Chapter 1

Maintenance scheduling and planning is very important in modern manufacturing system. In production scheduling, interactions between machines are needed to achieve common goal [1]. Whenever a machine breakdown or need to be serviced, it will alter the current production scheduling as the machine will not be able to continue to process task. Therefore, it is important to finish all machines repairing or maintaining as fast as possible. In a big organization, a lot of machines may need to be repaired and maintained at one time. In this case, it is needed to have a plan called Machine Repair Planning. The planning must take employee and spare part availability constraint into account. Strategic Employee Scheduling was proposed in 2006 to produce detailed daily schedules for individual employees [3]. Without knowing how much time needed to finish every single repair task, we need to consider employee’s skills, experiences and working period to make sure all task are done in a best period. The planning will be more complex and produce a balanced planning for each employee. This has similar properties with ‘nurse scheduling’ [4].

Among the variety if techniques which have been researched for the planning, the interest on AI techniques have received more attention as the have capability to produce optimal or near to optimal solution. Recently, lots of research has been done on strategies finding optimal maintenance schedules. Much less attention is turned to optimizing machine repair plan. However, Machine Repair Planning is important to maintenance service and extra efforts are required to achieve a better solution.

In the context of this paper, we present a approach to produce a Machine Repair Planning that will consider these constraint:

1. Employee's expertise. Each employee defines their own expertise. They have a better chance to work on repair order that have the similar type with their expertise.

2. Employee's performance. The number of time(s) and that each employee repair a specific machine and time needed to repair the machine are saved. This saved data are used to calculate performance of every employee.

3. Employee's group. Employees can be divided into several groups i.e. two groups such as mechanical and electrical.

4. Urgency of repair orders. We defined two type of urgency that are 'urgent' and 'normal'.

5. Spare part availability. Only repair order that have enough spare part needed will be added into our machine repair plan.

6. Group needed to work on each repair order i.e. repair order A needs any employee from mechanical group, repair order B needs any employee from electrical group or repair order C needs any employee from any group.

7. Number of employee(s) to work on each repair order. Each repair order can be done by one to four employee. For repair order that have more than one employee, all of employee are planned to work on the repair order at the same time.

In our approach, we use real-time hybrid Genetic Algorithm(GA) that is combined with Tabu-search algorithm. Machine Repair Planning is a real-time planning as machines can be breakdown or need to be maintained at any time. Although genetic algorithm generally has a long computation time, we use continuous assessment to produce this real-time planning where it is reschedule whenever a new repair order created or deleted, number of available employee changed or spare part constraint occured. In order to reduce the computation time, some of processes are run in parellel. We also use simple machine learning algorithm to produce employee's perfomance data that are useful for our GA. Next is presented how this paper is structed: in section 2, we present a description of the problem being address; in section 3, we present about the concept of real-time hybrid GA; our tabu-search algorithm and machine learning algorithm are described in section 4 and 5, respectively; followed by computational results in section 6; lastly the conclusion in section 7.


============

[1] Maintenance scheduling in manufacturing systems based on predicted
machine degradation
[2] A flexible model and a hybrid exact method for
integrated employee timetabling and production
scheduling
[3] Strategic Employee Scheduling
[4] Bard J.F., Purnomo H.W., Preference Scheduling for Nurses using Column Generation,
European Journal of Operational Research 164: 510 – 534, 2005.
[5] real time genetic scheduling of aircraft landing times