Quality fault locating method based on federated data driven production process

Abstract

The invention provides a quality fault locating method based on a federated data driven production process. The quality fault locating method includes the steps: extracting a quality cause and effect topological graph model; establishing a multimoding monitoring model driven by the federated data; based on the rate of contribution and the process knowledge, establishing a performance evaluation index for quality fault diagnosis; and according to the multimoding monitoring model, identifying a quality fault propagation path, and according to the performance evaluation index for quality fault diagnosis, locating a quality fault. Based on process monitoring of topological graph feature extraction and multivariate statistical data driving and machining learning, the quality fault locating method provides federated data driven fault diagnosis being suitable for quality monitoring, provides a new approach for quality fault diagnosis during a production process based on data and knowledge, makes up for the deficiency of traditional statistical process monitoring which is difficult to solve the problem of quality fault propagation path identification and the fault location problem, and can realize accurate and effective quality fault locating and diagnosis by means of quantification-qualitative-quantification federated data driving based on data and knowledge.

Description

technical field [0001] The invention belongs to the technical field of production process control and monitoring, and in particular relates to a method for locating quality faults based on a joint data-driven production process. Background technique [0002] Batch production is a process of division of labor and streamlining of modern production processes, and is widely used in machinery, hardware, plastics, auto parts and other industries. In recent years, in order to meet the market demand for multi-variety, multi-standard, high-quality functional products, the batch industrial process is developing in the direction of high efficiency, large scale and integration, and with the expansion of production scale and complexity, the production The requirements for process safety and reliability are also increasing. Modern complex intermittent processes often have many variables and control loops that are interrelated. Failure of one node will directly affect product quality and ...

AI Technical Summary

Problems solved by technology

These process variables and control loops influence and correlate with each other, and it is difficult to accurately and timely determine the cause of relevant faults when product quality (especially plate shape, tissue performance, etc.) fluctuates, causing some companies to often stop production for maintenance due to product quality and user returns (Often purposeless full-line maintenance)

[0004] The root cause of the above problems is that the complex batch process of hot strip rolling has inherent dynamic nonlinearity, strong coupling between variables and loops, multi-mode characteristics caused by multiple batches and multiple working conditions, and the time-dependent nature of the process. Due to the characteristics of variable characteristics and uncertainty caused by random noise, the causes of quality failures are diverse, the evolution process of failures is complex, the specific location and direction of failure are uncertain, the scope of failures is wide, and there are overlaps between failures and causes. Traditional process monitoring The method is too rough in the description of the process, and cannot fully exploit the prior knowledge of the process to monitor the quality failure. monitoring and judgment

Images