Six-Sigma: It’s a Dirty Job Andrew Gonce, McKinsey Bob Landel and Jitendra Gupta MBA ‘08, Darden Six-sigma approach Practical Problem Traditional Approach Statistical Problem Six-sigma Approach Statistical Solution Practical Solution Six-sigma is a systematic data-driven approach, which leads to a sustainable solution for any problem 2 Narrowing the Project Scope: F(x) 3 DMAIC Example – It’s a Dirty Job Define What are the customer expectations of the process? Purpose and scope of the project Reduce the Incidence of Dirt in the Primer Coat that occurs on the Hood of the vehicles at the Lexington Assembly Plant between the E-Coat Scuff Booth and the Prime Scuff after Oven station The outcomes with defects are identified as red in the population Six-sigma leaders have a mind-set for meeting customer needs 5 DMAIC Example – It’s a Dirty Job Key Deliverables of Define Phase Define • Project Scope and estimation of benefit based on customer requirements and bottom-line performance Y Dirt in Paint y Dirt in Primer spray area; Dirt in Ovens x Critical X’s to be determined in Analyze phase • A team charter with defined roles and responsibilities • A high-level process map Few of the applicable tools Baseline Performance for Y, Customer Survey Methods (focus groups, interviews, etc.) Project Risk Assessment, Stakeholder Analysis, High Level Project Plan 6 Narrowing the Project Scope: F(x) • Y = Dirt in Paint (F20)= f(x) {Prime, E-Coat, Base Coat, Clear Coat} o X = Dirt in Prime (37%) = f(w) {Agglomerates, Sealer, Fibers, Rust, Condensate, Pollen} W = Agglomerates in Prime (33%)= f(v) {Primer Spray Booth, Ovens} Critical “X” Contribution = 8-10% of F20 Calls 7 DMAIC Example – It’s a Dirty Job Measure • Perform Gauge R&R on Primary Measurement System • Evaluate Critical “X” Process capability • Determine controls in place for Critical “X” If you can’t measure it accurately, you can’t improve it! 8 DMAIC Example – It’s a Dirty Job Measure What is a defect? Is the measurement system capable of separating acceptable from defective parts? • For a continuous metric (such as distance, time etc) capture the specification limits (LSL, USL) and the target to determine the tolerance band (for acceptable parts). For discrete metric, identify the characteristics of a part that result in it being acceptable/defective • The total observed variation in the data is a sum of variation in the process and variation in the measurement system. If the latter is higher than a limit, we will not be able to differentiate between good and bad parts. A good measurement system has to be both repeatable and reproducible If you can’t measure it accurately, you can’t improve it! 9 Gauge R&R – Dirt Count at Spill Out • The Gauge R&R was conducted on the Hoods alone. – The Hood area is the easiest to see Dirt in Prime. – 15% of Warranty Verbatims call out the Hood as the location of Dirt. 10 DMAIC Example – It’s a Dirty Job Key Deliverables of Measure Phase Measure • Defined Performance standards (Spec limits and target) • Gauge R&R analysis of measurement system Few of the applicable tools GR&R, FMEA, Pareto analysis, Data collection plan 11 Baseline Data: Prime Dirt Count Attribute Control Chart Dirt in Prime Count Prior to Prime Scuff Average DPUs 5 4 DPU 3 UCL: 3.1 2 MEAN: 2.3 LCL: 1.1 1 4/ 10 /0 4/ 2 24 /0 2 5/ 8/ 0 5/ 2 22 /0 2 6/ 5/ 0 6/ 2 19 /0 2 7/ 3/ 0 7/ 2 17 /0 7/ 2 31 /0 8/ 2 14 /0 8/ 2 28 /0 9/ 2 11 /0 2 0 Time • The Dirt Analysts report on 20 unit samples before each scuff station in daily inspections 12 DMAIC Example – It’s a Dirty Job Measure Is a the process in statistical control? What is the current process capability? • The practical problem is converted to a statistical one. Capability is measured in terms of Z score and Cpk, which captures the mean and variation relative to specifications. 3.1 DPU Upper Spec. Limit 0 DPU Lower Spec. Limit Current Sigma Level: 1.33 Objective of six-sigma is to reduce variation and to center process 13 DMAIC Example – It’s a Dirty Job Measure Is a the process in statistical control? What is the current process capability? • The practical problem is converted to a statistical one. Capability is measured in terms of Z score and Cpk, which captures the mean and variation relative to specifications. 3.1 DPU Upper Spec. Limit 0 DPU Lower Spec. Limit Current Sigma Level: 1.33 Objective of six-sigma is to reduce variation and to center process 15 DMAIC Example – It’s a Dirty Job Analyze What is the current and desired process capability? Why, when and where do the defects occur? The fundamental objective of analyze phase is to identify those key process inputs (critical X’s) that are different for the good and the defective ones (or are statistically significant). Def Acceptable Critical X’s for Dirty Job example Factor 1: Temperature and Humidity Factor 2: Weekday Variability Factor 3: Prime Automation Equipment Factor 4: Prime Ovens Factor 5: Area Conditioning Tools Fish bone, normality test, Hypothesis testing (for mean, median and variation), Regression, chi-square 16 Factor 4: Prime Ovens A Dirt Count was conducted for 28 vehicles, immediately before and after the Prime Ovens The average increase in counted dirt was 10 defects per vehicle hood 18 Factor 4: Prime Ovens The 28 units that were counted were