A data-driven diagnosis system is developed for fault diagnosis of a fuel injection system in a heavy-duty diesel truck. Physical insights and standard component modeling are used to derive a structural model of the fuel injection system. Based on structural analysis, a set of data-driven residual generator candidates is derived, both linear and nonlinear models, and trained using nominal training data from a truck. Evaluations of different fault scenarios show that the proposed models can distinguish between different faults and show the potential of utilizing basic physical insights in data-driven fault diagnosis design. Copyright (c) 2024 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

Fault diagnosis of a certain class of hybrid systems referred to as structurally reconfigurable (SR) systems is complicated. This is because SR systems tend to switch their configuration, which may or may not be faulty. It is important to identify the mode of the SR system along with the corresponding fault if any, in order to facilitate a fault tolerant action. This paper combines discrete fault diagnosis with mode identification for SR systems to achieve two main objectives: Sensor selection for fault detection, isolation and mode identification, and residual selection for mode identification. The framework is built using a structural analysis-based approach to meet these objectives. This framework is demonstrated for a 10-speed Automatic Transmission, which is an illustrative example of SR systems.

Fault diagnosis of structurally re-configurable systems is complicated as the system structure changes when the system operates in different modes. It is important that faults can be detected and isolated in each operating mode. In model-based diagnosis, faults are detected and isolated by detecting inconsistencies between model predictions and sensor data. Thus, determining where to mount sensors is an important task to be able to detect and isolate faults, especially when faults can result in unexpected system re-configuration. For structurally re-configurable systems this means selecting a set of sensors that fulfills requirements in multiple models describing the different system modes. A sensor selection algorithm is proposed for structurally re-configurable systems which computes minimal sensor sets that make faults in all modes detectable and isolable. As a case study, the sensor selection algorithm is applied to determine sensor locations in an eight-speed automatic transmission.

Recent improvements in vehicle-to-everything (V2X) communication and onboard computing power have enabled the development of control algorithms that jointly optimize the vehicle velocity and powertrain control in Connected and Automated Vehicles (CAVs), commonly referred to as the Eco-Driving problem. This paper presents a novel and computationally efficient algorithm to optimize the velocity planning and energy management in a CAV with a hybrid electric powertrain. The Eco-Driving problem is formulated as a dynamic, constrained optimization problem in the spatial domain, where information about the upcoming speed limits and road topography is assumed known. This problem is solved by embedding an Equivalent Consumption Minimization Strategy (ECMS) into a Dynamic Programming (DP) optimization to obtain a sub-optimal solution that provides results close to the global optimum at a fraction of the computational cost. Further, a multi-layer hierarchical control architecture is proposed as a path to a causal, real-time implementation. The DP-ECMS algorithm is converted into a Model Predictive Control (MPC) framework by using principles of Approximate Dynamic Programming (ADP). This causal implementation is finally benchmarked to a global optimal solution obtained with DP for different scenarios.

Our society depends on advanced and complex technical systems and machines, for example, cars for transportation, industrial robots in production lines, satellites for communication, and power plants for energy production. Consequences of a fault in such a system can be severe and result in human casualties, environmentally harmful emissions, high repair costs, or economical losses caused by unexpected stops in production lines. Thus, a diagnosis system is important, and in some applications also required by legislations, to monitor the system health in order to take appropriate preventive actions when a fault occurs. Important properties of diagnosis systems are their capability of detecting and identifying faults, i.e., their fault detectability and isolability performance.

This thesis deals with quantitative analysis of fault detectability and isolability performance when taking model uncertainties and measurement noise into consideration. The goal is to analyze diagnosability performance given a mathematical model of the system to be monitored before a diagnosis system is developed. A measure of fault diagnosability performance, called distinguishability, is proposed based on the Kullback-Leibler divergence. For linear descriptor models with Gaussian noise, distinguishability gives an upper limit for the fault to noise ratio of any linear residual generator. Distinguishability is used to analyze fault detectability and isolability performance of a non-linear mean value engine model of gas flows in a heavy duty diesel engine by linearizing the model around different operating points.

It is also shown how distinguishability is used for determine sensor placement, i.e, where sensors should be placed in a system to achieve a required fault diagnosability performance. The sensor placement problem is formulated as an optimization problem, where minimum required diagnosability performance is used as a constraint. Results show that the required diagnosability performance greatly affects which sensors to use, which is not captured if not model uncertainties and measurement noise are taken into consideration.

Another problem considered here is the on-line sequential test selection problem. Distinguishability is used to quantify the performance of the different test quantities. The set of test quantities is changed on-line, depending on the output of the diagnosis system. Instead of using all test quantities the whole time, changing the set of active test quantities can be used to maintain a required diagnosability performance while reducing the computational cost of the diagnosis system. Results show that the number of used test quantities can be greatly reduced while maintaining a good fault isolability performance.

