“PROCESS QUALITY, PRODUCT

QUALITY AND SERVICE QUALITY”

with Case Study of 3D Printers

Shanmuga Sundaram Giri Haran - G1900663G

Shanmugam Vijayakumar - G1801616D
Puthan Veetil Sreerag - G1900645E
Devarassu Manojkumar - G1900825B
Introduction
 The purchase intentions of a customer, customer loyalty and
ultimately an organization's revenue are affected by customer
satisfaction.
 Customer satisfaction mainly depends on product quality and/or
service quality and price.
 What is Process Quality, Product Quality and Service Quality?
 Product quality is defined as the ability of the product to
fulfil and meet the requirements of the end user (User
Perspective). It is the collection of features and
characteristics of a product that contribute to its ability to
meet given specifications (Manufacturing Perspective).
 Process quality focuses on all activities and steps used to
achieve maximum acceptance.
 Service quality is a comparison of perceived expectations of
a service with perceived performance, giving rise to the
equation
Service Quality = Performance - Expectation
REFERENCES: https://en.wikipedia.org/wiki/Service_quality
Differences & Relationships between “Process, Product & Service
Quality”
Product Quality Service Quality
Product Quality depends on how Service Quality depends on timeliness,
materials are chosen and converted. responsiveness, communication
effectiveness, etc., of the service provider.
It focuses on tangible product It mainly focuses on non-tangible factors
features like conformance, like courtesy, empathy etc., Though, it can
performance, reliability & features. be quantified by waiting time, etc.,

We can have more control over We can have only less control over service
product quality on comparing with quality on comparing with product quality.
service quality.

 Example: Manufacturing sector (E.g.: Automobile manufacturers, Aircraft

manufactures, etc.,)
 Customer satisfaction is impacted by both product quality and service
quality.
 If greater attention is given to product quality during production, then,
customer services (for repairing, warranty claim, etc.,) to correct subsequent
problems will be fewer.
1) Statistical methods are mathematical equations, models and/or techniques that No VARIATION
are used in statistical analysis of set of data. It helps in determining product, process No Quality issue
and service quality quantitatively. In quality management following statistical But impossible
methods are used:
A. Statistical process control (SPC)
B. Hypothetical testing with Analysis of variance (ANOVA)
2) Other analytical method used: Failure mode effect analysis

1.A) Statistical Process Control (SPC)

 ‘Statistical Process Control’ also known as ‘SPC’ Charts, helps in quality
management, by controlling the process by indicating when to intervene and take
action or when to leave the process alone. 
 The purpose of SPC is to identify, troubleshoot and correct process variation,
especially non-random variation.
 Non random variation occurs when there are special causes. Examples: Jammed
equipment, broken tools, contamination, etc.,. If special causes of variation
exist, then the process needs to be corrected, since it is out of control. 
 SPC charts are simply the voice of the process, showing process STATISTICS helps
behaviour/performance over time. us to deal with the
variations
REFERENCES: https://www.linkedin.com/learning/search?keywords=spc%20chart&u=43752620
  It's a run chart, that has a
centreline, which represents the
mean, and a pair of control limits
above and below the centreline.
 Upper and lower control limits equals
average value plus or minus three
standard deviations. This likelihood is
so low, that it must be due to
something non-random; something
sporadic, assignable to special causes. 
Fig.1. SPC Chart showing mean & pair of control limits
Two types of SPC charts: 
i. SPC for variables using continuous data, such as dimensions and time
ii. SPC charts for attributes, using discrete data such as counts of defects and proportions. 
1.A.i) SPC charts for variables
 Include Individual or I-Chart which plots individual values. For example, processing time of invoices. 
 The individual moving range chart, I-MR chart, in addition to the individual value, the moving range of
processing times are also charted. The moving range is the current value minus previous value.
REFERENCES: https://www.linkedin.com/learning/search?keywords=spc%20chart&u=43752620
 1.A.ii) SPC charts for Attributes
There are four types of Attribute SPC Charts. 
i. P-chart, which plots the proportion of defective units. For example, the proportion of invoices processed
incorrectly. 
ii. NP-chart where the number of defective units are plotted. For example, number of invoices processed
incorrectly.
iii. C-chart plots the number of faults. For example, number of errors in invoices.
iv. U-chart plots the number of faults per unit. For example, number of errors per invoice. 
 
