Journal of Applied Research and Technology

www.jart.icat.unam.mx

Journal of Applied Research and Technology 22 (2024) 230-242

Original

Optimizing multi-skill call center staffing using queuing

models: A study of service level

N. A. Hassan N. M. S. Abdallah* R. A. Attwa

The University of Zagazig, Faculty of Science, Department of Mathematics, Zagazig, Egypt

Received 06 20 2023; accepted 02 22 2024

Available 04 30 2024

Abstract: This paper addresses the staffing problem of a multi-skill call center with prioritized and
patient customers based on a queuing model for minimizing staffing costs subject to service level
requirements and determining the optimal number of agents. The model includes three types of calls
and four groups of agents with different skills. Initially, a method for describing the state space is
presented, followed by obtaining state-transition rates of the study model from 𝑀𝑀/𝑀𝑀/𝑐𝑐 and 𝑀𝑀/𝑀𝑀/𝑐𝑐/𝑐𝑐
queuing systems. Equations for steady-state probabilities and service level computation are derived.
A staffing calculation model for optimal agent numbers in each group is then formulated. An algorithm
is proposed to find solutions, and its process is outlined. Finally, a numerical example is provided to
analyze the system's sensitivity to several factors.

Keywords: Queuing model, multi-skill call center, service level, staffing problem

∗
Corresponding author.
E-mail address: [email protected] (N. M. S. Abdallah).
Peer Review under the responsibility of Universidad Nacional Autónoma de México.
 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

1. Introduction
Call centers play a crucial role in the service industry,
Call centers are the preferred and important way for many where the adept handling of customer inquiries and requests
companies to communicate with their customers. The call is essential for ensuring customer satisfaction and overall
center industry has undergone rapid growth in recent years. business prosperity. Efficiently operating call centers must
Call centers are widely utilized in many fields, including e- strike a balance between the cost of service and service levels.
business, finance, banking, hotels, credit card companies, It is imperative for call centers to ensure the presence of an
information centers, care hotlines, airlines, emergency appropriate number of agents possessing the requisite skills
services, telemarketing, help desks, and various other at any given time, aligning with the anticipated demand level.
domains. A call center is characterized as a service system in This challenge is commonly known as the staffing problem. A
which agents (servers) serve customers (callers) through comprehensive overview of staffing problems was given (Aksin
telephone, fax, email, network communication software, as et al., 2007; Defraeye & Van Nieuwenhuyse, 2016; Koole & Pot,
well as other multimedia channels. This fast progress in 2006; Wallace & Witt, 2005). In (Li et al., 2020; Priyai &
technology, with the variety of contact methods and the size Rajendran, 2018), the staffing problem is discussed in an M-
of call centers, made the agents busier, their work more design multi-skill call center with impatient customers. Also, Li
complicated, and required them to master many skills. and Yue (2016) conducted a study on multi-skill call centers
However, it is important to note that each individual agent is with N-design, which involves partitioning agents into distinct
unable to acquire proficiency in all skills. This shift has led to groups based on their skills and utilizing skill-based routing to
the transformation of numerous call centers from a single-skill direct customers to appropriate agents. Dam et al. (2022)
model to a multi-skill approach. In multi-skill call centers, addressed staffing optimization in multi-skill call centers,
incoming calls may be categorized into distinct types, and aiming to minimize costs while meeting quality of service
agents are differentiated based on the specific types of calls (QoS) targets. They introduced a joint chance-constrained
they are capable of managing. Gans et al. (2003) illustrated an approach that considers correlations between call types and
exemplary example of an international call center adept at employed a combination of heuristics and simulation-based
managing calls in diverse languages. The study addressed methods to find staffing solutions, showing its effectiveness in
multi-skill call center scenarios, specifically those involving a complex call center scenarios with multiple call types and
limited number of skill groups—three groups or fewer. agent groups. Chevalier (2004) presented the staffing problem
Furthermore, the research offered thorough insights into of a block model. Other research about this problem, see, e.g.,
various design approaches, encompassing V-design, M- (Liao et al., 2012).
design, W-design, and N-design. Ormeci (2004) conducted Several researchers paid attention to applying heuristic
research on the dynamic admission control problem within an methods to solve the staffing problem. Atlason et al. (2004)
M-design multi-skill call center, examining scenarios involving introduced an iterative cutting plane method for minimizing
two distinct call classes and three stations. Bhulai et al. (2008), staffing costs in a single-skill call center model. Horng and Lin
Cezik and L’Ecuyer (2008) studied the cases of more call types (2020) proposed a method that combines ordinal
and more skill groups of the multi-skill call centers. optimization with elephant herding optimization to efficiently
The increasing significance and intricacy of contemporary address staffing problems, providing a viable solution within a
call centers have led to a surge in literature dedicated to them, reasonable computation time. In Lu et al. (2023), a personnel
often concentrating on queuing models. In the context of a call scheduling problem is described, and uncertain factors are
center queuing model, the call agents are analogous to analyzed and modelled. Avramidis et al. (2009) improved two-
servers, while the calls represent customers. This theoretical stage heuristic staffing search methods in the first stage and
framework is instrumental in tackling various challenges, local search techniques in the second stage for solution
including the efficient management of customer wait times, refinement. Through a comparison with the cutting-plane
optimization of agent utilization, and meeting service level method, the researchers determined that their heuristic not
agreements. For more related surveys, we refer to the only yielded superior solutions but also resulted in faster
following works (Gans & Zhou, 2003; Koole & Mandelbaum, execution. Pot et al. (2008) introduced a straightforward
2002; Kuhl et al., 2005). Aitchanov et al. (2021) presented heuristic approach with local search for staffing in a call
solutions for optimizing contact centers, addressing center. A comparative analysis with the cutting plane method
infrastructure and financial challenges, with a focus on led to the conclusion that the heuristic exhibited several
improving customer satisfaction by examining factors related desirable characteristics, including short computation time,
to employees. Their study offers practical insights into call high accuracy, and ease of implementation. Li et al. (2017)
center management models and the scalability of introduced a search algorithm grounded in the optimal
infrastructure for emergency situations. computing budget allocation method. Through application in

Vol. 22, No. 2, April 2024 231

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

a financial call center case study, the suggested approach respectively. Specialized agents 𝑁𝑁1 , 𝑁𝑁2 and 𝑁𝑁3 can only serve
demonstrated enhanced search efficiency and yielded calls of Type 1, Type 2, and Type 3 respectively while a flexible
satisfactory outcomes. agent 𝑁𝑁4 can serve calls of both Type 1, Type 2, and Type 3. The
The rest of this paper is organized as follows: Section 2 service times of agents in Group 1, Group 2, Group 3, and Group
describes a system model. Section 3 provides the state space, 4 are all exponentially distributed with means µ1 , µ2 , µ3 , and
the state-transition rates, the steady-state probabilities, and µ4 , respectively.
the service level formula. Section 4 outlines the formulation of
the staffing problem and discusses the proposal of the 2.3. Limitation of the waiting queue
adapted algorithm to solve the problem, and Section 5 The waiting spaces for three queues are infinite. Each type of
presents a numerical example. Finally, Section 6 concludes call has its own queue. The queues of both three call types are
the paper. independent of each other. For the same type of waiting calls,
they are served by a free agent of its own group (Group 1, Group
2. Model description 2, or Group 3) and, a free agent in Group 4 according to FCFS
(First-come First-served) discipline.
In our model, the calls are categorized into three types (Type
1, Type 2, and Type 3) and agents are split into four groups 2.4. Routing policy
(Group 1, Group 2, Group 3, and Group 4) with various skills. The routing policy in our model is skill-based routing and the
This queuing model is fully characterized by customer profiles importance of the three distinct types of calls. It is assumed
(arrival process), agent properties (service process), routing that calls of Type 1 are more important than calls of Type 2 and
policies, and the limitation of the waiting queue (queuing Type 3. Also, the calls of Type 2 are more important than calls
discipline). The model is shown in Figure 1. of Type 3. In other words, the calls of Type 1 have non-
preemptive calls of Type 2 and Type 3; and the calls of Type 2
2.1. Customer profiles have non-preemptive calls of Type 3. The waiting calls
Three call types arrive according to the Poisson process with (customers) of Type 1 are served by a free agent in Group 4; the
rates 𝜆𝜆1 , 𝜆𝜆2 , and 𝜆𝜆3 respectively. There are three queues waiting calls of Type 2 are served by a free agent in Group 4 if
(Queue 1, Queue 2, and Queue 3), the arriving calls are lined there are no waiting calls of Type 1 also, the waiting calls of
with it, which contain calls of Type 1, Type 2, and Type 3, Type 3 are served by a free agent in Group 4 if there are not
respectively. There are infinite waiting spaces for three queues. waiting calls of Type 1 and Type 2.

3. Steady-state probabilities

In this section, we first determined the states of our model, then

procured the state-transition rates in two cases (call arrival and
service completion) and found the equilibrium equations for
the steady-state probabilities of the model when the model is
stationary. Finally, we computed the service level.

3.1. State space description

We have three groups with various skills in our model wherein
each of them there are three states (idle, busy, and overload),
while the fourth group has two states (idle, and busy).
I- An idle state: In this case, at least one agent is idle. This
state is symbolized by 1.
II- A busy state: In this case, all agents in the group are busy
and there are no calls waiting for service served by this group.
This state is symbolized by 2.
Figure 1. The multi-skill call center queue model. III- An overload state: In this case, all agents in the group are
busy, and there is at least one call waiting for service by this
2.2. Agent properties group. This state is symbolized by 3.
There are four categories of agents. Group 1 (has Skill 1), Group We note that the state space of the system model consisted
2 (has Skill 2), Group 3 (has Skill 3), and Group 4 (have Skill 1, of 34 states according to the routing policy assumed above.
Skill 2, and Skill 3) consist of 𝑁𝑁1 , 𝑁𝑁2 , 𝑁𝑁3 , and 𝑁𝑁4 agents, Thus, the state space is given by:

