BITS, Pilani — Dubai Dubai International Academic City, Dubai IV Year (ALL) Second Semester, 2009-2010 Comprehensive Examination Course No: EA C482 Course Title: Fuzzy Logic and Applications Date: 19™' May 2010 Weightage: 40% 3 Hours Max. Marks. 80 | (Answer Parts A and B on separate answer books.) (Answer the questions in the sequential order.) (Answer all the parts of a question together.) PART-A 1. In the speed control of a DC shunt motor system the Series Resistance Rie, Armature Current (J,)and Speed (NN) are expressed in terms of fuzzy sets given below, Find the relation between Resistance (R,.) and Speed (N) using relations computed in part a and part b, using max-product composition. [4M] Rue = (0.3/30 + 0.7/60 + 1/100 + 0.2/120} a= {0.4/40 + 0.6/60 + 1,0/80 + 0.5/100 } and N = {0.2/500 + 0.6/1000 + 1.0/1500 + 0.1/1800} 2) Find the Fuzzy Cartesian product (R) of the Series Resistance (Ree) and Armature Current (I,) [4M] b) Find the Fuzzy Cartesian product (S) of the Armature Current ([,) and Speed (N) [4M] 2. A BAM network (with bipolar input vectors and bipolar target codes) is used to recognize two letters A and C represented using a 5x3 grid. (* represents 1 and empty space -1) @ ®) 2 «|e (a) Letter A —> target code (-1, 1) * * * (b) Letter C — target code (1, 1) a) Find the total weight matrix (4) ) Obtain the response of the net with A and C as the input. [4M] ©) Test the response of the net when it is presented with an noisy version of target code [0 1] [4m]3. a) Outline the steps involved in the Hebb Learning Algorithm for ANN. 4M] b) Determine the weights and bias to classify the 2-dimensional input patterns (letters 'E’ and 'F' ) represented in a 3x3 grid using the Hebb algorithm. Assume bipolar inputs and targets, with initial weights and bias as 0 (zero). [4M] ©) Also draw the diagram for the neural network, for the given input pattern along with the weights. PM] (b) [*|*[+ (@) Letter E — target code (1) *[*l (b) Letter F + target code (-1) * (* represent 1 and empty space -1) PART-B 4. a) A process control systems involves the monitoring of two linguistic parameters temperature (range from 134 to 138 F) and pressure (400 to 900 psi). The parameters are characterized by linguistic values having the following membership functions i) High Temperature = { 0/134 + 0,4/135 + 0.6/136 + 0.8/137 + 1.0/138} ii) High Pressure = {0/400 + 0.2/600 + 0. 4/700 + 0.6/800 + 1.0/900 } Find the membership functions for the following i) Very Very High Temperature ii) Dilation of High Temperature. iii) Intensification of High Pressure. [3x2M] b) For the fuzzy sets A, B, C defined on U={ 1,2, 3, 4, 5, 6 Fas A= {O11 +0.3/2 +0.5/3 + 0,6/4 + 0.4/5 +0.2/6 } B={_ 0.2/2 + 0.6/3 + 0.8/4 + 0.9/5 +0.95/6 } C={ 0.9/1 + 0.8/2 + 0.7/3 + 0.4/4 + 0.5/5 } Verify that the following properties hold i) Distributive law (any one) [aM] ii) Identity Law (for fuzzy set A only) PM] 5. Develop an ADALINE network to realize the OR Boolean function (as shown in the truth table below) using bipolar inputs and bipolar targets. The initial value of weights w ‘wz and bias are 0 (zero), and the learning parameter a is 0.2 Xt x | Target © it 1 1 i 1 =n 1 al a) Find the total error in the first epoch, using the inputs in the same order as shown in the truth table. [8M] b) The weights w1, wa and bias values after 6 epoch are 0.57, 0.64, and 0.50, Find the LMS error. [1M] 26. a) The mass (m) of an object is divided into 3 fuzzy sets Small Mass, Medium Mass and Large Mass, The Universe of Discourse of mass (m) is 0-500 kg. Triangular membership functions are used to represent the fuzzy sets Small, Medium and Large. The range of Small is 0-250 with membership value of 1 at 200kg, the range of Medium is 200- 300 with a membership value of I at 250kg and the range of Large is 250-500 with ‘membership value of 1 at 300kg respectively. 