The importance of decision support in manufacturing operations has been well documented. Traditional decision support systems (DSS) rely heavily on process models and prior knowledge of the process to provide the decision support to engineers and operators in a manufacturing facility. With recent advances in computers and the advent of artificial intelligence a relatively nascent field of data-based decision support systems has emerged in the manufacturing sector. Atypical decision support system features means of handling large amounts of data from various sources and provides supporting tools such data reconciliation and optimization.
A data-driven DSS emphasizes the access to and the manipulation historical and real-time process data. The analysis of the data is provided by computer software programs that are tailored to a particular operationís needs. Online tools for performing necessary analytical and statistical tasks are developed autonomously and integrated along with graphical tools to provide operators with the necessary support in a real time environment.
In this project, we are using Java Agent Development (JADE) technology to develop multi-agent systems with an aim to provide architecture for real-time decision support to plant operators, engineers, and management. JADE is a software framework developed in JAVA that simplifies the implementation of multi-agent systems through a middle-ware that supports debugging and deployment. Multi-agent systems are a sub-field of soft computing, where in a collection of agents are integrated to perform a task in a particular environment. An agent is an autonomous entity that is characterized by its humanistic properties, specifically belief, desire, and intention. In this project, we propose a novel architecture to develop multi-agent systems that will carry out operational tasks such as data acquisition, regulatory control, process monitoring, data reconciliation, fault detection and diagnosis, and supervisory control. These agents are developed in computing languages such as JAVA and C/C++ and integrated into a multi-agent system on the DCS (Distributed Control System) using JADE. This architecture is aimed at reducing the computation and implementation burden off the DCS while providing decentralized and autonomous algorithms to help an operator make decisions. The execution of these agents also provides a graphical, user-friendly environment for the operators hence reducing the complexity of the decision making.