A quantitative diagnosability analysis has been used during the design of an engine misfire detection algorithm based on the estimated torque at the flywheel. Decisions during the development of the misfire detection algorithm are motivated using quantitative analysis of the misfire detectability performance. Related to the misfire detection problem, a flywheel angular velocity model for misfire simulation is presented. An evaluation of the misfire detection algorithm show results of good detection performance as well as low false alarm rate.

Engine misfire detection is an important part of the On-Board Diagnostics (OBDII) legislations to reduce exhaust emissions and avoid damage to the catalytic converters. The flywheel angular velocity signal is analyzed, investigating how to use the signal in order to best detect misfires. An algorithm for engine misfire detection is proposed based on the flywheel angular velocity signal. The flywheel signal is used to estimate the torque at the flywheel and a test quantity is designed by weighting and thresholding the samples of estimated torque related to one combustion. During the development process, the Kullback-Leibler divergence is used to analyze the ability to detect a misfire given a test quantity and how the misfire detectability performance varies depending on, e.g., load and speed. The Kullback-Leibler divergence is also used for parameter optimization to maximize the difference between misfire data and fault-free data. Evaluation shows that the proposed misfire detection algorithm is able to have a low probability of false alarms while having a low probability of missed detections.

A flywheel angular velocity model for misfire and disturbance simulation is presented. Applications of the model are, for example, initial parameter calibration and robustness analysis of misfire detection algorithms. An analytical cylinder pressure model is used to model cylinder torque and a multi-body model with torsional flexibilities is used to model crankshaft and driveline oscillations. Misfires, cylinder variations, changes in auxiliary load, and flywheel manufacturing errors can be injected in the model and the resulting speed variations can be simulated. A qualitative validation of the model shows that simulated angular velocity captures the amplitude and oscillatory behavior of measurement data and the effects of different phenomena, such as misfire and flywheel manufacturing errors.

Analyzing fault diagnosability performance for a given model, before developing a diagnosis algorithm, can be used to answer questions like “How difficult is it to detect a fault f_i?” or “How difficult is it to isolate a fault f_i from a fault f_j?”. The main contributions are the derivation of a measure, distinguishability, and a method for analyzing fault diagnosability performance of discrete-time descriptor models. The method, based on the Kullback–Leibler divergence, utilizes a stochastic characterization of the different fault modes to quantify diagnosability performance. Another contribution is the relation between distinguishability and the fault to noise ratio of residual generators. It is also shown how to design residual generators with maximum fault to noise ratio if the noise is assumed to be i.i.d. Gaussian signals. Finally, the method is applied to a heavy duty diesel engine model to exemplify how to analyze diagnosability performance of non-linear dynamic models.

A sequential test selection algorithm is proposed which updates the set of active test quantities depending on the present minimal candidates. By sequentially updating the set of active test quantities, computational time and memory usage can be reduced. If test quantities are generated on-line, a sequential test selection algorithm gives information about which test quantities that should be created. The test selection problem is defined as an optimization problem where a set of active test quantities is chosen such that the cost is minimized while the set fulfills a required minimum detectability and isolability performance. A quantitative diagnosability measure, distinguishability, is used to quantify diagnosability performance of test quantities. The proposed test selection algorithm is applied to a DC-circuit where the diagnosis algorithm generates residuals on-line. Experiments show that the sequential test selection algorithm can significantly reduce the number of active test quantities during a scenario and still be able to identify the true faults.

A theory is developed for quantifying fault detectability and fault isolability properties of time discrete linear dynamic models. Based on the model, a stochastic characterization of system behavior in different fault modes is defined and a general measure, called distinguishability, based on the Kullback-Leibler information, is used to quantify the difference between the modes. An analysis of distinguishability as a function of the number of observations is discussed. This measure is also shown to be closely related to the fault to noise ratios in residual generators. Further, the distinguishability of the model is shown to give upper limits of the fault to noise ratios of residual generators.

A theory is developed for quantifying fault detectability and fault isolability properties of static linear stochastic models. Based on the model, a stochastic characterization of system behavior in different fault modes is defined and a general measure, based on the Kullback-Leibler information, is proposed to quantify the difference between the modes. This measure, called distinguishability, of the model is shown to give sharp upper limits of the fault to noise ratios of residual generators. Finally, a case-study of a diesel engine model shows how the general framework can be applied to a dynamic and nonlinear model.

A good placement of sensors is crucial to get good performance in detecting and isolating faults. Here, the sensor placement problem is cast as a minimal cost optimization problem. Previous works have considered this problem with qualitative detectability and isolability specifications. A key contribution here is that quantified detectability and isolability performance is considered in the optimization formulation. The search space for the posed optimization problem is exponential in size, and to handle complexity a greedy optimization algorithm that compute optimal sensor positions is proposed. Two examples illustrate how the optimal solution depends on the required quantified diagnosability performance and the results are compared to the solutions using a deterministic method.

