Advanced Solutions in Diagnostics and Fault Tolerant Control

In modern societies, the reliable and continuous operation of certain infrastructures plays a fundamental role in the quality of life, economic development and security of nations. This paper presents several approaches for diagnosis and fault-tolerant control of critical infrastructure systems (CIS) including: the analysis of these systems to understand the weaknesses and risks in case some fault occurs, fault diagnosis using analytical redundancy relation, fault tolerant control schemes and assessment of the fault tolerance and inclusion of health-aware mechanisms in the CIS control systems.

The main objective of the paper is to present selected results for fault-tolerant control for constrained dynamic systems. The paper starts with an integrated actuator fault-tolerant control scheme. Subsequently, sensor fault estimation and fault-tolerant is outlined. Based on this seminal work, the need for simultaneous multiple sensor and actuator fault estimation is justified. As a results, a recently developed multiple sensor and actuator scheme is outlined. Each of the reviewed schemes is accompanied by a set of illustrative examples. The final part of the paper contains a discussion about possible future research directions.

This paper deals with the problem of fault-tolerant control for non-linear systems under process faults. The proposed strategy is based on estimation strategy. The estimator is designed using the so-called quadratic boundedness approach. The control strategy in case of process fault is proposed. The final part shows an illustrative example with an implementation to the real laboratory multi-tank system.

The joined H$$_{2}$$/H$$_{\infty }$$ norm approach to virtual actuators design, intended to linear discrete-time systems, is presented in the paper. Using the joined norm principle, new conditions for virtual actuators design are formulated in terms of linear matrix inequalities. Related to the static output control under influence of single actuator faults, an example is presented to highlight the benefit of the proposed framework.

The paper considers the problem of control reconfiguration to retain fault tolerance in linear continuous-time systems. The full state control law is combined with the static output compensation for the reference specification in the so called cascade reconfiguration structure. Following the concept of reference model control, the main idea is to keep untouched the nominal parameters of such control structure. Formulating through the model reference principle, the extended Erzberger’s conditions are derived for a cascade reconfiguration structure and new design conditions are introduced and proven for systems with actuator faults. The obtained results are illustrated with a numerical example to note the effectiveness of the proposed approach.

The paper presents the results of the regulation strategy based on the reconfiguration of the control allocation module steering redundant actuators. As the reconfiguration criterion the required level of system reliability was assumed at the end of the operating time, estimated by the means of stochastic differential equations. In this case, two tasks become the key. The first concerns the choice of the method of determining the pseudo inverse matrix taking into account the constraints on the control signal. The second concerns the prioritization of the actuators. i.e. the distribution of the increased demand for the control signal so that, this increase, in minimum way, decreases the reliability.

The paper presents a robust Guaranteed Cost Control (GCC) for nonlinear system using Product Reduction Algorithm (PRA). The proposed approach starts with a general description of the nonlinear system with nonlinear term in the state equation and assumptions regarding to a nonlinear function. The subsequent part of the paper is concerned with the design of the robust controller using Linear Matrix Inequalities (LMIs). Next, an algorithm to solve linear optimization problem base on PRA is proposed. The final part presents results obtained for the two–tank system.

Diagnostic reasoning is often based on abduction. Abductive inference consists in generation of hypotheses which explain the current behavior of the system under investigation. Such a reasoning is based on accessible background knowledge and the results must be consistent with all auxiliary observations. Efficient abductive diagnosis is carried out as Model-Based Reasoning. The knowledge about the model defines the search-space for diagnostic hypotheses. Unfortunately, use of classical consistency-based reasoning leads to rough, qualitative results only, even if good knowledge of the correct model is available. In this paper and attempt to use Constraint Programming as a tool for diagnostic reasoning is presented. The ultimate goal is to provide more precise diagnoses. Two case studies, one concerning fault parameter evaluation, and the second concerning structural fault localization are presented.

Measurements play a crucial role in diagnosis of industrial processes. The main aim of this paper is to discuss selected problems connected with the role of measurements in model-based diagnosis. The issue of the separation of the diagnostics of measuring instruments and process diagnosis is discussed. The idea of self-diagnosable sensors of single faults is shown as a solution. A short review of existing techniques meeting the requirements for sensors of single faults is presented. We also discuss practical heuristic rules, that can be used in the selection of measuring instruments and diagnostic tests for diagnostic system.