List of out of control conditions for Attribute SPC Charts
1) Any point outside of the control limits. 
2) 9 points in a row on the same side of the centreline. 
3) 6 points in a row, either all increasing or decreasing. 
4) 14 points in a row, alternating up and down.
Fig.2. shows NP chart of a process. Condition one is present because
there is a point outside of the control limits. Condition two is present
because there are 9 points in a row on the same side of the
centerline. So, this process is out of control.
Fig.2. NP chart of a process
REFERENCES: https://www.linkedin.com/learning/search?keywords=spc%20chart&u=43752620
 1.B) Hypothetical testing with Analysis of variance (ANOVA)
 It is a data-driven method and is often referred to as the scientific method to determine quality of
product quantitatively. 
 A theory or hypothesis is proposed and data is used to run a statistical test to prove or disprove
that theory. 
 Hypothesis testing is set up as a pair of statements, a null hypothesis and an alternate hypothesis.
The null hypothesis is no difference between batches and the alternate hypothesis is there is a
difference between batches. 
  To carry out hypothesis testing, sample data are collected and statistical tests are conducted. 

Two errors in hypothesis testing

a) Type I error: Dismissing the null hypothesis when it is true.
The probability limit of a type I error to no more than 5%. This threshold, or maximum probability,
is denoted by alpha. Alpha is based on risk tolerance, application and industry needs. 
b) Type II error: Failing to dismiss the null hypothesis when it is false.
The probability of a type II error is called beta.
REFERENCES: https://www.linkedin.com/learning/learning-minitab/variation-plots?u=43752620
  True  False

Type I error Correct decision

  Reject Probability = Probability =
(1- = Power

Correct decision Type II error

  Fail to reject Probability = = Probability =
Confidence level

Fig.3. Summary Grid showing the four possible scenarios in hypothesis testing

 The power of the test is the probability of rejecting the null hypothesis when the
null hypothesis is, in fact, false. 
Power = 1 – β
 Typically, we want the power to be 90% or higher so that a correct decision can be
made.
 The confidence level of test is the probability of correct decision to be as high as
possible so that we can be confident in not rejecting the null hypothesis. Confidence
level equals one minus alpha, so if alpha is 5%, confidence level is 95%. 
REFERENCES: https://www.linkedin.com/learning/learning-minitab/variation-plots?u=43752620
1.B.i) Analysis of variance (ANOVA)
 To draw out conclusions about the equality of means of multiple populations, we use
a hypothesis-testing methodology called ANOVA, which is Analysis of variance.
 One-way ANOVA is for comparing means of observed responses of several different
levels of a single factor.
 ANOVA tests the hypothesis that the means of all populations are equal against the
alternative hypothesis that at least one mean differs from the others.
 To do this, ANOVA separates the total variability of the data into 2 parts: Variation
between groups & within groups.
 When total variation between the groups is smaller than the variation within groups, it
suggests that the populations are essentially the same. A relatively large variation
between groups suggests the existence of differences in the unknown population
mean.
F statistic
 The variation in the data is calculated as a sum of squared (SS) deviations from the
appropriate sample mean, and scaled as a variance measure, or "mean square" (MS).
Mean square between groups is divided by the mean square within groups, an F
statistic is calculated. If this value is larger than a critical value, then data suggest that
a difference in means exist.
REFERENCES: Managing for Quality and Performance Excellence by James R. Evans; William M. Lindsay - Edition
11 - ISBN: 9780357118252 Cengage Learning US (2020)
 p – Value
 The p-value is the probability of obtaining a value for the test statistic that is, as extreme or more
extreme than what is observed, given that the null hypothesis is true.
 If the actual value of quality of product is too far from its expected value, then the p-value will be
very low, indicating that the null hypothesis is unlikely, so compare the p-value to alpha, which is
the maximum risk willing to take in wrongly dismissing the null hypothesis. 
 If the statistical test returns a p-value that is less than or equal to alpha, reject the null
hypothesis. If the p-value exceeds alpha, then we fail to reject the null hypothesis.

Steps used to find p-value

 Look up your test statistic on the appropriate distribution.
 Find the probability that Z is beyond (more extreme than) your test statistic
• If Ha contains a less-than alternative, find the probability that Z is less than test statistic. This is the p-
value.
• If Ha contains a greater-than alternative, find the probability that Z is greater than test statistic. The
result is your p-value.
• If Ha contains a not-equal-to alternative, find the probability that Z is beyond your test statistic and
double it.
REFERENCES: https://www.linkedin.com/learning/learning-minitab/variation-plots?u=43752620
 2) Failure mode effect analysis
 This tool prioritises, anticipates and mitigates any risk of failure, thereby process steps can be
improved and process controls can be put in place to reduce the RPN (Risk Priority Number). 
 There are two types of FMEA: Design FMEA & Process FMEA
 The Design FMEA is for reducing the risk of failure associated with a product or service design such as,
when you are designing a new hotel or the next smartphone. 
 The Process FMEA is for reducing the risk of potential failures in a process. It works with a series of
questions.