Structural methods in model-based fault diagnosis applications are simple and efficient tools for finding candidates for residual generation. However, the structural methods do not take model uncertainties and information about fault behavior into consideration. This may result in selecting residual generators with bad performance to be included in the diagnosis system. By using the Kullback-Leibler divergence, the performance of different residual generators can be compared to find the best one. With the ability to quantify diagnostic performance, the design of residual generators can be optimized by, for example, combining several residual generators such that the diagnostic performance is maximized. The proposed method for residual generation selection is applied to a water tank system to show that the achieved residual performance is improved compared to only use a structural method.

With trends as IoT and increased connectivity, the availability of data is consistently increasing and its automated processing with, e.g., machine learning becomes more important. This is certainly true for the area of fault diagnostics and prognostics. However, for rare events like faults, the availability of meaningful data will stay inherently sparse making a pure data-driven approach more difficult. In this paper, the question when to use model-based, data-driven techniques, or a combined approach for fault diagnosis is discussed using real-world data of a permanent magnet synchronous machine. Key properties of the different approaches are discussed in a diagnosis context, performance quantified, and benefits of a combined approach are demonstrated.

To facilitate the use of advanced fault diagnosis analysis and design techniques to industrial sized systems, there is a need for computer support. This paper describes a Matlab toolbox and evaluates the software on a challenging industrial problem, air-path diagnosis in an automotive engine. The toolbox includes tools for analysis and design of model based diagnosis systems for large-scale differential algebraic models. The software package supports a complete tool-chain from modeling a system to generating C-code for residual generators. Major design steps supported by the tool are modeling, fault diagnosability analysis, sensor selection, residual generator analysis, test selection, and code generation. Structural methods based on efficient graph theoretical algorithms are used in several steps. In the automotive diagnosis example, a diagnosis system is generated and evaluated using measurement data, both in fault-free operation and with faults injected in the control-loop. The results clearly show the benefit of the toolbox in a model-based design of a diagnosis system. Latest version of the toolbox can be downloaded at faultdiagnosistoolbox.github.io. (C) 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.

Several technological challenges delay the adoption of electrified powertrains in the heavy-duty transport sector. For fuel-cell hybrid electric trucks, key issues include slow cold start, reduced cooling power during high ambient temperatures, and uncertainties regarding durability. In addition, the engineers must handle the complexity of the system. In this article, a Matlab/Simulink library is introduced, which has been developed to aid engineers in the design and optimization of energy management systems and strategies of this complex system that consider mechanical, electrochemical, and thermal energy flows. The library is introduced through five example vehicle models, and through case studies that highlight the various kinds of analysis that can be performed using the provided models. All library code is open source, open for commercial use, and runs in Matlab/Simulink without any need for external libraries.

A generalized fault isolability matrix is proposed for quantitative analysis of fault isolability properties. The original fault isolability matrix gives information about which faults that are isolable from each other. However, other relavant isolability properties are not visible which can be important, for example, information regarding alternative fault hypotheses and multiple-fault isolability. The result of the analysis can be presented in the same compact form as the existing fault isolability matrix which makes it simple to visualize. As a case study, a model of an internal combustion engine is analyzed and two different solutions to the test selection problem are compared.

Behavioral modeling of nonlinear dynamic systems for control design and system monitoring of technical systems is a non-trivial task. One example is fault diagnosis where the objective is to detect abnormal system behavior due to faults at an early stage and isolate the faulty component. Developing sufficiently accurate models for fault diagnosis applications can be a time-consuming process which has motivated the use of data-driven models and machine learning. However, data-driven fault diagnosis is complicated by the facts that faults are rare events, and that it is not always possible to collect data that is representative of all operating conditions and faulty behavior. One solution to incomplete training data is to take into consideration physical insights when designing the data-driven models. One such approach is grey-box recurrent neural networks where physical insights about the monitored system are incorporated into the neural network structure. In this work, an automated design methodology is developed for grey-box recurrent neural networks using a structural representation of the system. Data from an internal combustion engine test bench is used to illustrate the potentials of the proposed network design method to construct residual generators for fault detection and isolation.

Data-driven fault classification in industrial applications is complicated by unknown fault classes and limited training data. In addition, different faults can have similar effects on sensor outputs resulting in fault classification ambiguities, i.e., multiple fault hypotheses can explain the data. One solution is to identify and rank all plausible fault classes that give useful information, for example, at a workshop when performing troubleshooting. A probabilistic fault classification algorithm is proposed for residual data classification combining the Weibull-calibrated one-class support vector machines for fault class modeling and Bayesian filtering for time-series analysis. The fault classifier ranks different fault classes and can identify sequences from unknown fault realizations, i.e., faults not represented in training data. Real residual data computed from sensor data and model analysis of an internal combustion engine are used as a case study illustrating the usefulness of the proposed method.

