GEOLOGY, RAW MATERIALS AND RAW MIX Process Engineering Certification Program April 7 - 8, 2003 Holcim Statistical Process ControlSession 3: Statistical Process Control bs: 2 Holcim (US) Inc - CTS. Geology, Raw Materials and Raw Mix eee jolcimSPC- Using Optimum Control Strategies « Use the optimum sampling and testing frequency determined on the preceding slides. Calculate the process capability standard deviation based on the moving ranges of the measurements at the optimum sampling frequency. * Once this result has been computed for a few days of typical operation, use the estimate of the process capability standard deviation as a constant. Re- compute this result every 6 months or after every major process change or shut-down for equipment modifications or maintenance. Monitor the existing control strategy using the shift-away strategy and different sets of control rules to see which are most sensitive to detecting shifts in the process. 3 Holeim (US) Inc - CTS Geology, Raw Materials and Raw Mix 12:12.2005, Pecra2 2002 uh jolcimStatistical Process Control-Variogram = Calculating variation from time-dependent data - Use the moving range method of estimating the standard deviation. - Estimate S,,,, based on the current sampling frequency (j = 1x) as well as j = 2x, 3x, 4x, etc. - Obtain 25 samples at 1/4x. Estimate S,.; based on these data for j = 1/4x, 1/2x and 3/4x. * Since S,,,,, increases as the time between samples increases due to new sources of variation occurring in the data, estimate the time increment during which the most variation enters the system using the formula {Spay )? - (Sporger) P* « Find the reasons why this variation is so large and eliminate them. jolcim Holeim (US) Inc - CTS. Geology, Raw Materials and Raw Mix eeu C3S Variogram from 1 minute sampling frequency Spel() 50 40 30 20 10 C3S Variogram-sampling frequency 1 minute peeaeeeeee v 0 5 10 06150 202580 85 Lag j (minutes) jolcim 5| Holeim (US) Inc - CTS Geology, Raw Materials and Raw Mix yai2.2005,SPC- Interpretation of Variographs = The nugget effect is an extrapolation of the variograph to a zero time difference. It is estimated by the S,., which is 1.9% for CS. Note that this is close to the FSE of 2.2% obtained from the simulation on page 49. = The sill is where the variograph reaches a plateau or asymptote. Here it is an Syc1.49 Of about 47%. This is the maximum standard deviation that the flow of materials attain. It reflects all the sources of variation that affect this property. = The range is the length of time required to reach the sill. It follows, though it is not obvious, that this is the time interval between samples of the raw material that are statistically independent. By an independent analysis | determined that it took 30 minutes for consecutive samples to be statistically independent. = Sometimes there are cycles in the variograph. Such cycles can reflect Periodic occurrences in the raw material due to some natural phenomena. Thotcim 6 Holeim (US) Ine - CTS Geology, Raw Materials and Raw Mix 7222005 ecro2- 2002Exercise 16. LSF Variograph « The 200 hourly LSF measurements are in an Exceld file. = Determine the Variograph for the first 10 lags. = Analyze these results to find the Nugget Sill Range Cycles if any e Holcim / Holcim (US) Inc - CTS Geology, Raw Materials and Raw Mix eeSPC- Optimum Sampling Frequency Identify the critical specification for each measurement. Examine the historical data to determine the worst-case process average relative to this critical specification. = Set the sampling frequency at a value of j = 1/4x, 1/2x, ... 4x for which [Critical Spec - Worst-Case Process Average| > 3 S,.1; = If the process is not stable, then the sampling frequency should be variable. When the average is close to the critical spec the frequency must increase while the frequency can decrease when the average is further from the critical spec. If a constant sampling frequency must be established then the worst case scenario is to be used as indicated above. , jolcim Holeim (US) Inc - CTS Geology, Raw Materials and Raw Mix peorSPC- Western Electric Control Rules * Asbhift has occurred if 1. A single point is outside the current process average or Target + 3S... 2. 2 out of 3 consecutive points, both on the same side of tl e Target or average, are outside the current process average or Target + 2S, 3. 4 out of 5 consecutive points, where all 4 are on the same “side of the Target or average, are outside the current process average or Target + S,., 4. 8 consecutive points are either above or below the current process average or Target. = When a shift is detected, the average of the points involved in the rule violation determines the performance level to which the process has shifted. = When a shift occurs, a process correction can be made to bring the process back to the target from the performance level estimated by the points p involved in the rule violation. Holcim Holcim (US) Inc - CTS. Geology, Raw Materials and Raw Mix PeasSPC- Stop-Light Control = A shift has occurred if 1. A single point is outside the current process average or Target + 2S,,..; This is called the “red” region 2. 2 consecutive points, both on the same side of the Target or average, are outside the current process average or Target + 1.0S,,..; This is called the “yellow” region 3. The “green” region is everything within average or Target + 1.0S,.., No action is taken as long as the results are in this region. = When a shift is detected, the average of the points involved in the rule violation determines the performance level to which the process has shifted. « When a shift occurs, a process correction can be made to bring the process back to the target from the performance level estimated by the points p involved in the rule violation. Holcim S Holcim (US) Ine - CTS Geology, Raw Materials and Raw Mix peo? amsSPC- Trouble-Shooting Rules « Ashift has occurred if 1. A single point is outside the current process average or Target + 2S,,., 2. 2 consecutive points, both on the same side of the Target or average, are outside the current process average or Target + 1.5S,.,; 3. 5 consecutive points are on the same side of the current process average or Target. » When a shift is detected, the average of the points involved in the rule violation determines the performance level to which the process has shifted. » When a shift occurs, a process correction can be made to bring the process back to the target from the performance level estimated by the points involved in the rule violation. Vhotcim mM Holeim (US) Inc - CTS Geology, Raw Materials and Raw Mix pet 208uh SPC- Shift-Away Analysis The first five data points are used to determine the starting process average. = The process capability standard deviation, S, is determined using at least 25 data points. pel = When a shift is detected, the average of the points involved in the rule violation determines the performance level to which the process has shifted. = The centerline of the control chart is positioned at the new average but the same control limits used in the rules are superimposed about this average. = In this way times when shifts occur are identified as well as the size of the shift. = The cause of a shift can be determined by the process operators. Removal of the cause of shifts will ultimately result in process improvement as indicated by a reduced process capability standard deviation with an average more consistently on target. jolcim 2 Holcim (US) Inc - CTS. Geology, Raw Materials and Raw Mix eae 08Exercise 17. Shift-away analysis for LSF = Calculate the process capability standard deviation using the moving ranges for all the data. = The hourly LSF data are in an Excel® file. For this exercise, analyze the first 25 for shifts using the shift-away logic. = What do you conclude from this analysis? Vhotcim 13 Holeim (US) Inc - CTS Geology, Raw Materials and Raw Mix eae 08Summary of Session 3 The Variograph is a tool for: 1. Determining the optimum sampling frequency 2. Estimating the FSE 3. Establishing carryover from one sample to the next 4, Identifying periodic behavior in the raw material source 5. Quantifying the maximum variation. = The control chart is a tool for identifying when a process shifts and how large this shift is. = There are several different sets of control chart rules. The best set depends on the specific process to which the strategy is applied. This requires trial and error. Any set of rules that minimizes the process variation is optimum. = Operators as well as process controllers should use the process capability standard deviation as a basis for deciding if and when a process shifts and by how much. Corrective action should be taken only in cases where statistically significant shifts are detected. jolcim Holeim (US) Ine - CTS Geology, Raw Materials and Raw Mix 7222005 ecro2- 2002