2.B) Steps in FMEA

  For each step of the process, “What can possibly go wrong?” “In what ways can it fail?” are anticipated.
These are called potential failure modes.
 For each potential failure mode, “What is the effect?” “How severe is the potential effect?”, are
represented using a 1 to 10 scale, where 10 is the worst. That is called the severity score (SEV). 
 “What are the potential causes for each failure mode?”  “How likely is the occurrence of these potential
causes?” are represented on a 1 to 10 scale, where 10 is the most likely. That is called the occurrence score
(OCC). 

REFERENCES: https://www.linkedin.com/learning/six-sigma-green-belt/tests-for-means?u=43752620
  “What process controls are currently in place to detect the cause or the failure mode?”  “What is
the likelihood of detection?” are represented on a 1 to 10 scale, where 10 is the least likely to
detect. That is called the detection score (DET).
 Risk Priority Number (RPN) = SEV * OCC * DET where OCC – Occurrence score; SEV – Severity score;
DET – Detection score
 The RPNs can be used to prioritize the failure modes. The higher scoring failure modes are the first
targets for improvement.
Table 1. FMEA WORKSHEET
PROCESS FAILURE MODES AND EFFECTS ANALYSIS
Process Name: Process Team: Prepared by: Document No:

PFMEA Date: Revision:

CURRENT STATE PROCESS FMEA EVALUATION IMPROVEMENT FOLLOW-UP

Resp.
S O D R S O D R
Potential Potential Failure Current Process &
Process Step E Potential Causes C E P Recommendations Actions Taken E C E P
Failure Modes Effects Controls Targe
V C T N V C T N
t Date

Where OCC - Occurrence; SEV - Severity; DET - Detection; RPN - Risk Priority Number

REFERENCES: https://www.linkedin.com/learning/six-sigma-green-belt/tests-for-means?u=43752620
 Case study – 3D printer
Product Quality issue
 Some of the customers faced issue with their 3D printer. Customers reported that
‘User Interface’ showed error messages.
 After analysis, this issue was found to be with Lamp Module which showed false
alarm of glass break event, while printing job. This (false) alarm wouldn’t allow
printer to print further.
 Issue was then passed down from Field Service Engineer to Product Engineer.
 Engineer reviewed manufacturing issues’ log. He didn’t find any relevant
abnormalities.
 He prepared a printer to simulate the issue. He then, ran a long print job to list out
the causes associated with the issue. The causes are listed out in the Fishbone
diagram (shown in next slide).
 Finally, major cause is narrowed down to Lamp sub system.

Lamp sub system in 3D printer

 Detection of failure of Lamp module is done with the help of PCA. When
Lamp glass breaks (fails), PCA detects high resistance.
 Hence, PCA should be accurate enough to trigger the alarm at the right time,
i.e., when the resistance is high.
Fish bone diagram: Root cause analysis
Design Environment

Design margin Temperature

Non reliable design Humidity

Air Flow
Error in design
Glass
break
Event
Less skilled operators
Glass
PCA epoxy bonding process
Lack of process
knowledge Printer circuit board
PCA components

Improper assembly Wrong part

Workmanship Part issue

 Causes of failure of lamp module may be due to Design issue, Environment,

Workmanship and/or Part issue.
 Root cause for this issue was analysed to be the PCA epoxy bonding process
 Hence, epoxy bonding process was analysed further.
 Process Quality issue – NP chart

Sample No of
Month Mean UCL LCL
Size Defects
Jan 100 25 19.08333 30.87207 7.294594
Feb 100 35 19.08333 30.87207 7.294594
Mar 100 10 19.08333 30.87207 7.294594
Apr 100 15 19.08333 30.87207 7.294594
May 100 36 19.08333 30.87207 7.294594
Jun 100 20 19.08333 30.87207 7.294594
Jul 100 17 19.08333 30.87207 7.294594
Aug 100 40 19.08333 30.87207 7.294594
Sep 100 18 19.08333 30.87207 7.294594
Oct 100 19 19.08333 30.87207 7.294594
Nov 100 45 19.08333 30.87207 7.294594
Dec 100 6 19.08333 30.87207 7.294594

