Machining center efficiency optimization using artificial intelligence.

Dusevic, Hrvoje ; Car, Zlatan ; Barisic, Branimir 等

Abstract: This paper presents machining parameters optimization based on usage of artificial intelligence. To increase efficiency and productivity of machine tool, optimal cutting parameters have to be obtained. In order to find optimal cutting parameters, genetic algorithm (GA) was used as optimal solution finder. GA is optimization algorithm based on artificial intelligence. Optimization has to yield minimum machining time and minimum production cost, while considering technological and material constrains.

Key words: machining parameters, manufacturing, multi-objective optimization, artificial intelligence

1. INTRODUCTION

Today machining process planning has to yield such results that are going to give maximum productivity and ensure economy of manufacturing. Today market has ever changing demands for new products, which requires shorter development cycle. Important part of product development cycle is manufacturing process planning. Shorter time of process planning can lead to usage of machining parameters that are not optimal which can lead to greater cost of production. A human process planner selects proper machining parameters using his own experience and knowledge or from handbooks of technological requirements, machine tool, cutting tool and selected part material. That manual selection can be slow and does not have to give optimal results. To overcome that problem machining process planning went automated by usage of Computer-Aided Process Planning (CAPP) system. CAPP system should be able to automatically choose operation sequence and machining parameters as well as machine tool and cutting tool on dependence of part material. In this paper focus is given on cutting parameters optimization. Cutting parameters, such as cutting depth, number of passes, feed rate and machining speed, have influence on overall success of machining operation (Shunmugam et al., 2000; Tandon et al., 2002). To conduct optimization, mathematical model has to be defined. It is not always easy to define a model that can be expressed by pure analytical functions. Also cutting parameters optimization presents multi-objective optimization problem, so classical mathematical methods such as linear programming would not work with these input data. There is also problem of finding fake optimum that is in fact local extremity. To overcome these problems this paper shows usage of Genetic Algorithm (GA) in machining process optimization. GA is part of evolutionary algorithms that copies intelligence of nature in order to find global extremities on given function problem. These algorithms are based on stochastic operations. In nature, only entity that is able to adapt to its surrounding is going to survive and transfer it's qualities to next generation. Guided by that idea GA transfers that nature intelligence to mathematical model, where every result represents one entity (Robinson, 2001). Then quality of every entity is measured using goal function. Depending on quality measure of entity, proposed result is kept or deleted. Results or entities that survived selection then are combined by using GA operator. These operators again mimic natural processes of reproduction and mutation. New combined results then are transferred to next generation that now should consists of better results, closer to global optimum. Whole process is terminated when stopping criteria is met and global optimum is found. GA ensures that calculated result is global or near global optimum.

The given example is based on optimization of cutting parameters of face milling and turning process. These two processes are most common operations that are performed on modern CNC machines. In first case the goal function (see Fig. 1.) was minimum production time and on the second case minimal power consumption (Crljenko et al., 2006; Cus & Balic, 2003). First case is applicable when we have requirement for production of smaller series of product, which has to be manufactured in short time. Second case could be used when we want to define machining process which is going to meet cutting power limitations of our manufacturing equipment.

[FIGURE 1 OMITTED]

2. PROBLEM SETUP AND OPTIMIZATION RESULTS

Optimization problem is solved for two most common operations, face milling and turning. For each operation, optimization was carried out by defining two goal functions, minimum production time and minimum power consumptions. Firstly we define mathematical model that describes machining operation as combination of functions whose variables are cutting parameters. Mathematical model consists of tree separated models: 1) tool life model 2) cutting force model 3) cutting power model. Tool life model was integrated in goal function while cutting force and power model represented constrain functions, Fig. 2.

Model input data is machine and tool data, information on part material and shape and time and cost limit of manufacturing. Input data was prepared based on technical data sheets of manufacturer of machine and tool that was chosen. Parameters that consider type of part material is gathered as influence coefficients from analytical experiment data. Model output are cutting parameters which at the end of optimization process satisfy constrain functions and give optimal value of goal function, which is global minimum.