This paper proposes an $$ H_{-}/H_{\infty } $$ fault detection observer for discrete-time polytopic linear parameter-varying (LPV) system based on relative degree of output. First, a new output is generated by gathering the original output and its time shifted value. Then an $$ H_{-}/H_{\infty } $$ fault detection observer is designed for the new system such that the generated residual is robust against unknown disturbances and sensitive to actuator faults, simultaneously. In order to reduce some conservativeness, the observer is solved by an iterative LMIs algorithm. Simulations results are given to illustrate the effectiveness of the proposed method.

Recently due to increasing complexity of systems existing in real world there is growing demand for inspecting and then diagnosing them. Inspection and diagnostics of large objects, such as galleries of mines affected by catastrophes, large civil engineering objects, airports, railway lines, or even regions of catastrophes, are done mainly by human experts. Unfortunately, experts are often faced with hazards to their lives and health, or simply must carry out long-lasting and boring work. To this end, mobile agents and intelligent systems can significantly improve not only safety of inspectors, but also efficiency of their work.The paper deals with general issues concerning robotized inspection and diagnostics that consists in replacing human experts by direct acting on the scene of operation, and also by interpreting evidence collected, finding conclusions, and reporting the work. Moreover, several examples are given concerning applications of robotized inspection and diagnostics in contemporary technical systems.

The idea of fuzzy inference approach to fault diagnosis in decentralized, single-level diagnostic structures is introduced in this paper. This approach is particularly intended for large scale industrial systems. The novel and practicable on-line fuzzy fault isolation approach in single-level structure is proposed and discussed. The fuzzy approach allows among others the refinement of the diagnoses particularly when taking into account the uncertainty of the fault symptoms. The proposed approach is depicted in an example. The conclusions regarding expected benefits of the decentralized two-level diagnostic structures summarize the paper.

The article presents a comparative analysis of two variants of the Particle Kalman Filter designed by using two different ship motion models. The first filter bases only on the kinematic model of the ship and can be used in many types of vehicles, regardless of the vehicle dynamics model. The input value to the filter is the noisy position of the ship. The second filter makes use of the kinematic and dynamic models of the moving ship. The input values to the filter are the noisy ship position and the forces generated by ship propellers during manoeuvres. These filters are used as state observers. The output values from the filters are position and velocity vectors in three degrees of freedom in the global coordinate system. The simulation test results show that both filters reveal similar accuracy in state observer.

The vibroacoustic diagnostics is currently one of the most important methods used for assessment of the mechanical condition of power transformers’ active part (windings and core). The analysis of transformer’s tank vibrations is performed for the steady state (stable load or no load) and for the transient state, during the first couple of seconds after unloaded transformer energization. Vibrations signal is recorded with accelerometric sensor attached to the transformer tank. In the case of transient state it is very important to determine the envelope of tank vibrations recorded signal. There is very often used for this purpose amplitude detector AM-DSB, which algorithm is based on the Hilbert filter. However recorded signal of vibrations (proportional to the acceleration) does not fulfill conditions of AM-DSB signal. Using the standard envelope detection (Hilbert filter) leads to wrong conclusions. In this paper a modified algorithm of envelope detector is presented, which can be used each time, where there is a need to determine signal’s envelope, that does not fulfill conditions of amplitude modulation AM-DSB. The quality of proposed algorithm was experimentally verified on the example of two transformers: low and medium power (0.8 MVA and 16 MVA).

In this paper the issue of continuous systems estimation, insensitive to certain perturbations, is presented and discussed. Such an approach has rational advantages, especially when robust schemes are used to assist a target system responsible for industrial diagnostics. This requires that estimated model parameters are generated on-line, and their values are reliable and to a great extent accurate. Practical hints are suggested to challenge the consistency problem of estimates. Namely, the technique of instrumental variables can improve the asymptotic behavior of estimators. With a weighting mechanism, in turn, tracking the time-varying parameters of non-stationary processes is realistic. Yet, evident insensitivity to destructive outliers in the measurement data is guaranteed by the applied estimation routine in the sense of the least sum of absolute errors. Finally, premises for a proper selection of persistently exciting input signals, as well as the directions of further research are summarized in the paper.

The paper presents the concept of time series decomposition by splitting into components with linear filtering methods. The modified moving trend algorithm (MTF) allows for more precise specification of desired trend properties and periodic component extraction from seasonal time series. In the paper the time and frequency properties of classical and modified FIR filters are presented and confronted with 4th order Butterworth filter. Three examples of empirical, seasonal time series are treated with the analyzed filters. Advantages and drawbacks of the proposed filters concerning the cyclic component extraction efficiency are discussed on the base of the processing results shown in time and frequency domain. Recommendations for the appropriate moving-trend bases filter selection suitable for processed time series properties are presented.

