Você está na página 1de 11

Int J Adv Manuf Technol (2005) 25: 10781088 DOI 10.

1007/s00170-003-1939-9

ORIGINAL ARTICLE

N. Baskar P. Asokan R. Saravanan G. Prabhaharan

Optimization of Machining Parameters for Milling Operations Using Non-conventional Methods

Received: 10 June 2003 / Accepted: 27 August 2003 / Published online: 25 March 2004 Springer-Verlag London Limited 2004 Abstract In this paper, optimization procedures based on the genetic algorithm, tabu search, ant colony algorithm and particle swarm optimization Algorithm were developed for the optimization of machining parameters for milling operation. This paper describes development and utilization of an optimization system, which determines optimum machining parameters for milling operations. An objective function based on maximum prot in milling operation has been used. An example has been presented at the end of the paper to give a clear picture from the application of the system and its efciency. The results are compared and analysed using the method of feasible directions and handbook recommendations. Machining cost, cost of raw material per part, cost of a cutting tool ($) Cu Unit cost ($) d Cutter diameter (mm) e Machine tool efciency factor F Feed rate (mm/min) f Feed rate, (mm/tooth) Fc , Fc(per) Cutting force, Permitted cutting force (N) FF , FR , FT Feed, radial and tangential forces resulting from all active cutting teeth (N) G, g Slenderness ratio, exponent of slenderness ratio. K Distance to be travelled by the tool to perform the operation (mm) K i (i = 13) Coefcients carrying constant values Kp Power constant depending on the workpiece material la Lead (corner) angle of the tool m Number of machining operations required to produce the product N Spindle speed (rev/min) n Tool life exponent P, Pm Required power for the operation, motor power (kW) Pr Total prot rate ($/min) Q Contact proportion of cutting edge with workpiece per revolution R Sale price of the product excluding material, setup and tool changing costs ($) Ra , Ra(at) Arithmetic value of surface nish, and attainable surface nish (m) Sp Sale price of the product ($) T , Tu Tool life (min), Unit time (min) tm , ts , ttc Machining time, set-up time, tool changing time (min) V , Vhb , Vo pt Cutting speed, recommended by handbook, optimum (m/min) w Exponent of chip cross-sectional area W Tool wear factor z Number of cutting teeth of the tool cm , cmat , ct

Keywords Ant colony algorithm Genetic algorithm Optimization Multi-tool milling Particle swarm optimization algorithm Tabu search algorithm Nomenclature A Chip cross sectional area (mm2 ) a, arad Axial depth of cut, radial depth of cut (mm) C Constant in cutting speed equation ca Clearance angle of the tool (degrees) Ci (i = 18) Coefcients carrying constants values cl , co Labour cost, overhead cost ($/min)
N. Baskar (u) School of Mechanical Engineering, Shanmugha Arts, Science, Technology & Research Academy (SASTRA, Deemed University) Thanjavur, Tamilnadu, India, Pin 613 402 E-mail: baskarnaresh@yahoo.co.in P. Asokan G. Prabhaharan Department of Production Engineering, Regional Engineering College, Tiruchirappalli, India, Pin 620 015 R. Saravanan Department of Mechanical Engineering, J.J. College of Engineering & Technology, Tiruchirappalli, India, Pin 620 009

