Power electronics are widely used in the transport and energy sectors. Hence, the reliability of these power electronic components is critical to reducing the maintenance cost of these assets. It is vital that the health of these components is monitored for increasing the safety and availability of a system. The aim of this paper is to develop a prognostic technique for estimating the remaining useful life (RUL) of power electronic components. There is a need for an efficient prognostic algorithm that is embeddable and able to support on-board real-time decision-making. A time delay neural network (TDNN) is used in the development of failure modes for an insulated gate bipolar transistor (IGBT). Initially, the time delay neural network is constructed from training IGBTs' ageing samples. A stochastic process is performed for the estimation results to compute the probability of the health state during the degradation process. The proposed TDNN fusion with a statistical approach benefits the probability distribution function by improving the accuracy of the results of the TDDN in RUL prediction. The RUL (i.e., mean and confidence bounds) is then calculated from the simulation of the estimated degradation states. The prognostic results are evaluated using root mean square error (RMSE) and relative accuracy (RA) prognostic evaluation metrics. Index Terms-Insulated gate bipolar transistor (IGBT), power electronics, prognostics, probability distribution function, remaining useful life (RUL), time-delay neural network.
Power electronics are increasingly important in new generation vehicles as critical safety mechanical subsystems are being replaced with more electronic components. Hence, it is vital that the health of these power electronic components is monitored for safety and reliability on a platform. The aim of this paper is to develop a prognostic approach for predicting the remaining useful life of power electronic components. The developed algorithms must also be embeddable and computationally efficient to support on-board real-time decision making. Current state-of-the-art prognostic algorithms, notably those based on Markov models, are computationally intensive and not applicable to real-time embedded applications. In this paper, an isolated-gate bipolar transistor (IGBT) is used as a case study for prognostic development. The proposed approach is developed by analyzing failure mechanisms and statistics of IGBT degradation data obtained from an accelerated aging experiment. The approach explores various probability distributions for modeling discrete degradation profiles of the IGBT component. This allows the stochastic degradation model to be efficiently simulated, in this particular example ∼1000 times more efficiently than Markov approaches. Index Terms-Isolated-gate bipolar transistor (IGBT), Monte-Carlo simulation (MCS), power electronics, prognostics, remaining useful life (RUL).
Abstract-Within the field of Integrated SystemHealth management, there is still a lack of technological approaches suitable for the creation of adequate prognostic model for large applications whereby a number of similar or even identical subsystems and components are used. Existing similarity among a number of different systems, which are comprised of similar components but with different topologies, can be employed to assign the prognostics of one system to other systems using an inference engine. In the process of developing prognostics, this approach will thereby save resources and time. This paper presents a radically novel approach for building prognostic models based on system similarity in cases where duality principle in electrical systems is utilized. In this regard, unified damage model is created based on standard Tee/Pi models, prognostics model based on transfer functions, and RUL estimator based on how energy relaxation time of system is changed due to degradation. An advantage is; the prognostic model can be generalized such that a new system could be developed on the basis and principles of the prognostic model of other systems. Simple electronic circuits, DC-to-DC converters, are to be used as an experiment to exemplify the potential success of the proposed technique validated with prognostics models from particle filter.
Health management and reliability are fundamental aspects of the design and development cycle of power electronic products. This paper presents the prognostic evaluation of a power electronic IGBT module. To achieve this aim, a simple state-based prognostic (SSBP) method has been introduced and applied on the data which was extracted from an aged power electronic IGBT and its remaining useful life was determined.
Intelligent fault diagnosis methods have replaced time consuming and unreliable human analysis, increasing anomaly detection efficiency. Deep learning models are clear cut techniques for this purpose. This paper’s fundamental purpose is to automatically detect leakage in tanks during production with more reliability than a manual inspection, a common practice in industries. This research proposes an inspection system to predict tank leakage using hydrophone sensor data and deep learning algorithms after production. In this paper, leak detection was investigated using an experimental setup consisting of a plastic tank immersed underwater. Three different techniques for this purpose were implemented and compared with each other, including fast Fourier transform (FFT), wavelet transforms, and time-domain features, all of which are followed with 1D convolution neural network (1D-CNN). Applying FFT and converting the signal to a 1D image followed by 1D-CNN showed better results than other methods. Experimental results demonstrate the effectiveness and the superiority of the proposed methodology for detecting real-time leakage inaccuracy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2025 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.