Quality prediction system and quality prediction method

Abstract

Provided are a quality prediction system and a quality prediction method. The quality prediction system comprises sensors including a local sensor installed on a machine and an environment sensor installed in the surrounding environment of the machine, the local sensor collects data of the machine itself, and the environment sensor collects data of the surrounding environment; an intelligent computing unit which is in communication connection with the sensor and the machine, controls the sensor to acquire data, receives the data acquired by the sensor and preprocesses the data; a storage and processing device which is in communication connection with an intelligent processing unit and is used for receiving preprocessed data from the intelligent computing unit, storing data and processing preprocessed data, wherein in the training mode, wherein the storage and processing device stores the preprocessed data into a training data set; a training data set is input into a quality predictionmodel for training to obtain prediction model parameters, and in the prediction mode, the preprocessed data is input into the quality prediction model by a storage and processing device to give a quality probability.

Description

technical field [0001] The invention relates to product quality detection in the manufacturing field. More specifically, the present invention relates to a real-time quality prediction method based on deep learning. Background technique [0002] Traditionally, during the manufacturing process, parts are evaluated as "pass" or "fail" after a "make-inspect-judgment" process, and this is still the workflow for quality assurance in manufacturing today. The process dictates that parts must be "fully finished" at every step of machining before they can be inspected. In order to reduce waiting time as much as possible and improve production efficiency, the manufacturing process of parts is carried out one by one. This means that it is not clear whether a part that has just been manufactured is "good". This leads to the following two situations. In the first case, the line needs to be stopped and a quality check of the manufactured parts performed in order to make a judgment. A...

AI Technical Summary

Problems solved by technology

[0003] In addition, product traceability is often disconnected from the actual production process, that is, the quality of the part is not linked to the actual machine that made the part

This correlation is difficult to achieve in practice because of the sheer number of variables and elements that need to be tracked during the manufacturing process

Moreover, until now, even if traceability is achieved, it still needs to coordinate the time relationship among a large amount of monitoring data, machine data, part serial numbers and part quality, which is usually very difficult and laborious

Images