| Theme of masters work | Subsystem of automated technological complexes work optimization using multiple-way genetic algorithms |
Scientific adviser |
t.s.c. assosiate professor Sekirin Alexandr Ivanovich |
| Faculty | Computer siences and technology |
| Department | Automated control systems |
| Speciality | Information control systems |
Content
1. Paper objectives and tasks
2. Introduction. Justification of paper topicality
3. Multi-objective genetic algorithms
Providing high-performance automated technological complexes of machining and to maximize the hardware capabilities through the use of developed computer subsystem of multiobjective optimization.
• An analysis of methods, models, algorithms and tools for optimizing ATC
• Undertake the selection and justification of the method to optimize the production area, select the key criteria for evaluating the performance of ATC
• Develop a generalized model of ATC
• Develop and implement a software algorithm to produce sub-optimal scheduling of the equipment taking into account multiple criteria using ATC model
In the practice of human activity, whether it's professional field or everyday life, are constantly arising problems of selection, requering a desicion as a result. Only in some cases, the process of selecting (decision making) is intuitive, based on our own experience and common sense, and the solution of more complex problems requires a special approach, as in this case the problem of decision is, in fact, already the optimization problem.
In the process of managing complex technological, organizational and technical systems, one must constantly make difficult decisions associated with the view of many criteria for percolation processes and resource constraints. Without using the possibilities of modern computer technology it is difficult to make the best choice. It is therefore necessary to develop and implement a decision support system.
Need to implement management information systems more acute before the machine-building enterprises. Mechanical engineering is one of the most intensive, complex production industries. The level of it's development determines the level of development of the whole country.
Basis for the creation of computerized manufacturing systems are automated technological complexes (ATC). This is technological structure with a complete production cycle and varying degrees of automation. ATC includes automated equipment for machining parts, as well as auxiliary equipment: automated transport and automated warehouse system.
ATC are complex dynamic objects. To effectively use such a facility an optimal schedule of the equipment workloads should be built and dynamically adjust to changing situations in the production. The real production process affected by a number of external factors:
- disruptions in transportation and warehousing system
- introduction of the new parts to the production
- failures of technological equipment.
To develop a method for optimizing the ATC work model, which will provide close to the reality mapping of the process of the equipment, material and information flows is required
Genetic algorithm - an algorithm that allows you to find a satisfactory solution to the analytically intractable or difficult problems solvable by sequential selection and combination of the desired parameters using the mechanisms that resemble biological evolution.
When using the GA optimization problem is reformulated into the problem of finding the maximum of a function f(x1, x2, …, xn), calles fitness function. It must take no negative values on a limited domain (in order so we can find out fitness for each individual, and it's can not be negative), it does not require continuity and differentiability.
The work of the classical GA consists of three operations: selection, crossingover and mutation. On the figure they are marked as numbers 1, 2 and 3.
Inherent properties of genetic algorithms promote their effective use in solving multiobjective optimization problems, since the GA are based on a set of potential solutions - the population and the global search in several directions. Also, GA does not impose any requirements in the form of objective functions and constraints.
In the application of genetic algorithms for solving a specific problem it is necessary to choose or develop the basic components such as the encoding method of the potential solutions, genetic operators crossover and mutation, the method of selecting parents, to build a fitness function (fitness function), allowing you to evaluate potential solutions.
Since the multiobjective optimization is a natural evolution of conventional numerical or combinatorial optimization problems, many of the developed methods were extended to this more general case. When using GA for multiobjective optimization the central issue is the construction of the fitness function. Over the past decade has developed several approaches.
To optimize the ATC work the method of weighted functions is proposed. It is a natural development of classical optimization methods, where the new "general " objective function is constructed from the defined as a weighted sum:
Here, each objective function fi(x) is assigned a weight wi and the problem reduces to the scalar case. In this case, different weights wi give different solutions in the sense of Pareto.
Given the specifics of the application, the method is optimal, because evaluation criteria have different importance.