1079

1 Introduction
Optimum machining parameters are of great concern in manufacturing environments, where economy of machining operation plays a key role in competitiveness in the market. Due to high capital and machining costs of Numerical Control (NC) machines, there is an economic need to operate NC machines as efciently as possible in order to obtain the required pay back. Since the cost of machining on NC machines is sensitive to the machining parameters, optimal values have to be determined before a part is put into production. A number of researchers have dealt with the optimization of machining parameters, considering only turning operations and graphical methods to determine the optimum speed and feed [15]. Few researchers concentrated on multi-tool machining operations and solved by a constrained mathematical programming methods [6] (S.S. Rao). Recently different methods have been reported in the literature to optimize machining parameters of face milling operations. These methods include Genetic Algorithm (GA) [7] using real number coding, and Genetic Algorithm (GA) using binary coding, Simulated Annealing (SA) algorithm and hybrid algorithm [8]. Some researchers optimized machining parameter for lathe operations and compared results with conventional and non-conventional methods [16]. Many researchers optimized machining parameters based on multi-pass milling operation via geometric programming and dynamic programming [9]. A multi-tool machining analysis was carried out by some researchers [10, 11]. However it is obvious that the real optimum values of machining parameters cannot be achieved without considering the objective function and constraint equations. A direct search procedure was employed by some researchers [12]. In multi-tool milling operation, the optimum values of cutting conditions are searched for each tool and each pass. Each pass for each tool is optimized individually. Wang and Armarego [13] studied a method to optimize the machining parameter for cutting operations. However, this work was restricted only to the face milling operation. Some researchers considered power as the only constraint and ignored, other constraints, e.g., cutting force, tool life and surface nish. Signicant work has been made on the optimization of milling operations for multi-tool application [14] and CAD-CAM integration for milling operations has been developed using optimum machining conditions. The problem was solved by direct search method. Recently, optimization of machining parameters for the milling operation was carried out by authors using Particle Swarm Optimization (PSO) [15] algorithm. In the present work, a workpiece machined by CNC machine centre is evaluated according to the rules of the optimization problem. Different optimization procedures have been developed using non-traditional approaches such as Genetic Algorithm (GA), Tabu Search (TS) algorithm, Continuous Ant Colony Optimization (CACO) algorithm and Particle Swarm Optimization (PSO) algorithm to optimize machining parameters for milling operations. Three objective functions (minimum production cost, minimum production time and maximum prot

rate) have been considered for both single tool and multi-tool operations. Optimum machining parameters resulting from this work are intended for use by NC machines in order to improve machining efciencies. However, they can also be used by ordinary machines. In this multi-tool operation, the optimum values are obtained for each tool and each pass. The depth of cut is taken as the maximum permissible depth for given work piece cutting tool combination. Each pass for each tool is optimized individually. Finally an example taken from the literature [14] is used for comparing the results obtained by different approaches.

2 Mathematical model
Depth of cut, feed rate and cutting speed have the greatest effect on the success of the machining operation. Depth of cut is usually predetermined by the workpiece geometry and operation sequence. It is recommended to machine the workpiece with the required depth in one pass to keep machining time and cost low, when possible. Therefore the problem of determining the machining parameter is reduced to select the proper cutting speed and feed rate combination. The mathematical model developed by M. Tolouei-Rad et al. [14] is considered in this work. 2.1 Objective function The main focus of this work is to maximize the total prot rate and can be determined by Pr = S p Cu . Tu (1)

The unit cost can be represented by


m

Cu = cmat + (cl + co )ts +


i=1 m

(cl + co )K 1i Vi1 fi1


m

+
i=1

cti K 3i Vi

(1/n)1 [(w+g)/n]1 fi +

(cl + co ).
i=1

(2)

The unit time to produce a part in the case of multi-tool milling can be dened by
m

Tu = ts +
i=1

K 1i Vi1 f i1 +

ttci .
i=1

(3)

2.2 Constraints In practice, possible range of cutting speed and feed rate are limited by the following constraints 1. 2. 3. 4. Maximum machine power Surface nish requirement Maximum cutting force permitted by the rigidity of the tool Available feed rate and spindle speed on the machine tool

1080

2.2.1 Power The machining parameters should be selected such that maximum machine power is used. The required machining power should not exceed available motor power. Therefore the power constraint can be written as C5 V f 0.8 1. Where 0.78K p Wzarad a 0.8 Vf . C5 = 60dePm 2.2.2 Surface nish The required surface nish Ra , must not exceed the maximum attainable surface nish Ra(at) under the conditions. Therefore the surface nish for end milling becomes C6 f 1, where C6 = 318[tan(la) + cot(ca)]1 Ra(at) (7) (6) (5) (4)

2.2.5 Feed rate limits 1. 2. 3. 4. 5. Face milling: 0.050.4 mm/tooth Corner milling: 0.050.5 mm/tooth Pocket milling: 0.050.5 mm/tooth Slot milling1: 0.050.5 mm/tooth Slot milling2: 0.050.5 mm/tooth

