ME15S098

ASSIGNMENT-I
NAVEEN RAJ.R
Optimization Methods for Mechanical Design (ME7680)
Q.1
The optimum solution by trial and error was found to be,

MANG
O

DATE
0

NUMBE
R OF
MANGO
JUICES

LIME
3

1

9

TOTAL
NO. OF
JUICE
S
13

COST
145

(Pls refer attached excel macro.)

Q.2 You are asked to minimize the stress level induced in a cuboid by a
tensile force. You are free to decide the dimensions of the cuboid at your
will.
Is this a valid Optimization problem? Why?
Ans: This probably is not a valid optimization problem as stress is a function of force and
area. And both the parameter can’t be without bounds. Even if we fix the tensile force and
work on it, by increasing the area under action the stress is going to decrease so without real
bound or other constraints from the physical problem, the problem is not a valid optimization
problem.

Q.3
The function to be optimized is

7773 . clearly the function is not unimodal.7070 3.As we can see. the function has two extremum points.6000 -0.6000 bound_phase = 0. of experiments: 10 enter lower bound: 1 enter upper bound: 10 extended_search = 2. From Matlab.9000 num_steps_for_interval_halving = 8 interval_halving = 3. enter no.8000 4. the restluts obtaine were.

4 According to the question time(y) taken to reach the lecture hall by me will be varying as a function of x(no. Q.6991 3. it dint lie near the minima). while Fibonacci took 10. of idli’s).6875 3.e.fibonacci = 3.still it took a lot of function evaluations as it took 8 iterations to reach a tolerance of 1e-1 The Fibonacci and the gold section methods gave a good result.8125 noum_of_func_calc_in_golsec = 11 gold_sec = 3. The bounding phase method gave values which are not even close as the function is not unimodal and the initial guess point and the initial step size taken were bad enough(i. Y=60*log(x)+500/x^2 .7723 RESULT COMPARISION:     We see that the exhaustive search gave a bigger interval still the minimum point lied inside the interval. And from the given scenario. but golden section method took 11 experiments. The interval halving method gave a smaller interval with the minima in it.

1000 num_steps_for_interval_halving = 8 interval_halving = 4.0544 4.0000 4.1000 8. The interval halving method gave a smaller interval with the minima in it. as the intial guess was poor. Thus the working of bounding phase method relies on the guess value we choose.6000 bound_phase = 2. The bounding phase gives an even bigger interval.still it took a lot of function evaluations as it took 8 iterations to reach a tolerance of 1e-1 .MATLAB OUTPUT: enter no.1250 noum_of_func_calc_in_golsec = 11 gold_sec = 4. of experiments: 10 enter lower bound: 1 enter upper bound: 10 extended_search = 2.1289 fibonacci = 4.0586 4.8000 4.1276 RESULT COMPARISION:    We see that the exhaustive search gave a bigger interval still the minimum point lied inside the interval.

As it gives a better result in fewer number of experiments done.1*x)*(1-(x/100)))]+[1000/((. but golden section method took 11 experiments. According to the problem. Q.1*x))]+[1000/((. while Fibonacci took 10.5 From the above problems we find that FIBONACCI method is better than the others. The Fibonacci and the gold section methods gave a good result. the time taken(y)in seconds . So let us use the same for this problem. Y=[(1000/(. to complete the race is.1*x)*(1-(x/100)^2))] MATLAB OUT PUT: .