Vol. 22, No. 2, April 2024 232

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

𝑬𝑬 = {(𝟏𝟏 𝟏𝟏 𝟏𝟏 𝟏𝟏), (𝟏𝟏 𝟏𝟏 𝟐𝟐 𝟏𝟏), (𝟏𝟏 𝟏𝟏 𝟐𝟐 𝟐𝟐), (𝟏𝟏 𝟏𝟏 𝟑𝟑 𝟐𝟐) Group 1 is 𝑛𝑛1 = 𝑁𝑁1 − 1 for state 𝑆𝑆1 . Note that 𝑛𝑛1 < 𝑁𝑁1 ,
, (𝟏𝟏 𝟐𝟐 𝟏𝟏 𝟏𝟏), (𝟏𝟏 𝟐𝟐 𝟏𝟏 𝟐𝟐), (𝟏𝟏 𝟐𝟐 𝟐𝟐 𝟐𝟐), (𝟏𝟏 𝟐𝟐 𝟐𝟐 𝟏𝟏) 𝑛𝑛2 < 𝑁𝑁2 , and 𝑛𝑛3 < 𝑁𝑁3 (i.e., if the operation is in state 𝑆𝑆1 then
, (𝟏𝟏 𝟐𝟐 𝟑𝟑 𝟐𝟐), (𝟏𝟏 𝟑𝟑 𝟏𝟏 𝟐𝟐), (𝟏𝟏 𝟑𝟑 𝟐𝟐 𝟐𝟐), (𝟏𝟏 𝟑𝟑 𝟑𝟑 𝟐𝟐) the number of calls of either Type 1 or Type 2 or Type 3 is less
, (𝟐𝟐 𝟏𝟏 𝟏𝟏 𝟏𝟏), (𝟐𝟐 𝟏𝟏 𝟏𝟏 𝟐𝟐), (𝟐𝟐 𝟏𝟏 𝟐𝟐 𝟏𝟏), (𝟐𝟐 𝟏𝟏 𝟐𝟐 𝟐𝟐) than the number of agents either in Group 1 or Group 2 or
, (𝟐𝟐 𝟏𝟏 𝟑𝟑 𝟐𝟐), (𝟐𝟐 𝟐𝟐 𝟏𝟏 𝟏𝟏), (𝟐𝟐 𝟐𝟐 𝟏𝟏 𝟐𝟐), (𝟐𝟐 𝟐𝟐 𝟐𝟐 𝟐𝟐) Group 3), and that the three queues are independent of each
, (𝟐𝟐 𝟐𝟐 𝟐𝟐 𝟏𝟏), (𝟐𝟐 𝟐𝟐 𝟑𝟑 𝟐𝟐), (𝟐𝟐 𝟑𝟑 𝟏𝟏 𝟐𝟐), (𝟐𝟐 𝟑𝟑 𝟐𝟐 𝟐𝟐) other so the results of the 𝑀𝑀/𝑀𝑀/𝑐𝑐/𝑐𝑐 loss queuing system can
, (𝟐𝟐 𝟑𝟑 𝟑𝟑 𝟐𝟐), (𝟑𝟑 𝟏𝟏 𝟏𝟏 𝟐𝟐) , (𝟑𝟑 𝟏𝟏 𝟐𝟐 𝟐𝟐), (𝟑𝟑 𝟏𝟏 𝟑𝟑 𝟐𝟐)
be used. Consequently, we have
, (𝟑𝟑 𝟐𝟐 𝟏𝟏 𝟐𝟐), (𝟑𝟑 𝟐𝟐 𝟐𝟐 𝟐𝟐), (3 2 3 2), (3 3 1 2)
, (3 3 2 2), (3 3 3 2)} 𝑁𝑁 −1
𝜌𝜌1 1
𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) = 𝑗𝑗 , (2)
𝑁𝑁1 𝜌𝜌1
let 𝑆𝑆𝑖𝑖 , (𝑖𝑖 = 1, 2, 3, . .. , 34) be the ith state in the state (𝑁𝑁1 − 1)! ∑𝑗𝑗=0
𝑗𝑗!
space E. Let 𝑛𝑛𝑚𝑚 , (𝑚𝑚 = 1, 2, 3) be the number of calls waiting
in a queue that are assumed to be served by Group m plus the Where
busy agents in Group 𝑚𝑚, while 𝑛𝑛4 be the number of calls being 𝜆𝜆1 𝜆𝜆1
serviced by agents of Group 4 as no waiting calls in the queue 𝜌𝜌1 = , 𝑎𝑎1 = .
µ1 𝑁𝑁1 µ1
according to routing policy.
In a comparable way, we can acquire the other
3.2. The construction of the state-transition rates transition rates 𝑞𝑞(𝑖𝑖−j) occurred by call arrival as shown in
We derive the state-transition rates by using the results of Addendum A.1 of Appendix A.
𝑀𝑀/𝑀𝑀/𝑐𝑐 and 𝑀𝑀/𝑀𝑀/𝑐𝑐/𝑐𝑐 queueing systems (which are given by
Shortle et al. (2018). There are only two events that can make 3.2.2. The transfer of states due to the fulfillment of service
the state transfer: call arrival or service fulfillment. We will be We assume the state 𝑆𝑆2 = (1 1 2 1), and 𝑆𝑆1 =
debating the two cases separately to obtain how the state (1 1 1 1), for example. If the call of Type 3 finished the
transition rates are computed. service, then the set of states will be varied from state 𝑆𝑆2
to 𝑆𝑆1 . The diagram of the transition from state 𝑆𝑆2 to state
3.2.1. The transfer of states due to the call arrival 𝑆𝑆1 is shown in Figure 3.
We assume the state 𝑆𝑆1 = (1 1 1 1), and 𝑆𝑆13 = (2 1 1 1), for
example. The diagram of the transition from state 𝑆𝑆1 to state
𝑆𝑆13 is shown in Figure 2.

Figure 3. Diagram of the state-transition rate

from the state 𝑆𝑆2 to the state 𝑆𝑆1 .

In the state 𝑆𝑆2 , 𝑛𝑛3 = 𝑁𝑁3 and 𝑁𝑁3 µ3 is the service rate
Figure 2. Diagram of the state-transition rate for the call of Type 3. The trigger for the transfer from
from the state 𝑆𝑆1 to the state 𝑆𝑆13 .
state 𝑆𝑆2 to 𝑆𝑆1 is obtained as follows:
Let 𝑞𝑞(𝑖𝑖−j) , (𝑖𝑖, 𝑗𝑗 = 1, 2, 3, . .. , 34) denote the state- 𝑞𝑞2−1 = 𝑁𝑁3 µ3 . (3)
transition rate. The trigger for the transfer from the state 𝑆𝑆1 to
the state 𝑆𝑆13 is obtained as follows: In a comparable way, we can acquire the other
transition rates 𝑞𝑞(𝑖𝑖−j) caused by the completion of service
𝑞𝑞1−13 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1), (1) of a call as shown in Addendum A.2 of Appendix A.
The results of the 𝑀𝑀/𝑀𝑀/𝑐𝑐 queuing system can be
where 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) is the probability that the used to obtain the probabilities of 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1),
number of call Type 1 needing to be serviced by the agents in 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1), and 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) as follows:

Vol. 22, No. 2, April 2024 233

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

∑34
𝑖𝑖=1 𝑃𝑃𝑖𝑖 = 1. (19)
𝑁𝑁 +1
𝜌𝜌1 1
𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1) = 𝑃𝑃1 , (4)
𝑁𝑁1 (𝑁𝑁1 )! 0 Remark. The description of the 𝑋𝑋𝑦𝑦 , 𝑦𝑦 = 1, 2, 3, … ,68 is
𝑁𝑁 +1
mentioned in Addendum A.3 of Appendix A.
𝜌𝜌2 2
𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) = 𝑃𝑃2 , (5) We note that “Equation 19” means the total probability of
𝑁𝑁2 (𝑁𝑁2 )! 0
being in any of the steady- state states is equal to 1. Since the
𝑁𝑁 +1
𝜌𝜌3 3
above equations are linear, so by solving these equations,
𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) = 𝑃𝑃3 , (6) (using MATLAB software), we can obtain all the steady-state
𝑁𝑁3 (𝑁𝑁3 )! 0
where probabilities.

