Verification and Anomaly Detection for Event-Based Control of ManufacturingSystems.

Control Systems;Manufacturing Systems;Verification;Fault Detection;Discrete Event Systems;Electrical Engineering;Mechanical Engineering;Engineering;Electrical Engineering: Systems

Allen, Lindsay VictoriaMoyne, James R. ;

University of Michigan

关键词: Control Systems;    Manufacturing Systems;    Verification;    Fault Detection;    Discrete Event Systems;    Electrical Engineering;    Mechanical Engineering;    Engineering;    Electrical Engineering: Systems;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/78897/lzallen_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Many important systems can be described as discrete event systems, including a manufacturing cell and patient flow in a clinic. Faults often occur in these systems and addressing these faults is important to ensure proper functioning. There are two main ways to address faults. Faults can be prevented from ever occurring, or they can be detected at the time at which they occur. This work develops methods to address faults in event-based systems for which there is no formal, pre-existing model. A primary application is manufacturing systems, where reducing downtime is especially important and pre-existing formal models are not commonly available. There are three main contributions.The first contribution is formalizing input order robustness - inputs occurring in different orders and yielding the same final state and set of outputs - and creating a method for its verification for logic controllers and networks of controllers. Theory is developed for a class of networks of controllers to be verified modularly, reducing the computational complexity. Input order robustness guarantees determinism of the closed-loop system.The second contribution is an anomaly detection solution for event-based systems without a pre-existing formal model. This solution involves model generation, performance assessment, and anomaly detection itself. A new variation of Petri nets was created to model the systems in this solution that incorporates resources in a less restrictive way. The solution detects anomalies and provides information about when the anomaly was first observed to help with debugging.The third contribution is the identification and resolution of five inconsistencies found between typical academic assumptions and industry practice when applying the anomaly detection solution to an industrial system. Resolutions to the inconsistencies included working with industry collaborators to change logic, and developing new algorithms to incorporate into the anomaly detection solution. Through these resolutions, the anomaly detection solution was improved to make it easier to apply to industrial systems.These three contributions for handling faults will help reduce down-time in manufacturing systems, and hence increase productivity and decrease costs.