Table 2. Source data for NP chart

Figure 3. Source data for NP chart
Process Quality issue – NP chart
 Supplier builds & ships PCA every month.
 NP chart (Previous slide) is plotted for number of defects vs month.
 It was observed that the defective rate was following some trend that, every second
month of the quarter. They faced high failure during that period.
 Company follows distinctive date code for every month. With the help of date code, it
was found that Mr. Lim Chua built parts in every second month of the quarter.
 PCA manufacturer has a policy that one operator can not work for more than one
month in same place. They wanted operator to rotate jobs, every month and get
experience from other task.
 Process Quality issue is due to the fact that: Mr. Lim was recently appointed for
work. He is not fully trained operator. He was one of the multitasking operator (under
job rotation).
Process Quality issue – ANOVA F statistics
 Sample from Mr. Lim was taken for analysis. It was observed that Mr. Lim has
his own method of building PCA. Let’s name his 3 methods of building PCA as
Method A, Method B & Method C.
 ANOVA F – statistics was used to find whether the parts PCA produced by
methods A, B & C are same.
Process Quality issue – ANOVA F statistics
 Table A shows the resistance results of PCA built by 3 methods after quality checks.
 Null Hypothesis: At least one mean is different from the others.
 P value in Fig.1. is 0.000299 which is less than 0.05, so it is significant. Therefore, we
reject the null hypothesis.
 F value > F critical in Fig.1. also suggest that at least one mean is different from the
others.
 ANOVA results suggests that quality produced by 3 different methods are not same,
statistically. Hence, Mr. Lim doesn’t produce statistically identical quality.

Table 3. Resistance results of PCA built by 3

methods after quality check
 FMEA to find High RPN Process step
To find any failure in PCA Epoxy bonding process, mitigate the risk of failure and
take suitable action proactively, FMEA method is used as shown in table.
It was found out that, ‘direct application of epoxy on PCA’ has high risk priority
number and need to be prioritised and corrected soon. 

Table 4. Case study – FMEA WORKSHEET

PROCESS FAILURE MODES AND EFFECTS ANALYSIS

Process Name: Process Team: Prepared by: Document No:
Process Engineering
PCA Epoxy bonding process PFMEA Date: Revision:
Team
CURRENT STATE PROCESS FMEA EVALUATION IMPROVEMENT FOLLOW-UP
S O D R Resp.& S O D R
Potential Potential Failure Current Process
Process Step E Potential Causes C E P Recommendations Target Actions Taken E C E P
Failure Modes Effects Controls
V C T N Date V C T N
Directly apply the epoxy on the PCA Damage another Overall PCA performance 9 Operator use hand to apply 3 None 5 135 Use proper opening to One month Implemented in busiest 9 1 1 9
components. will affect the epoxy. accomplish perfect epoxy on from now location; roll out over next 2
the PCA months; daily checks on
resistance check
Apply Epoxy at the end of PCA assembly Operator may forgot to High Resistance will occur. 9 Long run the PCA will have 2 None 5 90 Design system controls to One week Implemented at the end of line 9 2 1 18
apply Epoxy prevent.
Service Quality Team
 Before finding the real root cause, service team were to face the
customers query regarding the faulty 3D printer.
 But after they found the real root-cause, service team shipped the FRU
(Field Replacement Unit) to fix customer side 3D printers.
 They created a schedule to replace the lamp module by the project
group members.
 Project team decided to change the lamp module together with printer
maintenance which is going to be potentially effective.
 Some printers are on the way to customer. Those printer normally would
reach the distribution centre, before delivery of printer to the customer.
 Hence, lamp module of those printers could be replaced with new Lamp
module at distribution centre.
 Service Quality Team
In order to resolve the issue from the customer side, serial number of the
3D printers was identified which were built by Mr. Lim.

 Printers at customer side : 110

 Printers on the way to customer ( air, ship ): 26
 Printers at manufacturing unit (WIP – Work In Progress): 20

Total number of printers affected : 156

Table 4. Lamp Module Service Activity
Lamp module service Activity
Who ( PIC) Quantity status Feb-20
Item Action Items
1-Feb 2-Feb 3-Feb 4-Feb 5-Feb 6-Feb 7-Feb 8-Feb 9-Feb 10-Feb 11-Feb 12-Feb 13-Feb 14-Feb 15-Feb 16-Feb
1 Replace Lamp module at customer side Service Engineer 110 ongoing
2 Printers on the way to customer ( air, ship ) Service Engineer 26 Pending
3 Printers at manufacturing WIP Manufacturing Engineer 20 Done
Number of Days 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Thank you