A short introduction to embedded optimization - Home - … · 2017-12-08 · A short introduction to embedded optimization ... Approximation Solutions for Stochastic Models 19. ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Tecnomatix Plant Simulation Worldwide User Conference 2016
Siemens Industry Software
A shortintroduction toembeddedoptimizationTecnomatix Plant SimulationWorldwide User ConferenceJune 22nd, 2016
1. The simulation is used as evaluation function ofoptimization
We are looking for optimal parameters for a model.By systematic changes of parameters we candetermine a satisfactory model configuration.
2. The optimization is an integrated component of asimulation.
A parent simulation is interrupted and an optimizationof the current state is performed.
A further structural connections is theuse of simulation results as starting values for anoptimization.Such optimization is possibly based on purecalculations.
Optimizer Model
Modification
Evaluation
Simulationtime
Optimizer
See Chapter 11: L. Iitzsche, P. Schmidt, S, Völker in Lothar März, Winfried Krug, Oliver Rose, Gerald Weigert (Hrsg.) Simulation und Optimierung in Produktion und Logistik
Characteristic properties of OptimizationProblemsIn most applications multiple targets are consideredwhich have opposite natureand must be described in a single numericalevaluation value.
Combination of several input parameter settings1. Determination of sequences2. Dimensioning of production resources.
We distinguish so called basic tasks1. Sequence tasks (find a numbering of a finite set)2. Allocation tasks (find a value of a finite range).
Solving of Trade-off
1. Reduce throughput timesby prevention of waiting time
2. Just-in-time delivery
3. Reduction of the warehouse stockand associated costs
Tecnomatix Plant Simulation Worldwide User Conference 2016
Exact and Heuristic Methods for OptimizationProblems
Basic tasks are well examined in OperationsResearch.The algorithms Branch & Bound and DynamicProgramming for the Traveling Salesman Problem aredescribed in theExample Collection (open via the Start Page).
For basic tasks with practical important problem sizeand combined tasks there are no efficient algorithms,which find the optimal solution in reasonablecomputational time.
Therefore heuristic methods are used, like Hill-Climbing, Simulated Annealing, Tabu Search andGenetic Algorithms (shorted by GA).
The Ideas and the Structure of Genetic AlgorithmsGenetic Algorithms model natural processes of theevolution.
The Fitness of the individuals describes, how well it isadapted to the environment. This evaluation isdetermined by the combination of their elementaryproperties.The Fitness can be determined by simulation orcalculation.
1. Recombination: Use of properties of previousgeneration (Selection of 2 Parents according to thefitness)
2. Mutation: Generate 2 children per familyby random application of Genetic Operations,like Crossover, Mutation, Inversion
3. Selection: Select one offspring per family (child or
Each new generated individual must be evaluated.The first generation has the Generation sizeindividuals.All following generations have Generation sizefamilies.Each family has 2 children, which must be evaluated.
Please note, that in a stochastic simulation study anevaluations needs more than 1 simulation runs(observations).
Number of evaluations = Generation size * (2 * Number ofGenerations - 1)
Tecnomatix Plant Simulation Worldwide User Conference 2016
Five Basic Objects for the Optimization by GAThe controller GAOptimization• Number and size of generations,• Direction of the optimization (minimum or maximum)• Termination condition• Definition of the selection of parents and offspring• Controls, such as for the fitness calculation• Recording of generated individuals
The other four objects are GA tables for basic tasks and its GeneticOperators.
GASequence: Sequence task for a given number of itemsGASelection: Selection task of a certain number of elements a givennumber of itemsGARangeAllocation: Determine an item of a range between two boundsGASetAllocation: Determine an item of a set of elements
We will find the sequence of the Delivery table of thesource,such that the throughput time for the given orders isminimal.The throughput time is determined by the setupmatrix.
For this example the resulting throughput time is8 + 1 + 6 + 2 + 2 + 3 + 3 + 9 + 4 + 4 = 42.The optimal solution is obvious.
The Plant Simulation model for adeterministic studyin English class libraryis generated by executing a method
Start Plant Simulation with the shortcut option/UILanguage:ENU.
Create a model with model language English.Select the Menu File > Preferences > Tab General.
The evaluation of a sequence is done by simulation.For the optimization we apply the GAwizard.
The fitness calculation can be definedas a weighted sum of multiple simulationresults in the table Fitness.Drag & Drop the Eventcontroller onto the GAwizard.
There are no general recommendations for theparameters of the Genetic Algorithm.For sequence tasks a generation size between 50 and 70 issuitable.
Tecnomatix Plant Simulation Worldwide User Conference 2016
Visualization the Optimization ProgressThe Performance Graph shows a typical appearanceofGenetic Algorithms. It is opened by the buttonEvolution.The worst individual in a generation also improvesif a better individual is found.
Scenario and TargetIn a line of two types of products are produced.The arrival process of the products is at random.To achieve an uniform utilization of followingresourceslong partial sequences of single products type shouldnot occur.During the production process the sequence isoptimized.
IdeaIn a buffer (modeled by the basic object Sorter)multiple products are collected.For this purpose the entrance and the exit of thebuffer are controlled by an Observer method for thevalue numMU.
Tecnomatix Plant Simulation Worldwide User Conference 2016
Online-OptimizationIf at the beginning of a simulation not all data areavailablethen multiple sequencings are necessary.This leads to smaller number of items of the sequencecompared to the full size of the problem.
Evaluation by RulesUnwanted sequences are evaluated by a bad fitness.The evaluation of a sequence is done by one or morecriteria and is performed by rules.The result of a rule describes the penaltyfor the considered sequence:
See Chapter 17: C. Heib, S. Nickel in Lothar März, Winfried Krug, Oliver Rose, Gerald Weigert (Hrsg.) Simulation und Optimierung in Produktion und Logistik, Springer
for local j := 1 to sequence.yDim - 1 loopif sequence[1, j].name = sequence[1, j+1].name then result :=