Method for predicting residual service life of equipment based on double attention network

Abstract

The invention discloses an equipment remaining service life prediction method based on a double attention network, and the method comprises the steps: 1, collecting and preprocessing sensor data, and constructing a data sample set through a sliding time window method; 2, building a double attention network, wherein the network structure comprises a space attention module, a bidirectional long and short term memory module, a time attention module and a full-connection network prediction module; 3, training a double attention network model, and optimizing model parameters; and 4, predicting the residual service life of the equipment by using the trained dual attention network model. According to the method, the multi-source sensor data can be adaptively fused, and the difference utilization of the characteristics at different time can be realized, so that the residual service life prediction effect is improved.

Description

technical field [0001] The invention belongs to the field of prediction of remaining service life, and in particular relates to a method for predicting remaining service life of equipment based on a double attention network. Background technique [0002] Prognostic and Health Management (PHM) technology is one of the core technologies of modern industrial development. Remaining Useful Life (RUL) prediction is an important part of PHM, which can provide basic decision-making information for enterprises to manage equipment health. The remaining service life prediction is based on the current or historical monitoring status data of the system to predict the time from the current moment to the failure of the system. Accurate remaining life prediction helps enterprises to take corresponding maintenance measures before the system fails, which is of great help to the guarantee system. Safety and reliability, reducing enterprise maintenance costs have great significance and applica...

AI Technical Summary

Problems solved by technology

However, most of these methods ignore the differences and relationships among multi-source sensors, lack an effective mechanism to automatically distinguish the importance of different sensors, and cannot make full use of the interaction information among multiple sensors.

On the other hand, the data monitored by sensors is essentially time-series data. Although the traditional long-short-term memory network can learn the time dependence in the monitored data, it still has certain limitations, and it is difficult to reflect the characteristics of different moments. Differences in service life prediction results

Images