3 Proposed methodology
Most of the researchers have used traditional optimization techniques for solving machining problems. The traditional methods of optimization and search do not fare well over a broad spectrum of problem domains. Recently many authors have been trying to bring out the utility and advantages of non-traditional optimization techniques, such as genetic algorithm, simulated annealing and particle swarm optimization. In this paper it is proposed to use the new evolutionary approach, viz. GA, TS, CACO and PSO for the machining optimization problem. Upon development of the optimization model and its constraints, the appropriate optimization method should be employed to solve the problem. The optimization model adopted in this work is a nonlinear, multi-variable and multi-constrained model of a complex nature. To determine the appropriate optimization method, a comprehensive literature survey has been carried out and different non-traditional methods were tested to achieve acceptable results.

and for end milling C7 f 2 1, where C7 = 318(4d)1 . Ra(at) (9) (8)

4 Case study
The component as shown in Fig. 1 is to be produced using a CNC milling machine. It is desirable to nd the optimum machining parameters, which result in the maximum prot rate. Specications of the machine, material and values of constants are given below. Also, the geometric information on the required operations and tools is presented in Tables 1 and 2. Constants: S p = $25 cmat = $0.50 co = $1.45 per min cl = $0.45 per min ts = 2 min tct = 0.5 min C = 33.98 for HSS tools w = 0.28 C = 100.05 for carbide tool K p = 2.24 W = 1.1 n = 0.15 for HSS tools n = 0.3 for carbide tool g = 0.14 Machine tool data: Type: Vertical CNC milling machine Pm = 8.5 kW, e = 95% Material data: Quality: 10L50 leaded steel. Hardness = 225 BHN

2.2.3 Cutting force The total cutting force Fc resulting from the machining operation must not exceed the permitted cutting force Fc (per) that the tool can withstand. The permitted cutting force for each tool has been considered as its maximum limit for cutting forces. Therefore the cutting force constraints becomes C8 Fc 1, where C8 = 1/Fc (per). 2.2.4 Speed limits 1. 2. 3. 4. 5. Face milling: 60120 m/min Corner milling: 4070 m/min Pocket milling: 4070 m/min Slot milling1: 3050 m/min Slot milling2: 3050 m/min (11) (10)

1081 Fig. 1. An example part

Table 1. Required machining operation Operation No. Operation type Tool Number 1 2 2 3 3 a (mm) K(mm) Ra (m)

5 Applications
The workpiece (Fig. 1) is to be manufactured in a CNC machine. The workpiece includes four machining features: Step, Pocket and two slots. Since different tools have been selected to machine these features, the problem becomes maximization of prot rate for multi-tool operation. The goal is to determine the cutting conditions of each feature so that the part can be machined with maximum prot rate. The problem is solved by GA, CACO, TS

1 2 3 4 5

Face milling Corner milling Pocket milling Slot milling Slot milling

10 5 10 10 5

450 90 450 32 84

2 6 5 1

1082 Table 2. Tools data Tool No. 1 2 3 Tool Type face mill end mill end mill Quality d (mm) 50 10 12 z Price ($) 49.50 7.55 7.55 la ca

Carbide HSS HSS

6 4 4

45 0 0

5 5 5

and PSO Algorithms. The Tables 1 to 4 represent unit cost, unit time, and total prot rate resulting from all the machining operations required to produce the product. These values have been determined based on optimum machining parameters in comparison with those resulting from handbook recommendations and the method of feasible direction.

termined according to the desired solution accuracy. Here 10 bits are chosen for depth of cut and feed rate. The strings (00000000) and (1111111111) would represents the points lower and upper limits for depth of cut and feed value. The GA operates on a generation and consists of three main operations. 1. Reproduction selection of copies of chromosomes proportionate to their tness value 2. Crossover an exchange of selections of the chromosomes and 3. Mutation a random modication of the chromosome. The chromosomes resulting from these three operations, often known as offspring or children, form the next generations population. The process is repeated for a desired number of generations, usually up to a point where the system converges to a signicant well performing sequence. The following parameters are used in the GA Sample Size 30 Crossover probability 0.6 Mutation Probability 0.05 Number of generation 100 6.1.3 Result of the genetic algorithm The following table shows the result of the genetic algorithm for 20 iteration. From the above graph, it is evident that the maximum prot rate is observed at the 20th iteration. The rise of the curve is due to a local optimum. The prot rate is gradually increasing up to end of the iteration. Table 4 represents unit cost, unit

