Environment. Technology. Resources.

Rezekne, Latvia
Proceedings of the 15th International Scientific and Practical Conference. Volume X, XX-XX

Statistical analysis of the quality of a

technological process for the production of
shafts for electric motors by roughness study
Kliment Georgiev
Technical University of Sofia,
branch Plovdiv
Faculty of Mechanical Engineering;
Department: Mechanical and
Instrument Engineering
Plovdiv, Bulgaria
k.georgiev@tu-plovdiv.bg

Abstract— This paper presents a statistical analysis The ability and regulation of the process is
of the quality of a technological process by investigating collectively called efficiency. The quality of a process
the surface roughness of shafts for electric motors is determined by its stability and efficiency.
obtained during machining. Repeated measurements
have been carried out using a portable roughness tester A.Process stability analysis
INSIZE ISR - C002 and the results are summarized.
Stability of the TP means preservation of the
Statistical analysis is applied to analyse the accuracy,
adjustability, and stability of technological process. The parameters of the distribution of deviations of the
stability of the process is presented by investigating the qualitative indicator according to a certain law for a
correlation between the experimental and theoretical sufficiently long time. The initial assessment of
curve of the quality index. Accuracy evaluation is stability can be performed by means of point
performed by calculating the accuracy coefficient. The diagrams. In stable processes, the variance of the
adjustability analysis is evaluated by the relative qualitative indicator is kept constant, and its meaning
position of the distribution curve of the obtained quality is either constant or changes in a regular manner. In
index with respect to the tolerance field. unstable processes the variance changes randomly.
Intermittent periods with qualitative or defective
Keywords— analysis, quality, quality control, surface
roughness, technological process, stability, accuracy, output are alternated. Stable processes are predictable,
adjustability. i.e. amenable to statistical control Fig. 1. Unstable
processes are unpredictable, i.e. unsuitable for
I. INTRODUCTION statistical control and quality regulation Fig. 2.
With the development of technology, the quality
requirements of the machined parts are increasing. The
quest for defect-free production of quality workpieces
leads to the need for in-depth technological process
(TP) research and analysis.
Statistical analysis methods can solve many
problems related to process quality. The three main
tasks presented are some of the most important in Fig. 1 Stable technological process
process analysis:
• Process stability analysis;
• Process capability (accuracy) analysis;
• Analysis on process regulation (tuning);

Fig. 2 Unstable technological process

Print ISSN 1691-5402

Online ISSN 2256-070X
https://dx.doi.org/10.17770/etr2024volX.XXXX
© 2024 The Author(s). Published by Rezekne Academy of Technologies.
This is an open access article under the Creative Commons Attribution 4.0 International License.
<##>
 First Author., et al. Paper Title

Process stability analysis is studied and evaluated

using two approaches:
• By examining the correspondence between
the experimental and theoretical qualitative
indicator curve;
• By examining the histogram of the process;
The first approach is used less frequently when the
process flow is not influenced by a dominant
systematic factor that shifts the distribution, i.e. the
average value of the qualitative indicator remains
constant over time (stationary TP).
The second approach is more universal and is
applied more often. It examines and analyses the
average value of the qualitative indicator using the
regression equation.
Fig. 3 The aspects of quality [1]

B. Statistical analysis of accuracy of a TP Quality of conformance implies that a

manufactured product or a service rendered must
This analysis applies only to proven stable meet the standards selected in the design phase. This
processes. Accuracy is assessed by the accuracy phase is concerned with the degree to which quality is
factor. controlled from the procurement of raw material to
If appropriate, the tolerance may be reduced. The the shipment of finished goods [1].
other solution is to keep the tolerance but intensify the
machining process by changing the mode parameters, Quality of performance is how a product functions
which will increase productivity. or service performs when used. It measures the degree
to which the product or service satisfies the user [1].

