Petroleum industry production systems are highly automatized. In this industry, all functions (e.g., planning, scheduling and maintenance) are automated and in order to remain competitive researchers attempt to design an adaptive control... more
Petroleum industry production systems are highly automatized. In this industry, all functions (e.g., planning, scheduling and maintenance) are automated and in order to remain competitive researchers attempt to design an adaptive control system which optimizes the process, but also able to adapt to rapidly evolving demands at a fixed cost. In this paper, we present a multi-agent approach for the dynamic task scheduling in petroleum industry production system. Agents simultaneously insure effective production scheduling and the continuous improvement of the solution quality by means of reinforcement learning, using the SARSA algorithm. Reinforcement learning allows the agents to adapt, learning the best behaviors for their various roles without reducing the performance or reactivity. To demonstrate the innovation of our approach, we include a computer simulation of our model and the results of experimentation applying our model to an Algerian petroleum refinery.
Flexible manufacturing systems as technological and automated structures have a high complexity for scheduling. The decision-making process is made difficult with interruptions that may occur in the system and these problems increase the... more
Flexible manufacturing systems as technological and automated structures have a high complexity for scheduling. The decision-making process is made difficult with interruptions that may occur in the system and these problems increase the complexity to define an optimal schedule. The research proposes a three-stage hybrid algorithm that allows the rescheduling of operations in an FMS. The novelty of the research is presented in two approaches: first is the integration of the techniques of Petri nets, discrete simulation, and memetic algorithms and second is the rescheduling environment with machine failures to optimize the makespan and Total Weighted Tardiness. The effectiveness of the proposed Soft computing approaches was validated with the bottleneck of heuristics and the dispatch rules. The results of the proposed algorithm show significant findings with the contrasting techniques. In the first stage (scheduling), improvements are obtained between 50 and 70% on performance indicators. In the second stage (failure), four scenarios are developed that improve the variability, flexibility, and robustness of the schedules. In the final stage (rescheduling), the results show that 78% of the instances have variations of less than 10% for the initial schedule. Furthermore, 88% of the instances support rescheduling with variations of less than 2% compared to the heuristics.
This paper describes development work on a knowledge based guidance system for the cold rolling of stainless steel strip on a 1670 mm Sendzimir Mill. Both the initial setup and the subsequent on-line control of the rolling mill are... more
This paper describes development work on a knowledge based guidance system for the cold rolling of stainless steel strip on a 1670 mm Sendzimir Mill. Both the initial setup and the subsequent on-line control of the rolling mill are complex and highly interactive. The expert rollers' experience of many physical interactions on the mill has enabled them to amass a number of loosely coupled qualitative and heuristic models relating to the behaviour of various stainless steel grades and the general rolling process. This system, SENDX, is an initial attempt to utilise shallow rule based knowledge of the expert mill operators coupled with deeper physical models from both the rolling theory and metallurgical domains. This turns out to be a useful vehicle for generating new compiled shallow knowledge-which has utility where the knowledge from multiple experts using different approaches can be accommodated. In formalising the overlapping expertise and offering new perspectives derived from the available deep knowledge, a single normalised knowledge base can be developed. This suggests a methodology to fulfil the desire of many complex manufacturing industries: to standardise diverse approaches to the control of complex processes without loosing any valuable expertise.
In today's real-life implementations, projects are executed under uncertainty in a dynamic environment. In addition to resource constraints, the baseline schedule is affected due to the unpredictability of the dynamic environment.... more
In today's real-life implementations, projects are executed under uncertainty in a dynamic environment. In addition to resource constraints, the baseline schedule is affected due to the unpredictability of the dynamic environment. Uncertainty-based dynamic events experienced during project execution may change the baseline schedule partially or substantially and require projects' rescheduling. In this study, a mixed-integer linear programming model is proposed for the dynamic resource-constrained project scheduling problem. Three dynamic situation scenarios are solved with the proposed model, including machine breakdown, worker sickness, and electricity power cut. Finally, generated reactive schedules are completed later than the baseline schedule.