6 Optimization procedure
Upon development of the optimization model and its constraints, the appropriate optimization procedure should be employed to solve the problem. Optimization model developed in this work is a nonlinear, multi-variable and multi-constrained models of a complex nature. To determine the appropriate optimization method, a comprehensive literature survey has been carried out and the methods identied in this work are GA, CACO, TS and PSO algorithms. 6.1 Genetic algorithm (GA) The genetic algorithm is an adaptive search and optimization algorithm that mimics principles of natural genetics. GAs are very different methods used in engineering design and other optimization problems. Because of their simplicity, easy of operations, minimum requirements and global perspective GAs have been successfully used in a wide variety of problem domains. 6.1.1 Algorithm Step 1: Choose a coding to represent problem parameter, a selection operator, a crossover operator and a mutation operator. Choose population size N, crossover probability pc , and mutation probability pm . Initialize a random population of strings of size of size 10. Set it = 0. Step 2: Evaluate each string in the population. Step 3: If it > itmax (or) other termination criteria is satised, terminate. Step 4: Perform reproduction on the population. Step 5: Perform crossover on the random pairs of strings. Step 6: Perform bit-wise mutation. Step 7: Evaluate strings in the new population. Set it = it + 1 and go to step 3. 6.1.2 Implementation of GA Binary coding is used in this work to represent the variables depth of cut and feed rate. The length of the string is usually de-

Table 3. Best GA prot rate Speed (V) (m/min) 119.941 41.642 40.557 31.017 44.311 Feed (f) (mm/tooth) 0.263 0.254 0.400 0.445 0.361 Unit cost (Cu ) ($) Unit time (Tu ) (min) Prot rate (Pr ) ($/min)

9.401

4.229

3.689

Fig. 2. GA graph

1083 Table 4. Comparison of unit cost, unit time and prot rate Sl. no 1. 2. 3. Handbook Method of feasible direction Genetic algorithm Improvement over handbook Improvement over method of feasible direction Improvement (handbook) Improvement (method of feasible direction) Unit cost Cu $18.36 $11.35 $9.401 $8.959 $1.949 48.7% 17.1% Unit time Tu 9.40 min 5.48 min 4.229 min 5.171 min 1.251 min 55% 22.8% Prot rate Pr $0.71 / min $2.49 / min $3.689 / min $2.979 / min $1.199 / min 419% 48%

time and total prot rate resulting from all the machining operations required to produce the product. These values have been determined based on optimum machining parameters and are compared with those resulting from the method of feasible direction and handbook recommendation. 6.2 Ant colony algorithm The nature metaphor on which ant algorithms are based is that of ant colonies. One of the problems which confronted scientists was to understand how almost blind ants could establish the shortest route from their nests to the food source and back. Their results revealed that these ants used their pheromone trails as a medium for communicating among themselves. A moving ant lays various quantities of pheromone on its route. To apply the ant colony optimization methodology for continuous function optimization problems, the domain has to be divided rst into a specic number of regions. The tness of the regions are rst evaluated and stored in descending tness. A total of A number of ants explore these regions and updating of the region is done locally and globally with the local search mechanism respectively. Thus these ants are divided into G global ants and L local ants. In the local search, the local ants have the capability of selecting regions proportional to the current pheromone values of these regions. The age of the region is another important parameter in the CACO algorithm. The size of the ant movement in the local search depends on the current age. The search radius is maximum for zero age and minimum for maximum age, with the variation in radius set to follow a linear relationship. The global search creates G new regions by replacing the weaker portions of the existing one. The natural metaphor on which ant algorithms are based is that of ant colonies. Researchers are fascinated by seeing the ability of the almost blind ants to establish the shortest route from their nests to the food source and back. These ants secrete a substance called pheromone and use its trails as a medium for communicating information among each other. The probability of the trail being followed by other ants is enhanced by increased trail deposition of others following this trail. This co-operative search behaviour of real ants inspired the new computational paradigm for optimizing real life systems and it is suited to solving large-scale optimization problems. ACO has also been applied to other optimization problem like the travelling salesman problem, scheduling etc. More recently, the

