Correct and useful system descriptions are central to model-based fault diagnosis, as they describe the structure and behaviour of the system. But so far, system descriptions were always interpreted as complete propositional logic models and were assumed to be given. However, with increasing use of data-driven methods that approximate system descriptions, it cannot be guaranteed that the system description is a complete model of the system. It cannot even be guaranteed that the system description contains all observations, components, and connections that the real system exhibits. This requires novel approaches to determine how well a system description models the real system and how well it can thus be used for fault diagnosis. We present a novel algorithm which takes a syntactic approach to calculate diagnosability of approximated system descriptions. This is different from previous diagnosability research, which determined the diagnosability of real systems, where the algorithm could rely on the model’s completeness and corresponding reliable observations. With approximated models, observations and models are (partially) disconnected. We also present two novel algorithms to calculate new metrics that determine how close two approximated system descriptions are to each other by heuristically solving the graph alignment problem. We believe that our new approach and corresponding algorithms benefit practitioners who want to evaluate the quality of their approximated models for fault diagnosis. We show the usefulness of our results on established benchmarks of tank-systems, the Tennessee Eastman Process, two spacecraft, and a benchmark of different Boolean circuits.

A long-term objective for many research fields, such as anomaly detection, discretization and root-cause diagnosis in Cyber-Phyiscal Production Systems is the realization of resilient and highly autonomous systems. Instances of those systems range from the control and operation of single production systems to controlling entire plants. While notable progress has been made, a key challenge remains unaddressed: the availability of comprehensive and standardized datasets necessary for advancing machine learning based solutions. Typical evaluation datasets comprise systems of (too) little complexity or are either suitable for only data-driven or only for symbolic methods. Yet, a comprehensive dataset and model for developing, training, and testing machine learning methods from anomaly detection to fault diagnosis in a structured and comparable manner does not exist. To bridge this gap, we present a benchmark specifically designed to support both data-driven methods and symbolic reasoning approaches, by extending the BerFiPl benchmark introduced by Ehrhardt et al. [1]. By providing hybrid data streams, labeled system states, and a modular, interpretable system structure, the dataset offers a unique opportunity to develop, train, test, and compare hybrid AI methods to bridge the gap between data-driven and symbolic paradigms.

In this paper, we tackle the problem of planning in complex numerical domains, where actions are indexed by control parameters, and their effects may be described by neural networks. We propose a lazy, hierarchical approach based on two ingredients. First, a Satisfiability Modulo Theory solver looks for an abstract plan where the neural networks in the model are abstracted into uninterpreted functions. Then, we attempt to concretize the abstract plan by querying the neural network to determine the control parameters. If the concretization fails and no valid control parameters could be found, suitable information to refine the abstraction is lifted to the Satisfiability Modulo Theory model. We contrast our work against the state of the art in NN-enriched numerical planning, where the neural network is eagerly and exactly represented as terms in Satisfiability Modulo Theories over nonlinear real arithmetic. Our systematic evaluation on four different planning domains shows that avoiding symbolic reasoning about the neural network not only leads to substantial efficiency improvements, but also enables their integration as black-box models.

Autonomously detecting anomalous behavior based on system observations is a fundamental task for Cyber-Physical Systems (CPS). Due to the high system complexity and large number of subsystems in modern CPS, rule- or knowledge-based approaches for anomaly detection are more and more replaced by Machine Learning (ML) approaches which leverage historical CPS data. Typically, ML approaches learn a system model based on the CPS data and identify anomalous behavior based on the distance of the real CPS behavior to the predicted model behavior. However, most classical ML approaches for anomaly detection are monolithic, meaning a single ML model is fitted on a global CPS observation, making them frail to spurious correlations and confounders that originate on CPS subsystem level. We hence propose a modular approach toward anomaly detection in CPS, specifically a novel Modular Neural Network (MNN) architecture. Our architecture not only models the behavior of individual CPS sub-systems in individual MNN modules, but additionally models the dependencies of the CPS subsystems into the MNN architecture. Thereby, we omit confounding effects and spurious correlations, enabling us to identify and allocate anomalies within the CPS at subsystem level. We benchmark our MNN architecture against monolithic Neural Networks and MNN architectures that do not explicitly model CPS subsystem dependencies using a real-world dataset of an industrial robot with different anomalies. We show that by modeling real-world dependencies into a MNN architecture, we can improve the performance of autonomous anomaly detection in CPS.