C. Tuning analysis of a TP The way the product or the service satisfies the
customer is a function of both quality of design and
This is the third task of statistical analysis (the first
quality of conformance [1].
for stability and the second for accuracy). Tunability is
characterized by the relative position of the
distribution curve or accuracy diagram with respect to In this paper, an analysis of a technological process
the tolerance field. is presented and its stability, accuracy and tunability
are investigated. The parameter used for quality
Tuning is a process where the process system is
assessment is surface roughness.
brought into a state to produce a certain value of a
quality metric. This value, which the output quality II. MATHERIALS AND METHODS
indicator should obtain at the beginning of operation
after tuning, is called the working setpoint. It is A. Matherials
defined differently for statistically stable and The details used for the experimental studies are
technologically stable processes. presented in Fig. 4. All experiments were conducted
Quality is a set of characteristics (indicators) of an in a laboratory environment under controlled
object that determine its ability to satisfy certain or temperature and humidity conditions. The measuring
supposed needs. An object can be a product, an instrument was pre-calibrated before the
activity, a process, an organization, a person. measurements were started.
According to Mitra A. (2016) the aspects of the
The shaft is processed in two operations. The first
quality are three: Quality of conformance, Quality of
Design and Quality of Performance (fig.3) [1]. machining is on the long side with a back saddle
support. The machine is a DMG CLX350 with the
following cutting modes:

• Rough machining: Vc = 115m/min, a =

1.5mm, F = 0.25 mm/rev;
• Fine machining: Vc = 200m/min, a =
0.5mm, F = 0.15 mm/rev;

<##>
 Environment. Technology. Resources. Rezekne, Latvia
Proceedings of the 15th International Scientific and Practical Conference. Volume X, XX-XX

The second processing is carried out on a Mazak

Qtn 200 MSY. The basis of this machining is on the
short side of the workpiece to the two-stand in the
following cutting modes:

• Rough machining: Vc = 100m/min, a =

1.5mm, F = 0.24 mm/rev;
• Fine machining: Vc = 185m/min, a = 0.4
mm, F = 0.15 mm/rev;

Fig 4. Measured detail

Fig 5. Schematic control chart [5]
B. Methods
When Control charts are applied, it is assumed
The technological process for manufacturing that the behavior of the process could be
shafts for electric motors is evaluated by statistical characterized by the level or one or several qualitative
analysis of the qualitative parameter roughness. values. These values are also known as regulated
Stability, accuracy, and process tuning analysis have values. Control charts are used in monitoring
been applied in the evaluation of the technological processes and when ascertaining the need for
process. Statistical analysis using X and R control corrections or changes in the process, to achieve a
charts was used to evaluate the process. better mean value of the process or to reduce
variability in the process. In control charts, the
The SPC is a basic statistical tool for verifying the horizontal axis contains the times when statistical
conformity of technical requirements of products. sampling of regulated values took place, and the
Even in robust manufacturing processes, quality vertical axis contains calculated values of the
characteristics are associated with randomness due to appropriate sample characteristics [3].
the presence of uncontrollable (or difficult/costly to
control) input variables [2]. To build the control charts X, it is necessary to
calculate: upper control limit (UCL), lower control
Gejdos P. (2015) says that the control charts are limit (LCL) and center line (CL)[6, 7].
the most frequently used tool in the statistical
regulation of processes. According to him they allow X chart is used to reflect changes in the average
us more accurate distinguishing of random from values of a process. To construct an X chart, it is
symmetric causes of the fluctuations in the value of a necessary to draw its center line (CL). This is
mark of quality. Control charts facilitate regulation achieved by collecting the sample data and
and improvement in the quality of the technological calculating the mean of 𝑥̅ [8, 9].
process [3]. 𝑥1 +𝑥2 +𝑥3 + 𝑥𝑛
𝑥̅ = (1)
𝑘
Process capability analysis is an important tool in
the DMAIC (Define, Measure, Analyze, Improve and CL = 𝑥̅ (2)
Control) process, with application in both the analyze
and improve steps [4]. To calculate the upper and lower control limits,
the following formulas are used:
According to Oakland J. (2007) the basis of the
theory of statistical process control is differentiation 𝑈𝐶𝐿 = 𝑥̅ + 𝑧𝜎𝑥̅ (3)
of the causes of variation during the operation of any
process. Certain variations belong to the category of 𝐿𝐶𝐿 = 𝑥̅ − 𝑧𝜎𝑥̅ (4),
chance or random variations, about which little may
be done, other than to revise the process [5]. Fig. 5 where:
shows a schematic control chart. 𝑥̅ - average value of the sample;