modied ACO is made as effective global optimization procedure by introducing bi-level search procedures called local and global search. The important aspect in ACO is that the articial ants select the solution to which they move with a selection probability proportional to the pheromone trial. 6.2.1 Schemes of the ant colony algorithm [1719] The objective function as mentioned previously is maximized subject to practical constraints. In the rst step, 200 solutions are generated randomly within parameter bounds satisfying the constraints. Then global and local search is performed. In this paper, the Continuous Ant Colony Algorithm (CACO) is implemented. The CACO scheme is explained in Fig. 3 and the distribution of ants is given in Fig. 4. 6.2.2 Global search The initial solutions are classied as superior and inferior solutions based on their tness values. Global updating is applied only on inferior regions. The global search for CACO is quite different from other non-traditional optimization techniques. The following three operators are to be performed on the randomly generated initial solution. a. Random Walk. b. Mutation. c. Trail Diffusion. a. Random walk (or) crossover The 90% of the solutions (randomly chosen) in the inferior solutions are replaced with randomly selected solutions from the superior solutions. The distribution of ants, as well as the selection of solutions is illustrated in Fig. 4. b. Mutation After the random walk step, randomly adding or subtracting a value to each variable of the newly created solutions in the inferior region with a probability equal to a suitably dened mutation probability. The mutation size is reduced as per the relation (T, R) = R(1 r (1T)b ),

1084

Fig. 4. Distribution of ants for local and global search

The probability of selecting the third option is set equal to the mutation probability while allotting equal probability of selecting the rst two steps. The trail value of the newly created child solutions is assigned a trail value lying between the values of the original parent solutions.
Fig. 3. Flow chart for ant colony algorithm scheme

6.2.3 Local search In the ACO algorithm, local (articial) ants select a region i with a probability. Pi (t) =
t

where r is a random number from [0, 1], R is the maximum step size, T is the ratio of the current iteration number to that of the total number of iterations, and b is a positive parameter controlling the degree of nonlinearity. The value of b considered in this work is 10, which is arrived at by a trial basis. c. Trail diffusion Trail diffusion, which is another element in global search, is applied on the inferior solutions that were not considered during the random walk and mutation stages. Here, two parents are selected at random from the present parent superior solutions. The variables of the childs position vector can have either 1. the value of the corresponding variable from the rst parent, 2. the corresponding value of the variable from the second parent, or 3. a combination arrived at from a weighted average of the above: x(child) = ().xi (parent1 ) + (1 ).xi (parent2 ) where is a uniform random number in the range [0, 1].

i (t) , k (t)

where i is the region index and i (k) is the pheromone trail on region i at time t. After selecting the region the ant moves through a short distance (nite random increment). The direction of movement is retained if the tness value improvement is observed, otherwise it is reverted. Correspondingly the solutions position vector is updated and the pheromone trail value is improved based on the tness value. The variables of this problem are velocity, feed rate and depth of cut, which can have any continuous value, subject to the limits available. In the continuous algorithm, the pheromone values are decreased after each iteration by: i (t + 1) = . i (t), where is the evaporation rate which is assumed to be 0.2 on a trial basis and i (t) is the trail associated with solution at time t.

1085

6.2.4 Result of the continuous ant colony algorithm The following table shows the result of the continuous ant colony algorithm.
Table 5. Best CACO prot rate Speed (V) (m/min) 117.474 64.721 51.901 46.480 43.931 Feed (f) (mm/tooth) 0.163 0.103 0.222 0.185 0.191 Unit cost (Cu ) ($) Unit time (Tu ) (min) Prot rate (Pr ) ($/min)

these positions may not be repeated while in the active list. The tabu search algorithm starts with an initial current solution obtained using a random function. Then, a neighbourhood structure is selected and all solutions in that neighbourhood, excepting those which are tabu, are searched. The best one obtained through the search is selected as the new current solution. The Tabu list is updated by recording the move that led to this new current solution. The process is continued in this way, until the number of iterations exceeds K max . The best solution is obtained from all the iterations and the overall best is found. 6.3.1 Algorithm 1. Find an initial solution P now by using a random function. Calculate the objective function value of this solution denoted by Z (P now ). Record P best = P now and Z (best) = Z(P now ) set the tabu list T = NULL. Set K = 0, where K denotes the current iteration number. Select maximum number of iteration K max . 2. Determine the neighbourhood N (T, P now ) as a subset of neighbourhood N (P now ) select the best prot rate. P next N (T, P now ) to maximize over N (T, P now ). Choose P next as the new seed and update the tabu list T . 3. If Z (P next ) > Z (best) then P best = P next and Z (best) = Z (P next ). 4. Set P now = P next , also set K = K + 1. Terminate if K > K max , otherwise go to step 2. 6.3.2 Implementation of the TS [20] In this problem, variables f and v are rst coded in some string structures: binary coding strings having 1s and 0s are mostly used. The length of the string is usually determined according to the desired solution accuracy. In the rst iteration, the neighbourhoods are found by a bit-changing method. The 1st bit is changed from the initial string solution for the 1st solution. Similarly, corresponding bits change the other string. If the bit is 1, it is changed to 0 or vice versa. Then, the other neighbourhood cutting speeds and feed rates are found. The corresponding decoded value of speeds and feed rates are calculated. Then the unit cost, unit time and prot rate are calculated.