This paper introduces approximate analytic quality criteria useful in assessing the efficiency of evolutionary multi-objective optimization (EMO) procedures. We present a summary of extensive research into computing. In the performed comparative study we take into account the various approaches of the state-of-the-art, in order to objectively assess the EMO performance in highly dimensional spaces; where some executive criteria, such as those based on the true Pareto front, are difficult to calculate. Whereas, on the other hand, the proposed approximated quality criteria are easy to implement, computationally inexpensive, and sufficiently effective.

A new method is proposed for black-box linear model identification of a dynamic system embedded at a nearly Gaussian noise. The Gaussian process can highlight areas of the output spaces where the prediction quality is poor, due to the lack of data or its complexity, by indicating the higher variance of the predicted mean; the input spaces in which we can reconstruct data represent the expected values. This paper proposed a new approach for the online system identification for non-zero initial conditions in the moving window.

The goal of this paper is to present a development of a low cost flight simulator, that allows to simulate flight controls failures. Cessna 172 has been chosen as an example of a general aviation aircraft and the flight model has been implemented in Simulink. The model allows for easy integration of an experimental autopilot, using various strategies. Aerodynamic coefficients have been calculated using software called DATCOM. Such approach reduces greatly the cost of development. An approximate inertia matrix has been calculated for proper rigid body dynamics, and a model of engine and propeller is based on actual data from producer. In the end evaluation has been performed, indicating that modeled airplane behaves in a proper way.

The paper addresses the problem of computational efficiency of the pipe-flow model used in leak detection and identification systems. Analysis of the model brings attention to its specific structure, where all matrices are sparse. With certain rearrangements, the model can be reduced to a set of equations with tridiagonal matrices. Such equations can be solved using the Thomas algorithm. This method provides almost the same values of the state vector and maintains stability for the same discretization grid, while the computational overhead is vastly reduced.

This work considers the problem of river floods, and presents a computer solution that can be used to predict such a threat. The basic tool, called Hydrologic Engineering Centers for River Analysis System (HEC-RAS), and created by the US Army Corps of Engineering, is well developed; and the models created in the system HEC-RAS are quite realistic. For the purpose of solving practical national flood problems, simplified static and dynamic models of the river Vistula were created in the system HEC-RAS. Furthermore, in this paper, the results of the performed simulations for flood problems in the Vistula basin are presented.

Traditionally posture deformity assessment for screening purposes is performed by visual examination of patient’s body by an expert. During further follow-up of the diagnosed deformity full spine, X-Ray 2D images are acquired. 3D medical imaging (Computer Tomography and MRI) is used when the spinal surgery is considered. Visual examination is subjective and is strongly dependent on expert knowledge. X-Ray and tomographic imaging exposure can be contraindicated in some cases (i.e. pregnancy). During last two decades, the dynamic development of methods and systems for 3D scanning and algorithms for measurement data analysis is observed. 3D scanning is successfully applied in modern industrial production lines, documentation of cultural heritage and human body analysis. Recently, algorithms for data analysis allow for 100% inspection of complex geometry and have increasing support for control of the technological process parameters on the base of calculated deviation between measurement and assumed 3D model. In this paper, we present an alternative approach for back posture analysis based on structured light 3D scanning. We present three different systems: mobile solution for prescreening of back shape, full body 3D scanner for monitoring of posture deformities and 4D scanner for dynamic analysis. Such a three-stage system fit into the idea of evidence-based medicine. Each of presented devices produces 3D geometry data representing the surface of patient’s body. Each of them has also accompanying software that has been developed for processing of geometry data into a final form that is easily interpreted by medical experts (angles, asymmetries, 3D models, changes in time of analyzed measures, etc.).

The paper presents BPD (Bronchopulmonary Dysplasia) prediction for extremely premature infants after their first week of life. In contrast to the most works where LR (Logit Regression) is used, the naive Bayes classifier was proposed. Data was collected thanks to the Neonatal Intensive Care Unit of The Department of Pediatrics at Jagiellonian University Medical College and includes 109 patients with birth weight less than or equal to 1500 g. Fourteen different features were considered and all $$2^{14}$$ of theirs combinations were analyzed. This paper also includes an accuracy and its deviation comparison with other prediction methods. It was possible because the calculations were performed on the very same data, which was used in previous works presenting LR and SVM forecasts.

