Manuscripts submitted to IMA Journal of Management Mathematics

Stochastic Model: NHPP and GRP, applied to the repair

warranty of an industrial asset.

Journal: IMA Journal of Management Mathematics

Fo
Manuscript ID: Draft

Manuscript Type: Manuscript

r
Date Submitted by the Author: n/a

Complete List of Authors: Gonzalez-Prida, Vicente; University of Seville, Industrial Management

Pe

Barbera, Luis; University of Seville, Industrial Management

Crespo, Adolfo; University of Seville, Industrial Management
Gómez Fernández, Juan Francisco; University of Seville, Industrial
Management
er

Parra, Carlos; University of Seville, Industrial Management

Keywords: LCCA, NHPP, GRP, After sales service, Warranty period

Re
vi
ew

http://mc.manuscriptcentral.com/imaman
Page 1 of 14 Manuscripts submitted to IMA Journal of Management Mathematics

1
2
3 Stochastic Model: NHPP and GRP, applied to the repair warranty of an
4 industrial asset
5
6 V. González-Prida*, L. Barberá**, A. Crespo**, J.F. Gómez**, C. Parra***
7
8 * General Dynamics – European Land Systems. Seville, Spain
9 (e-mail: vicente.gonzalezprida@gdels.com).
10 ** Department of Industrial Management, Escuela Superior de Ingenieros de Sevilla
11
(e-mail: lubarmar@etsi.us.es, adolfo@etsi.us.es, juan.gomez@iies.es)
12
*** IngeCon, Asesoría en Integral de Confiabilidad, Caracas, Venezuela
13
14 (e-mail: parrac37@gmail.com)
15
16
17 This paper seeks to address the decision process involved in setting a
18 warranty length for a product after successive repairs are performed. The
Fo
19 underlying stochastic failure models used are called: General Renewal
20
21
Process and Non-homogeneous Poisson Process both determined by a
22 Weibull distribution, distinguishing renewal or not of the lifetime
r
23 distribution by successive repairs respectively. This article makes a
24 comparison between two different stochastic methods, one of them applied
Pe

25 to simple cases when repairs are simple (NHPP) and the other used when
26 repairs are complex (GRP).For that purpose, the paper starts describing
27 briefly a reference framework proposal for the warranty management and
28 introducing the relevant literature related to LCCA. Then, the main aspects
er

29 of LCCA are defined in order to be applied for the calculation of the

30
31
warranty period in a product which has been sold and requires a technical
32 assistance for its repair. This work includes as a novelty the calculation of
Re

33 such period of time in relationship with the risk that the company is
34 willing to assume. The result is a procedure that may be crucial for a
35 maintenance company, not only to make better forecasts of future warranty
36 costs but also as an important marketing tool. With this goal, this paper
vi

37 describes important aspects of both stochastic models. The mathematical

38 development is illustrated with a case study divided in two exercises,
39
where the above mentioned concepts are applied to calculate the proper
ew

40
41
warranty period for a specific repaired product. Finally, the conclusions
42 are presented summarizing the main contributions of the paper.
43
44 Keywords: After-Sales Service; Life Cycle Cost Analysis (LCCA); Non-
45 homogeneous Poisson Process (NHPP); General Renewal Process (GRP);
46 Technical Assistance; Warranty Period.
47
48
49
50
51
52
53
54
55
56
57
58
59
60
http://mc.manuscriptcentral.com/imaman
 Manuscripts submitted to IMA Journal of Management Mathematics Page 2 of 14

1
2
3 1. Introduction
4
5 The life cycle cost in a physical asset is determined identifying the applicable
6
functions in each phase of the product life, calculating the cost of these functions and
7
8 applying the appropriate costs during the whole extension of the life cycle.
9 Therefore, this cost should include all those ones related to design, manufacturing
10 and production (Ahmed N.U. 1995), (Levy H. and Sarnat M. 1990). Part of these
11 costs, in the case of a product launched to the market, has to be faced by the own
12 buyer. Nevertheless, it has been here included all those costs involved in a product
13 life cycle, regardless the actor (manufacturer or user) who has to face such charges.
14 From the financial point of view, the costs generated along the life cycle of the asset
15 can be classified in two types of costs:
16
17
18 CAPEX: Capital costs (design, development, acquisition, installation, staff training,
Fo
19 manuals, documentation, tools and facilities for maintenance, replacement parts for
20 assurance, withdrawal).
21
22 OPEX: Operational costs (manpower, operations, planned maintenance, storage,
r
23 recruiting and corrective maintenance - penalizations for failure events / low
24
Reliability).
Pe

25
26
27 The aftersales support is frequently offered while the production lines are still open.
28 Therefore, the product engineering and manufacturing can be improved with the
er

29 feedback of warranty program data, reducing consequently the general costs of the
30 product life cycle. With an adequate reliability and availability assessment, is
31 possible to demonstrate in the first stages of the product, how requirements expressed
32
in initial technical specifications can be incompatible or even impossible to
Re

