Mechanical equipment is composed of several parts, the interaction between parts exists throughout whole life cycle leads to widespread phenomenon fault coupling and diagnosis independent faults cannot meet requirements health management mechanical under actual working conditions. In this paper, topological structure graph neural network used describe correlation multiple types are regarded as nodes respectively. The features fused extracted by convolution achieve classification faults....

Long short-term memory (LSTM) is widely applied in both academic and industrial fields. However, there no reliable criterion on selecting hyperparameters of LSTM. Currently, although some used classic methods such as random search grid have obtained success to extent, the problems local optimum convergence still exist. In this research, we propose use grey wolf optimizer (GWO) for Through method, superiority metaheuristic global optimization strength LSTM predicting are combined. model,...

In this article, the random vibration test and finite-element simulation analysis of SAC305 solder joints in BGA packaged electronic chips were carried out. A failure model for was established based on life data. And vulnerable parts analyzed. First, conducted. Second, a lognormal distribution by means data analyzed fracture mechanics theory. Finally, chip discussed. Failure characteristics at different positions compared mode joint discussed using result observation with scanning electron...

To tackle the problems of over-reliance on traditional experience, poor troubleshooting robustness, and slow response by maintenance personnel to changes in faults current aircraft health management field, this paper proposes use a knowledge graph. The graph represents new way. aim is improve correlation between fault data representing experience. source for study consists flight control system manual typical cases specific type. A construction approach proposed construct management....

The integrated-servo-actuator (ISA) is one of the most critical subsystems aircraft flight-control system, which controls speed, direction, displacement, and force load for an aircraft. degradation state ISA directly influences safety Hence, it valuable to predict ensure its operating reliability. However, as a kind complex mechanisms stress cannot be fully understood. Moreover, nonlinear process brings more difficulties establish accurate life prognosis model. To address these challenging...

High-cost equipment is often reused after maintenance, and whether the information before maintenance can be used for Remaining Useful Life (RUL) prediction directly determined by consistency of degradation pattern maintenance. Aiming at this problem, an RUL method based on test a Wiener process proposed. Firstly, parameters estimated Maximum Likelihood Estimation (MLE) are proved to biased, modified unbiased estimation proposed verified derivation simulations. Then, h statistic constructed...

Airborne fuel pump is an important part of aircraft system, for the lack fault samples, low diagnostic efficiency and high maintenance cost, in order to achieve more accurate reliable diagnosis airborne pump, experimental platform transfusion system developed a method based on empirical mode decomposition (EMD) probabilistic neural networks (PNN) proposed. Firstly, vibration signals pressure normal state six types representative faults are collected platform. Then EMD applied decompose...

Aiming at the problem of BGA solder joint failure under random vibration, vibration simulation and fatigue life prediction were studied, results verified by experiments. Firstly, three-dimensional finite element model was established ANSYS software. The characteristics obtained analyzing first five modes. Then analysis performed. It found that had 3 extremely vulnerable areas. After that, estimated to be 601mins using empirical formula three-band technique. Finally, a test conducted. show...

The new-generation avionics system is an electronically intensive device. Its complex structure, multiple failure modes and mechanism make it difficult to carry out effective fault prediction health management based on traditional methods. A prognostic (PHM) massive data mining of presented in this paper. Firstly, the characteristics new generation are analyzed from aspects service environment, source storage mode. Secondly, key technologies mass discussed, including preprocessing,...

The chip is a core functional component. Its reliability plays vital role in electronic equipment normal operation. As the typical cause for malfunction, solder joint degradation selected to study reliability. models of different failure modes are established through data-driven and physical model, model constructed based on mutually competing failures multiple joints. First, test finite element modeling(FEM) carried out under coupled environment stress. processes studied analysis data,...

A question recognition method based on BERT-BiLSTM-ATT-CRF is proposed to solve the problem of entity difficulties faced by Question answering in field aircraft PHM. First, BERT model used for preprocessing, and then information further processed BiLSTM-ATT-CRF. Then, Bayesian classifier obtain classification. Finally, Knowledge graph realize intelligent After experimental comparison, has better ability PHM, with accuracy, recall, F1 values 72.236%, 73.563%, 72.8995% PHM corpus,...

In response to the current problems in maintenance process of aircraft power systems, such as relying on traditional experience, poor troubleshooting robustness, and slow personnel fault changes. this paper, Knowledge graph is introduced into field aviation supply health management, experience presented form improve correlation ability between data. This paper proposes a method build for system management based data source certain manual typical cases: first, set according semi structural...

Parameter-Efficient Fine-Tuning is wildly used to transfer models between different domains. However, for some high-reliability-equipment, the degradation at a slow rate and continually fluctuates, making it difficult extract features effectively. Moreover, collecting an integrated source domain high-reliability-equipment tough due little sample of related dataset. Aiming problem time-series prediction model, this research proposed LSTM-fine-tune where parameters model are explicitly trained...

This paper develops a multistate degradation structure of the solder joints which can be used under various vibration conditions based on nonhomogeneous continuous-time hidden semi-Markov process. The parameters were estimated to illustrate stochastic relationship between process and monitoring indicator by using unsupervised learning methods. Random tests with different levels power spectral density fixed forms conducted real time electrical resistance examine suitability model. It was...

The degradation behavior of the solder joint under constant vibration loading is discussed, voltage value at both sides taken as characteristic. Considering heteroskedasticity characteristic signal loading, general time series model cannot describe heteroskedasticity, a Generalized Autoregressive Conditional Heteroskedastic (GARCH) proposed. Firstly, upper and lower quantiles are extracted. failure determined by using quantile process analyzed quantile. Secondly, Auto-Regressive(AR) built...

Due to the influence of working environment and stress, failure degradation process mechanical equipment is characterized by non-stationarity uncertainty. The pattern a large number obeys from steady accelerated degradation. rotating machinery wear results in gradual early stage while rough contact surface spall leads stage. However, change-point patterns are uncertain which ambiguity transition uncertainty remaining life prediction model. Aiming at problems model, SKF-EN method proposed...

Conventional methods for fault diagnosis typically require a substantial amount of training data. However, equipment with high reliability, it is arduous to form large‐scale well‐annotated dataset due the expense data acquisition and costly annotation. Besides, generated have large number redundant features which degraded performance models. To overcome this, we proposed feature transfer scenario that transfers knowledge from similar fields enhance accuracy small sample. reduces information,...