

This proof-of-concept paper proposes a method that integrates occupants' feedback and personal data to enhance the performance of fault detection and diagnostic (FDD) technology in buildings. The proposed approach uses a smartwatch app called Cozie to collect occupants' subjective feedback on thermal comfort and air quality, and the data is transferred to a cloud database through a data stream. Similarly, an application programming interface is used to retrieve sensor readings from the building automation system. FDD rulesets then use the data to determine process variable errors and report alarms. The results of an experiment show that occupant input accurately prove/disproved alarms and allowed for a faster and more accurate FDD. The occupants' feedback also helped prompt rule-based FDD algorithms and support the alarms triggered. Two independent surveys were also conducted. According to the first survey, occupants tend not to report discomfort to building operators through conventional methods unless the indoor temperature is more than 3 °C warmer or cooler than their desired temperature. The proposed method helped occupants report discomfort issues earlier, and the percentage of occupants choosing 3 °C warmer or colder conditions decreased by 29%. The paper concludes that the integration of occupants' feedback and personal data can significantly enhance FDD performance and provide more comfortable environments for building occupants.

Semi-Supervised Machine Learning for Fault Detection and Diagnosis of a Rooftop Unit

Most heating, ventilation, and air-conditioning (HVAC) systems operate with one or more faults that result in increased energy consumption and that could lead to system failure over time. Today, most building owners are performing reactive maintenance only and may be less concerned or less able to assess the health of the system until catastrophic failure occurs. This is mainly because the building owners do not previously have good tools to detect and diagnose these faults, determine their impact, and act on findings. Commercially available fault detection and diagnostics (FDD) tools have been developed to address this issue and have the potential to reduce equipment downtime, energy costs, maintenance costs, and improve occupant comfort and system reliability. However, many of these tools require an in-depth knowledge of system behavior and thermodynamic principles to interpret the results. In this paper, supervised and semi-supervised machine learning (ML) approaches are applied to datasets collected from an operating system in the field to develop new FDD methods and to help building owners see the value proposition of performing proactive maintenance. The study data was collected from one packaged rooftop unit (RTU) HVAC system running under normal operating conditions at an industrial facility in Connecticut. This paper compares three different approaches for fault classification for a real-time operating RTU using semi-supervised learning, achieving accuracies as high as 95.7% using few-shot learning.

A review of data-driven fault detection and diagnostics for building HVAC systems

With the wide adoption of building automation system, and the advancement of data, sensing, and machine learning techniques, data-driven fault detection and diagnostics (FDD) for building heating, ventilation, and air conditioning systems has gained increasing attention. In this paper, data-driven FDD is defined as those that are built or trained from data via machine learning or multivariate statistical analysis methods. Following this definition, this paper reviews and summarizes the literature on data-driven FDD from three aspects: process, systems studied (including the systems being investigated, the faults being identified, and the associated data sources), and evaluation metrics. A data-driven FDD process is further divided into the following steps: data collection, data cleansing, data preprocessing, baseline establishment, fault detection, fault diagnostics, and potential fault prognostics. Literature reported data-driven methods used in each step of an FDD process are firstly discussed. Applications of data-driven FDD in various HVAC systems/components and commonly used data source for FDD development are reviewed secondly, followed by a summary of typical metrics for evaluating FDD methods. Finally, this literature review concludes that despite the promising performance reported in the literature, data-driven FDD methods still face many challenges, such as real-building deployment, performance evaluation and benchmarking, scalability and transferability, interpretability, cyber security and data privacy, user experience, etc. Addressing these challenges is critical for a broad real-building adoption of data-driven FDD.