𝑁𝑁1 −1
𝑗𝑗
𝜌𝜌1
𝑁𝑁
𝑁𝑁1 𝜌𝜌1 1 3.4. The computation of service level
𝑃𝑃01 =[� + ]−1 , Calculating service level is used in our paper to evaluate the
𝑗𝑗! 𝑁𝑁1 ! (𝑁𝑁1 − 𝜌𝜌1 )
𝑗𝑗=0 performance of a multi-skill call center and is calculated by
using steady-state probabilities. The service level can be
𝑁𝑁2 −1
𝑗𝑗
𝜌𝜌2
𝑁𝑁
𝑁𝑁2 𝜌𝜌2 2 expressed as the percentage of the calls that should be
𝑃𝑃02 =[� + ]−1 , 𝑃𝑃
serviced within a given waiting time 𝑇𝑇𝑖𝑖 , denoted as 𝑖𝑖.
𝑗𝑗! 𝑁𝑁2 ! (𝑁𝑁2 − 𝜌𝜌2 ) 𝑇𝑇𝑖𝑖
𝑗𝑗=0
Let 𝑃𝑃𝑠𝑠𝑠𝑠𝑖𝑖 = 1 − 𝑃𝑃𝑛𝑛𝑛𝑛
𝑖𝑖
, 𝑖𝑖 = 1, 2, 3 be the probability that the call
𝑁𝑁3 −1
𝑗𝑗 𝑁𝑁 of Type 𝑖𝑖 is serviced in a fixed time 𝑇𝑇𝑖𝑖 .
𝜌𝜌 𝑁𝑁3 𝜌𝜌3 3 Consider the call of Type 1, for example. We suppose that
𝑃𝑃03 =[� 3+ ]−1 .
𝑗𝑗! 𝑁𝑁3 ! (𝑁𝑁3 − 𝜌𝜌3 ) the service level of the call of Type 1 is known as the probability
𝑗𝑗=0
of the calls are not serviced in a fixed time 𝑇𝑇1 . Calls of Type 1
3.3. The computation of steady-state probabilities has a queue only occur in the states 𝑆𝑆26 = (3 1 1 2), 𝑆𝑆27 =
The steady-state probabilities of the state procedure are (3 1 2 2), 𝑆𝑆28 = (3 1 3 2), 𝑆𝑆29 = (3 2 1 2), 𝑆𝑆30 =
denoted by 𝑃𝑃𝑖𝑖 , 𝑖𝑖 = 1, 2, 3, . .. , 34 and are shown for each (3 2 2 2), 𝑆𝑆31 = (3 2 3 2), 𝑆𝑆32 = (3 3 1 2), 𝑆𝑆33 =
state by applying the equilibrium equation of the system (rate (3 3 2 2), and 𝑆𝑆34 = (3 3 3 2). The service rate for the call of
in = rate out) as follows: Type 1 in each state of 𝑆𝑆26 , 𝑆𝑆27 , 𝑆𝑆28 , 𝑆𝑆29 , 𝑆𝑆30 , 𝑆𝑆31 , 𝑆𝑆32 , 𝑆𝑆33 , and
𝑆𝑆34 . is 𝑁𝑁1 µ1 + 𝑁𝑁4 µ4 as all agents in Group 1 are busy and there
𝑃𝑃1 𝑋𝑋1 = 𝑋𝑋2 , 𝑃𝑃2 𝑋𝑋3 = 𝑋𝑋4 , 𝑃𝑃3 𝑋𝑋5 = 𝑋𝑋6 , (7) are waiting calls of Type 1 in a queue if an agent in Group 4
finished its service, an agent will select the call of Type 1 in the
𝑃𝑃4 𝑋𝑋7 = 𝑋𝑋8 , 𝑃𝑃5 𝑋𝑋9 = 𝑋𝑋10 , 𝑃𝑃6 𝑋𝑋11 = 𝑋𝑋12 , (8) queue to serve instantly. Therefore, the number of calls that
could be served in time 𝑇𝑇1 is 𝑇𝑇1 (𝑁𝑁1 µ1 + 𝑁𝑁4 µ4 ). We can obtain
𝑃𝑃7 𝑋𝑋13 = 𝑋𝑋14 , 𝑃𝑃8 𝑋𝑋15 = 𝑋𝑋16 , 𝑃𝑃9 𝑋𝑋17 = 𝑋𝑋18 , (9) the probability of the call of Type 1 cannot be serviced in time
1
𝑇𝑇1 , 𝑃𝑃𝑛𝑛𝑛𝑛 , as follows:
𝑃𝑃10 𝑋𝑋19 = 𝑋𝑋20 , 𝑃𝑃11 𝑋𝑋21 = 𝑋𝑋22 , 𝑃𝑃12 𝑋𝑋23 = 𝑋𝑋24 , (10)
1
𝑃𝑃𝑛𝑛𝑛𝑛 = 𝑊𝑊1 ∑∞
𝑖𝑖=𝐾𝐾1 𝑃𝑃(𝑛𝑛1 = 𝑖𝑖), (20)
𝑃𝑃13 𝑋𝑋25 = 𝑋𝑋26 , 𝑃𝑃14 𝑋𝑋27 = 𝑋𝑋28 , 𝑃𝑃15 𝑋𝑋29 = 𝑋𝑋30 , (11)
Where
𝑃𝑃16 𝑋𝑋31 = 𝑋𝑋32 , 𝑃𝑃17 𝑋𝑋33 = 𝑋𝑋34 , 𝑃𝑃18 𝑋𝑋35 = 𝑋𝑋36 , (12)
𝑊𝑊1 = (𝑃𝑃26 + 𝑃𝑃27 + 𝑃𝑃28 + 𝑃𝑃29 + 𝑃𝑃30 + 𝑃𝑃31 + 𝑃𝑃32
𝑃𝑃19 𝑋𝑋37 = 𝑋𝑋38 , 𝑃𝑃20 𝑋𝑋39 = 𝑋𝑋40 , 𝑃𝑃21 𝑋𝑋41 = 𝑋𝑋42 , (13)
+ 𝑃𝑃33 + 𝑃𝑃34 ), (21)
𝑃𝑃22 𝑋𝑋43 = 𝑋𝑋44 , 𝑃𝑃23 𝑋𝑋45 = 𝑋𝑋46 , 𝑃𝑃24 𝑋𝑋47 = 𝑋𝑋48 , (14)
𝐾𝐾1 = 𝑁𝑁1 + 𝑁𝑁4 + 𝑇𝑇1 (𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁4 𝜇𝜇4 ). (22)
𝑃𝑃25 𝑋𝑋49 = 𝑋𝑋50 , 𝑃𝑃26 𝑋𝑋51 = 𝑋𝑋52 , 𝑃𝑃27 𝑋𝑋53 = 𝑋𝑋54 , (15)
Similarly, for a call of Type 2, calls have a queue only occur
𝑃𝑃28 𝑋𝑋55 = 𝑋𝑋56 , 𝑃𝑃29 𝑋𝑋57 = 𝑋𝑋58 , 𝑃𝑃30 𝑋𝑋59 = 𝑋𝑋60 , (16)
in the states 𝑆𝑆10 = (1 3 1 2), 𝑆𝑆11 = (1 3 2 2), 𝑆𝑆12 =
(1 3 3 2), 𝑆𝑆23 = (2 3 1 2), 𝑆𝑆24 = (2 3 2 2) and 𝑆𝑆25 =
𝑃𝑃31 𝑋𝑋61 = 𝑋𝑋62 , 𝑃𝑃32 𝑋𝑋63 = 𝑋𝑋64 , 𝑃𝑃33 𝑋𝑋65 = 𝑋𝑋66 , (17)
(2 3 3 2). Thus, we can obtain the probability of the call of
2
Type 2 cannot be serviced in time 𝑇𝑇2 , 𝑃𝑃𝑛𝑛𝑛𝑛 , as follows:
𝑃𝑃34 𝑋𝑋67 = 𝑋𝑋68 , (18) 2 ∞
𝑃𝑃𝑛𝑛𝑛𝑛 = 𝑊𝑊2 ∑𝑖𝑖=𝐾𝐾2 𝑃𝑃(𝑛𝑛2 = 𝑖𝑖), (23)

Vol. 22, No. 2, April 2024 234

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

where variables, namely, 𝑁𝑁1 , 𝑁𝑁2 , 𝑁𝑁3 , and 𝑁𝑁4 . The constraints pertain
to incoming calls aimed at meeting specific service level
𝑊𝑊2 = (𝑃𝑃10 + 𝑃𝑃11 + 𝑃𝑃12 + 𝑃𝑃23 + 𝑃𝑃24 + 𝑃𝑃25 ), (24) criteria. Notably, the constraints are intricately nonlinear, as
evident from the service level formula. Given the involvement
𝐾𝐾2 = 𝑁𝑁2 + 𝑁𝑁4 + 𝑇𝑇2 (𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁4 𝜇𝜇4 ). (25) of four variables in this model, an effective resolution can be
achieved through the application of the following algorithm
And, by the same analysis method above, we can obtain using MATLAB software to obtain the optimal numbers of
the probability of the call of Type 3 cannot be serviced in time every agent group and the minimum cost associated with it.
3
𝑇𝑇3 , 𝑃𝑃𝑛𝑛𝑛𝑛 , as follows:

3
𝑃𝑃𝑛𝑛𝑛𝑛 = 𝑊𝑊3 ∑∞
𝑖𝑖=𝐾𝐾3 𝑃𝑃(𝑛𝑛3 = 𝑖𝑖), (26)

where

𝑊𝑊3 = (𝑃𝑃4 + 𝑃𝑃9 + 𝑃𝑃17 + 𝑃𝑃22 ), (27)

𝐾𝐾3 = 𝑁𝑁3 + 𝑁𝑁4 + 𝑇𝑇3 (𝑁𝑁3 𝜇𝜇3 + 𝑁𝑁4 𝜇𝜇4 ). (28)

And, 𝑃𝑃(𝑛𝑛1 = 𝑖𝑖), 𝑃𝑃(𝑛𝑛2 = 𝑖𝑖), and 𝑃𝑃(𝑛𝑛3 = 𝑖𝑖) are the
probabilities that there are i customers in the 𝑀𝑀/𝑀𝑀/𝑁𝑁1 ,
𝑀𝑀/𝑀𝑀/𝑁𝑁2 , and 𝑀𝑀/𝑀𝑀/𝑁𝑁3 queueing system with arrival rate 𝜆𝜆1 ,
𝜆𝜆2 , and 𝜆𝜆3 and service rates µ1 , µ2 , and µ3 , respectively, which 5. Example
their formulas are given in (Shortle et al., 2018). By using
MATLAB software, we can get the service levels 𝑃𝑃𝑠𝑠𝑠𝑠1 , 𝑃𝑃𝑠𝑠𝑠𝑠2 , and 𝑃𝑃𝑠𝑠𝑠𝑠3 .
The purpose of this example is to indicate how these factors,
like service levels and the number of agents in each group,
4. Staffing optimization affect the entire system. We consider the same system model
introduced earlier in the above sections, but with the following
We offered the optimization model of the staffing problem to parameters. Arrival rates are 𝜆𝜆1 = 5, 𝜆𝜆2 = 4, and 𝜆𝜆3 = 3,
find the optimal number of agents in each group to minimize and service rates are µ1 = 0.8, µ2 = 0.6, µ3 = 0.4, µ4 =
the cost of the system while ensuring that the service level for 0.2, 𝛼𝛼1 = 0.8, 𝛼𝛼2 = 0.7, and 𝛼𝛼3 = 0.6. The fixed waiting
each call type meets the minimum service level requirement. times are 𝑇𝑇1 = 20, 𝑇𝑇2 = 30, and 𝑇𝑇3 = 40. The cost for each
The service rate of each call type is a key factor in determining agent in the four groups are 𝐶𝐶1 = 25, 𝐶𝐶2 = 20, 𝐶𝐶3 = 15,
the optimal number of agents in each group. Supposing that and 𝐶𝐶4 = 10. To show how the results in Tables (1, 2, 3) were
the cost of the agents’ Group 1 is 𝐶𝐶1 , the cost of the agents’ obtained, let us take case (1), for example, when 𝑁𝑁1 = 21,
Group 2 is 𝐶𝐶2 , the cost of the agents’ Group 3 is 𝐶𝐶3 , and the cost 𝑁𝑁2 = 16, 𝑁𝑁3 = 8, and 𝑁𝑁4 = 5. By giving the MATLAB
of the agents’ Group 4 is 𝐶𝐶4 .To minimize the system models’ software, the parameter settings are as follows: 𝜆𝜆1 = 5, 𝜆𝜆2 =
cost, we seek to get the optimal number of agents 𝑁𝑁1 , 𝑁𝑁2 , 𝑁𝑁3 , 4, 𝜆𝜆3 = 3, µ1 = 0.8, µ2 = 0.6, µ3 = 0.4, µ4 = 0.2, 𝑇𝑇1 =
and 𝑁𝑁4 . The staffing optimization can be expressed as follows: 20, 𝑇𝑇2 = 30, 𝑇𝑇3 = 40, 𝐶𝐶1 = 25, 𝐶𝐶2 = 20, 𝐶𝐶3 = 15,
𝐶𝐶4 = 10, (𝑎𝑎, 𝑏𝑏) = (2,100) and utilizing “Equations from
min 𝑧𝑧 = 𝐶𝐶1 𝑁𝑁1 + 𝐶𝐶2 𝑁𝑁2 + 𝐶𝐶3 𝑁𝑁3 + 𝐶𝐶4 𝑁𝑁4 , (A.1) to (A.89)” we get all the state-transition rates of both call
arrival and service fulfilment, then by substituting in
𝑠𝑠. 𝑡𝑡. 𝑃𝑃𝑠𝑠𝑠𝑠1 ≥ 𝛼𝛼1 , “Equations from (7) to (18)”, we get all the steady-state
𝑃𝑃𝑠𝑠𝑠𝑠2 ≥ 𝛼𝛼2 , (29) probabilities, and by using “Equations from (20) to (28)”, we
𝑃𝑃𝑠𝑠𝑠𝑠3 ≥ 𝛼𝛼3 , calculate the service levels, and from “Equation (29)” and
𝑎𝑎 ≤ 𝑁𝑁𝑖𝑖 ≤ 𝑏𝑏, 𝑎𝑎, 𝑏𝑏, 𝑁𝑁𝑖𝑖 ∈ 𝑍𝑍 + ; 𝑖𝑖 = 1,2,3,4. applying the above algorithm, we get the optimal number of
agents in each group. By the same manner, we can get the
where 𝛼𝛼1 , 𝛼𝛼2 , and 𝛼𝛼3 are the given service rate of the call other cases.
Type 1, the call Type 2, and the call Type 3, respectively, 𝑍𝑍 + Based on the results in Table (1), we can see that all steady-
denote the set of positive integers. The number of agents state probabilities are present in small-sized cases (50 or less
within each group may be chosen within the interval of (𝑎𝑎, 𝑏𝑏). agents), while in medium (51-200 agents) or large (more than
This problem is nonlinear integer programming with a linear 201 agents)-sized call centers, some steady-state probabilities
objective function. The optimization involves determining four are absent in each case, indicating a reduction in the complexity