Mathematical model was programmed in MATHLAB [TM]. As base for GA we used Matlab's GA toolbox [TM]. GA toolbox works with any given function that is formatted to have input and output matrix which consisting of input and output data. For GA, standard settings were taken. All settings that considered mathematical model of machining and GA optimization were stored in separate m-file (MathLAB standard). For each depth of cutting optimization process gave optimal cutting parameters. GA converges until stopping criteria is met (see Fig 3).

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

Optimization was carried out for range of cutting depths from minimal to maximum cutting depth that is allowed by chosen cutting tool. In this case cutting depth increase step was 0.5 mm.

For obtained optimal cutting parameters number of passes is decided. In this case binary linear programming optimization (LP) method was the best choice. LP gives number of tool passes that meet requirement of total roughing depth and minimal cutting time. Method outputs binary matrix from which we can read which row of optimization results matrix (see Fig.4) should take cutting parameters to satisfy goal function. In case shown above, LP chose depths of 4mm and 5mm that will give shortest machining time of 16.86s for total roughing depth of 9mm. Whole optimization process can be represented by next block diagram:

[FIGURE 5 OMITTED]

3. CONCLUSION

Optimization of machining process requires great knowledge about cutting process and optimization techniques. Cutting parameters optimization is mathematically hard problem, because of its multi-dimensionality and discontinuity of its functions. Here classical optimization method very easily can produce sub-optimal solution or give not any. New method based on artificial intelligence give us powerful optimization algorithms like GA which overcome most of mentioned problems. Manufacturing process which is running with optimized parameters will have better output with lower producing cost and time needed for production.

4. REFERENCES

Crljenko, D.; Barisic, B. & Car, Z. (2006). Optimization Of Tool Motion On A 2-Dimensional Surface By Means Of The Genetic Algorithm, Proceedings of the 5th International DAAAM Conference Advanced Technologies for Developing Countries-ATDC 2006, Mikac, T.; Katalinic, (Ed.). pp. 159-164, Rijeka, Croatia

Cus F. & Balic J. (2003). Optimization of cutting process by GA approach. Robotics and Computer-Integrated Manufacturing, Vol. 19, No. 1, pp. 113-121(9), ISSN 0736-5845

Robinson, A. (2001). Genetic Programming: Theory, implementation and evolution of unconstrained solution, Available from: http://csclab.ucsd.edu/~alan/genetic/ Accessed: 2007-05-25

Shunmugam M.S.; Bhaskara Reddy S.V. & Narendran T.T. (2000). Selection of optimal conditions in multi-pass face-milling using GA, International Journal of Machine Tools and Manufacture, Vol. 40., No 3., pp. 401-414(14) , ISSN: 0890-6955.

Tandon, V.; El-Mounayri, H. & Kishawy, H. (2002). NC End Milling Optimization Using Evolutionary Computation, International Journal of Machine Tools and Manufacture, Vol. 42, pp. 595-605, ISSN: 0890-6955.

Fig. 4. Optimal cutting parameters for various depths

 Cutting Machining Cutting
Feed speed time power
s [mm/okr] v [m/min] [t.sub.1][s] P[kW]

0,90 603 4,70 16,15
0,90 588 4,72 23,59
0,89 566 4,81 29,99
0,78 518 5,62 29,96
0,72 464 6,40 29,90
0,68 420 7,11 29,99
0,77 327 7,80 29,99
0,70 320 8,44 29,99
0,40 496 9,06 29,96

 Cutting Tool Cutting
Feed force frequency depth
s [mm/okr] [F.sub.z][N] n[[min.sup.-1]] v [mm]

0,90 1608,18 3998 1
0,90 2405,80 3987 1,5
0,89 3178,94 3919 2
0,78 3467,89 3669 2,5
0,72 3864,34 3360 3
0,68 4286,21 3109 3,5
0,77 5509,45 2476 4
0,70 5629,99 2483 4,5
0,40 3621,02 3953 5