Multifidelity Framework for the Efficient Identification of Damages in Complex Aerospace Systems

Next-generation aircraft require the development and integration of a deal of innovative green technologies to meet the ambitious sustainability goals set for aviation. Those transformational efforts are associated with a tremendous increase in the complexity of the onboard systems and their multiphysics-coupled behaviors and dynamics. A critical aspect relates to the identification of the coupled faults resulting from the integration of those green technologies, which introduce damage identifiability issues and demand new approaches for the efficient and accurate identification of non-nominal fault conditions. Model-based fault detection and identification (FDI) methodologies have been proven essential to identify onboard the damages affecting the systems from physical signal acquisitions, but existing methods are typically computationally expensive and fail to capture incipient coupled faults, which prevents their adoption and scaling with the increasing complexity of novel multiphysics systems. This work introduces a multifidelity framework to accelerate the identification of fault modes affecting complex systems. An original two-stage compression computes an optimally informative and highly reduced representation of the monitoring signals for the minimum demand of onboard resources. A multifidelity scheme for Bayesian inversion is developed to infer multidomain fault parameters from the compressed signals: variable cost and fidelity models are optimally queried for a major reduction of the overall computational expense. The framework is demonstrated and validated for aerospace electromechanical actuators affected by incipient multimodal faults. Remarkable accelerations of the FDI procedure are observed, and the exact identification of the incipient fault condition was achieved one order of magnitude faster than with standard algorithms.

The objective of this paper is to present the development and numerical testing of a robust fault detection and identification (FDI) system using artificial neural networks (ANNs), for incipient (slowly developing) faults occurring in process systems. The challenge in using ANNs in FDI systems arises because of one's desire to detect faults of varying severity, faults from noisy sensors, and multiple simultaneous faults. To address these issues, it becomes essential to have a learning algorithm that ensures quick convergence to a high level of accuracy. A recently developed accelerated learning algorithm, namely a form of an adaptive back propagation (ABP) algorithm, is used for this purpose. The ABP algorithm is used for the development of an FDI system for a process composed of a direct current motor, a centrifugal pump, and the associated piping system. Simulation studies indicate that the FDI system has significantly high sensitivity to incipient fault severity, while exhibiting insensitivity to sensor noise. For multiple simultaneous faults, the FDI system detects the fault with the predominant signature. The major limitation of the developed FDI system is encountered when it is subjected to simultaneous faults with similar signatures. During such faults, the inherent limitation of pattern-recognition-based FDI methodsmore » becomes apparent. Thus, alternate, more sophisticated FDI methods become necessary to address such problems. Even though the effectiveness of pattern-recognition-based FDI methods using ANNs has been demonstrated, further testing using real-world data is necessary.« less

Computational framework for real-time diagnostics and prognostics of aircraft actuation systems

Detection of incipient (slowly developing) faults is crucial in automated maintenance problems where early detection of worn equipment is required. In this paper, a general framework for model-based fault detection and diagnosis of a class of incipient faults is developed. The changes in the system dynamics due to the fault are modeled as nonlinear functions of the state and input variables, while the time profile of the failure is assumed to be exponentially developing. An automated fault diagnosis architecture using nonlinear online approximators with an adaptation scheme is designed and analyzed. A simulation example of a simple nonlinear mass-spring system is used to illustrate the results.

Electro Hydraulic Actuators (EHAs) keep their role as the leading solution for the control of current generation primary flight control systems: the main reason can be found in their high power to weight ratio, much better than other comparable technologies. To enhance efficiency and reliability of modern EHAs, it is possible to leverage the diagnostics and prognostics disciplines; these two tools allow reducing life cycle costs without losing reliability, and provide the bases for health management of integrated systems, in compliance with regulations. This paper is focused on the development of a fault detection algorithm able to identify the early signs of EHA faults, through the recognition of their precursors and related degradation patterns. Our methodology provides the advantage of anticipating incoming failures, triggering proper alerts for the maintenance team to schedule adequate corrective actions, such as the replacement of the degraded component. A new EHA model-based fault detection and identification (FDI) method is proposed; it is based on deterministic and heuristic solvers able to converge to the actual state of wear of the tested actuator. Three different progressive failure modes were chosen as test cases for the proposed FDI strategy: clogging of the first stage of the flapper-nozzle valve, spool-sleeve friction increase, and jack-cylinder friction increase. A dedicated simulation model was created for the purpose. The results highlighted that the method is adequate in robustness, since EHA malfunctions were identified with a low occurrence of false alarms or missed failures.

