ECMM136 Systems Analysis in Engineering 2021
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
ECMM136
2021
Systems Analysis in Engineering
SECTION A
Answer ALL the questions in this section.
Question 1 (5 marks)
In the context of sensitivity/uncertainty analysis for optimisation problem,
(a) describe at least one reason for converting the original primal problem into its “dual” form.
(2 marks)
(b) Convert the following primal optimisation problem into its “dual” form:
Maximise = 11 + 32 + 53 + 74
Subject to
21 + 42 + 63 + 84 ≤ 100
31 + 62 + 92 + 124 ≤ 150
41 + 82 + 123 + 164 ≤ 200
1, 2, 3, 4 ≥ 0
In the dual form, clearly identify the dual objective function, dual constraints and dual decision variables.
(3 marks)
Question 2 (5 marks)
In the context of Artificial Neural Network,
(a) the “Confusion Matrix” for a classification problem is given by Table Q2 below. Calculate the accuracy of the classifier.
|
Actual Class |
||
Car |
Van |
||
Predicted class |
Car |
9 |
5 |
Van |
1 |
15 |
Table Q2: Confusion matrix
(3 marks)
(b) Provide at least two statements based on the above “Confusion Matrix” .
(2 marks)
Question 3 (15 marks)
A generic “Single Layer Perceptron” (SLP) with 3 inputs is used for classification. Table Q3 below provides 5 known samples and their classification.
Sample |
Input 1 |
Input 2 |
Input 3 |
Classification (output) |
1 |
1 |
2 |
-2 |
-1 |
2 |
2 |
2 |
-3 |
-1 |
3 |
3 |
3 |
-2 |
1 |
4 |
4 |
4 |
-3 |
1 |
5 |
5 |
5 |
-4.9 |
1 |
Table Q3. Samples and its classification
The bias is given by = 0 , while the initial weights for the SLP are 1 = 0, 2 = 1 and 3 = 1, and the learning rate = 0.05. The signum activation function is given by
() = () = {+1 for ≥ 0
(a) Draw the structure of the modified neuron that represents the above classification problem. Clearly label all parts of the neuron (including the inputs, output, weights and activation function).
(5 marks)
(b) Using all samples given in the table above, check if the SLP correctly classifies the output. Otherwise, update the weights using the “Delta rule” .
(Remark: “Delta rule” is given by the following equation):
(+1) = + ( − )
(10 marks)
Question 4 (5 marks)
The variable have a nominal value of 3 and the uncertainty range from 2 to 4. Generate 5 random samples for the variable using the “Latin Hypercube” sampling method.
SECTION B
Answer any TWO out of the three questions in this section.
Question 5 (35 marks)
The following optimisation problem has two minimisation objectives 1 and 2 :
1 = ( − 0.3)2
2 = ( − 0.5)2
subject to
0 ≤ ≤ 1
where is the decision variable.
Assume an initial population with 4 individuals 0.2 , 0.4, 0.6, 0.8. Each individual represents a potential solution for the objective functions 1 and 2 .
(a) Sketch the two functions 1 and 2 on the same graph for the specified range of .
(5 marks)
(b) By using an appropriate encoding, use the ‘random selection’ method to select parents from the population. Then generate 4 children by applying both crossover and mutation operators, with mutation size of ±0. 1 (randomly applied). Evaluate the objective functions based on the combined pool of parents and children.
(6 marks)
(c) Draw a scatter plot and show the Pareto front based on the combined pool of population generated in part (b). Identify the best fitness value and select 4 solutions from the scatter plot into the next generation.
(6 marks)
(d) Continue to produce another generation of children and illustrate the final Pareto- optimal solution in a separate scatter plot. Keep the population size fixed at 4 . Provide detailed explanations for each process.
(13 marks)
(e) Based on the final Pareto front in part (d), discuss the potential minimum value of the two objective functions 1 and 2 and the value of its decision variable . Compare the results with the plot from part (a).
(5 marks)
Question 6 (35 marks)
A car production line receives raw materials each week which need to be processed into two types of cars: Sedan-x and Coupe-y. Both cars yield the same profits to the company. However, there are three production constraints which need to be satisfied, as stated in table Q6 below. The plant engineer must decide how much of each car to produce to maximize profit.
|
Product |
|
|
Resource |
Sedan-x |
Coupe-y |
Resource availability |
Storage |
2 |
3 |
24 |
Production time |
3 |
2 |
24 |
Staff |
2 |
2 |
18 |
Table Q6: Production constraints
(a) In the context of the optimisation problem, describe why the problem above can be solved using Linear Programming optimisation.
(2 marks)
(b) Convert the problem above into the Linear Programming formulation. Clearly
identify the decision variables, objective function and the constraint equations.
(3 marks)
(c) Using the graphical method, identify the feasible optimal solution. Clearly identify all the constraints lines and the feasible region.
(10 marks)
(d) Using the results from part (c), apply the “corner point method” to find the maximum profit and how many of each car the engineer must produce (while satisfying the constraints).
(3 marks)
(e) Using the results from part (c), apply the “iso-profit line” method to find the maximum profit and how many of each car the engineer must produce (while satisfying the constraints).
(12 marks)
(f) Using the results in part (c) to part (e), discuss if the solution to the problem above
is unique. Elaborate your reasoning.
(5 marks)
Question 7 (35 marks)
A simple model of a physical pendulum system is given by the following equation
2 () Equation Q7a
(2 ) 2 + ()() = ()
where the output () is the angle of the pendulum, () is the input torque and = 9.81 /2 is the gravity (constant). The variable is the mass of the pendulum and is the length of the pendulum rod. Both and have a nominal (default) value of 3 respectively, and both have uncertainty range from 2 to 4 respectively.
Assume that the output time response of the system is given by
() = (2 ) (1 − cos (√ ))
Remark: cos(. ) need to be evaluated in radian.
(a) From the context of sensitivity/uncertainty analysis, state whether the system represented in Equation Q7a is an open-loop or a closed-loop system? Justify your answer.
(2 marks)
(b) Use integrator, gains and summation/subtraction blocks, draw a Simulink block diagram that can be used to simulate Equation Q7a. Clearly label all blocks and signals.
(5 marks)
(c) Using the “one-factor-at-a-time” method and a “tornado graph” , determine the order of the sensitivities for the uncertain variables and to the system output response in Equation Q7b when time = 1 second.
(8 marks)
(d) Generate 5 random samples for the uncertain variables and using the “Latin Hypercube” sampling method.
(10 marks)
(e) Evaluate the function based on the 5 “Latin Hypercube” samples generated from part (d) and calculate the mean and variance of the output response in Equation Q7b, when time = 1 second.
(5 marks)
(f) Comment on the sensitivity of the model represented by the function given
above based on the one-factor-at-a-time method from part (c) and Latin Hypercube sampling method from part (e).
(5 marks)
2022-05-23