Research of Chemical Process Monitoring for Process Variables Characterized through a Time Series

Author:Ma Zuo

Supervisor:zhang lai bin hu zuo qiu

Database:Doctor

Degree Year:2018

Download:20

Pages:131

Size:10931K

Keyword:

In chemical industrial processes,process variable monitoring is a critical step to ganruantee the product quality and process safety.Statistical process control(SPC)is a widely used tool of real-time process control using mathematical and statistical methods.There has been much research work done on SPC by many academic groups and the SPC tools have also been widely applied to industrial processes to keep the process in control and facilitate the field operation.As process variable measurements are usually polluted by noise and outliers,a low signal-noise ratio in data would make the process control chart have a high false alarm rate and a low sensitivity to abnormal situation.Process variables are usually measured with a high sampling frequency,leading to a high autocorrelation level in data and a violation of the assumption of identical and independent distribution(lid)for traditional SPC tools.That would cause a high false alarm rate and a serious alarm delay for the SPC charts.To address the issue of process data filtering and serially relation,research works are done as follows:(1)A robust online filtering method(OLREMD)is proposed to implement online process data rectification with Empirical Mode Decomposition(EMD)as the basic algorithm.Tests with synthetic data show that OLREMD performs robustly with a lower sensitivity to parameters and improved performance on elimination of both noise and outliers.With a changing parameter setting,comparing with other online data filter,OLREMD has a narrow range of denoising error,which is reduced by 94%at most.When applied to an industrial de-ethanizing column,OLREMD is shown to enhance the process monitoring performance.The false alarm rate of control chart,which is constructed based on the filtered data,is reduced by 6%.(2)A new trend CUSUM chart is introduced as a way to detect the deviation of batch process variable trends,which may be caused by abnormal operations.The potential of these novel univariate control charts is demonstrated using the batch manufacture of polypropylene(PP).Results show that the trend CUSUM chart is capable of capturing trend abnormalities and alarms can be triggered in the trend CUSUM chart,about 94 samples ealier than the measured process variables exceed the control limits of their Shewhart individuals charts,allowing for the corrections to be made at an early stage of an abnormal situation(3)The effect of autocorrelation on Shewhart charts,including both the univariate and multivariate versions,is analyzed through simulations.After that,a new monitoring strategy for both univariate and multivariate Shewhart control charts is proposed aimed at reducing the effect of autocorrelation on chart performance.The advantages of the proposed approach are noted as being a model-free approach and having a performance consistent with that of the benchmark modified Shewhart chart.The improved chart has a false alarm rate at an expected level,which is 70%lower than that of the conventional Shewhart chart and detect the abnormities 20 samples earlier(4)The modified Shewhart individuals chart,an extension of the classical Shewhart individuals chart for time series,is known to be robust to low and moderate autocorrelation,yet for highly autocorrelated data its performance would be compromised with unexpected run lengths.A quantitative relation between in-control run length and false alarm rate for the Shewhart individuals chart is developed and is general for all types of time series data.The AR(1)process data is used to testify the proposed relation and the solution of control limit is simplified for AR(1)process.After the adjustment of control limit,the run length averaged in simulations is tested to be 98.5%of the expected ARL.It provides guidance for choosing appropriate control limits when the Shewhart chart is used for autocorrelated process monitoring.