Scandinavian Organisation of Logistics Engineers (SOLE) OPTIMISE - A web-services based platform for simulation-based optimisation: applications in production and logistics OPTIMisation: using Intelligent Simulation Tools (OPTIMIST) Amos H.C. Ng (PhD, MIET) Senior Lecturer Centre for Intelligent Automation University of Skövde, PO Box 408, 54128 Skövde, Sweden amos.ng@his.se A KKS HÖG 2004 project 1 April 2005 31 March 2008
Our research areas: Centre for Intelligent Automation Manufacturing machinery/machine systems simulation Integrated product and process development through Virtual Manufacturing and Digital Plants Simulation support for health care system design and planning Simulation-based scheduling/optimisation
Scandinavian Organisation of Logistics Engineers (SOLE) Presentation Agenda : 1. What is simulation-based optimisation and why 2. Why OPTIMISE: an industrial perspective 3. Why OPTIMISE: a research perspective 4. Overview of some industrial test cases 5. The PRiTSi test case with Posten
Scandinavian Organisation of Logistics Engineers (SOLE) Presentation Agenda : 1. What is simulation-based optimisation and why 2. Why OPTIMISE: an industrial perspective 3. Why OPTIMISE: a research perspective 4. Overview of some industrial test cases 5. The PRiTSi test case with Posten
Simulation is Not the Goal Cogito ergo sum! (I think, therefore I am!) René Descartes I simulate, therefore I optimise! A simulation engineer Perfection of means and confusion of goals seem to characterize our age. Albert Einstein
Simulation Optimisation Do you know that simulation is not an optimisation tool? Design Evaluative Model (Simulation) Performance measures e.g. no. of machines e.g. Simulation, queueing networks e.g. utilisation Are there any real optimisation tools? Objective Generative Model (AI) Optimal solution e.g. utilisation e.g. Expert system, mathematical programming e.g. no. of machines Problems must be abstracted and formulated formally based on unrealistic assumptions.
Simulation-based Optimisation Evaluative data Evaluative Model (Simulation) Generative Model (AI) Control parameters Requirements: Validated simulation models On-line system data for operational optimisation Intelligent optimisation engine Much much computing power/time
The Future of Simulation One of the disadvantages of simulation historically is that it was not an optimisation technique simulationbased optimisation is the most important new simulation technology in the last five years it is relatively new, but it will have a considerable impact on the practice of simulation in the future, particularly when computers become significantly faster. Averill Law 2002 WSC (Author of the book: Simulation modelling and analysis)
Scandinavian Organisation of Logistics Engineers (SOLE) Presentation Agenda : 1. What is simulation-based optimisation and why 2. Why OPTIMISE: an industrial perspective 3. Why OPTIMISE: a research perspective 4. Overview of some industrial test cases 5. The PRiTSi test case with Posten
Project Aim To leverage the effectiveness of the Swedish industrial and logistic sectors by introducing Simulation-Based Optimisation (SBO) to their system design and daily operations.
Objectives: OPTIMIST: Objectives Real-life industrial and logistic test cases. Gain and then spread the knowledge and experience of applying SBO and advanced simulation techniques in Sweden. OPTIMISE (OPTIMsation with Intelligent Simulation and Experimentation) - a software environment that tightly integrates Discrete-Event Simulation (DES) systems, soft-computing optimisation tools, realised on a Web-Services platform.
OPTIMISE Realisation of a Web-services based simulation platform Optimal or sub-optimal solution Data Analysis
http://optimise.its.his.se/optimise/service.asmx
Scandinavian Organisation of Logistics Engineers (SOLE) Presentation Agenda : 1. What is simulation-based optimisation and why 2. Why OPTIMISE: an industrial perspective 3. Why OPTIMISE: a research perspective 4. Overview of some industrial test cases 5. The PRiTSi test case with Posten
Some of Our Research Focuses Advanced search algorithms and methods that are, e.g. applicable for stochastic simulation (noisy optimisation), multi-objective, and/or more efficient for specific complex real-world production and logistic applications. Methods to handle imprecision/errors from the metamodels/surrogate models in SBO processes. Robust algorithms to search solutions that can sustain to input variations/uncertainies. Hybrid algorithms that interleave global and local search (Memetic Algorithm) or embed domain knowledge, e.g. shifting bottleneck detection.