10.202

5.438

2.721

Fig. 5. CACO graph

From the above graph the maximum prot rate result is observed at the 6th iteration. The corresponding speed, feed rate, unit cost and unit time are obtained in the above table. Table 6 presents the unit cost, unit time and total prot rate resulting from all the machining operations required to produce the product. These values have been determined based on optimum machining parameters and are compared with those resulting from method of feasible direction and handbook recommendation. 6.3 Tabu search method The tabu search describes the neighbourhood search with a list of recent search positions. Tabu comes from the fact that

Table 6. Comparison of unit cost, unit time and prot rate Sl. no 1. 2. 3. Handbook Method of feasible direction Ant colony algorithm Improvement over handbook Improvement over method of feasible direction Improvement (handbook) Improvement (method of feasible direction) Unit cost Cu $18.36 $11.35 $10.202 $8.158 $1.148 44.4% 10.1% Unit time Tu 9.40 min 5.48 min 5.438 min 3.962 min 0.042 min 42.1% 0.76% Prot rate Pr $0.71 / min $2.49 / min $2.721 / min $2.011 / min $0.231 / min 283% 0.92%

1086

Tabu value selection. The maximum prot rate is calculated as tabu and is stored in the tabu list. The string value of the tabu prot rate is taken as the initial string of the next iteration. The tabu is deactivated for the next 8 iterations. It will activate from the 9th iteration. The same procedure is repeated up to the end of the iterations. Then the overall best value is called the overall best tabu value. 6.3.3 Result of the tabu search The following table shows the result of the tabu search.
Table 7. Best tabu search prot rate Speed (V) (m/min) 80.469 56.510 62.522 40.616 32.952 Feed (f) (mm/tooth) 0.398 0.102 0.354 0.295 0.432 Unit cost (Cu ) ($) Unit time (Tu ) (min) Prot rate (Pr ) ($/min) Fig. 6. Tabu search graph

11.295

4.231

3.240

approaches that can be used across a wide range of applications as well as for specic applications focussed on a specic requirement. 6.4.1 Algorithm 1. Initialize each particle. 2. Calculate the tness value for each particle. If the tness value is better than the best tness value (pbest) in history. Set the current value as the new pbest. 3. Choose the particle with the best tness value of all the particles as the gbest. 4. For each particle, calculate the particle velocity according to the equation v[] = v[] + c1 rand() (pbest[] present[]) + c2 rand() (gbest[] present[]) present[] = present[] + v[], where v[] is the particle velocity, present[] is the current particle (solution) and rand() is a random number between (0,1). c1, c2 are learning factors. Usually c1 = c2 = 2. 5. Particle velocities in each dimension are clamped to a maximum velocity Vmax . If the sum of acceleration would cause the velocity on that dimension to exceed Vmax , which is a parameter specied by the user, the velocity in that dimension is limited to Vmax .

From the above graph the maximum prot rate and the optimum machining parameters are obtained in the 8th iteration. The prot rate is gradually increasing and decreasing in continuous iteration. Table 8 presents unit cost, unit time and total prot rate resulting from all the machining operations required to produce the product. These values have been determined based on optimum machining parameters and are compared with those resulting from the method of feasible direction and handbook recommendation.

