DSES-6620 Simulation Modeling and Analysis Navy Supply Chain Model Verification and Validation Rich Sewersky May 2, 2002 Background • • • • New business for Aerospace: Provide spare parts at fixed cost per flight hour called Performance Based Logistics (PBL) Key measure and risk is %Fill Rate = on time shipments/total requisitions Goal of model is to predict fill rate and other measures such as inventory to be able to complete business tradeoffs/contingency planning Model prepared in TaylorED by researcher, Sewersky was Subject Matter Expert – 14 repairable components (Main Gearbox Module chosen for project) – One year run period with specified starting position (contract terms) • • Project goal was to verify and validate model so it can be used for business Key steps – – – – • Research commercial/military V&V guidelines Review model logic and code and research report Explore model sensitivity to key inputs and output variability Calibrate output to known commercial simulation package Terms: – Ready for Issue (RFI) - Components in working order – Carcasses - Broken components TaylorED Modeling Paradigm (connect the dots) PoolInfoTb InfoTb BkOrderTb Uniform (1, flowtime) Uniform (1, deltime) Negexp (3.121) Results - Verification • Research found DOD has Defense Modeling and Simulation Office – has web site with guidance and educational materials, bibliography • www.dmso.mil/public/transition/vva/ – several working groups including formal V&V and Accreditation – key issue is model interoperability and contractor sell-off – many of these models support “war games” • Developed logic diagram based on model structure and code flow – (found one code error and corrected it - minor performance effect) • Checked effect of key variables on output measures – – – – – Variation turned off Demand interarrival rates Transport times Processing times Distribution fitting • Model reactions made sense Results - Validation • Researcher had checked model outputs against OPUS spares modeling tool - OPUS 91% Fill vs. Model 96% • Sewersky baseline runs averaged 93% with fairly high variation (stdev 9.2). (note that the run with all variation turned off was 90.6%) • Input data is also highly variable – Navy demand histories – Navy transportation times (returned “carcasses”) – Shop flow times (part shortage driven) Results Summary Baseline Run (with WIP distribution error) Baseline Run (with WIP distribution error corrected) Run with all variations removed Run with demand variation only (transport and processing at worst case fixed level) Run with demand variation only (transport and processing at average fixed level) "Wartime" Scenario (Worst Case Everything) Avg StDev Avg StDev Avg StDev Avg StDev Avg StDev Avg StDev Total BK Total Fill Order Order# Rate(%) 7.9 119.8 93.7 9.8 8.0 7.8 8.4 117.5 93.3 11.5 9.2 9.2 11 116 90.5 n/a n/a n/a 15.8 110.3 86.6 14.5 12.1 11.1 2.4 121.6 98.1 3.2 344.2 20.1 6.7 352.2 20.1 2.5 2.3 0.1 Avg WIP 16.8 1.1 18.6 1.5 20.7 n/a 20.2 1.8 11.7 0.6 57.0 3.1 Backorders Avg RFI Count Average Minimum Maximum in WH Duration Duration Duration 10.6 13.2 4.0 0.1 18.3 1.7 9.5 3.1 0.6 5.8 9.2 21.0 6.6 0.1 23.3 2.4 6.6 3.9 0.3 8.2 7.2 11 16.3 0.7 31.9 n/a n/a n/a n/a n/a 8.3 15.8 13.1 0.1 50.3 2.9 14.5 5.6 1.7 10.7 21.7 3.0 4.1 0.8 11.9 0.9 0.1 0.1 Most orders remained unprocessed so these statistics are misleading (only for those actually processed) 3.4 344.2 20.1 2.1 50.9 3.1 1.6 30.5 2.7 4.5 79.6 2.1