Tagir Fabarisov

Abstract

Recently we introduced a new model-based fault injection method implemented as a highly-customizable Simulink block called FIBlock. It supports the injection of typical faults of essential heterogeneous components of Cyber-Physical Systems, such as sensors, computing hardware, and network. The FIBlock allows to tune a fault type and configure multiple parameters to tune error magnitude, fault activation time, and fault exposure duration. We demonstrated its performance to generate highly adjustable and various CPS fault types on a Simulink case study representing a lower-limb EXO-LEGS exoskeleton, an assistive device for the elderly in everyday life. In our previous paper we discovered the spatial and temporal thresholds for different fault types. Upon exceeding said thresholds, the Dynamic Movement Primitives-based control system could no longer adequately compensate these errors. In this paper we proposed a Deep Learning-based approach for the system failure prevention. We employed the Long Short-Term Memory (LSTM) network for error detection and mitigation. Error detection was performed using data prediction. The LSTM models were mitigating the detected errors with computed predictions only when they were subject to the imminent failure (i.e., exceeded the aforementioned thresholds). To compare our approach with previous findings, we trained two LSTM models on angular position and on angular velocity signals. For evaluation, we performed fault injection experiments with varying fault effect parameters. The ’Stuck-at current value’ fault was injected into angular position sensor, and the ’Stuck-at 0’ fault was injected into angular velocity sensor. Presented Deep Learning-based approach prevented system failure even when the injected faults were substantially exceeding thresholds. In addition, a reasoning for data access point choice has been evaluated. We compared two options: (i) the input data for LSTM is provided from the controller output and (ii) from the sensor output. In the paper the pros and cons for both options are presented. We deployed the trained LSTM models on an Edge Tensor Processing Unit. For that, the models have been quantized, i.e. all the 32-bit floating-point numbers (such as weights and activation outputs) were converted to the nearest 8-bit fixed-point numbers and converted to the TensorFlow Lite models. The Coral USB Accelerator was coupled with a Raspberry Pi 4B for signal processing. The result proves the feasibility of the proposed method. Because the LSTM models were converted to the 8-bit integer TensorFlow Lite models, it allowed real-time error mitigation. Furthermore, the light weight of the system and a minimal power consumption allows its integration into wearable robotic systems.

Abstract

Online error detection helps to reduce the risk of failure of safety-critical systems. However, due to the increasing complexity of modern Cyber-Physical Systems and the sophisticated interaction of their heterogeneous components, it becomes harder to apply traditional error detection methods. Nowadays, the popularity of Deep Learning-based error detection snowballs. DL-based methods achieved significant progress along with better results. This paper introduces the KrakenBox, a deep learning-based error detector for industrial Cyber-Physical Systems (CPS). It provides conceptual and technical details of the KrakenBox hardware, software, and a case study. The KrakenBox hardware is based on NVIDIA Jetson AGX Xavier, designed to empower the deep learning-based application and the extended alarm module. The KrakenBox software consists of several programs capable of collecting, processing, storing, and analyzing time-series data. The KrakenBox can be connected to the networked automation system either via Ethernet or wirelessly. The paper presents the KrakenBox architecture and results of experiments that allow the evaluation of the error detection performance for varying error magnitude. The results of these experiments demonstrate that the KrakenBox is able to improve the safety of a networked automation system.

Abstract

Model-based fault injection methods are widely used for the evaluation of fault tolerance in safety-critical control systems. In this paper, we introduce a new model-based fault injection method implemented as a highlycustomizable Simulink block called FIBlock. It supports the injection of typical faults of essential heterogeneous components of Cyber-Physical Systems, such as sensors, computing hardware, and network. The FIBlock GUI allows the user to select a fault type and configure multiple parameters to tune error magnitude, fault activation time, and fault exposure duration. Additional trigger inputs and outputs of the block enable the modeling of conditional faults. Furthermore, two or more FIBlocks connected with these trigger signals can model chained errors. The proposed fault injection method is demonstrated with a lower-limb EXO-LEGS exoskeleton, an assistive device for the elderly in everyday life. The EXO-LEGS model-based dynamic control is realized in the Simulink environment and allows easy integration of the aforementioned FIBlocks. Exoskeletons, in general, being a complex CPS with multiple sensors and actuators, are prone to hardware and software faults. In the case study, three types of faults were investigated: 1) sensor freeze, 2) stuck-at-0, 3) bit-flip. The fault injection experiments helped to determine faults that have the most significant effects on the overall system reliability and identify the fine line for the critical fault duration after that the controller could no longer mitigate faults.