<##>
 First Author., et al. Paper Title

z – standard normal variable equal to 2 for According to what is written above σ is usually
95.44% confidence interval and 3 for 99.73% [10]; unknown and we can replace it with (s), µ or с 𝑥̅ .
𝜎
𝜎𝑥̅ = (5),
√𝑛 III. RESULTS AND DISCUSSION
where:
σ – process standard deviation; The experiments were conducted in a laboratory
n – number of tests per sample. environment under controlled conditions of
temperature and humidity. Twenty shafts were
Since σ is usually unknown, we must replace it measured, and each shaft was measured three times.
with an estimate. Most frequently the use of as an According to the submitted technical documentation,
estimate of σ, which is the sample standard deviation the nominal prescribed roughness was Ra = 1.25 µm
(s) [4]. with USL = 2.25 µm and LSL = 1 µm. The results of
the shaft roughness measurements are presented in
Process capability ratio is a frequently used way Table 1. The same results were used for the statistical
to express process capability. There are many ways to analyses.
Table 1 Measurement Data
express process capability and one of them is in terms
of the process capability ratio (PCR) Cp, which for a № Surface Roughness, µm Average
quality characteristic with both upper and lower 1 2 3 ̅
𝒙
specification limits (USL, LSL) is [4]. USL stands for 1 2.023 2.051 2.068 2.05
2 2.076 2.063 2.083 2.07
Upper specification limit, that can be also called
3 2.114 2.049 2.109 2.09
Upper Tolerance limit. The same goes to LSL, which 4 2.077 2.064 2.106 2.08
is Lower Specification limit, which is also known as 5 1.996 2.018 2.019 2.01
Lower Tolerance Limit: 6 2.059 2.055 2.046 2.05
7 2.072 2.012 2.021 2.04
𝑈𝑆𝐿−𝐿𝑆𝐿 8 2.061 2.043 2.037 2.05
𝐶𝑃 = (6) 9 2.061 2.054 2.049 2.05
6𝜎
10 2.063 2.068 2.007 2.05
11 2.034 2.063 2.078 2.06
According to Montgomery D. C (2009) the 6σ 12 2.086 2.073 2.095 2.08
spread of the process is the basic definition of process 13 2.123 2.061 2.121 2.10
capability. Since σ is usually unknown, we must 14 2.082 2.075 2.128 2.10
replace it with an estimate. Most frequently the use of 15 2.005 2.025 2.028 2.02
as an estimate of σ, resulting in an estimate 𝐶 ̂𝑝 of 𝐶𝑝 16 2.069 2.068 2.056 2.06
17 2.083 2.023 2.035 2.05
is “s”, so [4]: 18 2.075 2.052 2.052 2.06
19 2.085 2.065 2.066 2.07
𝜎̂ = 𝑠 (7) 20 2.073 2.085 2.025 2.06
Total average 2.06
Based on the above, we arrive at the formula for
̂𝑝 by the following
calculating the precision factor 𝐶 The standard deviation is calculated using the formula
formula: (12):

𝑈𝑆𝐿−𝐿𝑆𝐿 ∑𝑛
𝑖=1 (𝑥𝑖 − 𝑥̅ )
2
̂𝑝 =
𝐶 𝑠= √ (12)
̂
6𝜎 𝑛−1
(8)
For the data presented in Table 1, the value of the
The analysis of the process setup can be represented standard deviation is:
by calculating the coefficient 𝐶𝑝𝑘 :
s = 0,0229mm (12)
𝐶𝑝𝑘 = min(𝐶𝑝𝑢 , 𝐶𝑝𝑙 ) (9),
The X chart is constructed using formulas (3) and
where: (4) to calculate the UCL and LCL, replacing σ by the
value of s as described above. The confidence interval
𝑈𝑆𝐿− 𝜇 was chosen to be 3σ, respectively 3s, hence:
𝐶𝑝𝑢 = (10)
3𝜎
𝑈𝐶𝐿 = 𝑥̅ + 𝑧𝑠 (13)
𝜇−𝐿𝑆𝐿
𝐶𝑝𝑙 = (11) 𝐿𝐶𝐿 = 𝑥̅ − 𝑧𝑠 (14)
3𝜎

<##>
 Environment. Technology. Resources. Rezekne, Latvia
Proceedings of the 15th International Scientific and Practical Conference. Volume X, XX-XX

