María Ángeles Rodríguez1, Pedro Gomez-Gasquet1, Llanos Cuenca1 and M. M. E. Alem
María Ángeles Rodríguez1, Pedro Gomez-Gasquet1, Llanos Cuenca1 and M. M. E. Alemany1
1 Research Centre on Production Management and Engineering (CIGIP), Universitat Politècnica de València, Camino de Vera S/N, 46022 València, Spain.
<EMAIL_ADDRESS><EMAIL_ADDRESS><EMAIL_ADDRESS><EMAIL_ADDRESS>1. Introduction
For almost 80 years researchers have been analyzing problems related to production scheduling. It is well known that a large part of the problems addressed can be classified as NP-Complete, which often leads researchers to propose resolution methods that do not guarantee optimal solutions, such as heuristics or metaheuristics. The validation of a new algorithm should be accompanied by a design of experiments (DoE) that will be supported with statistical data related to the measure or measures that are used for its evaluation as is highlighted in [1][2]. This experimental study can have a double objective. On the one hand, carry out a parametric adjustment of the pro-posed method, since in most cases there are one or more parameters whose values are difficult to specify. On the other hand, they would allow algorithms to be compared with each other, with the idea of establishing a classification against a well contextualized problem. This paper propose a service that allows you to create experiments for a wide group of production scheduling problems, apply them to algorithms already available or to your own, and obtain the results when they are available without waiting and without consuming your own resources. Users of this service can use a class written in Python language. This allows initialize the problem, as first step in each execution of the experimenter as:
- Type: Machine, flowshop o job shop problem. Allows adapted some calculation methods to make them more efficient.
- And the following True/False parameter to define the problem context. Multi (True if a set of machines μij ⊆ {M1, . . . , Mm}where m>1), Setup (True if setup is considered), Ready (True if any machine is not available at the beginning), Release (True if any job is not available at the beginning), Duedate (True if due date is considered for jobs) and Weight (True if weights are considered for jobs).
2. Production Scheduling Web Application for Experiments
The web application (http://niotome.cigip.upv.es/) has been designed and implemented according to Web information system development methodology (WISDM) [3]. the web application is mainly designed to create experiments for a wide group of production scheduling problems, apply them to algorithms already available or to your own, and obtain the results.
This translates into a basic functionalities that are obtained due to the user requirements, which are described as follows:
- Initialize the problem (between a wide range of workshops).
- Instantiation of a concrete problem with data (randomly generated by machine or uploaded by user).
- Calculation of a schedule of the instantiated problem.
- Calculation any of the obje