Electric valve fault detection method based on LSTM model

Abstract

The invention discloses an electric valve fault detection method based on an LSTM model, and the method comprises the following steps: detecting the feature data of a valve during operation in real time, and collecting the feature data into a fault diagnosis processor through an eight-channel high-precision VGA signal collection card; setting a dynamic time window, extracting time series data withequal length as a sample of a data set, and making a diagnosis data training set; establishing an LSTM fault detection network model, and training by adopting a back propagation algorithm according to the LSTM fault detection model; and detecting the fault of the electric valve in real time, enabling the fault type corresponding to the maximum value of the output vector element to serve as the fault detection result of the electric valve, displaying the fault detection result on a liquid crystal display screen, and if detecting that a certain electric valve breaks down, triggering a buzzing alarm and a fault signal lamp at the same time to remind monitoring personnel. The intelligent level of the electric valve is improved, a large amount of work of manual detection is omitted, the real-time performance is high, and the industrial requirement is met.

Description

technical field [0001] The invention relates to the technical field of electric valves, in particular to an LSTM model-based fault detection method for electric valves. Background technique [0002] In most industrial fields, valves are one of the essential components. Nowadays, electric valves are widely favored in chemical, petroleum, natural gas and other industries due to their advantages of large torque and the ability to be used in high temperature and high pressure media control. However, in a real industrial environment, a considerable part of the valves will have internal and external leakage or leakage failure due to long-term use, wear, corrosion or other reasons. If it is not discovered and dealt with in time, it will lead to media leakage, environmental pollution and even catastrophic accidents such as fire and explosion. The traditional mechanism of valve fault diagnosis is to inspect and inspect the maintenance workers at a certain period of time. This metho...

AI Technical Summary

Problems solved by technology

Recurrent Neural Network (RNN) introduces the concept of time series into the design of network structure. The interconnection structure between hidden layers reflects the mutual influence relationship between time series, but RNN has gradient disappearance, gradient explosion and long-term memory capacity. And other issues

Images