Vol. 22, No. 2, April 2024 235

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

of the system as it scales. Comparing probabilities across cases Table 3. Numerical results of the optimal 𝑁𝑁1 , 𝑁𝑁2 , 𝑁𝑁3 , and 𝑁𝑁4 .
reveals patterns and variations in system behavior, providing
insights into resource utilization and performance.

Table 1. Numerical results of the steady-state probabilities.

6. Conclusion

The paper focuses on a multi-skill call center that operates on a

queuing model, specifically the exponential model. The call
center has three types of customers and four groups of agents,
with specialized agents attending to their respective customer
type and flexible agents managing both types. By examining the
state space of the system and using results from M/M/c/c and
M/M/c queuing systems, the transition rates of the state sets were
determined. Equilibrium equations for the system’s steady-state
equations were also formulated. The paper presents a formula for
computing the service level and discusses the technique used to
calculate the optimal number of agents in each group to solve the
staffing problem. Overall, these results can provide insights and
recommendations for call center managers to improve the
Table (2) demonstrates that the service level formula is service levels and optimize their staffing levels. The paper’s
applicable to call centers of any scale. In multi-skill call centers, methodology easily adapts to various multi-skill call centers and
the total number of agents remains constant, but the proves valuable for evaluating contact center performance.
composition of agent groups significantly impacts the overall Future extensions could include analyzing non-exponential
service level. Hence, optimizing the numbers of each agent models with general service times, impatient times, and skill-
group is crucial to achieve the desired service level while based routing without priority.
minimizing system costs.

Table 2. Numerical results of the service levels.

Conflict of interest

The authors have no conflict of interest to declare.

Acknowledgements

Authors are grateful for the reviewer’s valuable comments that

improved the paper.

Funding
As shown in Table (3). The table shows that the optimal
number of agents in each group varies for each case, and it also
The authors received no specific funding for this work.
affects the system cost. It demonstrates that the system cost
can be minimized by adjusting the number of agents in each
group based on the service rate and cost for each group.

Vol. 22, No. 2, April 2024 236

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

References Gans, N., Koole, G., & Mandelbaum, A. (2003). Telephone call
centers: Tutorial, review, and research
Aksin, Z., Armony, M., & Mehrotra, V. (2007). The modern call prospects. Manufacturing & Service Operations
center: a multi-disciplinary perspective on operations Management, 5(2), 79-141.
management research. Production and Operations https://doi.org/10.1287/msom.5.2.79.16071
Management, 16(6), 665-688.
https://doi.org/10.1111/j.1937-5956.2007.tb00288.x Gans, N., & Zhou, Y. P. (2003). A call-routing problem with
service-level constraints. Operations Research, 51(2), 255-271.
Aitchanov, B., Baimuratov, O., Zhussupekov, M., & Aitzhanov, https://doi.org/10.1287/opre.51.2.255.12787
T. (2021). Contact centers management models: analysis and
recommendations. In 2021 IEEE International Conference on Horng, S. C., & Lin, S. S. (2020). Staffing Optimization of One-
Smart Information Systems and Technologies (SIST) (pp. 1-5). Period Multi-Skill Call Center. In Proceedings of the 2020 the 3rd
IEEE. International Conference on Computers in Management and
Business (pp. 89-93).
Atlason, J., Epelman, M.A., and Henderson, S.G. (2004). Call https://doi.org/10.1145/3383845.3383849
center staffing with simulation and cutting plane methods.
Annals of Operations Research, 127, 333-358. Koole, G., & Mandelbaum, A. (2002). Queueing models of call
https://doi.org/10.1023/B:ANOR.0000019095.91642.bb centers: An introduction. Annals of Operations Research, 113,
41-59.
Avramidis, A. N., Chan, W., & L'Ecuyer, P. (2009). Staffing multi- https://doi.org/10.1023/A:1020949626017
skill call centers via search methods and a performance
approximation. Iie Transactions, 41(6), 483-497. Koole, G., & Pot, A. (2006). An overview of routing and staffing
https://doi.org/10.1080/07408170802322986 algorithms in multi-skill customer contact centers.

Bhulai, S., Koole, G., & Pot, A. (2008). Simple methods for shift Kuhl, M. E., Steiger, N. M., Armstrong, F. B., & Joines, J. A. (2005).
scheduling in multiskill call centers. Manufacturing & Service Modeling and simulation of call centers. In Winter Simulation
Operations Management, 10(3), 411-420. Conference (Vol. 38).
https://doi.org/10.1287/msom.1070.0172
Li, C., Sun, C., & Zhang, J. (2020). The staffing optimization
Cezik, M. T., & L'Ecuyer, P. (2008). Staffing multiskill call centers problem for the m-design multi-skill call center based on
via linear programming and simulation. Management queuing model. In Proceedings of the 11th International
Science, 54(2), 310-323. Conference on Modelling, Identification and Control
https://doi.org/10.1287/mnsc.1070.0824 (ICMIC2019) (pp. 585-595). Springer Singapore.
https://doi.org/10.1007/978-981-15-0474-7_55
Chevalier, P., Shumsky, R. A., & Tabordon, N. (2004). Routing
and staffing in large call centers with specialized and fully Li, C. Y., & Yue, D. Q. (2016). The staffing problem of the N-design
flexible servers. Submitted to Manufacturing and Service multi-skill call center based on queuing model. In 3rd
Operations Management. International Conference on Wireless Communication and
Sensor Networks (WCSN 2016) (pp. 427-432). Atlantis Press.
Dam, T. T., Ta, T. A., & Mai, T. (2022). Joint chance-constrained https://doi.org/10.2991/icwcsn-16.2017.90
staffing optimization in multi-skill call centers. Journal of
Combinatorial Optimization, 1-25. Li, N., Yu, X., & Matta, A. (2017). Modelling and workload
https://doi.org/10.1007/s10878-021-00830-1 reallocation of call centres with multi-type
customers. International Journal of Production
Defraeye, M., & Van Nieuwenhuyse, I. (2016). Staffing and Research, 55(19), 5664-5680.
scheduling under nonstationary demand for service: A https://doi.org/10.1080/00207543.2017.1329958
literature review. Omega, 58, 4-25.
https://doi.org/10.1016/j.omega.2015.04.002 Liao, S., Koole, G., van Delft, C., & Jouini, O. (2012). Staffing a call
center with uncertain non-stationary arrival rate and
flexibility. OR spectrum, 34, 691-721.
https://doi.org/10.1007/s00291-011-0257-0

Vol. 22, No. 2, April 2024 237

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

Lu, X., Wu, C., Yang, X., Zhang, M., & Zheng, Y. (2023). Adapted
water wave optimization for integrated bank customer service
representative scheduling. International Journal of Production
Research, 61(1), 320-335.
https://doi.org/10.1080/00207543.2021.1942284

Ormeci, E. L. (2004). Dynamic admission control in a call center

with one shared and two dedicated service facilities. IEEE
Transactions on Automatic Control, 49(7), 1157-1161.
https://doi.org/10.1109/TAC.2004.831133

Priya, K. B., & Rajendran, P. (2018). Nonlinear Programming

Problem for an M-Design Multi-Skill Call Center with
Impatience Based on Queueing Model Method. In Advances in
Algebra and Analysis: International Conference on Advances in
Mathematical Sciences, Vellore, India, December 2017-Volume
I (pp. 243-251). Springer International Publishing.
https://doi.org/10.1007/978-3-030-01120-8_28

Shortle, J. F., Thompson, J. M., Gross, D., & Harris, C. M.

(2018). Fundamentals of queueing theory (Vol. 399). John Wiley
& Sons.

Pot, A., Bhulai, S., & Koole, G. (2008). A simple staffing method
for multiskill call centers. Manufacturing & service operations
management, 10(3), 421-428.
https://doi.org/10.1287/msom.1070.0173

Wallace, R. B., & Whitt, W. (2005). A staffing algorithm for call

centers with skill-based routing. Manufacturing & Service
Operations Management, 7(4), 276-294.
https://doi.org/10.1287/msom.1050.0086

Vol. 22, No. 2, April 2024 238

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

Appendix 𝑞𝑞14−27 = 𝑞𝑞19−30 = 𝑞𝑞23−33 = 𝜆𝜆1 + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1)

