3 Gas Champion : MBB : IBM BCS PO : 2 BBc : : 20049 0/45
Define ~ Analyze Define VOB KBI R 250 O 2 2.2% CBR Gas Dome 1290 CTQ KCI VOC Measure Process Data USL Target LSL Mean Sample N StDev (Within) StDev (Overall) * * 0.800000 0.811947 119 0.0075351 0.0116898 Process Capability Analysis for heat_p LSL Within Overall Potential (Within) Capability Z.Bench 1.59 Z.USL * Z.LSL 1.59 Cpk 0.53 Cpm * Overall Capability Z.Bench 1.02 Z.USL * Z.LSL 1.02 Ppk 0.34 0.77 0.78 0.79 Observed Performance PPM < LSL 142857.14 PPM > USL * PPM Total 142857.14 0.80 0.81 0.82 0.83 Exp. "Within" Performance PPM < LSL 56432.53 PPM > USL * PPM Total 56432.53 0.84 0.85 Exp. "Overall" Performance PPM < LSL 153398.49 PPM > USL * PPM Total 153398.49 1/45
2/45 I I Others STS 4.5 2.5 4.0 4.2 4.5 5.8 9.4 65.1 4.5 2.5 4.0 4.2 4.5 5.8 9.4 65.1 100.0 95.5 93.0 89.0 84.8 80.3 74.5 65.1 100 50 0 100 80 60 40 20 0 Defect Count Percent Cum % Percent Count ('02) Others 0.7 2.1 14.2 83.1 0.7 2.1 14.2 83.0 100.0 99.3 97.2 83.0 100 50 0 100 80 60 40 20 0 Defect Count Percent Cum % Percent Count ('02) 13.75 86.25 13.8 86.3 100.0 86.3 100 50 0 100 80 60 40 20 0 Defect Count Percent Cum % Percent Count ('02) 65.1% 83.1% 86.3% 13.7%
Define ~ Analyze Analyze Vital Few AHP Analysis 44 17 Insight Variable Selection Regression Decision Tree 1. Mgas_cal 2. Cog_p 3. A_g 4. Mgas_w Vital Few ( ) Process (Regression, Multi Variate Analysis) Machine Learning (Decision Tree/Neural Network) Data 3/45
Process Overview Process HBV GAS BAV BGV (1100~1300) Process 1 4. 3BF., M-Gas 60. 3 2( Parallel ), CBV CHV 4/45
Model Regression Model (1) Model (2 ) Input (, ) Decision Tree (1) (2 ). DT Neural Network (1) Data Input Layer. (2) signal Multilayer Perceptron Hidden Layers Output Layer Hidden Unit 5/45
P_heat_p( )Vital Few Contour 6/45
Data Mining Vital Few V.Few (heat_p >83%) DM Mgas_cal (kcal/n) 930 976 COGMgas_w (%) (%) (N/D) 6.0 1.0 3,668,825 5.4 1.08 3,634,000? Problem : Data Mininer Local Solution OK, But Gobal Solution? Data Mining Pilot Test. Crytal Ball Based Simulator,. 7/45
Quick Win Pilot Test Quick Win 1 ( 1,4 ) EV Open #1 : Feedforward Guidance Simulator ( 1,2,4 ) #2 : ( 1,4 ) 10.15 ~ 10.20 ~ 8.4 ~ Quick Win 2 ( 4 ) Steam & N2 Line 9.4 ~ #3 : RV Open Program ( 4 ) #4 : & ( 5 ) #5 : MCV ( 4 ) #6 : BFG ( 1,2 ) 10.9 ~ 10.13 ~ 9.4 ~ 10.23 ~ Pilot Test 10.14~ 10.31 Quick Win 3 ( 3,4 ) #7 : Pattern ( 1,2,3,4 ) Calorie #8 : Mgas PID Gain Tunning( 2,3,4 ) 10.23 ~ 10.23 ~ 9.4 ~ Vital Fews 1:COG, 2:, 3:Mgas Calorie, 4:Mgas, 5: 8/45
(1) #1 : BFG & COGFeedforward Guidance Simulator Feedforward Guidance Simulator3 Meter Room - Neural Network Model : Guidance ( )COG Mgas. 9/45
(2) #2 : 1. BFG M-GAS 2. COG MIX-GAS M-GAS 3. : Program & Logic DOME DOME DOME 4. Gas Chro Calorie (H2%, CO%) : Over 10/45
(3) #4 : - No.34 No.34 Max : 189 Max : 187 Max : 187 T/C Max : 200 3 03.9.30 ~10.1, 11/45
(4) #7 : Pattern 3.. Pilot Test checker Pattern.. 1.10 1.05 1.00 0.95 0.90 Conventional Method 1.10 1.03 1.00 0.95 0.90 Proposed Method 99,588N/Hr 91,853N/Hr 0 10 20 30 40 50 0 10 20 30 40 50 3 96,687N/Hr (9) 5% 2Pattern : Mix Gas Calorie 12/45
(5) #8 : Mgas PID Gain Tunning PID( Proportional-Integral-Derivative) Controller. 3 (PV:Process Value) (SV :Set Value). P Gain Rising Time I Gain Setting Time. PID Gain Tunning Matlab Simulation. Tunning Tunning P Gain- Mgas_w I Gain- Mgas_w D Gain- Mgas_w 100 45 0??? 3 Mix Gas Gain Mix Gas Hunting : Mix Gas & COG 13/45
CTQ ( ) & Vital Few : 36.9 / : ( 81 87%) Gas Total : 13.96 / {( Gas Gas ) x } x Mixed Gas = {(397 359 Mcal/T-P)} x 8,587Ton/D x 32.46/Mcal x 365 x 0.98 = 36.9 / Gas Gas : : 396.9 396.9 359.7 359.7 Mcal/T-P Mcal/T-P :37.2 :37.2 Mcal/T-P Mcal/T-P (10% (10% ) ) CTQ CTQ ( ( ): ): 81.2 81.2 % 87.1% 87.1% : : 5.9 5.9 % % 14/45