6.4 Particle swarm optimization (PSO) [21] The particle swarm optimization is an evolutionary computation technique inspired by behaviour of bird ocking and sh schooling. The PSO is similar to GA, i.e., a population-based optimization tool. The system is initialized with a population of random solutions and searches for optima by updating generations. In PSO, the potential solutions, called particles, are own through the problem space by following the current optimum particles. Particle swarm optimization has been used for

Table 8. Comparison of unit cost, unit time and prot rate Sl. no 1. 2. 3. Handbook Method of feasible direction Tabu Search Algorithm Improvement over handbook Improvement over method of feasible direction Improvement (handbook) Improvement (method of feasible direction) Unit cost Cu $18.36 $11.35 $11.295 $7.065 $0.055 38% 0.48% Unit time Tu 9.40 min 5.48 min 4.231 min $5.169min $1.249min 54.9% 22.7% Prot rate Pr $0.71 / min $2.49 / min $3.240 / min $2.53 / min $0.75 / min 356% 30%

1087

6.4.2 Numerical illustration of PSO Calculation of cutting speed. Cutting speed is calculated randomly using the following formula: for face milling, pocket milling, slot 1 and slot 2 milling. Cutting speed = Vmin + (Vmax Vmin )rand() Similarly calculate the cutting feed Cutting feed = f min + ( f max f min )rand() These values are substitute in the cost equation, time equation and prot rate equation and also calculated prot rate for corresponding speed and feed. Calculation of Pbest balue. The maximum prot rate for each initial solution is consider as the Pbest value. This is the best value of the particular solution only. Calculation of Gbest value. The maximum prot rate for the initial solution or whole iteration is considered as the Gbest value. Calculation of rst iteration. For cutting speed V [] = C1rand(){Pbest[] Present[]} + C2rand(){Gbest[] Present[]}, where present [] is the cutting speed value for the rst particle of the initial random solution. Present[]1 = present[] + V []
The speed value is replaced by present[]1. Similarly the speed and feed values of the other milling operations are calculated.

6.4.3 Result of particle swarm optimization The following table shows the result of particle swarm optimization
Table 9. Best PSO prot rate Speed (V) (m/min) 119.367 58.533 40.006 37.524 39.702 Feed (f) (mm/tooth) 0.400 0.307 0.406 0.271 0.362 Unit cost (Cu ) ($) Unit time (Tu ) (min) Prot rate (Pr ) ($/min)

9.316

4.089

3.836

Fig. 7. PSO graph

Pbest value of the rst iteration. The Pbest value is the result of combination of the initial and rst iteration values. Gbest value of the rst iteration. The Gbest value is the overall maximum prot value of the initial and rst iteration. Gbest value of overall iteration. PSO has taken 45 iterations but the Gbest value is obtained at the 31st iteration.

From the above graph, it is evident that the maximum prot rate is obtained at the 31st iteration. The prot rate is gradually increasing up to the 31st iteration. Then the prot rate is constant for further iterations. Table 10 represents unit cost, unit time and total prot rate resulting from all the machining operations required to produce the product. These values have been determined based on optimum machining parameters and are compared with those resulting from the method of feasible direction and handbook recommendation.

7 Results and discussion


The Table 11 shows the result of the continuous ant colony algorithm, genetic algorithm, particle swarm optimization and tabu

Table 10. Comparison of unit cost, unit time and prot rate Sl. no 1. 2. 3. Handbook Method of feasible direction Particle swarm optimization Improvement over handbook Improvement over method of feasible direction Improvement (handbook) Improvement (method of feasible direction) Unit cost Cu $18.36 $11.35 $9.316 $9.044 $2.034 49% 17.9% Unit time Tu 9.40 min 5.48 min 4.089 min 5.311 min 1.391 min 56.5% 25% Prot rate Pr $0.71 / min $2.49 / min $3.836 / min $3.126 / min $1.346 / min 440% 54%

1088

search for the input values of cutter diameter, cutting length of workpiece, number of machining operation and number of cutting teeth of the tool. The method proposed in the present work based on the PSO algorithm always yields optimal result as compared to the other methods.

PSO for solving the optimization problem of the milling operations. Results obtained in this work are intended for use by NC machines. However they can also be used for manually operated machines. It is also observed that the procedures described in this work can be applied to similar machining operations as well as other nonlinear optimization problems.

