基于KPCA与PSO-WLSSVM的顶吹熔炼系统喷枪寿命预测研究

The research of lance service life prediction for the system of top-submerged smelting based on KPCA-PSOLSSVM

针对顶吹熔炼系统喷枪寿命预测原始生产数据含噪量大以及单一预测模型容易失效的问题,提出一种基于核主元分析法(KPCA)的PSO-WLSSVM组合预测方法。首先利用KPCA对原始生产数据进行去噪处理,通过贡献率对样本降维,提取样本中的非线性主元信息,然后用粒子群算法(PSO)优化WLSSVM的两大主要参数,从而避免了经验法的弊端。与PSO-WLSSVM和WLSSVM模型进行对比,实验结果表明,KPCA-PSO-WLSSVM模型对喷枪寿命预测的可信度和准确度较高,为下一步提出维修策略奠定了基础。

Aiming at the problems that there are lots of noises in the raw production data of the spray guns＇ life prediction of the top-blown melting systems and the single prediction models are easy to get failures, the PSO-WLSSVM combining prediction method based on the Kernel Principal Component Analysis （KPCA） is proposed. Firstly using KPCA to make raw production data denoising treatment, the samples are reduced in dimension by contribution rates, and the nonlinear principal component information of the samples is extracted. Then using the Particle Swarm Optimization （PSO） to optimize the two main parameters of WLS-SVM so that disadvan- tages of the empirical method are avoided. Comparing with the PSO-WLSSVM and the WLSSVM model, it＇s shown in the experimen- tal results that the KPCA-PSO-WLSSVM model is better on credibility and accuracy of the spray guns＇ life prediction, and it provides the next maintenance strategy＇s proposing with support.