Model-based diagnosis compares observations from a system with predictions using a mathematical model to detect and isolate faulty components. Analyzing which faults that can be detected and isolated given the model gives useful information when designing a diagnosis system. This information can be used, for example, to determine which residual generators can be generated or to select a sufficient set of sensors that can be used to detect and isolate the faults. With more information about the system taken into consideration during such an analysis, more accurate estimations can be computed of how good fault detectability and isolability that can be achieved.

Model uncertainties and measurement noise are the main reasons for reduced fault detection and isolation performance and can make it difficult to design a diagnosis system that fulfills given performance requirements. By taking information about different uncertainties into consideration early in the development process of a diagnosis system, it is possible to predict how good performance can be achieved by a diagnosis system and avoid bad design choices. This thesis deals with quantitative analysis of fault detectability and isolability performance when taking model uncertainties and measurement noise into consideration. The goal is to analyze fault detectability and isolability performance given a mathematical model of the monitored system before a diagnosis system is developed.

A quantitative measure of fault detectability and isolability performance for a given model, called distinguishability, is proposed based on the Kullback-Leibler divergence. The distinguishability measure answers questions like "How difficult is it to isolate a fault f_i from another fault f_j?. Different properties of the distinguishability measure are analyzed. It is shown for example, that for linear descriptor models with Gaussian noise, distinguishability gives an upper limit for the fault to noise ratio of any linear residual generator. The proposed measure is used for quantitative analysis of a nonlinear mean value model of gas flows in a heavy-duty diesel engine to analyze how fault diagnosability performance varies for different operating points. It is also used to formulate the sensor selection problem, i.e., to find a cheapest set of available sensors that should be used in a system to achieve required fault diagnosability performance.

As a case study, quantitative fault diagnosability analysis is used during the design of an engine misfire detection algorithm based on the crankshaft angular velocity measured at the flywheel. Decisions during the development of the misfire detection algorithm are motivated using quantitative analysis of the misfire detectability performance showing, for example, varying detection performance at different operating points and for different cylinders to identify when it is more difficult to detect misfires.

This thesis presents a framework for quantitative fault detectability and isolability analysis that is a useful tool during the design of a diagnosis system. The different applications show examples of how quantitate analysis can be applied during a design process either as feedback to an engineer or when formulating different design steps as optimization problems to assure that required performance can be achieved.

Feature selection is an important task in data-driven control applications to identify relevant features and remove non-informative ones, for example residual selection for fault diagnosis. For multi-class data, the objective is to find a minimal set of features that can distinguish data from all different classes. A distributed feature selection algorithm is derived using convex optimization and the Alternating Direction Method of Multipliers. The distributed algorithm scales well with increasing number of classes by utilizing parallel computations. Two case studies are used to evaluate the developed feature selection algorithm: fault classification of an internal combustion engine and the MNIST data set to illustrate a larger multi-class classification problem.

Design of fault diagnosis systems is complicated by limited training data and inaccuracies in physical-based models when designing fault classifiers. A hybrid fault diagnosis approach is proposed using model-based residuals as input to a set of data-driven fault classifiers. As a case study, sensor data from an internal combustion engine test bed is used where faults have been injected into the system and a physical-based mathematical model of the air flow through the engine is available. First, a feature selection algorithm is applied to find a minimal set of residuals that is able to separate the different fault modes. Then, two different fault classification approaches are discussed, Random Forests and one-class Support Vector Machines. A set of one-class Support Vector Machines is used to model data from each fault mode separately. The case study illustrates an advantage of using one-class classifiers, which makes it possible to detect unknown faults by identifying samples not belonging to any known fault mode. (C) 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.

Structural analysis is a useful tool for fault diagnosability analysis to handle systems that are described by a large set of non-linear differential algebraic equations. Distributed fault diagnosis is an attractive approach for complex systems to reduce computational complexity by partitioning the system into a set of smaller subsystems and perform fault diagnosis of each subsystem. Defining these subsystems requires methods to understand how fault diagnosis properties of each subsystem relates to the properties of the whole system. Another related problem is that large and complex systems are likely to be developed by several companies where each company is developing different subsystems that can be used in different system configurations. In these situations, each subsystem will have limited model information about the other subsystems, which complicates performing structural analysis of the whole system. The main contribution in this work is extending some of the existing results in structural analysis for one system model to a distributed set of connected subsystems. The results show the relationship between structural fault diagnosis properties of the whole system and properties of the set of individual subsystems.

Design space exploration of hybrid electric vehicles is an important multi-objective global optimization problem. One of the main objectives is to minimize fuel consumption while maintaining satisfactory driveability performance and vehicle cost. The design problem often includes multiple design options, including different driveline architectures and component sizes, where different candidates have various trade-offs between different, in many cases contradictory, performance requirements. Thus, there is no global optimum but a set of Pareto-optimal solutions to be explored. The objective functions can be expensive to evaluate, due to time-consuming simulations, which requires careful selection of which candidates to evaluate. A design space exploration algorithm is proposed for finding the set of Pareto-optimal solutions when the design search space includes multiple design options. As a case study, powertrain optimization is performed for a medium-sized series hybrid electric delivery truck.