HCI technology improves human-computer interaction. Such communication can be carried out with the use of emotions that are visible on the human face since birth [1]. In this paper the Emotion system for detecting and recognizing facial expressions, developed in the MSc work [2], is presented. The system recognizes emotion from webcam video in real time. It is based on color segmentation and morphological operations. The system uses a cascade of boosted classifiers based on Haar-like features, to locate the face and to reduce the searched area for characteristic points. For identification purposes, the Emotion system uses an expanded action unit EAU, based on a facial action coding system, FACS [3, 4].

Outlier detection in medical data covers a broad spectrum of medical research. In this paper, the authors propose a new approach to outlier detection based on the multi-objective genetic algorithm. In medical data, an outlier may be considered as a deviation which indicates the existence of cancerous cells in the breast. The paper presents the results of tests which were conducted on the set of medical data from the repository. The results of the study indicate that our method can be successfully applied to the medical problem in question.

The paper presents method of nuclei segmentation on cytological images based on the Convolutional Neural Network (CNN) and modified Hough Transform method. It approximates nuclei by ellipses fitted to nuclei regions segmented by CNN. As study data set 50 cytological RGB images were used, divided into training set (50 images) and test set (10 images). The first step is to create a CNN model for pixel-wise classification of cytological images. As training set for CNN, patches of size 28$$\,\times \,$$28 pixels were created based on images from training set and corresponding ground-truth labels. Using trained model, nuclei regions classification and segmentation from test set images was conducted. The reason of choosing the CNN for segmentation it’s better accuracy in separated overlapping nuclei than conventional methods such as for example Otsu thresholding etc. Subsequently, using Canny algorithm and Euclidean Distance Transform (EDT), edges and centers of segmented regions were extracted. Edges and centers of nuclei were extracted for reduce time computation for next step. Finally, finding nuclei using the modified Hough Transform by fitted ellipses was carried out.

In this work a new method of automatic detection of apnea–hypopnea episodes is presented. It combines snore/nonsnore classification with information about body and limbs movements. The snore/nonsnore detection is performed using Discrete Fourier Transform and energy calculation. The feature space is reduced using Linear Discriminant Analysis and a linear classifier was obtained. The feasibility of this method was tested on the set of 8 full-night polysomnography recordings of which 2 indicate sleep apnea syndrome. The result shows that the method is effective in detection of apneic events.

Floods are one of the most important natural hazards, and thus flood prediction and monitoring is an important research problem. Over several decades, the fault diagnosis community has established techniques for modelling, prediction, and uncertainty assessment. In this paper, we investigate whether these methods can be applied to the advantage of flood risk assessment expert systems. The paper contains a general description of the concept of a rule-based flood expert system. We show specifics of flood monitoring, including the main physical phenomena and available input data. Qualitative description of main processes leading to floods is provided, using the example of river floods. All of the problems and challenges are considered from the point of view of the fault diagnosis community. The analysis is promising, but more work is needed to apply the diagnostics approach efficiently to flood risk assessment.

This article investigates the effects of voltage dips on the operation of an electrical-motor-driven centrifugal gas compressors with major focus on time to surge key performance indicator behavior. Centrifugal and axial compressors are susceptible to phenomena described as surge. Surge is associated with operation in a specific region of the compressor operating area, located on the left-hand side of the so-called “surge line”, where an area of flow instability typically caused by compressor inducer stall which can cause severe damage to machine. The protection from compressor’ surge is one of the most critical turbomachinery control applications: this function is more important in electrical motor driven compressors, because the dynamics of the electrical driver are faster than the anti-surge control system. In order to prevent catastrophic failures of the electrical motor driven compressors is crucial to know the residual time before a surge phenomenon could manifest, also defined as the “time to surge” KPI.

Fires in coal mills are undesirable phenomenon, especially in the case of co-fired biomass. In order to face this problem the detection and prevention system of fire and ignition was created, which has two systems: identify the loss of fuel and fire detection. Logic signals which informs about loss of fuel might be used in control systems of the load and the amount of total air. The logic signal informing about fire detection of the mill is created when the defined threshold value is exceeded by the signal relating the difference in temperatures between the coal-air mixture and its model. It is then successfully used in the system protecting the mill against the propagation of a threat (fire), for example dosaging of chemically neutral gas (water spray)

The paper proposes a method to select time-periods in which the measured data conform to steady-state conditions. This selection is the first step of any monitoring method based on estimating parameters of a static model. The proposed method is based on a local polynomial modeling of the data evolution and deals with multivariate data by applying principal component analysis. This method is applied on data collected from an industrial heat exchanger to monitor its heat exchange capacity.

