Abstract: According to the shortcoming of total projection to latent structure (T-PLS) algorithm which has a higher false alarm rate and a higher missing report rate when it is used to detect fault in the nonlinear process,a new algorithm which is a combination of kernel function and T-PLS (KT-PLS) is proposed in this paper. Firstly,the low dimension process data is projected to the high dimension feature space through kernel function,which can resolve the nonlinear problem.Secondly,the feature space is divided into four subspaces by the guiding of quality variable.These four subspaces include the quality directly correlated subspace,the quality orthogonal subspace,the quality uncorrelated subspace and the residual subspace. Finally,statistics D and Q are used to detect fault respectively.Applying this algorithm to the Tennessee Eastman process (TEP),the simulation results of various fault modes show that KT-PLS is better than T-PLS in monitoring strong nonlinear system.

Key words: safety, computer simulation, dynamic simulation, chemical process, fault detect, KT-PLS

摘要： 针对全潜结构投影法（T-PLS）在检测非线性过程故障时误报率和漏报率较高的缺点，提出了基于核函数的全潜结构投影法（KT-PLS）。该算法通过核函数将过程数据从低维输入空间非线性地映射到高维特征空间，实现非线性问题的线性化；然后在质量变量的引导下将特征空间分为与质量直接相关、与质量正交、与质量无关和残差四个子空间；最后分别构建D和Q统计量进行故障检测。将该算法应用到Tennessee Eastman process（TEP），多种故障模式下的仿真结果表明，KT-PLS比T-PLS更适合监控具有强非线性的生产过程。

关键词: 安全, 计算机模拟, 动态仿真, 化工过程, 故障检测, KT-PLS