[2405.04336] Temporal and Heterogeneous Graph Neural Network for Remaining Useful Life Prediction

View the PDF file from the paper entitled Time and In -homogeneous nerve network to predict the remaining life, by Zhihao Wen and 5 other authors

PDF HTML (experimental) view

a summary:RUL plays a decisive role in the fake management and health management of industrial systems that involve a variety of interconnected sensors. Looking at a continuous flow of sensory data for the time series of these systems, deep learning models have increased to their emergence in determining the complex and non -linear dependencies in these data. In addition to the temporal dependencies of individual sensors, spatial dependencies appear as important relationships between these sensors, which can be naturally designed by a graphic graph that describes the changing spatial relationships over time. However, most current studies have relied on capturing separate shots for this temporal graph, which is the rough grain approach that leads to the loss of time information. Moreover, given a variety of heterogeneous sensors, it becomes necessary that the inherent insecure insecurity of the rules in the graphs of the timeline graphs be used. To capture the nuances of time and spatial relationships and heterogeneous characteristics in an interconnected graph of sensors, we offer a new model called THGNN. Specifically, Thgnn collects historical data from the contract adjacent to capturing time dynamics and spatial relationships accurately inside the sensor data flow in an accurate way. Moreover, the model benefits from the linear modification of the feature (film) to address the diversity of the types of sensors, which greatly improves the model’s ability to learn not homogeneity in data sources. Finally, we have verified the correctness of our approach through comprehensive experiences. Our experimental results show significant progress on the N-CMAPSS data collection, with improvements up to 19.2 % and 31.6 % in terms of different evaluation measures in its modern way.