also tracked by which Prime Oven they passed through There was a significant difference between the Oven Dirt Contribution, with Oven 2 adding the most defects 19 Factor 4: Prime Ovens The ANOVA Analysis for Smoke Primed Vehicles only shows that there is a greater than 95% significance between the change in dirt counts for each oven. 20 Factor 4: Oven Cleaning The Ovens are not covered in the Existing Work Order System – there is a gap 21 DMAIC Example – It’s a Dirty Job Improve How can we fix the process? • Identify the relationship of X’s on Y by developing the transfer function for Y=F(X), using tools such as DOE • Determine the optimal settings and tolerance limits for X’s inputs to achieve the desired Z-score for Y. • Run a test plan to confirm the causal relationship and to validate the improvement in Y Improvement plan for X’s No Description Improvement Plan 1 Temperature and Humidity Automatic booth balance 2 Weekday Variability Weekend PM schedule revisions 3 Prime Automation Equipment Tracking process initiated, PM revisions 4 Prime Ovens Oven cleaning 5 Area Conditioning Update PM sheets, follow procedures 22 Factor 4: Oven Cleaning BASF Recommendations Reduction or Elimination of Contaminants in all Systems – Lower Agitation, Overhead Structure, Air Seals and inside Burner Units Eliminate Mounds of charred dirt and Paint Chips visible inside of Conveyor Chain Track Eliminate Dirt from rear side of High Temperature Recirculating Filters Eliminate Rust and Dirt lying inside of Air Seals Eliminate Dirt blowing out of Lower Convection Hot Air Supply Ducts Eliminate Dirt and Fibers on rear side of Panel Filters Eliminate Rust and Dirt Particles falling off Overhead Ceiling and Hardware onto vehicles traveling through the ovens 23 Factor 4: Oven Cleaning Oven Cleaning is best conducted in a cycle that allows the oven to: be cleaned with dry ice, vacuum, and rags be heated to operating temperatures for 4-8 hours be inspected and re-cleaned and be re-heated for 4-8 hours prior to use Action Taken Discussion with Sam Lemay to standardize Oven Cleaning Procedure and sign-off. A Gap Analysis shows that the Prime Ovens do not have the level of standardized cleaning that the Prime Spray Booth, Sealer Deck and Vestibule have. 24 Factor 4: Oven Cleaning The 3rd Pass Oven Cleaning that occurred in October ’02 resulted in a measurable improvement in dirt count per 20 units Oven Cleanliness has a real effect on overall Dirt Count! 25 Factor 5: Area Conditioning A study was conducted in November 2002 following the Vehicle View through the entire Paint Process (BASF) A number of Maintenance, Cleaning and Repair items were documented and recommendations were made 26 Factor 5: Area Conditioning Recommendations Prime Spray Booth: Muck cleaning (grates and water) currently occurs annually. Entering this cleaning into the PM Work Order system is recommended. Develop plan for additional humidity and water flow. Trial an adhesive paper on the floor of the vestibule or oven entrance to trap airborne dirt and sprayed paint. Eliminate Cotton Mops, Newspapers and Contaminants from the Spray Areas, follow the Dress Codes. 27 DMAIC Example – It’s a Dirty Job How did we improve the process? Improve Actions already taken lowered the DPMO from 112,000 to 6000! Booth Balance, External Environment and a reduction in environmental variability lowers the problems due to Prime being out of spec. Regular cleaning and Maintenance reduces fiber count and additional dirt in paint from airborne contamination. Improvement plan for X’s No Description Improvement Plan 1 Temperature and Humidity Automatic booth balance 2 Weekday Variability Weekend PM schedule revisions 3 Prime Automation Equipment Tracking process initiated, PM revisions 4 Prime Ovens Oven cleaning 5 Area Conditioning Update PM sheets, follow procedures 28 DMAIC Example – It’s a Dirty Job Control How can we ensure that process stays fixed? • Establish post improvement capability and validate that the pre and post difference is statistically significant • Run the MSA on X’s and establish control plan for Y and X’s Pre-Improvement Post-Improvement Few of the applicable Tools Control charts, Hypothesis testing, Mistake Proofing,, FMEA 29 DMAIC Example – It’s a Dirty Job How can we ensure that process stays fixed? Control Control Chart: Defect Tracking “Y” 30 DMAIC Example – It’s a Dirty Job How can we ensure that process stays fixed? Control Control Chart: Action Plan “X” 31 DMAIC Example – It’s a Dirty Job How can we ensure that process stays fixed? Control Control Chart: Defect Tracking “Y” Control Chart: Action Plan “X” 32 Lessons Learned The walk-through by BASF, the dirt analysts, filter rep’s et al. was instrumental in discovering a number of system problems. This should be an annual occurrence to maintain the systems. Improvement efforts need to be quantified with data (dirt count, operator comments, efficiency etc.) in order for the results to be weighed. 33 Lessons Learned There are many, many factors that effect how clean a particular vehicle is on any given day. There are no easy, cheap or obvious solutions, all will take some effort to discover and some effort to resolve. The Sealer Deck and Prime personnel understand the issues that they face in producing clean vehicles. 34