| In the last decades,there has been a constant increase in the awareness of company manage-ment about the importance of formal techniques in industrial settings to address problems related to monitoring and reliability,fault diagnosis,and optimal use of resources,during the management of plants.Of particular relevance in this setting are the so-called Automated Manufacturing Systems(AMSs),which are characterized by complex technological cycles that must adapt to changing demands.Moderm AMSs are interconnected subsystems such as numerically controlled machines,assembly stations,automated guided vehicles,robots,conveyors and computer control systems.Manufacturers are using automated machines and controls to produce quality products faster and more efficiently.Meanwhile,these automat-ed systems can provide critical information to help managers make good business decisions.However,due to the high flexibility of AMSs,failures such as a wrong assembly or a part put in a wrong buffer may happen during the operat:ion of the system.Such failures may de-crease the productivity of the system which has an economical consequence and can cause a series of disturbing issues.As a result,the performance optimization in AMSs are imper-ative.The quantity of products which have to be stored or moved and the number and type of machines which operate the system have economical consequences.Once the resources arenot well assigned.the system may produce products with a low efficiency and even cause a deadlock.Therefore,the main problem for engineers or designers is to find an optimal mode of operations given a set of available resources or to find an optimal set of resources capable of meeting the required production constraints.As a powerful mathematical tool,timed Petri nets models have been extensively used to model,analyze,and control of AMSs.They can be used for performance analysis,tasks scheduling in real-time,and optimizing the use of resources.This thesis focuses on the performance evaluation and performance optimization of auto-mated manufacturing systems in the DES model of timed Petri nets.The main results of this research are as follows.1.For a class of deterministic timed Petri nets called timed weighted marked graph-s(TWMGs),which are extensively used to model and analyze cyclic AMSs.The marking optimization of deterministic TWMGs under single server semantics plays an important role in the manufacturing domain,which consists in finding an initial marking to minimize the cost of resources while the system's throughput is less than or equal to a given value.The existing results fail to provide practically effective and computationally efficient methods to analyze and solve the problem in such systems.We take the advantages of the net structure characteristics of a TWMG and utilize related knowledge of liveness of a TWMG to select a proper initial marking.Next,based on simulation a heuristic algorithm used to increase the system's throughput by iteratively adding tokens to some places is developed.Finally,a technique to reduce the cost of the obtained solution by taking the advantages of the previous works is proposed.2.From a physical point of view,the server semantics can be interpreted as the number of servers that can be used to execute an operation.Under single server semantics,the same operation can only be executed once at a time,while the same operation can be executed as many times as the number of available servers under infinite server se-mantics.As an extension of single server semantics,this study proposes two efficient heuristic methods for the marking optimization problem of deterministic TWMGs un-der infinite server semantics.These proposed algorithms can provide a near optimal solution step by step and also apply for the marking optimization of deterministic TWMGs under single server semantics by adding to each transition a self-loop place with one token.3.The cycle time optimization of deterministic TWMGs under single server semantics is originally studied in this research,which consists in finding an initial resource assign-ment to maximize the system's throughput while the cost of resources is less than or equal to a given value.We prove that a TWMG under single server semantics can be transformed into a series of equivalent timed marked graphs(TMGs)under the condi-tion that the initial marking is not given.Hence the problem to determine an optimal initial marking for a TWMG can be converted to determining an optimal initial mark-ing for a series of equivalent TMGs.A practically efficient algorithm is developed to solve the optimization problem based on solving a series of mixed integer linear pro-gramming problems(MILPPs),which can guarantee the convergence to the optimum.Finally,this approach is further extended to a generalized optimization problem which maximizes the system's throughput and minimizes the cost of the resources.4.Based on previous results,the cycle time optimization of deterministic TWMGs under infinite server semantics is studied.We consider the transformation of a given TWMG into an equivalent TMG under infinite server semantics and prove that this transfor-mation is periodical with regard to the initial marking.This allow us to transform a TWMG into a finite family of equivalent TMGs,each one valid for a partition of set of initial markings.Then,we present an MILPP to solve the optimization problem that requires finding an optimal allocation for the equivalent TMG under the constraint that the initial marking belongs to a particular partition.However,this procedure has a high computational complexity due to the fact that the number of partitions can in-crease exponentially with the number of places.In order to reduce the computational complexity,two sub-optimal approaches are proposed without enumerating the entire partitions.Finally,conclusions and future studies on performance evaluation and optimization for AMSs are prospected. |
PDF Full Text Request