OPTIMISE : A Research Platform
Scandinavian Organisation of Logistics Engineers (SOLE) Presentation Agenda : 1. What is simulation-based optimisation and why 2. Why OPTIMISE: an industrial perspective 3. Why OPTIMISE: a research perspective 4. Overview of some industrial test cases 5. The PRiTSi test case with Posten
8 Real-life Test Cases 1. Posten AB: Prosit 2. Posten AB: PRiSTi 3. Volvo Aero: Multi-Task Cell 4. Volvo Aero: Multi-Task Cell weekly planning 5. Volvo Cars Engine, Skövde: L-factory 6. Volvo Cars Engine, Skövde: H-factory 7. Volvo Powertrain, Skövde: D31 DOE 8. Volvo Powertrain, Skövde: D31 Optimisation
Real-life Test Cases
Optimal buffer allocation in Volvo Cars Engine Multi-objective optimisation for L-factory through buffer allocation: 7% higher throughput, decreased WIP and higher delivery performance.
OPTIMISE for camshaft machine scheduling
Cell Optimisation for Volvo Aero Volvo Aero Optimization of Multi-Task cell Higher utilization average 10%, decreased product delay time, decreased number of delayed products.
Simulation Model for Multi-Task Cell
Test case with Posten: Prosit Scheduling of automatic post sorting programs Objective: Search the optimal sorting programs schedule that can reduce cost, increase machine utilisation with minimal delay.
The OPTIMISE client for Prosit
Scandinavian Organisation of Logistics Engineers (SOLE) Presentation Agenda : 1. What is simulation-based optimisation and why 2. Why OPTIMISE: an industrial perspective 3. Why OPTIMISE: a research perspective 4. Overview of some industrial test cases 5. The PRiTSi test case with Posten
Case PRiTSi med Posten Optimering av transportupplägg över hela Sverige Att manuellt hitta optimala transportupplägg är ohyggligt komplex Att bara hitta ett hyfsat transportupplägg manuellt är mycket krävande Syfte med testcaset: Att ta fram en applikation som automatiskt genererar och optimerar transportupplägg
Brevdistributeringsprocess
Brevdistributeringsprocess 1. Collection Boxes 2. Mail Processing Facilities 3. Inter-regional Transportation 4. Mail Processing Facilities 5. Intra-regional Transportation 6. Mail Carrier Centres 7. Distribution to Recipents
Problemrepresentation Typ av kanter: Tåg Flyg Bil Lastbil Lastbil + släp Till varje kant associeras: c Kostnad p Miljöpåverkan t Tid v Maxvolym
Omlastningspunkter kat. 1
Transportupplägg En sträcka består av en eller flera delsträckor. En sträcka börjar alltid på en terminal eller en omlastningspunkt kategori 1 och slutar alltid på en terminal. En transport består av en sträcka och en starttid. Ett transportupplägg består av ett antal transporter. Transportupplägget är alla transporter som kommer att köras i simuleringen.
Målet av Optimering Målet med optimeringen är att ta fram det bästa transportupplägget med hänsyn till antal brev som kommer fram i tid, transportkostnad och miljöpåverkan Krav för ett giltigt transportupplägg: Hänsyn måste tas till de regler för samlastning som finns Bara de transportrelationer som finns definierade får användas Fordons maxkapacitet får ej överträdas (Observera att det ej är ett krav att samtliga deadlines hålls, men det är önskvärt) Utnyttjande av omlastningspunkter uppmuntras Om det inte är möjligt att få fram vissa brev i tid ges mer belöning ju närmare målet breven kommer
In- och utdataskal
Optimeringsalgoritmer Två optimeringsalgoritmer används Hill climber för lokal sökning Genetisk algoritm för global sökning Heuristiker används i algoritmerna Utnyttjande av omlastningspunkter uppmuntras Om det inte är möjligt att få fram vissa brev i tid ges mer belöning ju närmare målet breven kommer
Arbetsgång Problem: Det krävs en mycket stor mängd simuleringar för att hitta bra lösningar och varje simulering är tidskrävande Lösning: Grovsimulering Ger mycket snabbt ett ungefärligt värde på resultatet
OPTIMISE Clienten för PRiTSi
Scandinavian Organisation of Logistics Engineers (SOLE) Questions?