Prognostics and Health Monitoring (PHM) is a discipline aiming to determine in advance the Remaining Useful Life (RUL) of a system. To do so, the operation of the system is monitored in search of the early signs of degradation and incipient faults; then, a model for the propagation of faults is employed to estimate the propagation of damages and evaluate the RUL. Usually, a fault threshold is employed as a stopping criterion for the evaluation of damage propagation, but this is not a reliable method when dealing with multiple faults affecting the system at the same time. Specifically, the combined effect of two fault modes can cause the system not to meet its requirements well before the single faults reach their individual thresholds. In this work, we address a model-based strategy to estimate whether the system with incipient faults is still able to meet its performance requirements. The method is applied to aerospace actuators, and performance is evaluated in terms of dynamical response. This model-based algorithm is too slow to be evaluated in real-time, so a Support Vector Machine (SVM) is trained as a surrogate function to speed up the computation. The results and computational times of the full, physics based model and those of its surrogate are compared and discussed.

This paper presents a new model-based fault detection and estimation framework for a class of multi-input and multi-output (MIMO) nonlinear distributed parameter systems (DPS) described by partial differential equations (PDE) with actuator and sensor faults. The fault functions cover both abrupt and incipient faults. A Luenberger type observer is used to monitor the health of the DPS as a detection observer on the basis of the nonlinear PDE representation of the system with measured output vector. By taking the difference between measured and estimated outputs from this observer, a residual signal is generated for fault detection. If the detection residual exceeds a predefined threshold, a fault will be claimed to be active. Once an actuator or a sensor fault is detected and the fault type is identified, an appropriate fault parameter update law is developed to learn the fault dynamics online with the help of an additional output measurement. Eventually, the proposed detection and estimation framework is demonstrated on a nonlinear process.

Model-based fault detection and diagnosis is an interesting method, due to economical and safety related matters. However, in practice it is very difficult to achieve accurate models for complex nonlinear plants. If the plant structure is not precisely known, the diagnosis has to be based primarily on data or heuristic information. The inherent characteristics of fuzzy logic theory make it suitable for fault detection and isolation (FDI). In this paper is proposed a modification to the regularity criterion algorithm, aiming the reduction of computational time without loss of accuracy. The fuzzy models obtained using the modified regularity criterion algorithms are optimized by a real-coded genetic algorithm. An industrial valve simulator is used to obtain several abrupt and incipient faults in the system. These faults are some of the possible faults in the real system. The fuzzy models used in the FDI system were able to detect and isolate the simulated abrupt and incipient faults

In this paper, the controller synthesis problem for fault tolerant control systems (FTCS) with stochastic stability and H2 performance is studied. System faults of random nature are modelled by a Markov chain. Because the real system fault modes are not directly accessible in the context of FTCS, the controller is reconfigured based on the output of a fault detection and identification (FDI) process, which is modelled by another Markov chain. Then state feedback and output feedback control are developed to achieve the mean square stability (MSS) and the H2 performance for both continuous‐time and discrete‐time systems with model uncertainties. Copyright © 2006 John Wiley & Sons, Ltd.

Traditional feature extraction methods for analog circuits incipient fault diagnosis rely on signal processing technology and expert experience, the diagnostic accuracy for incipient faults is not satisfactory. To deal with these problems, a novel approach based on deep neural network is presented in this paper. The proposed method includes stacked autoencoders and a softmax classifier. The stacked autoencoders extract deep features from the raw time-domain responses, and then the softmax classifies the fault mode of analog circuits automatically. The experiment results on Sallen-Key band-pass filter and Leapfrog low-pass filter circuit illustrate that the proposed method can automatically and effectively extract incipient fault features and achieve high diagnostic accuracy.

