大型LNG工厂气动阀故障率分析预测

Failure Rate Analysis and Prediction for Pneumatic Valve in Large LNG Plant

为探究LNG工厂控制阀故障率发展变化规律,预测关键零部件维修损坏频次,指引日常维护检修与备品备件的采购库存,收集2014-2019年某工程控制阀故障情况数据,丰富问题分析宽度,运用灰色系统均值GM(1,1)模型进行相关问题表征。研究表明:1)GM(1,1)模型是基于累加生成和最小二乘法的指数拟合模型,具有“贫信息、小样本”的通用性优势,能在时间序列数据有限的前提下得到较高精度的预测结果;2)本次建模得出的2019年故障情况预测较为可靠,其中定位器维修频次平均模拟相对误差仅3.67%,但随着未来的发展,该模型的预测意义就越弱;3)下步将运用DGM模型或SDGM模型全面提升原始均值GM(1,1)模型的精确性;4)控制阀控制回路简单,但仅凭停工期间的阀门开度比对不能及时和全面地发现问题,且该研究结果时效性有限,需要制定严谨而简便的阀门检测方案,并使相关数据适应于GM(1,1)建模。本研究通过简约数学算法,极大预测后续可能发生的异常工况,在成本可靠的前提下辅助生产。

This study aims to explore the development,and changes over time,of control valve failure rate in LNG plant,predict the frequency of repair and damage to key components,and guide the routine maintenance,purchase and inventory of spare parts.Data of control valve failure from a development project between 2014 to 2019 were collected to widen the issue analysis,and the Gray system mean GM(1,1)model was used to characterize the relevant problems.The study shows that:1)GM(1,1)model is an exponential fitting model based on accumulation generation and least square method.It is versatile and can be used with“poor information and small sample”situation to get more accurate prediction results under the premise of limited time series data;2)The failure prediction in 2019 obtained by this modeling is relatively reliable.The relative error of simulated positioner average maintenance frequency is only 3.67%.However,with longer time frame,the model prediction becomes less meaningful;3)The next step is to use DGM model or SDGM model to comprehensively improve the accuracy of the original mean GM(1,1)model;4)Although the control loop driving the control valve is simple;comparing valve opening ratios during shutdowns alone will not provide timely detection of overall control valve issues.In addition,this method has limited prediction time frame,so a rigorous and yet simple valve testing program needs to be developed,and relevant data could be adapted to GM(1,1)modeling.This study makes use of simple mathematical algorithm to predict,to a large extent,possible abnormal production scenarios in future operation,to enhance operation uptime at a reliable cost.