After substituting the data in formula (13) and (14),

the following results are obtained: 2,25− 2,06
𝐶𝑝𝑢 = (10)
3×0,0229

𝐶𝑝𝑢 = 2,76
𝑈𝐶𝐿 = 2.06 + 3 × 0.0229 (13)

𝑈𝐶𝐿 = 2.128 µm; 2,06−1

𝐶𝑝𝑙 = (11)
3×0,0229
𝐿𝐶𝐿 = 2.06 − 3 × 0.0229 (14)

𝐿𝐶𝐿 = 1.991 µm; 𝐶𝑝𝑙 = 15,43

The value of 𝑥̅ is taken as the central line. The It follows from the calculations that for the
histogram in Figure 6 graphically presents the results coefficient 𝐶𝑝𝑘 must be chosen the value of 𝐶𝑝𝑢 =
of the study. 2,76, which is the minimum. The value adopted for
𝐶𝑝𝑘 = 2,76 is greater than that recommended in the
criterion (𝐶𝑝𝑘 > 1,67).

IV. CONCLUSIONS
1. The process is set, and no upper or lower
control limit crossing is observed.
2. The calculated process accuracy factor Cp =
9,1 indicates that the process is accurate.
3. The process tuning evaluation factor 𝐶𝑝𝑢 =
2,76 indicates that the process is well tuned.
4. The set nominal roughness is not met, but
nevertheless the measurement data and
Fig 6. Histogram statistical analysis show that the process is
set, accurate and adjustable.
The plot in Fig. 6 shows the distribution of the
value of 𝑥̅ versus the upper and lower bounds. AKNOWLEDGEMENTS
Although all the values of 𝑥̅ are outside the The author/s would like to thank the Research and
tolerances, the process does not cross the control Development Sector at the Technical University of
limits. Sofia for the financial support."

The calculation of the accuracy coefficient was

carried out using formula (8). REFERENCES
[1] Mitra, A. (2016). Fundamentals of quality control and
2,25−1
̂𝑝 =
𝐶 (8) improvement. John Wiley & Sons.
6×0,0229 [2] Hristov, M. H. (2023, September). SPC as an Instrument for
Implementation and MSA as an Instrument for Evaluation of
Non-Linear Object Based Temperature Compensation into
̂𝑝 = 9,1
𝐶 Shop Floor CMM. In 2023 15th Electrical Engineering
Faculty Conference (BulEF) (pp. 1-9). IEEE.
[3] Gejdoš, P. (2015). Continuous quality improvement by
According to the standard used, the values of the statistical process control. Procedia Economics and
resulting accuracy factor Cp must be greater than 1.67 Finance, 34, 565-572.
to satisfy the process accuracy conditions. The [4] Montgomery D. C., Introduction to Statistical Quality
resulting value for the coefficient Cp = 9,1, which is Control, Sixth Edition, Copyright 2009 by John Wiley &
Sons, Inc., ISBN 978-0-470-16992-6
greater than the 1.67 that is specified in the standard.
[5] Oakland, J., & Oakland, J. S. (2007). Statistical process
control. Routledge.
The estimation of the tunability of the [6] H.R. Radev, “Metrologiya i izmervatelna tehnika”, vol. II, S.,
technological process was made by using the Softtreyd, 2010, ISBN: 978-954-334-093-4
coefficient 𝐶𝑝𝑘 , formula (9). Calculate the values of [7] Дюкенджиев Г., Р.Йорданов, Контрол и управление на
the criteria in the upper limit deviation (Cpu) (10) and качеството, София., Софттрейд, 2002.
the value of the criteria in the lower limit deviation [8] Statistical process control (SPC), Reference manual,
Copyright 1992, Chrysler Corporation, Ford Motor Company,
(Cpl) (11), The criteria then require the minimum and General Motors Corporation
value obtained to be assumed to be selected.
<##>
 First Author., et al. Paper Title

[9] Allen T., Introduction to Engineering Statistics and Lean Six

Sigma, Statistical Quality Control and Design of Experiments
and Systems, Third Edition Springer-Verlag London Limited
2006, ISBN 978-1-4471-7420-2 (eBook),
https://doi.org/10.1007/978-1-4471-7420-2
[10] Qiu P., Introduction to Statistical Process Control, 2014 by
Taylor & Francis Group, LLC, ISBN-13: 978-1-4822-2041-4
(eBook - PDF)

<##>