(A.28)
A.1. Addendum 𝑞𝑞6−11 = 𝑞𝑞19−24 = 𝑞𝑞29−33 = 𝜆𝜆2 + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1)
(A.29)
𝑞𝑞19−33 = 𝜆𝜆1 + 𝜆𝜆2 + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.30)
In this addendum to Appendix A, we present the other
𝑞𝑞3−22 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) + 𝜆𝜆3
transition rates 𝑞𝑞(𝑖𝑖−𝑗𝑗) occurred by call arrival. The results are
(A.31)
re-listed below.
𝑞𝑞6−24 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1)
(A.32)
𝑞𝑞1−13 = 𝑞𝑞2−15 = 𝑞𝑞3−16 = 𝑞𝑞4−17 = 𝑞𝑞5−18 = 𝑞𝑞6−19 𝑞𝑞14−30 = 𝜆𝜆1 + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1)
= 𝑞𝑞7−20 = 𝑞𝑞8−21 = 𝑞𝑞9−22 = 𝑞𝑞10−23 = 𝑞𝑞11−24
+ 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.33)
= 𝑞𝑞12−25 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) (A.1)
𝑞𝑞5−19 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 𝑃𝑃1 (𝑛𝑛4 = 𝑁𝑁4 − 1)
𝑞𝑞1−5 = 𝑞𝑞2−8 = 𝑞𝑞3−7 = 𝑞𝑞4−9 = 𝑞𝑞13−18 = 𝑞𝑞14−19
(A.34)
= 𝑞𝑞15−21 = 𝑞𝑞16−20 = 𝑞𝑞17−22 = 𝑞𝑞26−29 = 𝑞𝑞27−30
𝑞𝑞2−16 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆3 𝑃𝑃2 (𝑛𝑛4 = 𝑁𝑁4 − 1)
= 𝑞𝑞28−31 = 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) (A.2)
(A.35)
𝑞𝑞1−2 = 𝑞𝑞5−8 = 𝑞𝑞6−7 = 𝑞𝑞10−11 = 𝑞𝑞13−15 = 𝑞𝑞14−16
𝑞𝑞2−7 = 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) + 𝜆𝜆3 𝑃𝑃2 (𝑛𝑛4 = 𝑁𝑁4 − 1)
= 𝑞𝑞18−21 = 𝑞𝑞19−20 = 𝑞𝑞23−24 = 𝑞𝑞26−27 = 𝑞𝑞29−30
(A.36)
= 𝑞𝑞32−33 = 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.3)
𝑞𝑞13−19 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1)
𝑞𝑞14−26 = 𝑞𝑞16−27 = 𝑞𝑞17−28 = 𝑞𝑞19−29 = 𝑞𝑞20−30
(A.37)
= 𝑞𝑞22−31 = 𝑞𝑞23−32 = 𝑞𝑞24−33 = 𝑞𝑞25−34
= 𝜆𝜆1 (A.4)
𝑞𝑞6−10 = 𝑞𝑞7−11 = 𝑞𝑞9−12 = 𝑞𝑞19−23 = 𝑞𝑞20−24 = 𝑞𝑞22−25 𝑞𝑞5−7 = 𝜆𝜆2 𝑃𝑃1 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1)
= 𝑞𝑞29−32 = 𝑞𝑞30−33 = 𝑞𝑞31−34 = 𝜆𝜆2 (A.5) (A.38)
𝑞𝑞3−4 = 𝑞𝑞7−9 = 𝑞𝑞11−12 = 𝑞𝑞16−17 = 𝑞𝑞20−22 = 𝑞𝑞24−25 𝑞𝑞13−16 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1)
= 𝑞𝑞27−28 = 𝑞𝑞30−31 = 𝑞𝑞33−34 = 𝜆𝜆3 (A.6) (A.39)
𝑞𝑞1−18 = 𝑞𝑞2−21 = 𝑞𝑞3−20 = 𝑞𝑞4−22 𝑞𝑞2−20 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1)
= 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) (A.7) + 𝜆𝜆3 𝑃𝑃2 (𝑛𝑛4 = 𝑁𝑁4 − 1) (A.40)
𝑞𝑞1−15 = 𝑞𝑞5−21 = 𝑞𝑞6−20 = 𝑞𝑞10−24 𝑞𝑞5−20 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 𝑃𝑃1 (𝑛𝑛4 = 𝑁𝑁4 − 1)
= 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.8) + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.41)
𝑞𝑞1−8 = 𝑞𝑞13−21 = 𝑞𝑞14−20 = 𝑞𝑞26−30 𝑞𝑞13−20 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1)
= 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.9) + 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.42)
𝑞𝑞1−21 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) 𝑞𝑞8−22 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 𝑃𝑃1 (𝑛𝑛4 = 𝑁𝑁4 − 1)
+ 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.10) + 𝜆𝜆3 (A.43)
𝑞𝑞15−22 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1)
𝑞𝑞19−32 = 𝑞𝑞20−33 = 𝑞𝑞22−34 = 𝜆𝜆1 + 𝜆𝜆2 (A.11)
+ 𝜆𝜆3 (A.44)
𝑞𝑞16−28 = 𝑞𝑞20−31 = 𝑞𝑞24−34 = 𝜆𝜆1 + 𝜆𝜆3 (A.12)
𝑞𝑞18−24 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆2
𝑞𝑞7−12 = 𝑞𝑞20−25 = 𝑞𝑞30−34 = 𝜆𝜆2 + 𝜆𝜆3 (A.13)
+ 𝜆𝜆3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) (A.45)
𝑞𝑞20−34 = 𝜆𝜆1 + 𝜆𝜆2 + 𝜆𝜆3 (A.14)
Where
𝑞𝑞13−14 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) (A.15) 𝑁𝑁 −1
𝜌𝜌2 2
𝑞𝑞5−6 = 𝜆𝜆2 𝑃𝑃1 (𝑛𝑛4 = 𝑁𝑁4 − 1) (A.16) 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) = ,
𝑗𝑗
𝑞𝑞2−3 = 𝜆𝜆3 𝑃𝑃2 (𝑛𝑛4 = 𝑁𝑁4 − 1) (A.17) 𝜌𝜌2
(𝑁𝑁2 − 1)! ∑𝑁𝑁 2
𝑗𝑗=0 𝑗𝑗!
𝑞𝑞18−23 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆2 (A.18)
𝜆𝜆2 𝜆𝜆2
𝑞𝑞8−9 = 𝜆𝜆2 𝑃𝑃1 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆3 (A.19) 𝜌𝜌2 = , 𝑎𝑎2 =
µ2
.
𝑁𝑁2 µ2
(A.46)
𝑞𝑞15−17 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆3 (A.20) 𝑁𝑁 −1
𝜌𝜌3 3
𝑞𝑞21−25 = 𝜆𝜆1 𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) + 𝜆𝜆2 + 𝜆𝜆3 (A.21) 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 − 1) = 𝑗𝑗
,
𝑞𝑞6−23 = 𝑞𝑞7−24 = 𝑞𝑞9−25 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) 𝑁𝑁3 𝜌𝜌3
(𝑁𝑁3 − 1)! ∑𝑗𝑗=0
+ 𝜆𝜆2 (A.22) 𝑗𝑗!
𝜆𝜆3 𝜆𝜆3
𝑞𝑞3−17 = 𝑞𝑞7−22 = 𝑞𝑞11−25 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) 𝜌𝜌3 = , 𝑎𝑎3 = . (A.47)
µ3 𝑁𝑁3 µ3
+ 𝜆𝜆3 (A.23) 𝑁𝑁 −1
𝜌𝜌4 4
𝑞𝑞7−25 = 𝜆𝜆1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 − 1) + 𝜆𝜆2 + 𝜆𝜆3 (A.24) 𝑃𝑃1 (𝑛𝑛4 = 𝑁𝑁4 − 1) = 𝑗𝑗
𝑞𝑞14−29 = 𝑞𝑞16−30 = 𝑞𝑞17−31 = 𝜆𝜆1 + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) 𝜌𝜌4
(𝑁𝑁4 − 1)! ∑𝑁𝑁4

(A.25)
𝑗𝑗=0 𝑗𝑗!
𝜆𝜆2
𝑞𝑞3−9 = 𝑞𝑞16−22 = 𝑞𝑞27−31 = 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) + 𝜆𝜆3 𝜌𝜌4 = . (A.48)
µ4
(A.26)
𝑞𝑞16−31 = 𝜆𝜆1 + 𝜆𝜆2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 − 1) + 𝜆𝜆3 (A.27)

