CN-121979133-A - Manufacturing process sensing and quality control method and equipment based on multiple modes

CN121979133ACN 121979133 ACN121979133 ACN 121979133ACN-121979133-A

Abstract

The invention discloses a method and equipment for sensing and controlling quality in a manufacturing process based on multiple modes, wherein the method comprises the steps of collecting multiple mode sensing data in the manufacturing process, wherein the multiple mode sensing data comprises image data, sound or vibration data and thermal imaging data; the method comprises the steps of extracting process characteristics corresponding to each mode based on multi-mode sensing data, adaptively fusing the process characteristics of each mode based on an attention mechanism to generate fused characteristic vectors, inputting the fused characteristic vectors and current process parameters into a quality prediction model to obtain a prediction result of a preset quality index, determining an influence relation of the process parameters on the preset quality index based on the quality prediction model, and optimizing and adjusting the process parameters based on the prediction result, the influence relation and a preset control target to perform feedforward control on the manufacturing process. The online feedforward closed-loop control of the manufacturing process is realized through multi-mode sensing, feature fusion and model prediction, and the consistency of the product quality and the production efficiency are improved.

Inventors

Lei huan
ZHONG ZHENYU
WU LIANGSHENG
CHEN ZAILI
HUANG TIANLUN

Assignees

广东省科学院智能制造研究所

Dates

Publication Date: 20260505
Application Date: 20251230

Claims (10)

1. A multi-modal based manufacturing process sensing and quality control method, the method comprising: Collecting multi-modal awareness data during a manufacturing process, the multi-modal awareness data including image data, sound or vibration data, and thermal imaging data; extracting process features corresponding to each mode based on the multi-mode sensing data, and adaptively fusing the process features of each mode based on an attention mechanism to generate fusion feature vectors; Inputting the fusion feature vector and the current process parameter into a quality prediction model to obtain a prediction result of a preset quality index, and determining an influence relationship of the process parameter on the preset quality index based on the quality prediction model, wherein the quality prediction model is obtained based on multi-mode fusion feature and process parameter historical data training; And optimizing and adjusting the technological parameters based on the prediction result, the influence relation and a preset control target to perform feedforward control on the manufacturing process, wherein the preset control target is to enable the prediction quality to approach a preset threshold value.
2. The method of claim 1, wherein the collecting multi-modal awareness data in a manufacturing process comprises: Data acquisition based on visual sensors, acoustic or vibration sensors, and infrared thermal imaging sensors during the manufacturing process; Based on a hardware synchronization mechanism, data from different sensors is made to align in time.
3. The method of claim 1, wherein extracting process features corresponding to each modality based on the multi-modality awareness data comprises: Extracting visual features related to surface defects and geometric dimensions from the image data through a convolutional neural network; extracting frequency domain features related to the running state of the equipment from the sound or vibration data through time-frequency analysis; thermal characteristics associated with the temperature field distribution are extracted from the thermal imaging data by statistical analysis.
4. The method of claim 1, wherein the adaptively fusing the process features of each modality based on an attention mechanism to generate a fused feature vector comprises: Acquiring working condition context information of the current manufacturing process; Dynamic weights are respectively distributed to the process characteristics of each mode based on the working condition context information; And fusing the weighted process characteristics of each mode based on the distributed dynamic weights to generate the fused characteristic vector.
5. The method of claim 1, wherein the quality prediction model is a hybrid prediction model comprising a data-driven sub-model trained based on historical data for learning nonlinear mappings and a physical mechanism sub-model constructed based on physical principles of a manufacturing process for providing prediction constraints.
6. The method of claim 5, wherein optimizing the process parameters based on the prediction results, the influence relationships, and a preset control objective to feed forward control of a manufacturing process comprises: Constructing a rolling time domain optimization problem by taking the deviation between the minimized prediction quality and a preset threshold value and the smoothness of the control action as optimization targets; based on the influence relation, solving the rolling time domain optimization problem by using a model predictive control algorithm to obtain an optimal process parameter adjustment sequence within a preset duration; And issuing the instant set value in the optimal process parameter adjustment sequence to a manufacturing execution system.
7. The method of claim 1, wherein the method further comprises: obtaining an actual quality measurement result produced in the manufacturing process; comparing the actual quality measurement result with the prediction result to generate a prediction deviation; and carrying out parameter on-line fine adjustment and self-adaptive calibration on the quality prediction model based on the prediction deviation.
8. A multi-modal based manufacturing process sensing and quality control apparatus, the apparatus comprising: the acquisition module is used for acquiring multi-modal sensing data in the manufacturing process, wherein the multi-modal sensing data comprises image data, sound or vibration data and thermal imaging data; The extraction module is used for extracting the process characteristics corresponding to each mode based on the multi-mode sensing data, and carrying out self-adaptive fusion on the process characteristics of each mode based on an attention mechanism to generate fusion characteristic vectors; the prediction module is used for inputting the fusion feature vector and the current process parameter into a quality prediction model, obtaining a prediction result of a preset quality index, and determining an influence relationship of the process parameter on the preset quality index based on the quality prediction model, wherein the quality prediction model is obtained based on multi-mode fusion feature and process parameter historical data training; and the control module is used for optimally adjusting the process parameters based on the prediction result, the influence relation and a preset control target so as to perform feedforward control on the manufacturing process, wherein the preset control target is used for enabling the prediction quality to approach a preset threshold value.
9. A computer storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of the method according to any one of claims 1 to 7.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method according to any one of claims 1 to 7 when the program is executed.