4) Draw the membership function diagram representing the fuzzy sets Small, Medium and Large [BM] ii) Find the fuzzified value at a crisp value of 260 kg in the membership diagram [3M] b) Due to the imprecise nature of measurement devices, the measured data will not be identical. To obtain the accurate value of the measured variable it is decided to take the union of the outputs and employ defuzzification method (Center of Sums). Shown below is the temperature output obtained from two such measurement devices. Find the defuzzified value of the temperature 0123456T o1234S56T ‘Temperature from Device 1 ‘Temperature from Device 2 [6M] 7. Compare the Biological and Artificial Neural network based on the following aspects a) Speed b) Processing c) Fault Tolerance d) Information storage ¢) Energy efficiency [S*2M=10] aaaaeBITS, Pilani — Dubai Dubai International Academic City, Dubai IV Year (ALL) Second Semester, 2009-2010 Test 2 (Open Book) Course No: EA C482 Course Title: Fuzzy Logie and Applications Date: 08" Apr 2010 Weightage: 20% n: 50 minutes Max, Marks. 40 (Answer the questions in the sequential order) (Answer all the parts of a question together) 1, Consider a fuzzy logic based washing machine, which selects the Washing Cycle (WC) based on the Laundry Quantity (1.Q) and Laundry Softness (LS), The LQ and LS are fuzzy inputs, each of which is characterized by 2 triangular fuzzy sets viz. Low and High. The fuzzy output WC is characterized by 3 triangular fuzzy sets viz. Low, Medium and High Low High Low Medium High UX. | 02 4 6 8 we 0 2 4 6 ims The fuzzy rule base is as follows i) if LQ is Low and LS is Low then WC is Low ii) if LQ is Low and LS is High then WC is Medium ili) if LQ is High and LS is Low then WC is Medium iv) if LQ is High and LS is High then WC is High If the LQ is 2.3 and LS is 5.6 , obtain the defirzzified value of the WC. (use max- min method for inference and Mean of Maxima method for defuzzification). [15 Marks} 2. The difference in concentrations (DC) of inlet and outlet streams of CSTR and PFR are given with the membership values and show in the figure below. Find the most representative value (defuzzified value) for the union of CSTR and PFR (which has both the characteristics of CSTR and PER for the fluidized bed design), using controid method. [10 Marks] P.T.O6 pe 3. For 2 inputs x1 and xz, realize the function y= (x AND (NOT x2)) using the McCulloch-Pitt’s neuron model. Initialization of weights and threshold values are to be chosen so that the neuron realizes the function. [5 Marks] 4, For the neural network shown McCulloch-Pitt net shown in figure. a) Define the response of neuron Ns at time # in terms of the activations of the input neurons, N; and N> at the appropriate times, [5 Marks} b) Show the activation of Ns at time t=O that results from the input signal of Ni = Land Np =O. [5 Marks] (Assume that the firing of the neurons will take place at intervals of t-2, -1, 1) eeeeeBITS, Pilani — Dubai Dubai International Academic City, Dubai IV Year (ALL) Second Semester, 2009-2010 Test 1 (Closed Book) Course No: EA C482 Course Title: Fuzzy Logic and Applications Date: 21* Feb 2010 Weightage: 25% 50 minutes Max, Marks, 25 (Answer the questions in the sequential order) (Answer all the parts of a question together) 1. Forthe fuzzy set A defined on U= {1, 2,3, 4, 5,6} as A= 0/1 + 0.1/2 +0,35/3 + 0.4/4 + 0.55/5 + 0.6/6, 2) Find the a-cut decomposition of A (2.0 marks) b) Find the Fuzzy Cardinality of A (2.0 marks) ©) Find the Scalar Cardinality of A (1.0 marks) 2. For the fuzzy set A, B defined on U = (2, 3, 4, 5} as A= 1/2 + 0.5/3 + 0.3/4 + 0.2/5, B= 0.5/2 + 0.7/3 + 0.2/4 + 0.4/5 Find the following a) Complement of A, Complement of B (2.0 marks) b) AUB (2.0 marks) 2 AnB (2.0 marks) 4) Difference A—B and B~ A (2.0 marks) ©) Verify De Morgan’s Law (A \B)'= A'UB! (2.0 marks) 3. The relationship R between temperature and maximum operating frequency depends on various factors for a given electronic circuit. Let T be a temperature fuzzy set (in °F) and F represent frequency fuxzzy set (in MHz) on the following universes of discourse. T={-50,0 50} and F = { 8, 16, 25,33} ‘The fuzzy Cartesian product of T and F results in the following relation R. R 50 0 50 sfo2]or]io] . 