Fault detection and diagnosis based on particle filters combined with interactive multiple-model estimation in dynamic process systems

This paper presents a simple yet efficient adaptive Kalman filter for tracking targets expected to perform varying turn maneuvers. The process noise covariance level of a second order Kalman filter is adjusted at each time step according to the calculated turn rate. The turning rate is estimated from the magnitude of the calculated acceleration divided by the estimated speed of the target. At each scan the previous and current velocity estimates are used to calculate the acceleration. The comparison of the performance of the proposed algorithm is made with that of an interacting multiple model (IMM) algorithm, employing three models with different levels of process noise covariance and also to that of a second order Kalman filter. Two different assumptions have been made for selecting the process noise values for the the IMM and Kalman filter algorithms, in the first case it was assumed that there was no prior information about the target motion whereas in the second case it was assumed that the largest turn rate that the target of interest could perform was known. The IMM algorithm utilizing three models gives slightly better estimates during the nonmaneuvering periods, but the proposed algorithm is superior to the IMM algorithm during maneuvering periods in terms of estimation errors. Also the proposed algorithm requires 78% less computation, almost the same number of calculations required by a single fixed process noise Kalman filter, than the IMM algorithm.

The problem of fault detection and diagnosis (FDD) in dynamic systems has received considerable attention in last decades due to the growing complexity of modern engineering systems and ever increasing demand for fault tolerance, cost efficiency, and reliability (Willsky, 1976; Basseville, 1988). Existing FDD approaches can be roughly divided into two major categories including model-based and knowledge-based approaches (Venkatasubramanian et al., 2003a; Venkatasubramanian et al., 2003b). Model-based approaches make use of the quantitative analytical model of a physical system. Knowledgebased approaches do not need full analytical modeling and allow one to use qualitative models based on the available information and knowledge of a physical system. Whenever the mathematical models describing the system are available, analytical model-based methods are preferred because they are more amenable to performance analysis. Generally, there are two steps in the procedure of model-based FDD. First, on the basis of the available observations and a mathematical model of the system, the state variable x and test statistics are required to be obtained. Then, based on the generated test statistics, it is required to decide on the potential occurrence of a fault. For linear and Gaussian systems, the Kalman filter (KF) is known to be optimal and employed for state estimation. The innovations from the KF are used as the test statistics, based on which hypothesis tests can be carried out for fault detection (Belcastro & Weinstein, 2002). In reality, however, the models representing the evolution of the system and the noise in observations typically exhibit complex nonlinearity and non-Gaussian distributions, thus precluding analytical solution. One popular strategy for estimating the state of such a system as a set of observations becomes available online is to use sequential Monte-Carlo (SMC) methods, also known as particle filters (PFs) (Doucet et al., 2001). These methods allow for a complete representation of the posterior probability distribution function (PDF) of the states by particles (Guo & Wang, 2004; Li & Kadirkamanathan, 2001). The aforementioned FDD strategies are single-model-based. However, a single-model-based FDD approach is not adequate to handle complex failure scenarios. One way to treat this

Control of the time between sawtooth crashes, necessary for ITER and DEMO, requires real-time detection of the moment of the sawtooth crash. In this paper, estimation of sawtooth crash times is demonstrated using the model-based interacting multiple model (IMM) estimator, based on simplified models for the sawtooth crash. In contrast to previous detectors, this detector uses the spatial extent of the sawtooth crash as detection characteristic. The IMM estimator is tuned and applied to multiple ECE channels at once. A model for the sawtooth crash is introduced, which is used in the IMM algorithm. The IMM algorithm is applied to seven datasets from the ASDEX Upgrade tokamak. Five crash models with different mixing radii are used. All sawtooth crashes that have been identified beforehand by visual inspection of the data, are detected by the algorithm. A few additional detections are made, which upon closer inspection are seen to be sawtooth crashes, which show a partial reconnection. A closer inspection of the detected normal crashes shows that about 42% are not well fitted by any of the full reconnection models and show some characteristics of a partial reconnection. In some case, the measurement time is during the sawtooth crashes, which also results in an incorrect estimate of the mixing radius. For data provided at a sampling rate of 1 kHz, the run time of the IMM estimator is below 1 ms, thereby fulfilling real-time requirements.