Description

Manufacturing process sensing and quality control method and equipment based on multiple modes Technical Field The invention relates to the technical field of precision manufacturing, in particular to a method and equipment for sensing and controlling quality of a manufacturing process based on multiple modes. Background In the process sensing and quality control technology in the traditional precision manufacturing field, a technical mode of single sensing, off-line detection and open-loop regulation is currently adopted, and obvious limitations exist in the aspects of process sensing, quality control and system coordination. In the process sensing aspect, the prior art mainly relies on single type sensors such as X-ray, beta-ray or laser thickness measurement, and the like, and can only detect local quality parameters such as surface density, thickness and the like, and cannot comprehensively obtain multidimensional process information such as slurry state of a coating die head, substrate transmission vibration, drying temperature field distribution and the like. Due to the lack of synchronous acquisition and fusion of multi-mode data such as vision, acoustics, heat and the like, a 'perception blind area' exists on key characteristics such as surface texture defects, positive and negative alignment degree, abnormal voiceprint of equipment, pole piece temperature gradient and the like of a product, and comprehensive digital characterization of the state in the manufacturing process is difficult to realize. At the quality control level, quality judgment is severely dependent on off-line sampling detection. This approach results in a lag in process parameter adjustment, often times up to several hours, and is not compatible with high speed coating conditions above 60 m/min. Meanwhile, as the dimension of the sensing data is single and is disjointed with the control system, a real-time and accurate correlation model between the equipment running state, the process parameter fluctuation and the product quality deviation is difficult to establish, so that the process quality problems of poor surface density uniformity, large alignment error and the like are frequent, and the product quality consistency and the production yield are difficult to guarantee. At the system integration level, the existing sensing module is usually simply spliced with manufacturing equipment in a 'plug-in' mode, and all subsystems (such as sensing, analysis and control) operate independently. The data transmission and processing chain is long and has high delay, and the deep coordination and real-time closed loop from sensing, analysis and decision-making to execution can not be realized, and the system essentially belongs to an open loop or post-regulation system. In summary, the prior art has the inherent defects of single sensing dimension, lag in quality control and poor system cooperativity, and cannot meet the real-time sensing and online cooperative control requirements of high-speed and high-precision intelligent manufacturing under complex working conditions. Disclosure of Invention Based on the problems, the embodiment of the invention provides a multi-mode-based manufacturing process sensing and quality control method and equipment, which can realize online feedforward closed-loop control of the manufacturing process through multi-mode sensing, feature fusion and model prediction, and improve the consistency of the product quality and the production efficiency. The embodiment of the invention provides a multi-mode sensing and quality control method based on a manufacturing process, which comprises the steps of collecting multi-mode sensing data in the manufacturing process, wherein the multi-mode sensing data comprise image data, sound or vibration data and thermal imaging data, extracting process characteristics corresponding to all modes based on the multi-mode sensing data, adaptively fusing the process characteristics of all modes based on an attention mechanism to generate fusion characteristic vectors, inputting the fusion characteristic vectors and current process parameters into a quality prediction model to obtain a prediction result of a preset quality index, and determining an influence relation of the process parameters on the preset quality index based on the quality prediction model, wherein the quality prediction model is obtained by training based on the multi-mode fusion characteristic and the process parameter historical data, and optimally adjusting the process parameters based on the prediction result, the influence relation and a preset control target to perform feedforward control on the manufacturing process, wherein the preset control target is to enable the prediction quality to approach a preset threshold. In one possible embodiment, collecting multi-modal awareness data in a manufacturing process includes: Data acquisition based on visual sensors, acoustic or vibration sensors, and infrared therma