CN-121999333-A - Stainless steel tube machining control method and device based on defect detection

CN121999333ACN 121999333 ACN121999333 ACN 121999333ACN-121999333-A

Abstract

The invention discloses a stainless steel tube processing control method and device based on defect detection, and relates to the technical field of intelligent processing, wherein the method comprises the steps of collecting multi-mode data in the stainless steel tube processing process in real time and preprocessing; based on the preprocessed multi-mode data, the surface defect characteristics, the internal structure abnormal characteristics and the thermal stress change characteristics of the stainless steel tube are extracted, defect description vectors are generated through multi-mode fusion and attention mechanism fusion, DNN is used for identifying defect types, influence degree scoring of machining rates on occurrence probability of different defect types is predicted, and an optimal machining rate adjusting instruction is generated and executed through DQN. The invention optimizes the production parameters by dynamically adjusting the processing rate, reduces the defect generation and improves the production efficiency and the product quality.

Inventors

FANG LEMING
OuYang Linzhi
GUAN JIANJUN

Assignees

湖南鑫立为科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251231

Claims (10)

1. A stainless steel tube processing control method based on defect detection is characterized by comprising the following steps of, Acquiring multi-mode data in the stainless steel tube processing process in real time, and preprocessing; Based on the preprocessed multi-mode data, extracting surface defect characteristics, internal structure abnormal characteristics and thermal stress change characteristics of the stainless steel tube, and generating defect description vectors through multi-mode fusion and attention mechanism fusion; Identifying defect types by using DNN, and predicting the influence degree scoring of the processing speed on the occurrence probability of different defect types; And generating and executing the optimal machining rate adjustment instruction through the DQN.
2. The method for controlling the processing of the stainless steel tube based on the defect detection as set forth in claim 1, wherein the multi-modal data comprises image data, a radiogram, temperature distribution data, and ultrasonic signals during the processing of the stainless steel tube.
3. The method for controlling the processing of the stainless steel tube based on the defect detection according to claim 2, wherein the pretreatment comprises, Adopting DTW to align the multi-mode data with different time sequences to the same time reference; Removing random noise and interference signals in the sensor multi-modal data by using wavelet transformation; The dimension difference in the multi-modal data is eliminated using Min-Max Normalization.
4. The method for controlling the processing of the stainless steel tube based on the defect detection of claim 3, wherein the method for controlling the processing of the stainless steel tube based on the preprocessed multi-mode data extracts the surface defect characteristics, the internal structure abnormality characteristics and the thermal stress change characteristics of the stainless steel tube, specifically comprises the following steps of, The convolution neural network is used, local texture features in the image data are extracted through the convolution layer, the pooling layer reduces noise, and finally surface defect features of the stainless steel tube are extracted; extracting internal defect characteristics of the stainless steel tube by using a transducer based on a ray perspective view; Carrying out multi-layer feature extraction on the ray perspective view through a Residual Block, preventing gradient disappearance through Skip Connection, and extracting internal structural abnormal features of the stainless steel tube; and Autoencoder, compressing the temperature distribution data to a low latitude space through an encoder, reconstructing the temperature distribution data through a decoder, and extracting the thermal stress change characteristics.
5. The method for controlling the processing of the stainless steel tube based on the defect detection of claim 4, wherein the defect description vector is generated by multimode fusion and attention mechanism fusion, specifically comprising the following steps of, Performing dimension alignment on the surface defect characteristics, the internal structure abnormal characteristics and the thermal stress change characteristics through normalization treatment; Inputting all the aligned multi-modal features into an MTN, and constructing an interaction diagram among the multi-modal features by adopting a method of combining a graph neural network and an attention mechanism; Based on an interaction diagram among the multi-modal features, adopting a SimCLR-based contrast learning method to analyze the correlation among different features, dynamically adjusting the feature representation based on the correlation, and generating enhanced multi-modal features; taking each enhanced modal feature as a node, and taking the correlation among the enhanced modal features as an edge; According to the nodes and the edges, constructing characteristic interaction relations among modes through GAT; Based on interaction relations among modes, dynamically adjusting importance weights of the characteristics of the modes by adopting a self-attention mechanism, identifying global attention among the modes through a transducer structure according to the importance weights among the modes, and constructing a global interaction matrix M; based on the global interaction matrix M, nonlinear mapping is carried out through MLP and ReLU, and multi-mode features are fused to generate a defect description vector; The multi-modal features include surface defect features, internal structural anomalies features, and thermal stress variation features.
6. The method for controlling the processing of the stainless steel tube based on the defect detection according to claim 5, wherein the step of using DNN to identify the defect type and predicting the degree of influence score of the processing rate on the occurrence probability of different defect types comprises the following steps, The input layer of DNN receives the defect description vector; The hidden layer adopts nonlinear activation to perform feature analysis, and part of neurons are randomly discarded through Dropout, so that the generalization capability of DNN is improved; Converting the processing rate into a processing rate feature vector through a full connection layer of DNN, and fusing the processing rate with the defect description vector; The output layer calculates probability distribution of each defect type by adopting Softmax, and predicts the influence degree of the processing rate on occurrence probability of different defect types by combining the processing rate characteristic vector.
7. The method for controlling machining of a stainless steel pipe based on defect detection according to claim 6, wherein the generating and executing of the optimal machining rate adjustment command by DQN is performed by the following expression: Defining a machining working condition state and a machining rate adjusting action based on the influence degree scores of the machining rates on the occurrence probabilities of different defect types; defining a reward function Q based on the occurrence probability of defects in the historical data and the production efficiency; Calculating a score of the current action based on long-term return after performing the machining rate adjustment action in the current machining working condition state by using a reward function Q based on the DQN; Selecting the current working condition state The following pairs of actions based on long-term rewards The action with the largest score is used as the optimal processing rate adjusting instruction and is executed.
8. The stainless steel tube machining control device based on defect detection is based on the stainless steel tube machining control method based on defect detection according to any one of claims 1-7, and is characterized by comprising a data acquisition module, a data fusion module, a defect probability scoring module and an adjustment instruction generation module; The data acquisition module is used for acquiring multi-mode data in the stainless steel tube processing process in real time and preprocessing the multi-mode data; The data fusion module is used for extracting the surface defect characteristics, the internal structure abnormal characteristics and the thermal stress change characteristics of the stainless steel pipe based on the preprocessed multi-mode data, and generating a defect description vector through multi-mode fusion and attention mechanism fusion; The defect probability scoring module is used for identifying defect types by using DNN and predicting the influence degree scoring of the processing rate on the occurrence probability of different defect types; and the adjustment instruction generation module is used for generating and executing an optimal machining rate adjustment instruction through the DQN.
9. A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is characterized in that the processor realizes the steps of the stainless steel tube processing control method based on defect detection according to any one of claims 1-7 when executing the computer program.
10. A computer-readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the method for controlling the processing of a stainless steel pipe based on defect detection as set forth in any one of claims 1 to 7.

