Industrial control system design of optimized fopid controller with anomaly detection based on walrus optimization and Bi-LSTM

Industrial control systems (ICSs) play a role in many forms of social infrastructure, and they are essential for the accomplishment of control goals and security maintenance. Consequently, industrial control systems that have security vulnerabilities are susceptible to being attacked by external attackers. To overcome the security issues, the proposed optimized FOPID controller with anomaly detection for ICS is designed on the basis of walrus optimization and Bi-LSTM in the control system design. At first, data is preprocessed by utilizing the rough k means centroid based imputation technique to replace missing values in the data, and then followed by Z‑score normalization to normalize each value in a dataset for a certain range based on the standard deviation and mean. Following that the preprocessed data is fed into the Bi-LSTM classifier to predict the attack as normal or anomaly. The Walrus Optimization Algorithm (WaOA) finds the ideal value of gains to make the process efficient, which enhances the performance of the FOPID controller. The FOPID generates an ON/OFF pulse to the valve to control the water flows depending on the error signal. The performance metrics are examined and contrasted with existing methodologies in terms of accuracy, positive predictive value, true positive rate, kappa, F1-score, and error, resulting in values of 96.2%, 92.4%, 94.7%, 88.4%, 93.5%, and 3.79% for the attack prediction of the proposed model. Furthermore, the rising time, peak time, and peak overshoot of the system responses and also the bode plot are evaluated and provides the system stability output response. The results of this process demonstrates that, in comparison to other traditional integer order controllers, the proposed WaOA optimized FOPID controller showed superior set point and smooth controller response.