| Statistical quality control uses statistical methods to identify random cause ofvariation and special cause of variation in the process of production, so as tomonitor stability of the process, and achieve the goal of improving product quality.Control charts are the major tools, which were introduced by Shewhart in the1920s,and the commonly used charts are the mean-range control chart and the mean-standard deviation control chart. Statistical quality control had successfully beenused in machinery manufacturing, Ford and General Motors, etc, and in recentyears, the research is focused on drawing control charts with data that don’t obeynormal distribution, small batch production, multiple quality characteristics, and theautocorrelation process, etc.It is of great importance to control the quality of the molten iron, as it is themiddle product of the integrated iron and steel enterprises, the quality of whichaffects the quality and profitability of all kinds of final products such as rail,seamless pipe, steel coil and wire rod etc. However, most enterprises adoptchemical reagent analysis to analyze the content of hot metal as well as usingvarious mathematical models to predict quality of molten iron, but the deviation ofSi and S exceed the tolerance of molten iron sometimes. So this article aims toexplore a suitable molten iron quality control chart method. Molten iron, which hasmultiple attributes such as the temperature, content of Fe, C, Si, Mn, P, S etc, is themain product of blast furnace. And blast furnace iron-making is a process ofproduction which includes continuous blast, periodic loading and periodic tappingiron, making the test data with autocorrelation.At first, this paper chose1#blast furnace’s data under steady state, tooksingle value control chart （1stchart） as benchmark, to analyze the effect of themoving time grouping quality control chart （2ndchart）.1) We selected elementS and Si content in molten iron as the key quality characteristics of blast furnaceiron-making process, tested the distribution of the samples, respectively drew single value control chart for S and Si element.2) We took SQL database commands togroup sample values by move time, calculated the control limit according to thegrouped data, and drew the moving time group control chart. Comparative analysisshowed that moving time group control chart （2ndchart） can identify outliersearlier than single value control chart （1stchart）; With the increase of movingtime window length, the control chart was easy to monitor product quality variationtrend. But moving time group control chart was only considered a qualitycharacteristic value, could not fully monitor multiple quality characteristics of hotmetal.This article took company A1#,2#,4#blast furnace steady state samplesand3#blast furnace unstable state samples, with dual T2control chart （3rdchart） as benchmark, to analyze the effect of the dual autocorrelation residual T2control chart （4thchart）. Analyzing the relationship between S and Si by dataanalysis software SPSS, using the time series model to calculate the correspondingresidual value, drew control charts of the1#,2#,4#blast furnace under steadystate and the unstable state of3#blast furnace, calculated the cry wolf rate and theaverage run chain length (ARL) of control chart and unstable state leakage alarmrate and ARL. Comparative analysis showed that ARL of the binary residual T2control chart was longer than the binary T2control chart under the stable state, andARL of the residual T2control chart was shorter than traditional T2control chartunder the unstable state.Above all, dual autocorrelation residual T2control chart (4thChart) kept theadvantage of the moving time group control chart （2ndchart） to monitor thetrend of the process, and the advantages of3rdChart to monitor multiple attributes,so4thchart can effectively control the blast furnace iron-making production process. |
PDF Full Text Request