Automated Manufacturing Systems With Complex Structure And Their Robust Supervisors

Owing to the mass customization and cruel market competition,automated manufacturing systems(AMSs)have experienced substantial changes.AMSs can significantly benefit enterprises by reducing cost,improving quality,and increasing productivity to satisfy fierce market competition.In general,an AMS consisting of a set of concurrent production routes processes a variety of parts by employing a set of highly autonomous resources such as automated handling devices,buffers,robots,and numerically-controlled machines.To obtain final products,resources and production routes can form a complex interaction.When resources are unreasonably allocated,some parts wait or request resources that are occupied by other parts in the same set.These parts cannot use the resources but remain in their current positions.This leads these parts into a circular waiting state.Thus,an unfavorable situation,i.e.,deadlock,appears and causes the stoppage of the partial or whole system.This can cause enormous economic loss for enterprises.AMSs can be classified as a kind of discrete event systems(DESs),whose dynamics are achieved by concurrently and asynchronously enabling or disabling a series of events in supervisory control theory.By modeling AMSs with finite state automata or Petri nets(PNs),researchers have developed various supervisory control policies to achieve desired objectives such as deadlock-free operations.Over the last three decades,researchers have made remarkable achievements in handling deadlock issues in AMSs.However,the majority of control policies are based on the assumption that allocated resources do not fail.Practicing manufacturing researchers know that the failures of resources are a common problem in a practical system,which are derived from a variety of causes,including component malfunctions,part defects,sensor faults,and tool breakages.When any resource fails to work,the existing deadlock control policies may not be applicable to the changed system.The controlled system cannot operate until the failed resource is recovered.In practice,this case can give rise to catastrophic results in large-scale AMSs.Therefore,excluding lower level control problems in systems' behavior,the robustness is another important issue in the control of a system.A system is referred to be robust if an AMS can still process and produce parts even if some resource failures occur.According to whether they are subject to failures,resources are divided into reliable and unreliable resources.It is necessary for researchers to develop robust supervisory control policies for AMSs with unreliable resources such that the controlled systems can accommodate unpredicted resource failures.Because of considerable concurrency,compactness,and constructibility,PNs are used to model the considered AMSs in this thesis.This thesis focuses on investigating robust deadlock control issues for AMSs with complex structure by using PNs.The main results of the thesis are summarized as follows:1.This work focuses on AMSs with assembly operations and multiple unreliable resources to address the problem of resource failures such that the systems can continue to operate smoothly even if some unreliable resources fail.At each progressing stage of AMSs,multiple types of multiple quantity resources are allowed to be acquired.First,based on the minimal resource requirements of processes,a resource capacity constraint policy is proposed,which is used to conduct resource allocation in AMSs such that the available resources are always sufficient to support processes not using any failed resources to operate.Second,by combining the above resource capacity constraint policy with a look-ahead control strategy,a robust supervisory control policy is synthesized for AMSs with assembly operations allowing resource failures.The control objective is to advance parts requiring failed resources in their remaining routes into a special position so as to release shared resources in case some unreliable resources fail to work.Consequently,those parts not necessarily requiring any failed resource can keep progressing all the time.2.The existing robust supervisory control policies only are applicable to either AMSs with assembly operations or AMSs with flexible routes,while there is no method to address both of them.Therefore,this work develops a robust supervisory control policy for a class of AMSs with both assembly operations and flexible routes to synthesize a robust supervisor.It can control resource allocation and route choice such that the stagnant parts requiring the failed resources in their remaining routes cannot affect the continual operations of parts not requiring the failed resources in their remaining routes.To improve the system's permissiveness compared with the above method,this work first proposes an improved shared resource constraint policy which can allow parts requiring the failed resource in their remaining routes to distribute among the buffers of shared resources.Next,by combining the shared resource constraint policy with a look-ahead control strategy,a robust supervisory control policy is synthesized to control resource allocation such that parts requiring the failed resource in their remaining routes stop at current positions and parts not requiring the failed resource in their remaining routes can continue to be processed when any unreliable resource malfunctions.3.In an AMS,each resource is modeled as a workstation consisting of a machine and some buffers.A resource failure implies that the machine is not working,while the buffers can still store parts.Based on the way of modeling system resources,this work studies the robust deadlock issue for AMSs with an unreliable resource in an offline way.The considered system allows multi-quantity and multi-type of resource acquisitions at each processing state.First,based on the above shared resource constraint policy,a set of linear constraints denoted by a set of inequalities are developed.Each acquired inequality is designed a monitor with its control variable computed by an integer linear programming problem to control the token distribution in a set of places.Second,by adding the obtained monitors to the original system,an augmented system net is obtained.By enumerating all strict minimal siphons(SMSs)in the augmented system net,a monitor is designed for each SMS to ensure the system's liveness.Thus,one can synthesize a robust prevention supervisor,which can guarantee that the controlled system can operate smoothly when an unreliable resource works well,and parts that do not require any failed resource in their remaining route can continue to be processed even if an unreliable resource malfunctions.4.Each token in each resource of an AMS is modeled as a machine.A resource failure can only remove the current failed machine while the remaining machines can continue to process parts.Based on the manner to model resources,this work studies a robust supervisory control issue in AMSs with multiple unreliable resources.By analysis,the prior research is based on the enumeration of either siphons or perfect resource transition circuits whose number exponentially increases with the system scale.This means that the synthesized supervisor has a much complex structure.Thus,by using a deadlock detection strategy,the thesis proposes a new robust deadlock control method,which can avoid enumerating siphons or perfect resource transition circuits.Based on a special kind of circuits at a deadlock marking detected by using a set of mathematical formulations,an effective and efficient method is proposed for AMSs with multiple unreliable resources to iteratively control deadlocks such that the controlled system can continue to operate smoothly even if some unreliable resources fail.