EE 765: Assignment 2

Reet Singh Solanki 21d070054

September 12, 2024

Question 1

a. Percentage yield of the PCB making process

 
Number of Healthy Microprocessors
Yield = × 100
Total Number of Microprocessors

289
Yield = × 100 = 96.33%
300
b. Supplier which is least likely to have supplied a non-defective
microprocessor
Assuming that the probability of taking any microprocessor from A, B
and C is equal
Probability of a microprocessor being non-defective from supplier:

• Supplier A: 148
150
= 0.9867
• Supplier B: 45
50
= 0.9
• Supplier C: 96
100
= 0.96

Supplier B is least likely to have supplied that microprocessor as the

probability of it having a non-defective microprocessor is lowest.
c. Supplier which is most likely to have supplied a defective mi-
croprocessor

Probability of a microprocessor being defective from the supplier:

1
 • Supplier A: 2
150
= 0.0133
• Supplier B: 5
50
= 0.1
• Supplier C: 4
100
= 0.04

Supplier B is most likely to have supplied a defective microprocessor

as it has the highest probability of having a defective microprocessor.
d. Are the answers to questions B and C the same? Justify with
qualitative reasoning
Yes, the answers to B and C are the same.
Supplier B is the least likely to supply a non-defective microproces-
sor and the most likely to supply a defective one. This is because a
low probability of a non-defective microprocessor (0.9) means a higher
chance of a defective one (0.1), as both probabilities must add up to 1.
So, if Supplier B is less likely to provide a working microprocessor, it’s
more likely to provide a faulty one.
e. Maximum possible yield of this process assuming the suppli-
ers’ performance in terms of the quality is the same as given
in the dataset
To maximize the yield under these constraints, let vA , vB , and vC de-
note the volume fractions from Supplier A, B, and C, respectively. The
constraints are:
vA + vB + vC = 1
0.1 ≤ vA ≤ 0.6
0.1 ≤ vB ≤ 0.6
0.1 ≤ vC ≤ 0.6

To achieve the highest yield, we choose:

vA = 0.6 (maximum allowed)

vB = 0.1 (minimum allowed)

vC = 0.3 (remaining volume fraction)

The yield is calculated as:

Yield = vA · (1 − 0.0133) + vB · (1 − 0.1) + vC · (1 − 0.04)

2
 Yield = 0.6 · 0.9867 + 0.1 · 0.9 + 0.3 · 0.96
Yield = 0.59202 + 0.09 + 0.288 = 0.97002 or 97.00%

Therefore, the maximum possible yield of the process, assuming the

suppliers’ quality performance remains as specified, is 97.00%.

Question 2
Given:-

• Total capacitors = 2000

• Total passed by manufacturer = 850

• Total passed by customer = 1000

• Products passed by both manufacturer and customer = 800

• Products passed by manufacturer but failed by customer = 50

• Products failed by both manufacturer and customer = 950

• Products failed by manufacturer but passed by customer = 200

a. Calculate the TPR, FNR, TNR, FPR.

Products passed by customers are considered actual good products,
and decisions made by the manufacturers are considered predictions.

Predicted Positive Predicted Negative

Actual Positive 800 (TP) 200 (FN)
Actual Negative 50 (FP) 950 (TN)

Table 1: Confusion Matrix

The performance metrics are calculated as follows:

• True Positive Rate (TPR):

TP 800
TPR = = = 0.80
TP + FN 800 + 200

3
 • False Negative Rate (FNR):
FN 200
FNR = = = 0.20
TP + FN 800 + 200
• True Negative Rate (TNR):
TN 950
TNR = = = 0.95
TN + FP 950 + 50
• False Positive Rate (FPR):
FP 50
FPR = = = 0.05
TN + FP 950 + 50
b. Probability that a capacitor that fails at the customer’s facil-
ity will also fail at your manufacturing facility

We can calculate the total number of failures at each facility:

Total failed by manufacturer = 2000 − 850 = 1150
Total failed by customer = 2000 − 1000 = 1000

Definitions:
Let A = event that a capacitor fails at the manufacturer’s facility
(OQC).
Let B = event that a capacitor fails at the customer’s facility (IQC).

We are also provided with the following information:

A ∩ B = 950 (both failed by manufacturer and customer)
Ac ∩ B = 50 (passed by manufacturer but failed by customer)
A ∩ B c = 200 (failed by manufacturer but passed by customer)

This is the conditional probability P (A|B):

P (A ∩ B) No. of products failed by manufacturer and customer
P (A|B) = =
P (B) Total failed by customer

950
P (A|B) = = 0.95
1000
So, the probability that a capacitor that fails at the customer’s facility
will also fail at the manufacturer’s facility is 0.95.