Vol. 22, No. 2, April 2024 239

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

𝑁𝑁 −1
𝜌𝜌5 4 𝑞𝑞25−21 = 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1)
𝑃𝑃2 (𝑛𝑛4 = 𝑁𝑁4 − 1) = 𝑗𝑗
, + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) + 𝑁𝑁4 𝜇𝜇4 (A.68)
𝑁𝑁4 𝜌𝜌5
(𝑁𝑁4 − 1)! ∑𝑗𝑗=0 𝑞𝑞23−6 = 𝑞𝑞24−7 = 𝑞𝑞25−9 = 𝑁𝑁1 𝜇𝜇1
𝑗𝑗!
𝜆𝜆3 + 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) (A.69)
𝜌𝜌5 = . (A.49) 𝑞𝑞17−3 = 𝑞𝑞22−7 = 𝑞𝑞25−11 = 𝑁𝑁1 𝜇𝜇1
µ4
𝑁𝑁 −1
𝜌𝜌6 4 + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.70)
𝑃𝑃3 (𝑛𝑛4 = 𝑁𝑁4 − 1) = 𝑗𝑗
, 𝑞𝑞25−7 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1)
𝑁𝑁4 𝜌𝜌6
(𝑁𝑁4 − 1)! ∑𝑗𝑗=0 + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.71)
𝑗𝑗!
𝜆𝜆1 𝑞𝑞29−14 = 𝑞𝑞30−16 = 𝑞𝑞31−17 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1)
𝜌𝜌6 = . (A.50) + 𝑁𝑁2 𝜇𝜇2 (A.72)
µ4
𝑞𝑞9−3 = 𝑞𝑞22−16 = 𝑞𝑞31−27 = 𝑁𝑁2 𝜇𝜇2
A. 2. Addendum + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.73)
𝑞𝑞31−16 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1) + 𝑁𝑁2 𝜇𝜇2
In this addendum to Appendix A, we present the other + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.74)
transition rates 𝑞𝑞(𝑖𝑖−𝑗𝑗) occurred by fulfillment of service of a 𝑞𝑞27−14 = 𝑞𝑞30−19 = 𝑞𝑞33−23 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1)
call. The results are re-listed below. + 𝑁𝑁3 𝜇𝜇3 (A.75)
𝑞𝑞11−6 = 𝑞𝑞24−19 = 𝑞𝑞33−29 = 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1)
𝑞𝑞13−1 = 𝑞𝑞15−2 = 𝑞𝑞16−3 = 𝑞𝑞17−4 = 𝑞𝑞18−5 = 𝑞𝑞19−6 + 𝑁𝑁3 𝜇𝜇3 (A.76)
= 𝑞𝑞20−7 = 𝑞𝑞21−8 = 𝑞𝑞22−9 = 𝑞𝑞23−10 = 𝑞𝑞24−11 𝑞𝑞33−19 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1)
= 𝑞𝑞25−12 = 𝑁𝑁1 𝜇𝜇1 (A.51) + 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) + 𝑁𝑁3 𝜇𝜇3 (A.77)
𝑞𝑞5−1 = 𝑞𝑞8−2 = 𝑞𝑞7−3 = 𝑞𝑞9−4 = 𝑞𝑞18−13 = 𝑞𝑞19−14 𝑞𝑞22−3 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.78)
= 𝑞𝑞21−15 = 𝑞𝑞20−16 = 𝑞𝑞22−17 = 𝑞𝑞29−26 = 𝑞𝑞30−27 𝑞𝑞24−6 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) + 𝑁𝑁3 𝜇𝜇3 (A.79)
= 𝑞𝑞31−28 = 𝑁𝑁2 𝜇𝜇2 (A.52) 𝑞𝑞30−14 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1) + 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁3 𝜇𝜇3 (A.80)
𝑞𝑞2−1 = 𝑞𝑞8−5 = 𝑞𝑞7−6 = 𝑞𝑞11−10 = 𝑞𝑞15−13 = 𝑞𝑞16−14 𝑞𝑞16−2 = 𝑞𝑞19−5 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁4 𝜇𝜇4 (A.81)
= 𝑞𝑞21−18 = 𝑞𝑞20−19 = 𝑞𝑞24−23 = 𝑞𝑞27−26 = 𝑞𝑞30−29 𝑞𝑞7−2 = 𝑞𝑞19−13 = 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁4 𝜇𝜇4 (A.82)
= 𝑞𝑞33−32 = 𝑁𝑁3 𝜇𝜇3 (A.53) 𝑞𝑞7−5 = 𝑞𝑞16−13 = 𝑁𝑁3 𝜇𝜇3 + 𝑁𝑁4 𝜇𝜇4 (A.83)
𝑞𝑞26−14 = 𝑞𝑞27−16 = 𝑞𝑞28−17 = 𝑞𝑞29−19 = 𝑞𝑞30−20 𝑞𝑞20−2 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁4 𝜇𝜇4 (A.84)
= 𝑞𝑞31−22 = 𝑞𝑞32−33 = 𝑞𝑞33−24 = 𝑞𝑞34−25 𝑞𝑞20−5 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁3 𝜇𝜇3 + 𝑁𝑁4 𝜇𝜇4 (A.85)
= 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1) (A.54) 𝑞𝑞20−13 = 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁3 𝜇𝜇3 + 𝑁𝑁4 𝜇𝜇4 (A.86)
𝑞𝑞10−6 = 𝑞𝑞11−7 = 𝑞𝑞12−9 = 𝑞𝑞23−19 = 𝑞𝑞24−20 = 𝑞𝑞25−22 𝑞𝑞22−8 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) + 𝑁𝑁4 𝜇𝜇4 (A.87)
= 𝑞𝑞32−29 = 𝑞𝑞33−30 = 𝑞𝑞34−31 𝑞𝑞22−15 = 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) + 𝑁𝑁4 𝜇𝜇4 (A.88)
= 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) (A.55) 𝑞𝑞24−18 = 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) + 𝑁𝑁3 𝜇𝜇3 + 𝑁𝑁4 𝜇𝜇4 (A.89)
𝑞𝑞4−3 = 𝑞𝑞9−7 = 𝑞𝑞12−11 = 𝑞𝑞17−16 = 𝑞𝑞22−20 = 𝑞𝑞25−24
= 𝑞𝑞28−27 = 𝑞𝑞31−30 = 𝑞𝑞34−33 A. 3. Addendum
= 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.56)
𝑞𝑞18−1 = 𝑞𝑞21−2 = 𝑞𝑞20−3 = 𝑞𝑞22−4 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁2 𝜇𝜇2 (A.57)
In this addendum to Appendix A, we present the description
𝑞𝑞15−1 = 𝑞𝑞21−5 = 𝑞𝑞20−6 = 𝑞𝑞24−10 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁3 𝜇𝜇3
of the X’s that are mentioned in Section 3.3.
(A.58)
𝑞𝑞8−1 = 𝑞𝑞21−13 = 𝑞𝑞20−14 = 𝑞𝑞30−26 = 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁3 𝜇𝜇3
𝑋𝑋1 = 𝑞𝑞1−2 + 𝑞𝑞1−5 + 𝑞𝑞1−8 + 𝑞𝑞1−13 + 𝑞𝑞1−15 + 𝑞𝑞1−18
(A.59)
+ 𝑞𝑞1−21 (A.90)
𝑞𝑞21−1 = 𝑁𝑁1 𝜇𝜇1 + 𝑁𝑁2 𝜇𝜇2 + 𝑁𝑁3 𝜇𝜇3
𝑋𝑋2 = 𝑃𝑃2 𝑞𝑞2−1 + 𝑃𝑃5 𝑞𝑞5−1 + 𝑃𝑃8 𝑞𝑞8−1 + 𝑃𝑃13 𝑞𝑞13−1
(A.60)
+ 𝑃𝑃15 𝑞𝑞15−1 + 𝑃𝑃18 𝑞𝑞18−1 + 𝑃𝑃21 𝑞𝑞21−1 (A.91)
𝑞𝑞32−19 = 𝑞𝑞33−20 = 𝑞𝑞34−22 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1)
𝑋𝑋3 = 𝑞𝑞2−1 + 𝑞𝑞2−3 + 𝑞𝑞2−7 + 𝑞𝑞2−8 + 𝑞𝑞2−15 + 𝑞𝑞2−16
+ 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) (A.61)
+ 𝑞𝑞2−20 + 𝑞𝑞2−21 (A.92)
𝑞𝑞28−16 = 𝑞𝑞31−20 = 𝑞𝑞34−24 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1)
𝑋𝑋4 = 𝑃𝑃1 𝑞𝑞1−2 + 𝑃𝑃3 𝑞𝑞3−2 + 𝑃𝑃7 𝑞𝑞7−2 + 𝑃𝑃8 𝑞𝑞8−2
+ 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.62)
+ 𝑃𝑃15 𝑞𝑞15−2 + 𝑃𝑃16 𝑞𝑞16−2 + 𝑃𝑃20 𝑞𝑞20−2 + 𝑃𝑃21 𝑞𝑞21−2
𝑞𝑞12−7 = 𝑞𝑞25−20 = 𝑞𝑞34−30 = 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1)
(A.93)
+ 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.63)
𝑋𝑋5 = 𝑞𝑞3−2 + 𝑞𝑞3−4 + 𝑞𝑞3−7 + 𝑞𝑞3−9 + 𝑞𝑞3−16 + 𝑞𝑞3−17
𝑞𝑞34−20 = 𝑁𝑁1 𝜇𝜇1 𝑃𝑃(𝑛𝑛1 = 𝑁𝑁1 + 1)
+ 𝑞𝑞3−20 + 𝑞𝑞3−22 (A.94)
+ 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1)
𝑋𝑋6 = 𝑃𝑃2 𝑞𝑞2−3 + 𝑃𝑃4 𝑞𝑞4−3 + 𝑃𝑃7 𝑞𝑞7−3 + 𝑃𝑃9 𝑞𝑞9−3
+ 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) (A.64)
+ 𝑃𝑃16 𝑞𝑞16−3 + 𝑃𝑃17 𝑞𝑞17−3 + 𝑃𝑃20 𝑞𝑞20−3 + 𝑃𝑃22 𝑞𝑞22−3
𝑞𝑞3−2 = 𝑞𝑞6−5 = 𝑞𝑞14−13 = 𝑁𝑁4 𝜇𝜇4 (A.65)
(A.95)
𝑞𝑞23−18 = 𝑁𝑁2 𝜇𝜇2 𝑃𝑃(𝑛𝑛2 = 𝑁𝑁2 + 1) + 𝑁𝑁4 𝜇𝜇4 (A.66)
𝑋𝑋7 = 𝑞𝑞4−3 + 𝑞𝑞4−9 + 𝑞𝑞4−17 + 𝑞𝑞4−22 (A.96)
𝑞𝑞9−8 = 𝑞𝑞17−15 = 𝑁𝑁3 𝜇𝜇3 𝑃𝑃(𝑛𝑛3 = 𝑁𝑁3 + 1) + 𝑁𝑁4 𝜇𝜇4 (A.67)

Vol. 22, No. 2, April 2024 240

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

+ 𝑃𝑃17 𝑞𝑞17−15 + 𝑃𝑃21 𝑞𝑞21−15 + 𝑃𝑃22 𝑞𝑞22−15 (A.119)