The interacting multiple model (IMM) algorithm is a widely accepted state estimation scheme for solving maneuvering target tracking problems, which are generally nonlinear. During the IMM filtering process, serious errors can arise when a Gaussian mixture of posterior probability density functions is approximated by a single Gaussian. Particle filters (PFs) are effective in dealing with nonlinearity and non-Gaussianity. This work considers an IMM algorithm that includes a constant velocity model, a constant acceleration model and a 3D turning rate (3DTR) model for tracking three-dimensional (3D) target motion, using various combinations of nonlinear filters. In existing literature on combining IMM and particle filtering techniques to tackle difficult target maneuvers, a PF is usually used in every model In comparison, simulation results show that by using a computationally economical PF in the 3DTR model and Kalman filters in the remaining models, superior performance can be achieved with significant reduction in computational costs.

In order to effectively monitor condition and detect fault types of high nonlinear system, and extract the features of system state under strong noise background, this paper proposes a novel fault detection and diagnosis (FDD) method based on modified particle filter (PF). The artificial neural network is incorporated in PF for adaptively adjusting weight of particle. In the modified PF, the large weight particles are split into several small weight particles, the particles with smaller weight is adjusted by using artificial neural network. By which the particles in the low probability density region are adjusted to the high probability density region, and the problem of particle leanness is solved effectively. Moreover, this paper also uses time-varying auto regressive (TVAR) and Akaike information criterion (AIC) methods to establish state space model for state estimation. Finally, the proposed method is implemented for fault diagnosis on a roller bearing. Good results are obtained, and the bearing faults, such as the outer race, the inner race and the roller element defects, have been effectively discriminated.

Hybrid systems consist of both continuous state evolutions and discrete state (or mode) transitions. In many hybrid systems, mode transitions are governed by guard conditions that are dependent on the continuous state. These kind of mode transitions can be described by continuous-state-dependent mode transition probabilities. However, most multiple model Kalman filter based algorithms, such as the well-known interacting multiple model (IMM) algorithm, assume constant mode transition probabilities. We propose a hybrid estimation algorithm based on the IMM approach for the stochastic hybrid systems with continuous-state-dependent transitions. We first develop two models to describe the guard conditions for various stochastic hybrid systems. From the models of the guard conditions, we derive the corresponding continuous-state-dependent mode transition probabilities, which are expressed in terms of Gaussian probability density functions (pdfs) or Gaussian cumulative density functions (cdfs). We then propose a hybrid estimation algorithm for the stochastic hybrid systems based on the IMM algorithm. Like the IMM algorithm, our algorithm utilizes Gaussian distributions to represent the continuous state pdfs which are updated analytically at each time step. We also use analytical integration methods to reduce the computational cost of updating the mode probabilities. As a result, the proposed algorithm is computationally efficient and is applicable to high-dimensional problems. The performance of the proposed algorithm is illustrated with two examples in air traffic control tracking applications.

Fault detection and diagnosis (FDD) are increasingly important for wheeled mobile robots (WMRs), especially those in unknown environments such as planetary exploration. There are many kinds of fault diagnosis methods available for mobile robots, including multiple model-based approaches, particle filter based approaches, sensor fusion based approaches. Currently, all of these methods are designed for complete models. However, completely modeling a system is difficult, even impossible. In this paper, particle filter and neural network are integrated to diagnose complex systems with imperfect models. Two features are extracted from particles: the sum of sample weights, and the maximal a posteriori probability. These features are further feed to a neural network to decide whether the estimation given by the particle filter is credible or not. An incredible estimation indicates that the true state isn’t included in the state space, i.e. it is a novel state (or an unknown fault). This method preserves the merits of particle filter and can diagnose known faults as well as detect unknown faults. It is testified on a real mobile robot.KeywordsMobile RobotFault DiagnosisParticle FilterBelief StateMulti Layer PerceptronThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Robust Fault Detection and Diagnosis for Multiple-Model Systems with Uncertainties

Dynamic kriging based fault detection and diagnosis approach for nonlinear noisy dynamic processes

Variable Structure Multiple Model (VSMM) estimation generalizes Multiple Model (MM) estimation by assuming that the set of models used for MM estimation is time varying. By using VSMM estimation, a large model set which may cover all possible target maneuvers can be used without significant increase of computational load, while maintaining a reasonable estimation accuracy. Various implementable VSMM algorithms, like Model Group Switching (MGS), Likely Model Set (LMS) and Minimal Sub-Model-Set Switching (MSMSS) using the Interacting Multiple Model (IMM) algorithm with a sub-model-set adaptation logic have appeared in the recent literature. However, the use of these algorithms for tracking maneuvering target in clutter has not been explored. In presence of clutter, one need to use data association technique to differentiate target originated measurement from clutter. The probabilistic data association (PDA) has been popularly adopted to many algorithms for tracking in clutter. In this paper, we integrate PDA technique with MSMSS and propose a VSIMM-PDA algorithm for maneuvering target tracking in clutter. A new grating technique to account for potential model errors is used. A numeric example via multiple Monte Carlo runs, which compares the performance of the new algorithm to a standard IMM-PDA in terms of Root Mean Squared error (RMS) and percentage of track loss, is presented.© (2003) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

