基于优化支持向量机的带钢延伸量软测量研究

带钢退火过程中存在多变量非线性主导因素和数据噪声,难以用数学模型精确描述退火炉内带钢的延伸量.针对这一问题,提出基于核主元分析(KPCA)与免疫粒子群(ICPSO)优化最小二乘支持向量机(LSSVM)的炉内带钢延伸量软测量方法.采用ICPSO算法避免了粒子群算法易陷入局部最优的缺陷,利用ICPSO对LSSVM进行参数寻优,通过KPCA去除样本噪声,提取输入数据样本中的非线性主元信息,建立ICPSO-LSSVM软测量模型.此方法用于退火炉内带钢延伸量预测,通过现场生产数据仿真实验进行非线性函数估计;对比其他几种现有算法,实验结果表明本文方法具有较高的预测精度.

A Soft Sensor Based on Optimized LSSVM for Elongation Prediction of Strip Steel

The strip elongation is difficult to predict accurately with mathematical model， which related with multi-variable nonlinear factors and data noise in the annealing process. Thus， the optimal soft-sensing method was proposed based on kernel principal component analysis (KPCA) and optimized least squares support vector machine (LSSVM) by immune clone particle swarm optimization (ICPSO). ICPSO can avoid the particles sinking into premature convergence and running into local optimization in the iterative process which was generated by particle swarm optimization (PSO) algorithm， and can also optimize the parameters of LSSVM. Then， KPCA was used to denoise the input data set and capture the high-dimensional nonlinear principal components in input data space， and the principal components were input into the ICPSO-LSSVM model to establish the soft-sensing prediction model. The proposed method was successfully applied to the strip elongation prediction in annealing furnace. The simulation results show that the KPCA and ICPSO-LSSVM model have higher prediction accuracy， compared with other algorithms.