A literature review and future research agenda on fault detection and diagnosis studies in marine machinery systems

Fault detection and diagnostics (FDD) have great potential to enable safety, efficiency, and reliability measures of critical machinery systems. However, it is clear that there is a lack of systematic literature review to identify and classify the FDD studies conducted within the scope of marine engineering. This paper offers a systematic review of FDD models particular to marine machinery and systems. The numbers of 72 core articles were highlighted through a comprehensive literature review conducted in the 2002–2022 period. The studies are classified based on the mostly utilized methods such as data-driven, model-based, knowledge-based, and new generation-hybrid. In addition, new generation and hybrid methods are discussed in detail. The experimental environment (i.e. shipboard, labs, simulator) and technical details of the conducted studies are extensively discussed. While 56.94% of the examined studies are related to the main engine, 43.06% of them are related to auxiliary engines. In addition, the main and auxiliary engine studies are also divided into subject headings and examined in detail. Given the recent developments in green and smart maritime concepts, a future research agenda of the FDD studies on marine machinery systems is then pinpointed. Consequently, the study stimulates scholars interested in FDD while it enables innovative ideas for marine engineers, technology providers, ship operators, and maritime entrepreneurs.

Implementation and test of an automated control hunting fault correction algorithm in a fault detection and diagnostics tool

Control hunting due to improper proportional–integral–derivative (PID) parameters in the building automation system (BAS) is one of the most common faults identified in commercial buildings. It can cause suboptimal performance and early failure of heating, ventilation, and air conditioning (HVAC) equipment. Commercial fault detection and diagnostics (FDD) software represents one of the fastest growing market segments in smart building technologies in the United States. Implementation of PID retuning procedures as an auto-correction algorithm and integration into FDD software has the potential to mitigate control hunting across a heterogeneous portfolio of buildings with different BAS in a scalable way. This paper presents the development, implementation, and field testing of an automated control hunting fault correction algorithm based on lambda tuning open-loop rules. The algorithm was developed in a commercial FDD software and successfully tested among nine variable air volume boxes in an office building in the United States. The paper shows the feasibility of using FDD tools to automatically correct control hunting faults, discusses scalability considerations, and proposes a path forward for the HVAC industry and academia to further improve this technology.

Norm-based data labelling in supervised learning for fault detection and diagnostics of rotating elements towards maintenance servitisation

In the current industrial context, the wide availability of data from the shopfloor is enabling companies to develop condition-based maintenance (CBM) and predictive maintenance (PdM) solutions towards production performance improvements. However, the path is not straightforward and several technological and managerial challenges have to be faced. Specifically, the current challenge is to mix the high-performance, yet difficult-to-interpret results, of AI (Artificial Intelligence) algorithms with the vast available domain knowledge provided by scientific literature and norms. It is the goal of this work to propose a norm-based data labelling to implement a supervised model which leverages on time-domain features to guarantee the interpretability of results for maintenance operators and technicians for FDD (Fault Detection and Diagnostics). The proposed approach is tested and a complete CBM solution is deployed in a case of an OEM (Original Equipment Manufacturer) of rotating elements. Through it, the company could move towards a fully-fledged maintenance service offering, already integrating norm-related knowledge.

Data driven fault detection and diagnostics for hydronic and monitoring systems in a residential building

Buildings are responsible for 40% of the global energy use and associated with up to 30% of the total CO2 emissions. The drive to reduce the environmental impact of the built environment was the catalyst to the increasing installation of meters and sensors to monitor the energy use and environmental monitoring. This is key to cost effective Fault Detection and Diagnostics (FDD) which guarantees enhanced thermal comfort for the occupants and reduction in energy use. Most of FDD research work in buildings was focused on the commercial buildings due to the higher consumption and the higher saving potential, while limited work was directed towards residential buildings. This paper investigates the usage of two supervised machine learning algorithms, namely Random Forest, K nearest neighbour, to detect and diagnose twelve faults in both the monitoring system of the indoor/outdoor conditions, and the hydronic circuit inside an apartment located in Milan using minimal features that are easy to access and inexpensive to monitor to cut down in both computational and financial costs. The thermal zones are being conditioned using an electric air to water heat pump connected to fan coils for cooling and radiant floor for heating. The faults include valve leakage, faulty temperature sensors and recirculating pump’s inadequate flow rate. The faults were modelled in a Modelica based detailed model of the apartment. After tuning the hyper-parameters of all three algorithms, the Receiver Operator Characteristics curve for each fault were compared for each algorithm to compare the optimal one to be used. The Random Forest algorithms showed the highest accuracy with almost 89% across the twelve faults. Generalization of the trained algorithm across different weathers were tested but the results were not promising.