Table 11. Comparison of CACO, GA, PSO and tabu search Variable/ Algorithm Speed (V) (m/min) Ant colony Algorithm 117.474 64.721 51.901 46.480 43.931 0.163 0.103 0.222 0.185 0.191 10.202 5.438 2.721 Genetic Algorithm 119.941 41.642 40.557 31.017 44.311 0.263 0.254 0.400 0.445 0.361 9.401 4.229 3.689 Particle swarm Optimization 119.367 58.533 40.006 37.524 39.702 0.400 0.307 0.406 0.271 0.362 9.316 4.089 3.836 Tabu Search 80.469 56.510 62.522 40.616 32.952 0.398 0.102 0.354 0.295 0.432 11.295 4.231 3.240

References
1. Brewer RC, Reuda RAA (1963) A simplied approach to the optimum selection of machining parameters. Eng Dig 24(9):131151 2. Colding BN (1969) Machining economics and industrial data manuals. Ann CIRP 17:279288 3. Ermer DS (1971) Optimization of the Constrained machining economics problem by geometric programming. Trans ASME J Eng Ind 93:10671072 4. Lwata K, Murotsa Y, Jwotsubo T, Fuji S (1972) A probabilistic approach to the determination of the optimum cutting conditions. Trans ASME J Eng Ind 94:10991107 5. Gopalakrishnan B, Faiz AK (1991) Machining parameter selection for turning with constraints: an analytical approach based on geometric programming. Int J Prod Res 29:18971908 6. Rao SS, Hati SK (1978) Computerized Selection of Optimum Machining Conditions for a job Requiring multiple operations. Trans ASME J Eng Ind 100:356362 7. Shanmugham MS, Bhaskara Reddy SV, Narendran TT (2000) Selection of Optimal Conditions in Multi-Pass Face Milling using a genetic algorithm. Int J Mach Tool Manuf 40:401414 8. Baskar N, Asokan P, Saravanan R, Prabaharan G (2002) Selection of Optimal conditions in Multi-Pass Face Milling using Non Conventional Methods. Proceedings of the 20th All India Manufacturing Technology, Design and Research Conference 9. Ihsan Sonmez A et al. (1999) Dynamic optimization of multipass milling operations via genetic programming. Int J Mach Tool Manuf 39:297320 10. Zompi A, Levi R, Ravig Nani GL (1979) Multi-Tool Machining Analysis, Part I. Tool Feature Patterns Implications 101:230236 11. Ravignani GL, Zompi A, Levi R (1979) Multi-Tool Machining Analysis, Part 2. Economic Evaluation in view of Tool life Scatter 101: 237240 12. Cakir MC, Gurarda A (2000) Optimization of machining conditions for multi-tool milling operations. Int J Prod Res 38:35373552 13. Wang J, Armarego EJA (1995) Optimization Strategies and CAM software for multiple constraint face milling operations. 6th Int. Conference on Manufacturing Engineering (ICME95), 29 Nov1 Dec; Melbourne, Australia, pp 535540 14. Tolouei-Rad M et.al (1997) On the optimization of machining parameters for milling operations. Int J Mach Tool Manuf 37(1):116 15. Baskar N, Asokan P, Saravanan R, Prabaharan G (2003) Optimization of machining parameters for Milling operations using Particle Swarm Optimization algorithm. Proc MOSIM 2003, D1321 16. Saravanan R (2001) Optimization of operating parameters for CNC manufacturing systems using conventional and non-conventional techniques (GA). PhD thesis, Regional Engineering College, Bharathidasan University, Tiruchirappalli 17. Jayaram VK et al. (2000) Ant colony frame work for optimal design and scheduling of batch plants. Int J Comput Chem Eng 24:19011912 18. Dorigo M et al. (1996) The Ant system: Optimization by a colony of cooperating agents. IEEE Trans Syst, Man and cybernetics-part B 26(1):113 19. Dorigo M et al. (1997) Ant Colony System: A Cooperative learning approach to the traveling Sales man Problem. IEEE Trans Evol Comput 1:5366 20. Glover F (1990) Tabu Search. Part II ORSA J Comput 2:432 21. Kennedy J et al. (1995) Particle swarm optimization. Proc. IEEE Intl Conf on Neural Networks, IV, 19421948

Feed(f) (mm/tooth)

Unit cost (Cu ) ($) Unit time (Tu ) (min) Prot rate (Pr ) ($/min)

Fig. 8. Comparison graph

8 Conclusion
In this work, different non-conventional optimization techniques such as GA, CACO, TS and PSO algorithm have been studied for the optimization of machining parameters in milling operations. Optimization procedures based on the above approaches have been developed and are successfully implemented. Signicant improvement is obtained with the above techniques in comparison to the results by handbook and method of feasible direction. This is because of the limitations of enumerative search followed by the direct search method. In the use of non-conventional techniques, PSO always yields better results. So it is suggested to use

Você também pode gostar