This paper discusses optimal control policies to achieve resilience in a class of serial manufacturing networks. Resilience is the ability of a system or enterprise to minimize the effects of a disruption. Although resilient manufacturing enterprises have been investigated in many different areas, there has been little analytical investigation into real-time control policies for resiliency. Our goal is to study the real-time resilient control for manufacturing systems through mathematical analysis. To achieve this, we start with developing a model for a general manufacturing network. Then, the optimal control problem is developed for a simple type of serial network called a Decreasing Storage Cost and Decreasing Capacity (DSCDC) network given disruptions with advance warning. It is then shown that the results of the DSCDC network can be generalized to more general serial networks.

We consider modeling of manufacturing enterprises as networks which process, store, and transport materials between enterprise nodes in order to feed customer demand. Each node within the network is represented as a dynamic model with associated costs of production and inventory. Examples of disruptions for such an enterprise could be weather events, material shortages, equipment disasters, or labor events. Using the dynamic model, we consider major disruptions within this network. We present aggregation methods which can support the analysis to evaluate the impact of these disruptions, and develop control strategies that reduce the impact of the disruption.

A condition system is a collection of Petri nets that interact with each other and the external environment through condition signals. Some of these condition signals may be unobservable. In previous work, fault diagnosis was defined in terms of observed behavior versus expected behavior of subsystem models, where the expected behavior is defined through condition system models, and approximate methods were presented for detection and diagnosis. We have also presented a method to determine a best possible diagnosis within the constraints of observability. However this method requires significant state space exploration. In this paper, we wish to exploit the causal structure imposed on the system by a partition of subsystem models in order to reduce (in certain situations) the amount of work required to perform a diagnosis.

Characterizing uncontrollable reachability is a central issue in forbidden state control of discrete event systems. In this paper, we present methods for building expressions which estimate uncontrollable reachability in a general class of Petri nets and which characterize the control sets which ensure future markings will not be forbidden. These expressions are determined by constructing an abstract syntax tree from an analysis of the Petri net model of the system. We show that these expressions represent bounds that are useful for evaluating uncontrollable reachability and for evaluating control actions.

This paper presents a new way of avoiding and getting around forbidden states in condition models that are used in automated control synthesis of manufacturing systems. The desired high-level behavior of the system is specified by a condition system Petri net called the "Specnet". The forbidden states are defined by a condition system Petri net called "Forbidden condition net". Our manufacturing system is defined by a set of interacting components, which are specified by component models. The high level Spec net is analyzed by a software tool, which converts it to an executable file, which is input to the controller, to control the manufacturing system. In this paper we introduce a new method to avoid and get around forbidden states, by taking the Specnet and the Forbidden condition net as inputs and analyzing the Specnet for the reachability of Forbidden markings, making necessary changes automatically (if needed) in the Specnet, to avoid reaching Forbidden markings, while maintaining the desired high level behavior of the system as defined by the original Specnet.

This paper considers the issues of synthesizing multi-layer state observers for classes of condition systems. A discussion of state observable and condition observable systems is given, and we present methods for constructing multi-layer observers through the analysis of the system component models. An example illustrates the results.

A condition system is a collection of Petri nets that interact with each other and the external environment through condition signals. Some of these condition signals may be unobservable. In this paper, a system failure is defined in terms of observed behavior versus expected behavior, where the expected behavior is defined through condition system models. A. diagnosis of this failure localizes the subsystem that is the source of the discrepancy between output and expected observations. We show that the structure of the interacting subsystems define a diagnostic causal model that captures the causal structure of subsystem dependencies. The diagnostic causal model can then be used to determine a set of subsystems that might be the source of a failure.

Many manufacturing systems are controlled using discrete sensors and actuators. The changes in values of the corresponding control I/O signals are events. The timing and sequencing relationships of these events can be used to determine whether a system is operating as expected, or whether a fault may have occurred. We present a method of learning intervened timing relationships using observations from a correctly operating system. The sample statistics of the observation characteristic of correct system operation are used to create a confidence space of possible timing relationships of the underlying system. Any timing relationship used as a specification of correct observation for fault monitoring will result in some level of false alarms and missed detections among all the possible relationships in the confidence space; the timing relationships can be chosen to minimize the worst case total of the false alarm and missed detection costs over the confidence space. Simulations are used to evaluate the performance of the chosen timing relationship over a range of perturbed systems.

Automated control synthesis methods for discrete-event systems promise to reduce the time required to develop, debug, and modify control software. Such methods must be able to translate high-level control goals into detailed sequences of actuation and sensing signals. In this paper, we present such a technique. It relies on analysis of a system model, defined as a set of interacting components, each represented as a form of condition system Petri net. Control logic modules, called taskblocks, are synthesized from these individual models. These then interact hierarchically and sequentially to drive the system through specified control goals. The resulting controller is automatically converted to executable control code. The paper concludes with a discussion of a set of software tools developed to demonstrate the techniques on a small manufacturing system.

This paper considers the problem of synthesizing a controller to drive a discrete state system through a sequence of desired outputs. The systems considered interact with the controller through discrete condition signals. The controller is constructed from a sequential composition of ActionBlocks, which are control structures to drive the plant system to individual condition outputs. The paper presents a procedure for automatically constructing the ActionBlocks from an analysis of the system model. A class of system models is identified for which the closed loop system achieves the desired goals.

This paper considers the problem of monitoring automated manufacturing systems with discrete input and output signals. In such systems, the timing and sequencing of signal changes (events) can be an indication of whether the process is operating correctly. We introduce condition templates as a model to represent the expected timing and sequencing relationships of these events. Condition templates can be distributed over multiple processors in the manufacturing system, allowing a distributed and robust method of fault monitoring. We present the condition templates monitoring method and develop a technique for deriving condition template models from timed automata models.