关键词：传感器；故障检测；核偏最小二乘；KPLS
摘 要：Time-delayed variables is of potentially critical impact on product quality within the processes of Metallurgical industry,Aviation industry and Chemical industry.It is very important to detect the sensor fault and estimate the values of time-delayed variables immediately as soon as faults occurred.Any fault of the time-delayed variable implies that a detrimental effect on the process will go unnoticed until the next sample become available.During this time,of course,poor quality product continues to be manufactured,a situation which is clearly unacceptable from a cost-effectiveness point of view.For nonlinear process with time-delayed variables,kernel partial least squares is introduced to monitor and estimate the fault of the Tennessee Eastman benchmark.Simulation results demonstrate the effectiveness of the proposed algorithm.