Optimization and determination of the parameters for a PID based ventilation system for smoke control in tunnel fires: Comparative study between a genetic algorithm and an analytical trial-and-error method

Abstract

A PID-based longitudinal ventilation system for smoke control in tunnel fires is numerically illustrated to yield excellent performance. However, the three parameters in the PID algorithm significantly affect the performance. Two methods are proposed to optimize the parameters: (1) an analytical trial-and-error method (ATEM), based on an order-of-magnitude analysis; (2) a heuristic method: system identification and genetic algorithm (SI-GA). An SI-GA without a differential controller was also used for the comparative study, labeled as SI-GA*. The results show that all methods can obtain near-optimum parameters under their optimization goals. The ATEM is more time consuming, but it is more straightforward to manipulate the performance metrics directly than with the SI-GA method. The optimized parameters are tested with different fire sizes and different tunnel dimensions. The results show that the system with the optimized parameters can control the smoke well for large enough fire. In general, the system performs better for the stabilization time of the smoke front when using the parameters obtained from ATEM, but it achieves shorter maximum back-layering distances with the parameters obtained from SI-GA and SI-GA*. The latter two also give stronger fluctuations, but lower average values, of the ventilation velocity. Nevertheless, SI-GA* has more application potential due to the absence of a differential controller.