\n \nIn this note, fault detection techniques based on finite dimensional\nresults are extended and applied to a class of infinite dimensional\ndynamical systems. This special class of systems assumes linear\nplant dynamics having an abrupt additive perturbation as the fault.\nThis fault is assumed to be linear in the (unknown) constant (and possibly\nfunctional) parameters.\nAn observer-based model estimate is proposed which serves\nto monitor the system's dynamics for unanticipated failures,\nand its well posedness is summarized.\nUsing a Lyapunov synthesis approach extended and applied to infinite\ndimensional systems, a stable adaptive fault diagnosis\n(fault parameter learning) scheme is developed. The resulting parameter\nadaptation rule is able to “sense” the instance of the fault occurrence.\nIn addition, it identifies the fault parameters using the additional\nassumption of persistence of excitation. Extension of the adaptive\nmonitoring scheme to incipient faults (time varying faults) is summarized.\nSimulations studies are used to illustrate the applicability\nof the theoretical results.\n\n

The signal acquisition system (SAS) operated by battery is designed in this paper. SAS includes signal acquisition and statistics function based on movement joints of basketball player. SAS is a recording of the electrical activity of the brain and pulse from the scalp and the recorded waveforms provide insights into the dynamic aspects of brain activity. The amplified SAS signals are digitized by an A/D converter. The digitized signal is transmitted to PC by a wireless serial port or stored in secure digital memory card. Experimental result shows that the system could implement the acquisition and storage of the foot compressive mechanical characteristics signals efficiently. This system would be of benefit to all involved in the use of SAS for sports training.

In order to further improve the timeliness and quality of wireless communication, an optimization method of wireless sensing signal fusion processing based on BPNN and DS evidence theory is proposed. The basic idea of the method: PSO is used as the link between BPNN and DS evidence theory in the field of wireless sensing signals, and achieves the global optimal solution process in signal processing and wireless communication. In the signal acquisition stage of wireless sensor nodes, BPNN is used to realize the artificial intelligence acquisition process of signal output, and PSO is used to realize the global optimal allocation of signal weight from signal acquisition input to signal output analysis; in the information fusion stage of wireless communication, DS evidence theory is used to realize the information fusion process, and PSO is used to realize the global optimal solution of information discrimination evidence synthesis. By proposing the joint innovation mode of the above intelligent algorithms, the main contributions of this paper are as follows: with the BPNN with PSO method, redundant signal compression processing and signal analysis efficiency can be improved; with the DS evidence theory with PSO method, the accuracy of information discrimination for wireless communication can be improved. This scientifically reconciles the contradiction between signal compression and information extraction.

An efficient radial distribution network reconfiguration method for minimization of energy losses over intervals of time is presented in this paper. In order to develop a simple and general method to solve the problem of network dynamic reconfiguration, the concept of system state characterization and its relevant applications are introduced in this paper, in which the computational effort required by traditional optimization method is decreased greatly through some appropriate linearization and stepwise correction techniques. The proposed method is based on loop-analysis theory and superimpose theorem, in which the objective is to minimize energy loss while keeping the feeder voltage profile within pre-specified limits. A case study of IEEE-33 bus system demonstrates the rationality and validity of strategies proposed in this paper, and the results illustrate that the proposed approach has a substantial reduction in computational expense while maintaining a high level of accuracy.

A new algorithm based on model reduction using the Krylov subspace technique is proposed to compute the DC solution of large nonlinear circuits. The proposed method combines continuation methods with model reduction techniques. Thus it enables the application of the continuation methods to an equivalent reduced-order set of nonlinear equations instead of the original system. This results in a significant reduction in the computational expense as the size of the reduced equations is much less than that of the original system. The reduced order system is obtained by projecting the set of nonlinear equations, whose solution represents the DC operating point, into a subspace of a much lower dimension. It is also shown that both the reduced-order system and the original system share the first q derivatives w.r.t. the circuit variable used to parameterize the family of the solution trajectories generated by the continuation method.