The paper presents a study on health diagnosis and prognosis of an industrial diesel motor. Two well-known approaches, Hidden Markov Model (HMM) and particle filter (PF), are applied from real recorded data with different measurements. The recorded data is firstly pre-processed and health indicator is then chosen before implementing each used approach. The obtained results are analyzed and discussed. The use and advantages of each approach are finally highlighted.

A system which checks the quality of automatic control system and diagnoses power unit’s operation is presented. The system is cooperates with existing DCS systems. Using the received, registered and processed data, the analysis results are presented. They allow the performance evaluation of the control systems and power unit (technological parameters excess and its influence on efficiency, characteristics of executive elements and directive to its linearization, valves tightness, valves and pumps cavitation, start-up costs) in a given period. The received information enables the operators to take preventive action in advance. Consequently, it is possible to obtain more efficient power station’s operation and the exploitation cost reduction.

Cyber security is an integral part of security system of any advanced country. Given the fact that the number of cyber attacks constantly increase with concurrent increase of their technological complexity, the paper proposes a new classifier structure to speed up detection of unauthorized interference while maintaining the established accuracy parameters. Method of reducing input data-flow dimensions is the basis for the designed structure of cyber attacks classifier. Unlike other well-known classifier principles, this one is based on a binary type classification of event patterns and two-stage scheme of network connection input data classification. The classifier is verified on the basis of real data and compared with advanced world standards. The results have confirmed the ability of the classifier to quickly detect and classify cyber attacks without loss of accuracy.

In this paper authors presents report about current Android applications security. OWASP’s top 10 mobile security risks was used, to verify level of security. Authors also design concept tool which will detect potential risks in code of Android application after reverse engineering of it.

The aim of this article is to identify and discuss some issues that can be encountered in designing the industrial automation and control systems (IACS) for implementing safety functions. In a functional safety standard IEC 61511 for the process industry such systems are named the basic process control systems (BPCS) and the safety instrumented systems (SIS). In a generic functional safety standard IEC 61508 they are depicted as the electric/electronic/programmable electronic systems (E/E/PES). The role of these systems is to implement safety functions for effective reducing and controlling the individual risk and/or societal risk in life cycle in relation to tolerable risk levels defined for given hazardous plant. Some aspects of potential influence of danger failures of the E/E/PES or SIS on the plant safety are considered. The influence of common cause failures (CCF) in verifying the safety integrity levels (SIL) achieved by safety functions is evaluated and discussed.

Human factors and cognitive engineering are considered nowadays as important multidisciplinary domains that focus on improving the relations between humans, technology and systems to be supervised and operated. The industrial automation and control systems (IACS) in hazardous plants are increasingly computerized and perform various safety functions. These are usually designed and implemented according to the functional safety requirements. The objective is to maintain high performance and productivity of the plant, and reduce risks related to identified hazards and threats. An approach is proposed to apply selected cognitive engineering methods for verifying the design of safety systems to be implemented in hazardous plant in context of defined safety functions, operator interfaces, procedures and other factors influencing risks.

Cybercrime becomes a real problem in the everyday operation of production plants, industrial control systems, and other technical devices. The purpose of this article is to demonstrate that on-line diagnostics is an effective way to recognize cyberattacks. The security assurance system against cyberattacks for Industrial Control Systems is layered. The article discusses the possibility of the usage of known methods of process diagnostics to recognize cyberattacks in the Industrial Control Systems as an additional protection layer. Cyberattacks manifest with a variety of changes in the operation of the control system and the process flow deviating from its normal state. The discussed concept is to detect such changes based on models, evaluate them and on this basis conclude about the primary reason of abnormal operation, including the detection of cyberattack. Simple examples of such detection system are also presented.

Advanced techniques of monitoring, diagnostics or remote supervision of technical devices based on current and archival measurements are more often used for simple and chip devices. It raises the need to collect and store measurements from such devices, preferably in the cloud system. This paper analyses performance of different, SQL as well as NoSQL databases used as a storage layer for such system, designed to keep and process time series. An efficiency in sense of disk usage and processing time for three most popular open source database engines (MySQL, Postgres, MongoDB) was described.