Abstract

Industry 4.0 is the current trend of automation and data exchange in manufacturing technologies that is focusing on the creation of smart factories with the modular structured Cyber-Physical Systems (CPS), in tight cooperation with humans. This trend also implies that the systems become more complex, heterogeneous, and distributed especially their network and software parts. This makes the CPS highly critical subject to failures at different levels, including software, hardware, and human operators. Consequently, ensuring reliable and safe operation under the presence of non-avoidable threats also becomes a more complicated task. The proper analysis of the CPS requires thorough comprehension of both the dependability properties of system components and their interactions as well as structural and behavioral aspects of the complete system. Such an analysis of complex and mutually interlinked system properties puts considerable challenges on appropriate methods for modeling and analysis, as well as, on the related applied software tools. The Dual-graph Error Propagation Model (DEPM), developed in our lab, is a mathematical abstraction of the main future system’s properties, which are vital for the determination of the error propagation processes. It is a useful analytical instrument for the evaluation of the influence of particular faults and errors to the overall system behavior. OpenErrorPro is our analytical software tool for stochastic error propagation analysis that supports the DEPM framework. Using OpenErrorPro, a DTMC model could be automatically generated from a DEPM, and the reliability metrics, in addition to, error propagation path, can be computed. This could be implemented for the analysis of the heterogeneous CPS components. The necessary steps for the DEPM framework extension, required for such an implementation, are discussed in this paper.

Abstract

Model-Based System Engineering (MBSE) is a popular mathematical and visual approach to the design of complex control, signal processing, and communication systems. It is used in safety critical industrial domains including aerospace, automotive, transportation, medical and robotics applications. Our group develops methods and tools for model-based system reliability and safety analysis with the main focuses on stochastic modelling of error propagation processes. This article is devoted to the optimisation and extensions to our analytical toolset. We have investigated the key modeling paradigms, requirements and industrial needs and have formulated the list of particular extensions.

Abstract

A cyber-physical system consists of sensors, micro-controller, networks, and actuators that interact with each other, generate a substantial amount of data, and form extremely complex system operational profiles. These heterogeneous components are subject to errors, e.g. spikes, off-sets, or delays, that may result in system failures. As the complexity of modern systems increases, it becomes a challenge to apply traditional fault detection and isolation methods to such complex systems. Deep learning based methods have surpassed traditional methods in terms of performance as the data size and complexity increase. The signals of cyber-physical systems are mainly time-series data. In this paper, we propose a new on-line error detection and mitigation approach for common sensor, computing hardware, and network errors of cyber-physical systems using deep learning based methods. More specifically, we train a Long Short-Term Memory (LSTM) network as a single step prediction model for the detection and mitigation of errors, like spikes, or offsets. In order to detect the long-duration errors that show no sharp change (a sudden drop or rise) between two successive data samples when errors occurred, e.g. network delays, we train an LSTM encoder-decoder as a multi-step prediction model. We also introduce the on-line error mitigation approach. Automatic recovery is achieved by replacing the detected errors with the predicted values. Finally, we demonstrate on-line error detection and mitigation capabilities of the trained single step and multi-step predictors using representative case studies.

Abstract

Dual-graph Error Propagation Model (DEPM) is a stochastic framework developed in our lab that captures system properties relevant to error propagation processes such as control and data flow structures and reliability characteristics of single components. The DEPM helps to estimate the impact of a fault of a particular component on the overall system reliability. The probability of a system failure, Mean Time Between Failures and Mean Time to Repair, and the expected number of failures are the quantitative results of the DEPM-based analysis. In addition, the DEPM provides an access to the powerful techniques from Probabilistic Model Checking (PMC) that allow the evaluation of advanced, highly customizable, time-related reliability properties, e.g., “the probability of a system failure during certain time units after the occurrence of a certain combination of errors in certain system components”. For this reason, the DEPM models are automatically transformed to discrete-time Markov chain (DTMC) models and are analyzed using the state of the art probabilistic model checker PRISM. However, the new promising model checker Storm has been recently released. This paper presents the results of the efficiency comparison of these two model checkers based on the automatically generated set of the DTMC models that are typical for the error propagation analysis tasks. Several computation engines that are supported by both model checkers have been taking into account. The paper also gives the general description of the DEPM workflow with the focus on the PMC interface.