Toward Adaptive Manufacturing: Scheduling problems in the Context of Industry 4.0.

Toward Adaptive Manufacturing: Scheduling problems in the Context of Industry 4.0.

Adaptive manufacturing and Cyber-Physical Systems (CPS) have recently emerged as main topics in academia and the industry since recent projects on a governmental level have been launched to investigate and propose innovations under the title “Smart Industry”. This movement has been even expressed as the trigger toward the fourth industrial revolution. The concepts of smart industry represent the future form of industrial network, in which the physical elements of manufacturing environments are coupled with IT-services to achieve a cyber representation of the real manufacturing environments. Industrial environments exhibit naturally high structural complexity since an enormous number of factors related to market changes, heterogeneity of customer demands and the internal operating procedures are highly dynamic. Although achieving a cyber-physical system promises a manufacturing environment with plenty of competitive advantages from a monitoring point of view, extracting knowledge from the massive amount of collected data is an extremely difficult task. In this work, a general overview of the optimization potential of different industry 4.0 concepts will be presented. Eventually, we focus on analyzing the optimization potential of scheduling policies for the management of the shop floor to answer this central question: What is the impact of industry 4.0 concepts on the traditional optimization methods for solving scheduling problems?

In this research, we investigate integrating machine breakdowns from a cyber manufacturing system and their impact on the used optimization techniques for setting scheduling policies in a manufacturing environment. We presented two hybrid algorithms and compared their performance against Genetic algorithms for solving four-stage Hybrid Flow Shop scheduling problem. In the first experiments, we reduced the overall production capacity of the machines by ten percent to take into consideration the machine breakdowns on an aggregated level. Reasons for that are to facilitate a higher quality analysis and comparison between the solution approaches. Then, we applied stochastic breakdowns on the obtained solution and recorded 100 replications to observe the robustness of the solutions against disturbances.

The experiments are dedicated to answering the research questions and shed a light on the required computational effort by the solution approaches if we want to find an optimized schedule under machine breakdowns constraint. The experiments are conducted on eight problem instances. All machines on the four processing stages are subject to the generated stochastic breakdowns. All developed and adopted optimization techniques report penalties in the delivery dates of jobs, even if we consider the machine breakdowns during the optimization. Of course, the obtained solution without considering breakdowns during the optimization is relatively worse in terms of the total tardiness and recorded penalties.

However, the required computational effort for obtaining the solutions witnessed a significant increase since evaluating a solution candidate requires recording at least ten replications. Moreover, a significance outperformance of the hybrid approaches against the Genetic algorithms can be noticed, when machine breakdowns are taken into consideration.Therefore, it is plausible to consider investing more effort in developing hybrid solution strategies to address the new requirements of Industry 4.0 concepts to deal with scheduling policies. Based on the presented analysis, more research effort might be invested in hybrid optimization strategies that combine the strength and robustness of the metaheuristics approaches and the light required computational effort of heuristic procedures. Thus, in the context of Industry 4.0 optimization processes must be conducted on near real-time to support decision-making processes instantly. With the capabilities of the CPS, monitoring data might be analyzed to derive different machine breakdown patterns dependent on product type or specific machine types. Those patterns can be then used during the optimization to propose more accurate scheduling policies. It is still matter of extensive research to decide whether we need to design rapid solution methodologies and conduct optimization periodically using small intervals or we need to design online optimization mechanisms.