Description

Stainless steel tube machining control method and device based on defect detection Technical Field The invention relates to the technical field of intelligent machining, in particular to a stainless steel tube machining control method and device based on defect detection. Background With the wide application of stainless steel pipes in the fields of petrochemical industry, aerospace, construction and the like, the quality control of the stainless steel pipes is becoming increasingly important. In the processing process of stainless steel pipes, surface defects, internal defects and structural anomalies are main factors affecting the quality of products. Traditionally, detection of these defects has relied primarily on manual visual inspection and some simple physical methods such as ultrasonic inspection or X-ray imaging. However, as the level of industrial automation increases, the quality requirements for stainless steel pipes become more stringent, which requires more efficient and accurate inspection techniques to ensure high quality standards for the products. In recent years, by introducing a machine learning algorithm and a multi-mode data processing technology, the accuracy and efficiency of defect detection are remarkably improved. For example, convolutional Neural Networks (CNNs) are used for image analysis to identify surface defects, or Deep Neural Networks (DNNs) are used to classify internal defects. Nevertheless, the prior art still has shortcomings in terms of real-time, accuracy and intelligence. On the one hand, although the existing defect detection technology can identify specific types of defects, when processing data from different sources, an effective fusion mechanism is often lacking, so that the information utilization rate is low. On the other hand, in terms of predicting the influence of the machining rate on the occurrence probability of defects, the existing method is generally based on a static model and cannot be dynamically adjusted to adapt to different production conditions. This not only limits the improvement of production efficiency, but also may increase the probability of occurrence of defects due to improper process parameter settings. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a stainless steel pipe processing control method based on defect detection, which solves the problems of low multisource data fusion efficiency and limited production efficiency and quality caused by static adjustment of processing parameters. In order to solve the technical problems, the invention provides the following technical scheme: The invention provides a stainless steel tube machining control method based on defect detection, which comprises the steps of collecting multi-mode data in the stainless steel tube machining process in real time, preprocessing, extracting surface defect characteristics, internal structure abnormal characteristics and thermal stress change characteristics of the stainless steel tube based on the preprocessed multi-mode data, generating defect description vectors through multi-mode fusion and attention mechanism fusion, identifying defect types by using DNN, predicting influence degree scoring of machining rates on occurrence probability of different defect types, and generating and executing an optimal machining rate adjustment instruction through DQN. As a preferable scheme of the stainless steel tube processing control method based on defect detection, the multi-mode data comprise image data, ray perspective view data, temperature distribution data and ultrasonic signals in the processing process of the stainless steel tube. As a preferable scheme of the stainless steel tube processing control method based on defect detection, the pretreatment comprises, Adopting DTW to align the multi-mode data with different time sequences to the same time reference; Removing random noise and interference signals in the sensor multi-modal data by using wavelet transformation; The dimension difference in the multi-modal data is eliminated using Min-Max Normalization. The method for controlling the processing of the stainless steel pipe based on the defect detection is used for extracting the surface defect characteristics, the internal structure abnormal characteristics and the thermal stress change characteristics of the stainless steel pipe based on the preprocessed multi-mode data, and comprises the following specific steps of, The convolution neural network is used, local texture features in the image data are extracted through the convolution layer, the pooling layer reduces noise, and finally surface defect features of the stainless steel tube are extracted; extracting internal defect characteristics of the stainless steel tube by using a transducer based on a ray perspective view; Carrying out multi-layer feature extraction on the ray perspective vi