气化炉耐火砖健康状态诊断和剩余寿命预测
Diagnosis of health state and prediction of residual life of gasifier refractory bricks
-
摘要: 作为气化炉的关键部件,耐火砖的有效厚度及剩余寿命决定着气化炉的性能和运行安全。因此需要为耐火砖制定合理的维护计划,而视情维修(CBM)能降低维修风险和成本,保证气化炉的可靠性,为此聚焦于耐火砖健康状态的准确诊断和剩余寿命的精准预测。构建了气化炉耐火砖健康状态诊断和剩余寿命预测模型,并对某公司气化炉燃烧室实际工况下的耐火砖进行了相关研究。研究过程如下:①用K-均值算法确定模型初始参数,用改进的Baum-Welch算法训练耐火砖各状态对应的模型,用改进的Viterbi算法计算测试序列和各状态模型的对数似然概率并比较大小,其中最大的状态模型对应的状态即为测试序列的状态;②用耐火砖全寿命周期数据训练全寿命周期模型,根据重估后的模型参数预测各状态的剩余寿命;③根据健康状态诊断结果预测耐火砖当前剩余寿命。并将基于HSMM得到的结果与基于HMM得到的结果以及实际值进行了对比分析。结果表明,基于HSMM对耐火砖进行健康状态诊断的正确率达到了95%,大于基于HMM的正确率92.5%;基于HSMM得到的耐火砖各状态剩余寿命预测值全部落入实际剩余寿命区间内,平均正确率达到了81.36%,高于基于HMM的平均正确率71.22%。说明所构建的耐火砖健康状态诊断和剩余寿命预测模型符合气化炉耐火砖的实际退化规律,取得了良好的效果。Abstract: As a vital part of a gasifier,the residual thickness and residual life of refractory bricks determine the performance and operational safety of the gasifier. Therefore,it is necessary to formulate a reasonable maintenance plan for the refractory bricks. Condition-based Maintenance (CBM) can reduce maintenance risks and costs,and ensure the reliability of the gasifier. For this purpose,the focus was on the accurate diagnosis of health state and the accurate prediction of the residual life of refractory bricks in this study. A health state diagnosis model and a residual life prediction model for gasifier refractory bricks were constructed,and a related study was conducted on the refractory bricks lining the combustion chamber of a company's gasifier under actual working conditions. The research process is as follows:① The initial parameters of the model were determined by the K-means algorithm. The model corresponding to the each state of the refractory bricks was trained by the improved Baum-Welch algorithm. The loglikelihood probabilities of the test sequence and each state model were calculated and compared in size by the improved Viterbi algorithm,where the state corresponding to the state model that got the largest log-likelihood probability was the state of the test sequence. ② The whole life cycle model was trained with the refractory brick whole life data and the residual life of each state was predicted based on the revalued model parameters. ③ The current residual life of the refractory brick was predicted based on the health state diagnosis results. The results obtained based on the HSMM were also compared and analyzed with those obtained based on the HMM as well as the actual values. The results show that the HSMM-based health state diagnosis of refractory bricks achieves a correct rate of 95%,which is greater than the HMM-based rate of 92.5%. The residual life prediction for each state of refractory bricks obtained from the HSMM all fall within the actual residual life interval,with an average correct rate of 81. 36%,which is higher than the HMM-based average correct rate of 71.22%. It indicates that the constructed refractory brick health state diagnosis and residual life prediction models are in line with the actual degradation pattern of gasifier refractory bricks and have achieved good results.