Hybrid electric vehicles (HEV) are essential for reducing fuel consumption and emissions. However, when analyzing different segments of the transportation industry, for example, public transportation or different sizes of delivery trucks and how the HEV are used, it is clear that one powertrain may not be optimal in all situations. Choosing a hybrid powertrain architecture and proper component sizes for different applications is an important task to find the optimal trade-off between fuel economy, drivability, and vehicle cost. However, exploring and evaluating all possible architectures and component sizes is a time-consuming task. A search algorithm, using Gaussian Processes, is proposed that simultaneously explores multiple architecture options, to identify the Pareto-optimal solutions. The search algorithm is designed to carefully select the candidate in each iteration which is most likely to be Pareto-optimal, based on the results from previous candidates, to reduce computational time. The powertrain of a medium-sized series plugin hybrid electric delivery truck with a range extender is optimized for different driving missions. Three different powertrain architectures are included in the design space exploration and the fuel economy is evaluated using a simulation model of the powertrain and Dynamic Programming. Results from the analysis show which ranges of powertrain component sizes are recommended for the different types of driving scenarios.

Finding the cheapest, or smallest, set of sensors such that a specified level of diagnosis performance is maintained is important to decrease cost while controlling performance. Algorithms have been developed to find sets of sensors that make faults detectable and isolable under ideal circumstances. However, due to model uncertainties and measurement noise, different sets of sensors result in different achievable diagnosability performance in practice. In this paper, the sensor selection problem is formulated to ensure that the set of sensors fulfils required performance specifications when model uncertainties and measurement noise are taken into consideration. However, the algorithms for finding the guaranteed global optimal solution are intractable without exhaustive search. To overcome this problem, a greedy stochastic search algorithm is proposed to solve the sensor selection problem. A case study demonstrates the effectiveness of the greedy stochastic search in finding sets close to the global optimum in short computational time.

A model-based misfire detection algorithm is proposed. The algorithm is able to detect misfires and identify the failing cylinder during different conditions, such as cylinder-to-cylinder variations, cold starts, and different engine behavior in different operating points. Also, a method is proposed for automatic tuning of the algorithm based on training data. The misfire detection algorithm is evaluated using data from several vehicles on the road and the results show that a low misclassification rate is achieved even during difficult conditions. (C) 2014 Elsevier Ltd. All rights reserved.

In model-based diagnosis there are often more candidate residual generators than what is needed and residual selection is therefore an important step in the design of model-based diagnosis systems. The availability of computer-aided tools for automatic generation of residual generators have made it easier to generate a large set of candidate residual generators for fault detection and isolation. Fault detection performance varies significantly between different candidates due to the impact of model uncertainties and measurement noise. Thus, to achieve satisfactory fault detection and isolation performance, these factors must be taken into consideration when formulating the residual selection problem. Here, a convex optimization problem is formulated as a residual selection approach, utilizing both structural information about the different residuals and training data from different fault scenarios. The optimal solution corresponds to a minimal set of residual generators with guaranteed performance. Measurement data and residual generators from an internal combustion engine test-bed is used as a case study to illustrate the usefulness of the proposed method.

A commonly used signal for engine misfire detection is the crankshaft angular velocity measured at the flywheel. However, flywheel manufacturing errors result in vehicle-to-vehicle variations in the measurements and have a negative impact on the misfire detection performance, where the negative impact is quantified for a number of vehicles. A misfire detection algorithm is proposed with flywheel error adaptation in order to increase robustness and reduce the number of mis-classifications. Since the available computational power is limited in a vehicle, a filter with low computational load, a Constant Gain Extended Kalman Filter, is proposed to estimate the flywheel errors. Evaluations using measurements from vehicles on the road show that the number of mis-classifications is significantly reduced when taking the estimated flywheel errors into consideration.

To be able to evaluate quantitative fault diagnosability performance in model-based diagnosis is useful during the design of a diagnosis system. Different fault realizations are more or less likely to occur and the fault diagnosis problem is complicated by model uncertainties and noise. Thus, it is not obvious how to evaluate performance when all of this information is taken into consideration. Four candidates for quantifying fault diagnosability performance between fault modes are discussed. The proposed measure is called expected distinguishability and is based of the previous distinguishability measure and two methods to compute expected distinguishability are presented.

Detecting changes in residuals is important for fault detection and is commonly performed by thresholding the residual using, for example, a CUSUM test. However, detecting variations in the residual distribution, not causing a change of bias or increased variance, is difficult using these methods. A plug-and-play residual change detection approach is proposed based on sequential quantile estimation to detect changes in the residual cumulative density function. An advantage of the proposed algorithm is that it is non-parametric and has low computational cost and memory usage which makes it suitable for on-line implementations where computational power is limited.