This paper presents a new scheme for fault detection and diagnosis in spacecraft Attitude Determination (AD) sensors. An integrated attitude determination system, which includes measurements of rate and angular position using rate gyros and vector sensors, is developed. Measurement data from all sensors are fused by a linearized Kalman filter, which is designed based on the system kinematics, to provide attitude estimation and the values of the gyro bias. Using this information the erroneous sensor measurements are corrected, and unbounded sensor measurement errors are avoided. The resulting bias-free data are used in the Fault Detection and Diagnosis (FDD) scheme. The FDD algorithm uses model-based state estimation, combining the information from the rotational dynamics and kinematics of a spacecraft with the sensor measurements to predict the future sensor outputs. Fault isolation is performed through Extended Kalman Filters (EKFs). The innovation sequences of EKFs are monitored by several statistical tests to detect the presence of a failure and to localize the failures in all AD sensors. The isolation procedure is developed in two phases. In the first phase, two EKFs are designed, which use subsets of measurements to provide state estimates and form residuals, which are used to verify the source of the fault. In the second phase of isolation, testing of multiple hypotheses is performed. The generalized likelihood ratio test is utilized to identify the faulty components. In the scheme developed in this paper a relatively small number of hypotheses is used, which results in faster isolation and highly distinguishable fault signatures. An important feature of the developed FDD scheme is that it can provide attitude estimations in presence of fault in one type of sensors.

Correct knowledge of noise statistics is essential for an effective estimator in maneuvering target tracking. In practice, however, the noise statistics are usually unknown or not perfectly known. To deal with the estimation problem in linear discrete-time systems with Markov jump parameters, where the measurement noise covariance is unknown, a novel approach is presented in this paper. This approach is based on the interacting multiple model (IMM) framework. An H ∞ filter is employed to construct a noise statistics estimator to obtain the information which is necessary for the IMM algorithm. In our approach, even the priori knowledge of noise statistics is not needed. The noise statistics loss problem is solved while the merits of IMM algorithm is reserved. The effectiveness of the proposed approach is demonstrated in comparison with single-model H ∞ filter through Monte Carlo simulation for maneuvering target tracking.

A fault-tolerant dead reckoning system is presented for a modular vehicle, which consists of one chassis unit and several wheel units. The units locally estimate the vehicle position based on their own internal sensors. The local estimates are exchanged among the units via an inter-communication system, and they are fused in a decentralized manner. The units can then determine the vehicle position accurately. The decentralized dead reckoning algorithm is formulated based on the information filter and the covariance Intersection method. For enhancing the reliability of the dead reckoning a multi-model based fault detection and diagnosis (FDD) of the internal sensors is incorporated into the dead reckoning system. The units diagnose their sensors with the FDD system, and they apply only the normal sensors for the vehicle localization. In this paper two fault modes (hard fault and noise fault modes) of the sensors are handled; on the hard fault the sensor output is stuck at a constant value. On the noise fault it is disturbed by a large noise. The FDD algorithm is based on the variable structure interacting multiple-model estimator. The fault-tolerant dead reckoning algorithm was implemented on our indoor test-vehicle, which consists of one chassis unit and four wheel units. Experimental results show that our dead reckoning provided better localization accuracy than the conventional one (i.e., the dead reckoning without sensor FDD system) did even though the sensors partially failed.

Hybrid modeling based double-granularity fault detection and diagnosis for quadrotor helicopter

Electric Vehicles (EVs) are emerging as the preferred mode of transportation owing to various advantages and to reach the Sustainable Development Goals (SDGs). However, the commercial usage of EVs is in its initial stage, due to multiple factors, among which faults caused by the operational environment and component failures are one. Furthermore, early diagnosis of such faults is necessary to protect from severe damage and catastrophic failures. This paper presents various fault modes in EV components, their impact on EV performance, and Fault Detection and Diagnosis (FDD) practices. Later sections of the paper focus on the use of Artificial Intelligence (AI) and Machine Learning (ML) techniques in FDD. Also, it provides a comparative analysis of model-based, signal-based, data-driven, and hybrid FDD approaches. Additionally, the paper explores component-specific FDD techniques, modelling of drive faults, current research trends, and suggestions on the selection of feature extraction techniques. Furthermore, the paper highlights the need to develop a hybrid FDD method to enhance the accuracy, reliability, and speed of FDD in EVs. The future scope of EV FDD techniques, considering various aspects, is also presented.