𝑋𝑋8 = 𝑃𝑃3 𝑞𝑞3−4 + 𝑃𝑃9 𝑞𝑞9−4 + 𝑃𝑃17 𝑞𝑞17−4 + 𝑃𝑃22 𝑞𝑞22−4 𝑋𝑋31 = 𝑞𝑞16−2 + 𝑞𝑞16−3 + 𝑞𝑞16−13 + 𝑞𝑞16−14 + 𝑞𝑞16−17
(A.97) + 𝑞𝑞16−20 + 𝑞𝑞16−22 + 𝑞𝑞16−27 + 𝑞𝑞16−28 + 𝑞𝑞16−30
𝑋𝑋9 = 𝑞𝑞5−1 + 𝑞𝑞5−6 + 𝑞𝑞5−7 + 𝑞𝑞5−8 + 𝑞𝑞5−18 + 𝑞𝑞5−19 + 𝑞𝑞16−31 (A.120)
+ 𝑞𝑞5−20 + 𝑞𝑞5−21 (A.98) 𝑋𝑋32 = 𝑃𝑃2 𝑞𝑞2−16 + 𝑃𝑃3 𝑞𝑞3−16 + 𝑃𝑃13 𝑞𝑞13−16
𝑋𝑋10 = 𝑃𝑃1 𝑞𝑞1−5 + 𝑃𝑃6 𝑞𝑞6−5 + 𝑃𝑃7 𝑞𝑞7−5 + 𝑃𝑃8 𝑞𝑞8−5 + 𝑃𝑃14 𝑞𝑞14−16 + 𝑃𝑃17 𝑞𝑞17−16 + 𝑃𝑃20 𝑞𝑞20−16
+ 𝑃𝑃18 𝑞𝑞18−5 + 𝑃𝑃19 𝑞𝑞19−5 + 𝑃𝑃20 𝑞𝑞20−5 + 𝑃𝑃21 𝑞𝑞21−5 + 𝑃𝑃22 𝑞𝑞22−16 + 𝑃𝑃27 𝑞𝑞27−16 + 𝑃𝑃28 𝑞𝑞28−16
(A.99) + 𝑃𝑃30 𝑞𝑞30−16 + 𝑃𝑃31 𝑞𝑞31−16 (A.121)
𝑋𝑋11 = 𝑞𝑞6−5 + 𝑞𝑞6−7 + 𝑞𝑞6−10 + 𝑞𝑞6−11 + 𝑞𝑞6−19 + 𝑞𝑞6−20 𝑋𝑋33 = 𝑞𝑞17−3 + 𝑞𝑞17−4 + 𝑞𝑞17−15 + 𝑞𝑞17−16 + 𝑞𝑞17−22
+ 𝑞𝑞6−23 + 𝑞𝑞6−24 (A.100) + 𝑞𝑞17−28 + 𝑞𝑞17−31 (A.122)
𝑋𝑋12 = 𝑃𝑃5 𝑞𝑞5−6 + 𝑃𝑃7 𝑞𝑞7−6 + 𝑃𝑃10 𝑞𝑞10−6 + 𝑃𝑃11 𝑞𝑞11−6 𝑋𝑋34 = 𝑃𝑃3 𝑞𝑞3−17 + 𝑃𝑃4 𝑞𝑞4−17 + 𝑃𝑃15 𝑞𝑞15−17
+ 𝑃𝑃19 𝑞𝑞19−6 + 𝑃𝑃20 𝑞𝑞20−6 + 𝑃𝑃23 𝑞𝑞23−6 + 𝑃𝑃24 𝑞𝑞24−6 + 𝑃𝑃16 𝑞𝑞16−17 + 𝑃𝑃22 𝑞𝑞22−17 + 𝑃𝑃28 𝑞𝑞28−17
(A.101) + 𝑃𝑃31 𝑞𝑞31−17 (A.123)
𝑋𝑋13 = 𝑞𝑞7−2 + 𝑞𝑞7−3 + 𝑞𝑞7−5 + 𝑞𝑞7−6 + 𝑞𝑞7−9 + 𝑞𝑞7−11 𝑋𝑋35 = 𝑞𝑞18−1 + 𝑞𝑞18−5 + 𝑞𝑞18−13 + 𝑞𝑞18−21 + 𝑞𝑞18−23
+ 𝑞𝑞7−12 + 𝑞𝑞7−20 + 𝑞𝑞7−22 + 𝑞𝑞7−24 + 𝑞𝑞7−25 + 𝑞𝑞18−24 (A.124)
(A.102) 𝑋𝑋36 = 𝑃𝑃1 𝑞𝑞1−18 + 𝑃𝑃5 𝑞𝑞5−18 + 𝑃𝑃13 𝑞𝑞13−18
𝑋𝑋14 = 𝑃𝑃2 𝑞𝑞2−7 + 𝑃𝑃3 𝑞𝑞3−7 + 𝑃𝑃5 𝑞𝑞5−7 + 𝑃𝑃6 𝑞𝑞6−7 + 𝑃𝑃9 𝑞𝑞9−7 + 𝑃𝑃21 𝑞𝑞21−18 + 𝑃𝑃23 𝑞𝑞23−18 + 𝑃𝑃24 𝑞𝑞24−18
+ 𝑃𝑃11 𝑞𝑞11−7 + 𝑃𝑃12 𝑞𝑞12−7 + 𝑃𝑃20 𝑞𝑞20−7 + 𝑃𝑃22 𝑞𝑞22−7 (A.125)
+ 𝑃𝑃24 𝑞𝑞24−7 + 𝑃𝑃25 𝑞𝑞25−7 𝑋𝑋37 = 𝑞𝑞19−5 + 𝑞𝑞19−6 + 𝑞𝑞19−13 + 𝑞𝑞19−14 + 𝑞𝑞19−20
(A.103) + 𝑞𝑞19−23 + 𝑞𝑞19−24 + 𝑞𝑞19−29 + 𝑞𝑞19−30 + 𝑞𝑞19−32
𝑋𝑋15 = 𝑞𝑞8−1 + 𝑞𝑞8−2 + 𝑞𝑞8−5 + 𝑞𝑞8−9 + 𝑞𝑞8−21 + 𝑞𝑞8−22 + 𝑞𝑞19−33 (A.126)
(A.104) 𝑋𝑋38 = 𝑃𝑃5 𝑞𝑞5−19 + 𝑃𝑃6 𝑞𝑞6−19 + 𝑃𝑃13 𝑞𝑞13−19 + 𝑃𝑃14 𝑞𝑞14−19
𝑋𝑋16 = 𝑃𝑃1 𝑞𝑞1−8 + 𝑃𝑃2 𝑞𝑞2−8 + 𝑃𝑃5 𝑞𝑞5−8 + 𝑃𝑃9 𝑞𝑞9−8 + 𝑃𝑃20 𝑞𝑞20−19 + 𝑃𝑃23 𝑞𝑞23−19 + 𝑃𝑃24 𝑞𝑞24−19
+ 𝑃𝑃21 𝑞𝑞21−8 + 𝑃𝑃22 𝑞𝑞22−8 + 𝑃𝑃29 𝑞𝑞29−19 + 𝑃𝑃30 𝑞𝑞30−19 + 𝑃𝑃32 𝑞𝑞32−19
(A.105) + 𝑃𝑃33 𝑞𝑞33−19 (A.127)
𝑋𝑋17 = 𝑞𝑞9−3 + 𝑞𝑞9−4 + 𝑞𝑞9−7 + 𝑞𝑞9−8 + 𝑞𝑞9−12 + 𝑞𝑞9−22 𝑋𝑋39 = 𝑞𝑞20−2 + 𝑞𝑞20−3 + 𝑞𝑞20−5 + 𝑞𝑞20−6 + 𝑞𝑞20−7
+ 𝑞𝑞9−25 + 𝑞𝑞20−13 + 𝑞𝑞20−14 + 𝑞𝑞20−16 + 𝑞𝑞20−19
(A.106) + 𝑞𝑞20−22 + 𝑞𝑞20−24 + 𝑞𝑞20−25 + 𝑞𝑞20−30 + 𝑞𝑞20−31
𝑋𝑋18 = 𝑃𝑃3 𝑞𝑞3−9 + 𝑃𝑃4 𝑞𝑞4−9 + 𝑃𝑃7 𝑞𝑞7−9 + 𝑃𝑃8 𝑞𝑞8−9 + 𝑞𝑞20−33 + 𝑞𝑞20−34 (A.128)
+ 𝑃𝑃12 𝑞𝑞12−9 + 𝑃𝑃22 𝑞𝑞22−9 + 𝑃𝑃25 𝑞𝑞25−9 𝑋𝑋40 = 𝑃𝑃2 𝑞𝑞2−20 + 𝑃𝑃3 𝑞𝑞3−20 + 𝑃𝑃5 𝑞𝑞5−20 + 𝑃𝑃6 𝑞𝑞6−20
(A.107) + 𝑃𝑃7 𝑞𝑞7−20 + 𝑃𝑃13 𝑞𝑞13−20 + 𝑃𝑃14 𝑞𝑞14−20 + 𝑃𝑃16 𝑞𝑞16−20
𝑋𝑋19 = 𝑞𝑞10−6 + 𝑞𝑞10−11 + 𝑞𝑞10−23 + 𝑞𝑞10−24 (A.108) + 𝑃𝑃19 𝑞𝑞19−20 + 𝑃𝑃22 𝑞𝑞22−20 + 𝑃𝑃24 𝑞𝑞24−20
𝑋𝑋20 = 𝑃𝑃6 𝑞𝑞6−10 + 𝑃𝑃11 𝑞𝑞11−10 + 𝑃𝑃23 𝑞𝑞23−10 + 𝑃𝑃24 𝑞𝑞24−10 + 𝑃𝑃25 𝑞𝑞25−20 + 𝑃𝑃30 𝑞𝑞30−20 + 𝑃𝑃31 𝑞𝑞31−20
(A.109) +𝑃𝑃33 𝑞𝑞33−20 + 𝑃𝑃34 𝑞𝑞34−20 (A.129)
𝑋𝑋21 = 𝑞𝑞11−6 + 𝑞𝑞11−7 + 𝑞𝑞11−10 + 𝑞𝑞11−12 + 𝑞𝑞11−24 𝑋𝑋41 = 𝑞𝑞21−1 + 𝑞𝑞21−2 + 𝑞𝑞21−5 + 𝑞𝑞21−8 + 𝑞𝑞21−13
+ 𝑞𝑞11−25 (A.110) + 𝑞𝑞21−15 + 𝑞𝑞21−18 + 𝑞𝑞21−25 (A.130)
𝑋𝑋22 = 𝑃𝑃6 𝑞𝑞6−11 + 𝑃𝑃7 𝑞𝑞7−11 + 𝑃𝑃10 𝑞𝑞10−11 + 𝑃𝑃12 𝑞𝑞12−11 𝑋𝑋42 = 𝑃𝑃1 𝑞𝑞1−21 + 𝑃𝑃2 𝑞𝑞2−21 + 𝑃𝑃5 𝑞𝑞5−21 + 𝑃𝑃8 𝑞𝑞8−21
+ 𝑃𝑃24 𝑞𝑞24−11 + 𝑃𝑃25 𝑞𝑞25−11 (A.111) + 𝑃𝑃13 𝑞𝑞13−21 + 𝑃𝑃15 𝑞𝑞15−21 + 𝑃𝑃18 𝑞𝑞18−21
𝑋𝑋23 = 𝑞𝑞12−7 + 𝑞𝑞12−9 + 𝑞𝑞12−11 + 𝑞𝑞12−25 (A.112) + 𝑃𝑃25 𝑞𝑞25−21 (A.131)
𝑋𝑋24 = 𝑃𝑃7 𝑞𝑞7−12 + 𝑃𝑃9 𝑞𝑞9−12 + 𝑃𝑃11 𝑞𝑞11−12 + 𝑃𝑃25 𝑞𝑞25−12 𝑋𝑋43 = 𝑞𝑞22−3 + 𝑞𝑞22−4 + 𝑞𝑞22−7 + 𝑞𝑞22−8 + 𝑞𝑞22−9
(A.113) + 𝑞𝑞22−15 + 𝑞𝑞22−16 + 𝑞𝑞22−17 + 𝑞𝑞22−20 + 𝑞𝑞22−25
𝑋𝑋25 = 𝑞𝑞13−1 + 𝑞𝑞13−14 + 𝑞𝑞13−15 + 𝑞𝑞13−16 + 𝑞𝑞13−18 + 𝑞𝑞22−31 + 𝑞𝑞22−34 (A.132)
+ 𝑞𝑞13−19 + 𝑞𝑞13−20 + 𝑞𝑞13−21 (A.114) 𝑋𝑋44 = 𝑃𝑃3 𝑞𝑞3−22 + 𝑃𝑃4 𝑞𝑞4−22 + 𝑃𝑃7 𝑞𝑞7−22 + 𝑃𝑃8 𝑞𝑞8−22
𝑋𝑋26 = 𝑃𝑃1 𝑞𝑞1−13 + 𝑃𝑃14 𝑞𝑞14−13 + 𝑃𝑃15 𝑞𝑞15−13 + 𝑃𝑃16 𝑞𝑞16−13 + 𝑃𝑃9 𝑞𝑞9−22 + 𝑃𝑃15 𝑞𝑞15−22 + 𝑃𝑃16 𝑞𝑞16−22
+ 𝑃𝑃18 𝑞𝑞18−13 + 𝑃𝑃19 𝑞𝑞19−13 + 𝑃𝑃20 𝑞𝑞20−13 + 𝑃𝑃17 𝑞𝑞17−22 + 𝑃𝑃20 𝑞𝑞20−22 + 𝑃𝑃25 𝑞𝑞25−22
+ 𝑃𝑃21 𝑞𝑞21−13 (A.115) + 𝑃𝑃31 𝑞𝑞31−22 + 𝑃𝑃34 𝑞𝑞34−22 (A.133)
𝑋𝑋27 = 𝑞𝑞14−13 + 𝑞𝑞14−16 + 𝑞𝑞14−19 + 𝑞𝑞14−20 + 𝑞𝑞14−26 𝑋𝑋45 = 𝑞𝑞23−6 + 𝑞𝑞23−10 + 𝑞𝑞23−18 + 𝑞𝑞23−19 + 𝑞𝑞23−24
+ 𝑞𝑞14−27 + 𝑞𝑞14−29 + 𝑞𝑞14−30 (A.116) + 𝑞𝑞23−32 + 𝑞𝑞23−33 (A.134)
𝑋𝑋28 = 𝑃𝑃13 𝑞𝑞13−14 + 𝑃𝑃16 𝑞𝑞16−14 + 𝑃𝑃19 𝑞𝑞19−14 𝑋𝑋46 = 𝑃𝑃6 𝑞𝑞6−23 + 𝑃𝑃10 𝑞𝑞10−23 + 𝑃𝑃18 𝑞𝑞18−23 + 𝑃𝑃19 𝑞𝑞19−23
+ 𝑃𝑃20 𝑞𝑞20−14 + 𝑃𝑃26 𝑞𝑞26−14 + 𝑃𝑃27 𝑞𝑞27−14 + 𝑃𝑃24 𝑞𝑞24−23 + 𝑃𝑃32 𝑞𝑞32−23 + 𝑃𝑃33 𝑞𝑞33−23 (A.135)
+ 𝑃𝑃29 𝑞𝑞29−14 + 𝑃𝑃30 𝑞𝑞30−14 (A.117) 𝑋𝑋47 = 𝑞𝑞24−6 + 𝑞𝑞24−7 + 𝑞𝑞24−10 + 𝑞𝑞24−11 + 𝑞𝑞24−18
𝑋𝑋29 = 𝑞𝑞15−1 + 𝑞𝑞15−2 + 𝑞𝑞15−13 + 𝑞𝑞15−17 + 𝑞𝑞15−21 + 𝑞𝑞24−19 + 𝑞𝑞24−20 + 𝑞𝑞24−23 + 𝑞𝑞24−25 + 𝑞𝑞24−33
+ 𝑞𝑞15−22 (A.118) + 𝑞𝑞24−34 (A.136)
𝑋𝑋30 = 𝑃𝑃1 𝑞𝑞1−15 + 𝑃𝑃2 𝑞𝑞2−15 + 𝑃𝑃13 𝑞𝑞13−15 𝑋𝑋48 = 𝑃𝑃6 𝑞𝑞6−24 + 𝑃𝑃7 𝑞𝑞7−24 + 𝑃𝑃10 𝑞𝑞10−24 + 𝑃𝑃11 𝑞𝑞11−24
+ 𝑃𝑃18 𝑞𝑞18−24 + 𝑃𝑃19 𝑞𝑞19−24 + 𝑃𝑃20 𝑞𝑞20−24

