Welding quality real-time detection method and system based on LSTM and residual distribution

Abstract

The invention discloses a welding quality real-time detection method and system based on LSTM and residual distribution. The method comprises the steps of building a residual matrix for an extremely short time sequence window based on a pre-trained LSTM model, measuring the distribution similarity of the residual matrix in a single time window and an overall residual matrix, and achieving welding quality defect detection in a short time according the judgement about whether the residual matrix exceeds a threshold value or not. According to the method and the system, on the basis of high-frequency time sequence signal data collected by non-invasive equipment, the residual matrix of real data and predicted data is constructed by utilizing LSTM, the distribution similarity of the residual matrix is measured, and an abnormal time sequence threshold value is determined to identify welding quality defects, so that the technical problems that welding images, spectrums, sounds and other data are difficult to collect, defect data are difficult to label, and model robustness is insufficient are solved, and real-time detection of the welding quality defects can be achieved.

Description

technical field [0001] The invention relates to a welding quality real-time detection method and system based on LSTM (long-short-term memory artificial neural network) and residual distribution, belonging to the technical field of automatic welding. Background technique [0002] In recent years, with the rapid development of industries such as automobiles, aerospace, construction, and transportation, the requirements for the technology and quality of industrial equipment are getting higher and higher. Welding quality inspection technology has been widely used in many fields in recent years. Welding quality can be divided into direct welding quality and indirect welding quality. The main contents of welding joints of general welding products include mechanical properties, internal and external defects, and geometric dimensions of welded products. The so-called indirect welding quality refers to the relevant factors that can be detected by the welder's senses or special senso...

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: Aiming at the problems existing in the prior art, the present invention provides a welding quality real-time detection method and system based on LSTM and residual error distribution, based on the collected high-frequency time-series signal data, using LSTM to construct real data and predicted data The residual matrix is ​​used to measure the similarity of the distribution of the residual matrix, and determine the abnormal timing threshold to identify welding quality defects, so as to solve technical problems such as difficult data collection of welding images, spectra, and sounds, difficulty in labeling defect data, and insufficient robustness of the model

Images