Most model-based diagnosis approaches reported in the literature adopt a generic architecture and approach. However, the fault hypotheses generated by these methods may differ. This is not only due to the methods, but also on the basic assumptions made by different diagnostic algorithms on fault manifestation and evolution. While comparing different diagnosis approaches, the assumptions made in each case will have a significant effect on fault diagnosability performance and must therefore also be taken into consideration. Thus, to make a fair comparison, the different approaches should be designed based on the same assumptions. This paper studies the relation between a set of commonly made assumptions and fault isolability performance in order to compare different diagnosis approaches. As a case study, five developed diagnosis systems for a wind turbine benchmark problem are evaluated to analyze the type of assumptions that are applied in the different designs.

Fault diagnosis is important for automotive systems, e.g., to reduce emissions and improve system reliability. Developing diagnosis systems is complicated by model inaccuracies and limited training data from relevant operating conditions, especially for new products and models. One solution is the use of hybrid fault diagnosis techniques combining model-based and data-driven methods. In this work, data-driven residual generation for fault detection and isolation is investigated for a system injecting urea into the aftertreatment system of a heavy-duty truck. A set of recurrent neural network-based residual generators is designed using a structural model of the system. The performance of this approach is compared to a baseline model-based approach using data collected from a heavy-duty truck during different fault scenarions with promising results.

Data-driven modeling and machine learning have received a lot of attention in fault diagnosis and system monitoring research. Since faults are rare events, conventional multi-class classification is complicated by incomplete training data and unknown faults. One solution is anomaly classification which can be used to detect abnormal behavior when only training data from the nominal operation is available. However, data-driven fault isolation is still a non-trivial task when training data is not representative of nominal and faulty behavior. In this work, the importance of redundancy for a set of known variables that are fed to a data-driven anomaly classification is discussed. It is shown that residual-based anomaly detection can be used to reject the nominal class which is not possible with one-class classifiers, such as one-class support vector machines. Based on these results, it is also discussed how data-driven residuals can be integrated with model-based fault isolation logic.

Machine learning can be used to automatically process sensor data and create data-driven models for prediction and classification. However, in applications such as fault diagnosis, faults are rare events and learning models for fault classification is complicated because of lack of relevant training data. This paper proposes a hybrid diagnosis system design which combines model-based residuals with incremental anomaly classifiers. The proposed method is able to identify unknown faults and also classify multiple-faults using only single-fault training data. The proposed method is verified using a physical model and data collected from an internal combustion engine.

Selecting residual generators for detecting and isolating faults in a system is an important step when designing model-based diagnosis systems. However, finding a suitable set of residual generators to fulfill performance requirements is complicated by model uncertainties and measurement noise that have negative impact on fault detection performance. The main contribution is an algorithm for residual selection that combines model-based and data-driven methods to find a set of residual generators that maximizes fault detection and isolation performance. Based on the solution from the residual selection algorithm, a generalized diagnosis system design is proposed where test quantities are designed using multivariate residual information to improve detection performance. To illustrate the usefulness of the proposed residual selection algorithm, it is applied to find a set of residual generators to monitor the air path through an internal combustion engine.

The rapid increase in electric vehicles (EVs) and installed photovoltaic systems (PV) has resulted in new challenges for electric systems, e.g., voltage variations in low-voltage grids. Grid owners cannot directly control the power consumption of the end consumers. However, by the design of transparent tariffs, economic incentives are introduced for the end consumers to adjust their EV charging patterns. In this work, the main objective is to design a time-of-use pricing tariff to reduce the voltage variations in a low-voltage grid when introducing PVs and EVs with smart charging. Data from an existing low-voltage grid and hourly data from household power consumption, together with models of PV and EV charging, are used to simulate the voltage fluctuations based on the modified electric consumption. The results show that a time-of-use pricing tariff taking into consideration maximum peak power is important to reduce grid voltage variations. Another observation is that the use of economic incentives, such as subsidies when selling power from the household, combined with V2G technology can be economical for households but increases the voltage variations in the grid.

Data-driven fault diagnosis of dynamic systems is complicated by incomplete training data, unknown faults, and overlapping classes. Many existing machine learning models and data-driven classifiers are not expected to perform well if training data is not representative of all relevant fault realizations. In this work, a data-driven model, called a flexi-pipe model, is proposed to capture the variability of data in residual space from a few realizations of each fault class. A diagnosis system is developed as an open set classification algorithm that can handle both incomplete training data and overlapping fault classes. Data from different fault scenarios in an engine test bench is used to evaluate the performance of the proposed methods. Results show that the proposed fault class models generalize to new fault realizations when training data only contains a few realizations of each fault class.