16/04] 07 | 08 2s| 06 | 08 | 06 33/09 | 08 | 04 | P.T-0The reliability of the electronic circuit is related to the maximum operating temperature and is expressed using a relation S which is the Cartesian product of Temperature T and Reliability Index M= {1, 2, 4} Ss ae eat ee: so [1 fos] 06] of o7{ t [or] solos | os) 1 Find the relationship between Frequency F and Reliability Index M, using (3.0 marks) a) max-min composition b) max-product composition (3.0 marks) (4.0 marks c) max-average composition (Write the expression for each type of composition with at least I solution step.) eaeeeBITS, PILANI - DUBAI SECOND SEMESTER 2009 - 2010 FOURTH YEAR Quiz 2 Course Code: EA C482 Date: 04.05.10 Course Title: Fuzzy Logic and Applications Max Marks: 14 Duration: 20 minutes Weightage: 7% cesses Sees Prog: Name: Instructions: Wiite your answers in the blank space provided affer each question. I z 1) Give the weight updation rule for MADALINE Neural network 2M) 2) If the net input to an output neuron is 0.64, calculate its output when the activation function isa) binary sigmoidal b) bipoloar sigmoidal [IM + 1M] 3) Identify the type of learning adopted in a) Perceptron NN [1M] b) Hebb NN [iM] [PTO]4) A single layer 2 input I output perceptron neural network, with bipolar inputs, and bipolar output is used to realise the boolean AND function. The weights w1, w2 and bias are 1, 1, 1 respectively. 2) Draw the decision boundary diagram, showing the equation of the boundary. [1M] b) Discuss on linear separability for part-a PM) 5) Fora single layer NN with binary inputs, bipolar targets, without bias and threshold value = 0. a) Will the network converge? [My ») Justify the answer in part -2 ina sentence or two [1M] [PTO]6) ABAM network (with bipolar input vectors and bipolar target codes) is used to recognize two letters L and C represented using @ 3x2 grid, Find the total information stored in the synapses with 2 inputs (3M] (a) () L] («Ik (a) Letter L_— target code (-1, 1) et | l* (b) Letter C — target code (1, 1) aaaesBITS, PILANI - DUBAI Version A SECOND SEMESTER 2008 ~ 2010 FOURTH YEAR Quiz 1 Course Code: EA C482 Date: 10.03.10 Course Title: Fuzzy Logic and Applications Max Marks: 16 ‘Weightage: 8% Duration: 20 minutes «+ Sec/Prog: ... Name: [ instructions: Write your answers in the blank space provided after each question.. Pp, q, }. Draw the 1, Risa fuzzy relation on U x V where U = {a, b, ¢}, V Sagittal diagram for the relation R. pPqer a 1.0 04 05 R= b03 00 07 606 08 02 [2 mart 2. Risa BER onU, where U={ 1, 2,3, 4} 03 03 05 10 R= 04 10 01 06 05 04 09 04 08 07 02 04 a, Write any one membership grade R(x,y) which satisfies the reflexive property. [1 marks] b. Write any one membership grade R(xy) which satisfies the symmetric [1 marks] property. [PTO]3. List down all the possible projections and the corresponding domains of a Ternary Fuzzy Relation T on AxBxC [2 marks} 4, R, a BFR on Ux V, where U= {a,b } and V= (x, y, 2}, is given by R=05/(a, x) + 0.1ay) + 0.9/(a, 2) + 03/(b, x) + 0.8/(b,y) + 0.2b,z) Find projections Rr and Ry [1.0 + 1.0 mark] 5. Fora fuzzy relation R on UxVxWxX, write the expression for projection Ras4 [2 mark] 6. For a fuzzy relation R on UxVxWxX, write the expression for cylindrical extension Ras! [2 mark][PTO] 7. Fora fuzzy relation R on UxVxWxX, write the expression for cylindrical closure of the projections Ri , Ros , Riss [2 marks] 8. Find out whether the given relation R on U = {a, b, ¢} satisfies transitive property. 0.20504 R= 0.00503 0.00504 [2 mark] aaaas