Improvement of optimization methods in indoor time-variant source parameters estimation combining unsteady adjoint equations and flow field information

Abstract

When hazardous gas leaks, accurate and prompt identification of leakage location and leakage rate is essential for ensuring indoor air quality and environmental safety. In the process of source parameters estimation, precise establishment of source-detector relationship is critical. Although computational fluid dynamics (CFD) simulation has high accuracy, it demands massive computing time. Prior studies mainly concentrate on estimating constant release sources, and few of them focus on time-variant sources peculiarly under indoor complex flow. Therefore, this paper further develops optimization methods based on fixed-detector networks by combining unsteady adjoint equations and flow field information. With only three fixed detectors, the effectiveness of proposed method is tested by estimating three time-variant source parameters under complex indoor flow. The results indicate that the application of unsteady adjoint equations not only saves a huge amount of computation but also preserves the high accuracy of CFD simulation, making it possible to estimate time-variant sources promptly and accurately under indoor complex flow. For the convenience of practical application, performances of five commonly used optimization methods and three common cost functions are compared carefully. The results show that the cost function gradient determines their different estimation performances. In addition, the reason for the formation of local optimal areas is analyzed and the flow field information is utilized to avoid the problem that optimization methods may converge to local optimal values, which improves the accuracy of estimated results significantly.