A hybrid diagnosis system design is proposed that combines model-based and data-driven diagnosis methods for fault isolation. A set of residuals are used to detect if there is a fault in the system and a consistency-based fault isolation algorithm is used to compute all diagnosis candidates that can explain the triggered residuals. To improve fault isolation, diagnosis candidates are ranked by evaluating the residuals using a set of one-class support vector machines trained using data from different faults. The proposed diagnosis system design is evaluated using simulations of a model describing the air-flow in an internal combustion engine.

The increasing complexity and size of cyber-physical systems (e.g., aircraft, manufacturing processes, and power generation plants) is making it hard to develop centralized diagnosers that are reliable and efficient. In addition, advances in networking technology, along with the availability of inexpensive sensors and processors, are causing a shift in focus from centralized to more distributed diagnosers. This paper develops two structural approaches for distributed fault detection and isolation. The first method uses redundant equation sets for residual generation, referred to as minimal structurally-over-determined sets, and the second is based on the original model equations. We compare the diagnosis performance of the two algorithms and clarify the pros and cons of each method. A case study is used to demonstrate the two methods, and the results are discussed together with directions for future work.

This paper presents a framework for distributed fault detection and isolation in dynamic systems. Our approach uses the dynamic model of each subsystem to derive a set of independent, local diagnosers. If needed, the subsystem model is extended to include measurements and model equations from its immediate neighbors to compute its diagnosis. Our approach is designed to ensure that each subsystem diagnoser provides the correct results, therefore, a local diagnosis result is equivalent to the results that would be produced by a global system diagnoser. We discuss the distribute diagnosis algorithm, and illustrate its application using a multi-tank system.

Model-based approaches to fault detection and isolation (FDI) rely on accurate models of the plant and a sufficient number of reliable measurements for residual generation and analysis. However, in realistic situations, there can be uncertainties in the plant models and measurements, which have a negative impact on the diagnosability performance that depends on the system state. In other words, the impact of the uncertainties can be larger in some operating regions as compared to others. To achieve acceptable performance in practice, it is necessary to find a set of residuals that are sufficiently sensitive to faults but robust to uncertainties across all operating conditions. In this paper, a quantitative measure, called detectability ratio, is used to evaluate and quantify detectability performance of different residuals in different operating regions. This measure is used to find a minimal residual set that fulfills a set of desired diagnosability performance requirements. The proposed method is demonstrated and validated through a case study.

A number of residual generation methods have been developed for robust model-based fault detection and isolation (FDI). There have also been a number of offline (i.e., design-time) methods that focus on optimizing FDI performance (e.g., trading off detection performance versus cost). However, design-time algorithms are not tuned to optimize performance for different operating regions of system behavior. To do this, would need to define online measures of sensitivity and robustness, and use them to select the best residual set online as system behavior transitions between operating regions. In this paper we develop a quantitative measure of residual performance, called the detectability ratio that applies to additive and multiplicative uncertainties when determining the best residual set in different operating regions. We discuss this methodology and demonstrate its effectiveness using a case study.

Clustering is an important tool in data-driven fault diagnosis to make use of unlabeled data. Collecting representative data for fault diagnosis is a difficult task since faults are rare events. In addition, using data collected from the field, e.g., logged operational data and data from different workshops about replaced components, can result in labelling uncertainties. A common approach for fault diagnosis of dynamic systems is to use residual-based features that filter out system dynamics while being sensitive to faults. The use of conventional clustering algorithms is complicated by that the distribution of residual data from one fault class varies for different realizations and system operating conditions. In this work, a clustering algorithm is proposed for residual data that clusters data by estimating fault signatures in residual space. The proposed clustering algorithm can be used on time-series data by clustering batches of data from the same fault scenario instead of clustering data sample-by-sample. The usefulness of the proposed clustering algorithm is illustrated using residual data from different fault scenarios collected from an internal combustion engine test bench. Copyright (C) 2022 The Authors.

Fault diagnosis of dynamic systems is done by detecting changes in time-series data, for example residuals, caused by system degradation and faulty components. The use of general-purpose multi-class classification methods for fault diagnosis is complicated by imbalanced training data and unknown fault classes. Another complicating factor is that different fault classes can result in similar residual outputs, especially for small faults, which causes classification ambiguities. In this work, a framework for data-driven analysis and open-set classification is developed for fault diagnosis applications using the Kullback-Leibler divergence. A data-driven fault classification algorithm is proposed which can handle imbalanced datasets, class overlapping, and unknown faults. In addition, an algorithm is proposed to estimate the size of the fault when training data contains information from known fault realizations. An advantage of the proposed framework is that it can also be used for quantitative analysis of fault diagnosis performance, for example, to analyze how easy it is to classify faults of different magnitudes. To evaluate the usefulness of the proposed methods, multiple datasets from different fault scenarios have been collected from an internal combustion engine test bench to illustrate the design process of a data-driven diagnosis system, including quantitative fault diagnosis analysis and evaluation of the developed open set fault classification algorithm.