Vol. 22, No. 2, April 2024 241

 N. A. Hassan et al. / Journal of Applied Research and Technology 230-242

+ 𝑃𝑃23 𝑞𝑞23−24 + 𝑃𝑃25 𝑞𝑞25−24 + 𝑃𝑃33 𝑞𝑞33−24

+ 𝑃𝑃34 𝑞𝑞34−24 (A.137)
𝑋𝑋49 = 𝑞𝑞25−7 + 𝑞𝑞25−9 + 𝑞𝑞25−11 + 𝑞𝑞25−12 + 𝑞𝑞25−20
+ 𝑞𝑞25−21 + 𝑞𝑞25−22 + 𝑞𝑞25−24 + 𝑞𝑞25−34
(A.138)
𝑋𝑋50 = 𝑃𝑃7 𝑞𝑞7−25 + 𝑃𝑃9 𝑞𝑞9−25 + 𝑃𝑃11 𝑞𝑞11−25 + 𝑃𝑃12 𝑞𝑞12−25
+ 𝑃𝑃20 𝑞𝑞20−25 + 𝑃𝑃21 𝑞𝑞21−25 + 𝑃𝑃22 𝑞𝑞22−25
+ 𝑃𝑃24 𝑞𝑞24−25 + 𝑃𝑃34 𝑞𝑞34−25 (A.139)
𝑋𝑋51 = 𝑞𝑞26−14 + 𝑞𝑞26−27 + 𝑞𝑞26−29 + 𝑞𝑞26−30 (A.140)
𝑋𝑋52 = 𝑃𝑃14 𝑞𝑞14−26 + 𝑃𝑃27 𝑞𝑞27−26 + 𝑃𝑃29 𝑞𝑞29−26
+ 𝑃𝑃30 𝑞𝑞30−26 (A.141)
𝑋𝑋53 = 𝑞𝑞27−14 + 𝑞𝑞27−16 + 𝑞𝑞27−26 + 𝑞𝑞27−28 + 𝑞𝑞27−30
+ 𝑞𝑞27−31 (A.142)
𝑋𝑋54 = 𝑃𝑃14 𝑞𝑞14−27 + 𝑃𝑃16 𝑞𝑞16−27 + 𝑃𝑃26 𝑞𝑞26−27
+ 𝑃𝑃28 𝑞𝑞28−27 + 𝑃𝑃30 𝑞𝑞30−27 + 𝑃𝑃31 𝑞𝑞31−27 (A.143)
𝑋𝑋55 = 𝑞𝑞28−16 + 𝑞𝑞28−17 + 𝑞𝑞28−27 + 𝑞𝑞28−31 (A.144)
𝑋𝑋56 = 𝑃𝑃16 𝑞𝑞16−28 + 𝑃𝑃17 𝑞𝑞17−28 + 𝑃𝑃27 𝑞𝑞27−28
+ 𝑃𝑃31 𝑞𝑞31−28 (A.145)
𝑋𝑋57 = 𝑞𝑞29−14 + 𝑞𝑞29−19 + 𝑞𝑞29−26 + 𝑞𝑞29−30 + 𝑞𝑞29−32
+ 𝑞𝑞29−33 (A.146)
𝑋𝑋58 = 𝑃𝑃14 𝑞𝑞14−29 + 𝑃𝑃19 𝑞𝑞19−29 + 𝑃𝑃26 𝑞𝑞26−29
+ 𝑃𝑃30 𝑞𝑞30−29 + 𝑃𝑃32 𝑞𝑞32−29 + 𝑃𝑃33 𝑞𝑞33−29 (A.147)

𝑋𝑋59 = 𝑞𝑞30−14 + 𝑞𝑞30−16 + 𝑞𝑞30−19 + 𝑞𝑞30−20 + 𝑞𝑞30−26

+ 𝑞𝑞30−27 + 𝑞𝑞30−29 + 𝑞𝑞30−31 + 𝑞𝑞30−33 + 𝑞𝑞30−34
(A.148)
𝑋𝑋60 = 𝑃𝑃14 𝑞𝑞14−30 + 𝑃𝑃16 𝑞𝑞16−30 + 𝑃𝑃19 𝑞𝑞19−30
+ 𝑃𝑃20 𝑞𝑞20−30 + 𝑃𝑃26 𝑞𝑞26−30 + 𝑃𝑃27 𝑞𝑞27−30
+ 𝑃𝑃29 𝑞𝑞29−30 + 𝑃𝑃31 𝑞𝑞31−30 + 𝑃𝑃33 𝑞𝑞33−30
+ 𝑃𝑃34 𝑞𝑞34−30 (A.149)
𝑋𝑋61 = 𝑞𝑞31−16 + 𝑞𝑞31−17 + 𝑞𝑞31−20 + 𝑞𝑞31−22 + 𝑞𝑞31−27
+ 𝑞𝑞31−28 + 𝑞𝑞31−30 + 𝑞𝑞31−34 (A.150)
𝑋𝑋62 = 𝑃𝑃16 𝑞𝑞16−31 + 𝑃𝑃17 𝑞𝑞17−31 + 𝑃𝑃20 𝑞𝑞20−31
+ 𝑃𝑃22 𝑞𝑞22−31 + 𝑃𝑃27 𝑞𝑞27−31 + 𝑃𝑃28 𝑞𝑞28−31
+ 𝑃𝑃30 𝑞𝑞30−31 + 𝑃𝑃34 𝑞𝑞34−31 (A.151)
𝑋𝑋63 = 𝑞𝑞32−19 + 𝑞𝑞32−23 + 𝑞𝑞32−29 + 𝑞𝑞32−33 (A.152)
𝑋𝑋64 = 𝑃𝑃19 𝑞𝑞19−32 + 𝑃𝑃23 𝑞𝑞23−32 + 𝑃𝑃29 𝑞𝑞29−32
+ 𝑃𝑃33 𝑞𝑞33−32 (A.153)
𝑋𝑋65 = 𝑞𝑞33−19 + 𝑞𝑞33−20 + 𝑞𝑞33−23 + 𝑞𝑞33−24 + 𝑞𝑞33−29
+ 𝑞𝑞33−30 + 𝑞𝑞33−32 + 𝑞𝑞33−34 (A.154)
𝑋𝑋66 = 𝑃𝑃19 𝑞𝑞19−33 + 𝑃𝑃20 𝑞𝑞20−33 + 𝑃𝑃23 𝑞𝑞23−33
+ 𝑃𝑃24 𝑞𝑞24−33 + 𝑃𝑃29 𝑞𝑞29−33 + 𝑃𝑃30 𝑞𝑞30−33
+ 𝑃𝑃32 𝑞𝑞32−33 + 𝑃𝑃34 𝑞𝑞34−33 (A.155)
𝑋𝑋67 = 𝑞𝑞34−20 + 𝑞𝑞34−22 + 𝑞𝑞34−24 + 𝑞𝑞34−25 + 𝑞𝑞34−30
+ 𝑞𝑞34−31 + 𝑞𝑞34−33 (A.156)
𝑋𝑋68 = 𝑃𝑃20 𝑞𝑞20−34 + 𝑃𝑃22 𝑞𝑞22−34 + 𝑃𝑃24 𝑞𝑞24−34
+ 𝑃𝑃25 𝑞𝑞25−34 + 𝑃𝑃30 𝑞𝑞30−34 + 𝑃𝑃31 𝑞𝑞31−34
+ 𝑃𝑃33 𝑞𝑞33−34 (A.157)

Vol. 22, No. 2, April 2024 242