4
c. Probability that a capacitor found to have failed at your fa-
cility will also fail at the customer’s facility?

This is the conditional probability P (B|A):

P (A ∩ B) No. of products failed by manufacturer and customer

P (B|A) = =
P (A) Total failed by manufacturer

Thus,
950
P (B|A) = ≈ 0.826
1150
So, the probability that a capacitor that fails at your facility will also
fail at the customer’s facility is approximately 0.826.

d. Probability that the capacitors accepted at your facility will

fail at the customer’s facility
This is the conditional probability P (B|Ac ):

P (Ac ∩ B) No. of products passed by manufacturer but not by customer

P (B|Ac ) = c
=
P (A ) Total failed by manufacturer

Thus:
50
P (B|Ac ) = ≈ 0.0588
850
So, the probability that capacitors accepted at your facility will fail at
the customer’s facility is approximately 0.0588 or 5.88%.

5
Question 3
a. Part failure rate for capacitors using the Parts Stress Model.

λp = λb · πT · πC · πV · πQ · πE
Where:
λb = 0.00076, πE = 7, πT = 1.3, πQ = 10, πV = 1, πC = 0.81 (for 0.1
µF capacitor)
Substituting the values:
λp = 0.00076 × 1.3 × 0.81 × 1 × 10 × 7 ≈ 0.05602 failures per 106 hours

b. Part failure rate for resistors using the Parts Stress Model.
λp = λb × πT × πS × πP × πQ × πE

Where: λb = 0.0037, πE = 12, πT = 1.1, πS = 1.2, πQ = 10, πP = 0.76

Substituting the values:
λp = 0.0037×1.1×1.2×0.76×10×12 ≈ 0.44542 failures per 106 hours

c. Failure rate for PCB using the Parts Stress Model

To calculate the failure rate of the given circuit, we assume it to be in
a series combination. Therefore, the overall failure rate can be given
as:

Circuit failure rate = 3 × λr + 2 × λv

λPCB = 3 × 0.44542 + 2 × 0.05602 ≈ 1.4483 failures per 106 hours
d. If the company ships out 100,000 such PCBs, how many
failures should one expect at the end of five years?
Assuming CFR model,
Total failures = 100,000 × (1 − exp (−λPCB × t))
t (in hours) = 5×365×24 = 43800 hours
Total failures = 100,000 × 1 − exp −1.4483 × 10− 6 × 43800

≈ 6146failures

6
 Code for part A:
1 import pandas as pd
2

3 # Load the Excel file

4 file_path = ’ P r o c e s s o r s _ Q u a l i t y _ C h e c k . xlsx ’
5 df = pd . read_excel ( file_path )
6

7 supplier_column = ’ Supplier ’
8 status_column = ’ Quality ␣ Check ’
9

10 # Group by supplier and count working and non - working

11 components
12 status_counts = df . groupby ( supplier_column )[ status_column ].
13 value_counts (). unstack ( fill_value =0)
14

15 # Rename the columns to be more descriptive

16 status_counts . columns = [ ’Non - Working ’ , ’ Working ’]
17

18 print ( status_counts )

7
 Code for Part B:
1 import pandas as pd
2 from sklearn . metrics import confusion_matrix ,
3 cla s s i f i c a t i o n _ r e p o r t
4 import seaborn as sns
5 import matplotlib . pyplot as plt
6

7 # Load the Excel file

8 file_path = ’ A s s i g n m e n t _ 2 _ P r o b l e m _ 2 . csv ’
9 df = pd . read_csv ( file_path )
10

11 # Extract the true values and obtained values

12 y_true = df [ ’ Cu stomer _IQC_T est ’]
13 y_pred = df [ ’ M a n u f a c t u r i n g _ O Q C _ C h e c k ’]
14

15 # Create a confusion matrix

16 conf_matrix = confusion_matrix ( y_true , y_pred )
17

18 # Plot the confusion matrix using seaborn

19 plt . figure ( figsize =(8 , 6))
20 sns . heatmap ( conf_matrix , annot = True , fmt = " d " ,
21 cmap = " Blues " , cbar = False )
22 plt . title ( ’ Confusion ␣ Matrix ’)
23 plt . xlabel ( ’ Predicted ␣ Values ’)
24 plt . ylabel ( ’ True ␣ Values ’)
25 plt . show ()
26

27 print ( " Classification ␣ Report :\ n " )

28 print ( c l a s s i f i c a t i o n _ r e p o r t ( y_true , y_pred ))