Sensor cost-effectiveness analysis for data-driven fault detection and diagnostics in commercial buildings

Data-driven building fault detection and diagnostics (FDD) is heavily dependent on sensors. However, common sensors from Building Automation Systems are not optimized to maximize accuracy in FDD. Installing additional sensors that provide more detailed building system information is key to maximizing the performance of FDD solutions. In this paper, we present a sensor cost analysis workflow to quantify the economic implications of installing new sensors for FDD using the concept of sensor threshold marginal cost (STMC). STMC does not represent actual sensor cost. Rather, it represents a target cost based on the economic benefit that would be realized through improved FDD performance and one or more specified economic criteria. We calculate STMCs for multiple possible fault types and use fault prevalence information to aggregate STMCs into a single dollar value to determine the cost-effectiveness of a potential sensor investment. We conducted a case study using Oak Ridge National Laboratory's Flexible Research Platform (FRP) test facility as a reference. The case study demonstrates the feasibility of the analysis and highlights the key cost considerations in sensor selection for FDD. The results also indicate that identifying and installing the few key sensor(s) is critical to cost-effectively improve FDD performance.

Building Analytics Tool Deployment at Scale: Benefits, Costs, and Deployment Practices

Buildings are becoming more data-rich. Building analytics tools, including energy information systems (EIS) and fault detection and diagnostic (FDD) tools, have emerged to enable building operators to translate large amounts of time-series data into actionable findings to achieve energy and non-energy benefits. To expedite data analytics adoption and facilitate technology innovation, building owners, technology developers, and researchers need reliable cost–benefit data and evidence-based guidance on deployment practices. This paper fulfills these needs with the energy use and survey data from a wide-ranging research and industry partnership program that covers thousands of buildings installed with analytics tools. The paper indicates that after two years of implementation, organizations using FDD tools and EIS tools achieved 9% and 3% median annual energy savings, respectively. The median base cost and annual recurring cost for FDD are USD 0.65 per square meter (m2) (USD 0.06 per square foot [ft2]) and USD 0.22 per m2 (USD 0.02 per ft2), and are USD 0.11 per m2 (USD 0.01 per ft2) and USD 0.11 per m2 (USD 0.01 per ft2) for EIS. The common metrics and analyses that are used in the tools to support the discovery of energy efficiency measures are summarized in detail. Two best practice examples identified to maximize the benefits of tool implementation are also presented. Opportunities to advance the state of technology include simplified data integration and management, and more efficient processes for acting on analytics outputs. Compared with previous efforts in the literature, the findings presented in this paper demonstrate the effectiveness of building analytics tools with the largest known dataset.

A holistic sequential fault detection and diagnostics framework for multiple zone variable air volume air handling unit systems.

A holistic fault detection and diagnostics (FDD) method should explicitly consider the dependencies between faults at the system- and zone-level to isolate the root cause. A system-level fault can trigger false alarms at the zone-level, while concealing the presence of a zone-level fault. However, most FDD methods have focused on a single component/equipment without considering the importance of the interactions between zone- and system-level devices. This paper proposes a holistic hierarchical framework for FDD, combining the process of detection and diagnosis of controls hardware and sequencing logic faults affecting the actuators at the system- and zone-level. The proposed framework follows a holistic sequential procedure to diagnose faults and suppress false alarms in this order: hard faults in air handling units (AHUs), hard faults in variable air volume (VAV) zones, sequencing logic faults in AHUs, and sequencing logic faults in VAV zones. The detection of faults is performed by visualizing the discrepancies between the expected and measured operational behaviour of AHUs and VAV boxes. Examples demonstrating the framework are provided with data from 10 different VAV AHU systems.Practical application: This paper provides a sequential hierarchical FDD framework to address two main issues in VAV AHU systems: detectability and significance. Regarding detectability, the framework prioritizes hard faults over sequencing logic faults to avoid false positives and false negatives; about significance, system-level faults are prioritized over zone-level faults to triage high-impact faults in the system. The detection of faults is performed via visualizing the biases from the expected behaviour of AHU and VAV characteristics to provide an envisioning interpretation for the experts in facilities management in commercial buildings to find the root cause of the fault and fix them on-site.