Single machine preemptive scheduling Considering Energy Consumption and Predicting Machine failures with Data Mining Approach

Abstract

Prediction of unexpected incidents and energy consumption are some industry issues and problems. Single machine scheduling with preemption and considering failures has been pointed out in this study. Its aim is to minimize earliness and tardiness penalties by using job expansion or compression methods. The present study solves this problem in two parts. The first part predicts failures and obtains some rules to correct the process, and the second includes the sequence of single-machine scheduling operations. The failure time is predicted using some machine learning algorithms includes: Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), Naïve Bayes, and k-nearest neighbors. Results of comparing the algorithms, indicate that the decision tree algorithm outperformed other algorithms with a probability of 70% in predicting failure. In the second part, the problem is scheduled considering these failures and machine idleness in a single-machine scheduling manner to achieve an optimal sequence, minimize energy consumption, and reduce failures. The mathematical model for this problem has been presented by considering processing time, machine idleness, release time, rotational speed and torque, failure time, and machine availability after repair and maintenance. The results of the model solving, concluded that the relevant mathematical model could schedule up to 8 jobs within a reasonable time and achieve an optimal sequence, which could reduce costs, energy consumption, and failures. Moreover, it is suggested that further studies use this approach for other types of scheduling, including parallel machine scheduling and flow job shop scheduling. Metaheuristic algorithms can be used for larger dimensions.

Keywords

• energy
• just-in-time
• machine failure
• single machine preemptive scheduling
• data mining