Data-driven fault diagnosis requires training data that is representative of the different operating conditions of the system to capture its behavior. If training data is limited, one solution is to incorporate physical insights into machine learning models to improve their effectiveness. However, while previous works show the usefulness of hybrid approaches for isolation of faults, the impact of training data must be taken into consideration when drawing conclusions from data-driven residuals in a consistency-based diagnosis framework. By giving an understanding of the physical interaction between the signals, a hybrid fault diagnosis approach, can enforce model properties of residual generators to isolate faults that are not represented in training data. The objective of this work is to analyze the impact of limited training data when training neural network-based residual generators. It is also investigated how the use of structural information when selecting the network structure is a solution to limited training data and how to ameliorate the performance of hybrid approaches in face of this challenge.

Data-driven modeling and machine learning are widely used to model the behavior of dynamic systems. One application is the residual evaluation of technical systems where model predictions are compared with measurement data to create residuals for fault diagnosis applications. While recurrent neural network models have been shown capable of modeling complex non-linear dynamic systems, they are limited to fixed steps discrete-time simulation. Modeling using neural ordinary differential equations, however, make it possible to evaluate the state variables at specific times, compute gradients when training the model and use standard numerical solvers to explicitly model the underlying dynamic of the time-series data. Here, the effect of solver selection on the performance of neural ordinary differential equation residuals during training and evaluation is investigated. The paper includes a case study of a heavy-duty truck's after-treatment system to highlight the potential of these techniques for improving fault diagnosis performance.

Batteries are useful in Fuel Cell Hybrid Electric Vehicles (FCHEV) to fulfill transient demands and for regenerative braking. Efficient energy management strategies paired with optimal powertrain design further improves the efficiency. In this paper, a new methodology to simultaneously size the propulsive elements and optimize the power-split strategy of a Range Extended Battery Electric Vehicle (REBEV), using a Polymer Electron Membrane Fuel Cell (PEMFC), is proposed and preliminary studies on the effects of the driving mission profile and the auxiliary power loads on the sizing and optimal performance of the powertrain design are carried out. Dynamic Programming is used to compute the optimal energy management strategy for a given driving mission profile, providing a global optimal solution. The component sizing problem is performed using a machine learning based, guided design space exploration to find the set of Pareto-optimal solutions that give the best trade-offs between the different objectives. The powertrain model includes the dynamic behavior of the fuel cell system compressor and a battery lumped parameter thermal model along with the quasi-static semi-empirical model of the fuel cell and a zero-order battery model. Initial results indicate an increase in the Pareto-optimal sizes with the inclusion of thermal management.

The present paper proposes an advanced approach for Polymer Electrolyte Membrane Fuel Cell (PEMFC) systems fault detection and isolation through a model-based diagnostic algorithm. The considered algorithm is developed upon a lumped parameter model simulating a whole PEMFC system oriented towards automotive applications. This model is inspired by other models available in the literature, with further attention to stack thermal dynamics and water management. The developed model is analysed by means of Structural Analysis, to identify the correlations among involved physical variables, defined equations and a set of faults which may occur in the system (related to both auxiliary components malfunctions and stack degradation phenomena). Residual generators are designed by means of Causal Computation analysis and the maximum theoretical fault isolability, achievable with a minimal number of installed sensors, is investigated. The achieved results proved the capability of the algorithm to theoretically detect and isolate almost all faults with the only use of stack voltage and temperature sensors, with significant advantages from an industrial point of view. The effective fault isolability is proved through fault simulations at a specific fault magnitude with an advanced residual evaluation technique, to consider quantitative residual deviations from normal conditions and achieve univocal fault isolation.

Security risk analysis and quantification for automotive systems is a challenging task. This challenge is exacerbated when physical systems are integrated with computation and communication networks to form Cyber-Physical Systems (CPS). The complexity arises from the multitude of attack possibilities within the overall system. This work proposes an attack index based on redundancy in the system and the computational sequence of residual generators. This work considers a nonlinear dynamic model of an automotive system with a communication network. The approach involves using system dynamics to model attack vectors, which are based on the vulnerabilities in the system that are exploited through open network components (like On-Board-Diagnosis (OBD-II)), network segmentation (due to improper gateway implementation), and sensors that are susceptible to adversarial attacks. The redundant and non-redundant parts of the system are identified by considering the sensor configuration and unknown variables. Then, an attack index is derived by analyzing the placement of attack vectors in relation to the redundant and non-redundant parts, using the canonical decomposition of the structural model. The security implications of the residuals are determined by analyzing the computational sequence and the placement of the sensors. Thus, this work promotes the notion of security by design by proposing sensor placement strategies to enhance the overall security index. Finally, it is verified how the proposed attack index and its analysis could be used to enhance automotive security using Model-In-Loop (MIL) simulations.