With the increasing complexity of highly automated cyber-physical systems (CPS), monitoring their behavior has become crucial. Failures in these systems can be costly, halt production, or even pose risks to human safety. Effective diagnosis depends on understanding the system's components, connections, and the influences among them, knowledge typically provided by experts. However, the shift towards self-diagnosing systems necessitates this knowledge be machine-readable and interpretable. This paper introduces a novel methodology that utilizes an ontology to encode knowledge about cyber-physical systems and systematically generate propositional logical expressions. These expressions can then be evaluated using state-of-the-art diagnostic algorithms to identify failure causes. Our methodology was validated using an established AI benchmark for diagnostics. We constructed an ontology description for the underlying cyber-physical system, deduced influences of system sensors from data, and successfully diagnosed induced failures, demonstrating the efficacy and applicability of our approach.

Root-cause diagnosis techniques, such as consistency-based and abductive diagnosis, offer essential support in explaining symptoms in a cyber-physical system. Developing and maintaining the required (detailed or abstract) models can be a serious challenge. Related issues include the complexity of the required knowledge and the dynamic changes we see in a system over its life cycle. This raises the question regarding strategies and the feasibility of utilizing unsupervised machine learning to learn diagnostic system models based on available time series data in order to address this challenge. This paper presents the novel methodology Discret2DeepDive for automated learning of diagnostic system models for root-cause diagnosis, focussing on the use of automata for state mapping over time and explores advancements related to the handling of dynamic time series data. These advancements are incorporated into both the discretization process and the generation of residuals. The findings demonstrate a notable enhancement in discretizing time series data into modes and residual generation for anomaly detection in sequential data, thereby providing a substantial value for diagnosing faults.

To improve the autonomy of Cyber-Physical Production Systems (CPPS), a growing number of approaches in Artificial Intelligence (AI) is developed. However, implementations of such approaches are often validated on individual use-cases, offering little to no comparability. Though CPPS automation includes a variety of problem domains, existing benchmarks usually focus on single or partial problems. Additionally, they often neglect to test for AI-specific performance indicators, like asymptotic complexity scenarios or runtimes. Within this paper we identify minimum common set requirements for AI benchmarks in the domain of CPPS and introduce a comprehensive benchmark, offering applicability on diagnosis, reconfiguration, and planning approaches from AI. The benchmark consists of a grid of datasets derived from 16 simulations of modular CPPS from process engineering, featuring multiple functionalities, complexities, and individual and superposed faults. We evaluate the benchmark on state-of-the-art AI approaches in diagnosis, reconfiguration, and planning. The benchmark is made publicly available on GitHub.

AI methods for fault handling in Cyber-Physical Production Systems (CPPS) such as production plants and tank systems are an emerging research topic. In the last years many methods for the detection of anomalies and faults, the diagnosis of the root cause and the automated repair have been developed. However, most of the methods are barely evaluated using a wide range of systems but applicability is shown using single use cases. In this paper, an ensemble of simulated benchmark systems is presented, which allows for a broad evaluation of AI methods for fault handling. The ensemble consists of seven different tank systems from process engineering with varying sizes and complexities and is made publicly available on Github. The suitability of the ensemble is shown using AI methods for fault handling such as anomaly detection, diagnosis and reconfiguration.

Currently, detecting and isolating faults in hybrid systems is often done manually with the help of human operators. In this paper we present a novel model-based diagnosis approach for automatically diagnosing hybrid systems. The approach has two parts: First, modelling dynamic system behaviour is done through well-known state space models using differential equations. Second, from the state space models we calculate Boolean residuals through an observer-pattern. The novelty lies in implementing the observer pattern through the use of a symbolic system description specified in satisfiability theory modulo linear arithmetic. With this, we create a static situation for the diagnosis algorithm and decouple modelling and diagnosis. Evaluating the system description generates one Boolean residual for each component. These residuals constitute the fault symptoms. To find the minimum cardinality diagnosis from these symptoms we employ Reiter’s diagnosis lattice.For the experimental evaluation we use a simulation of the Tennessee Eastman process and a simulation of a four-tank model. We show that the presented approach is able to identify all injected faults.

Trends in novel manufacturing systems lead to an increased level of data availability and smart usage of these data. Nowadays, many approaches are available to use the data, but because of an increased flexibility of the systems the interaction between machines and humans has become a challenge. Humans have to browse through a huge amount of data, need knowledge about the machine and underlying algorithms to interpret the results; they cannot use their known terms for communication, we call it the conceptual gap. The user should be enabled to communicate with the machine on a more abstract level and in a more natural way. Therefore, a natural language layer is introduced to provide users with a familiar interaction interface. Underlying layers contain knowledge about the domain, the machines and how data can be accessed and processed. This enables users' questions such as “Are there any anomalies in the system?” to be answered. Answers are provided in natural language and evaluated with a test set of 204 questions.

S.280-287