33
34
accomplish for determined product configurations (González-Prida V. et al. 2009),
35 (Crespo A. and Iung B. 2007). If the product is already launched, this analysis can
36 help to take quickly the necessary measures to correct and/or improve the product,
vi

37 foreseeing also probable claims from the users due to the real lack of reliability on
38 the product, in comparison with the previous reliability, sold a priori. As already
39 commented, the typical life cycle cost analysis includes costs for planning, research
ew

40 and development, production, operation, warranty and disposal (Parra C. et al. 2007).
41 From the consumer’s point of view, the life cycle cost will suppose the acquisition
42
costs, purchase price, costs of operation and maintenance, etc. That means, in general
43
44
terms, the total cost of the item ownership. In any case, the life cycle cost regarding
45 warranty issues, is highly influenced on the values for reliability and failure rate, cost
46 of spares, repair times, and component costs, as well as other attributes related to the
47 asset’s maintainability (Moreu P. et al 2012). Normally, a low budget for product
48 engineering leads to high warranty costs in the future (Constantino F. et al. 2012).
49 Those customer complaints related to important or costly failures, should be soon
50 attended and the failures fully analysed to identify not only further tasks to proceed
51 with the repair, but also preventive actions which can avoid or at least, decrease
52
future claims due to similar reasons. Therefore, this consideration involves
53
54
performing a root cause failure analysis (included in step 1 of the proposed reference
55 framework (Gonzalez-Prida and Crespo, 2012)). An overall review of all the
56 warranty complaints can be helpful to show, for example, repetitive failures and
57 trends related to vendor/buyer problems, quality issues, manufacturing conditions,
58 product design, etc. As a continuation to a previous research (Gonzalez-Prida V. et al
59
60
http://mc.manuscriptcentral.com/imaman
Page 3 of 14 Manuscripts submitted to IMA Journal of Management Mathematics

1
2
3 2012), this paper is intended now to compare two different stochastic methods, the
4 NHPP and the GRP. The first one is usually applied to simple repair cases, and the
5 second one is usually implemented when repairs are complex. With that aim, the
6 paper is structured as follow: First of all, there is a mathematical development on
7 how to calculate an appropriate warranty period, applying both methods (Section 2).
8
After that, both methods are implemented in a case scenario (Section 3). That
9
10 scenario refers to repairs with a certain level of complexity; therefore, we will see
11 how results from the GRP are more realistic than the ones obtained with the NHPP
12 (Section 4).
13
14
15 2. Calculation of the warranty period
16
17
18 In order to calculate the appropriate warranty period, is needed to apply suitable
Fo
19 measures which should be defined during the strategy phase of the warranty
20 program. The measures must enable the comparison of reliability data and the
21 inclusion of the life cycle cost assessment. The company should therefore establish
22 and use a standard and repeatable method for collecting and analyzing data and
r
23 interpreting results, which may be based on corporate or industry factors. The results
24
should be used to support and justify enhancements. This section reviews the
Pe

25
26
mathematics of Non-Homogeneous Poisson Processes and General Renewal Process.
27
28
er

29 2.1 NHPP Model proposed for the warranty period assessment

30
31
This subsection presents an algorithmic breakdown considering the warranty period
32
choice. That means that, here below those steps to estimate the minimal time for the
Re

33
34 warranty period (tMTW) are described according to NHPP. As starting data, it is here
35 necessary to know the intervals of time ti (i = 1,... , n), when failure and repair events
36 have taken place. With these data, it is possible to follow the following steps:
vi

37
38
39
ew

40
41
42
43
44
45
46
47
48
49 Fig. 1. Conditional probability of occurrence of failure
50
51
52 a) Calculation of parameters:
53
54
55 Considering Tn as the total accumulated time:
56
57
58
59
60
http://mc.manuscriptcentral.com/imaman
 Manuscripts submitted to IMA Journal of Management Mathematics Page 4 of 14

1
2
3 n
4 Tn = ∑ ti
5 i =1
6
(1)
7
8 The parameters α and β of the Weibull distribution in the (n)th event will be then:
9
10 Tn
11 αˆ n = 1
12
13 nβ (2)
14 n
15 βˆn = n−1
T 
16 ∑ ln  n 
17 i =1  Ti  (3)
18
Fo
19 Where Ti is the time at which the (i)th failure occurs, Tn is the total time where the
20
21
last failure occurred, and n is the total number of failures, and deriving by the
22 maximum likelihood approach (Maximum likelihood Estimation, MLE) (Mettas and
r
23 Wenbiao, 2005).
24
Pe

25 n  β
26  1  n  t 
LK = log L{ dataα ,⋅β } =   ⋅ β n ⋅ ∏ exp − n  ⋅ t β −1
27  β
28 α  i = 1  α β  i

 
er

29 (4)
30
31 b) Calculation of expected time till next failure (TNF):
32
Re

33
The expected time till next failure, taking into account the Weibull parameters and
34
35 the total accumulated time, will be given by the following expression:
36
TNFn = {[(α ) β + (Tn ) β ](1/ β ) } − Tn
vi

37
38 (5)
39
ew

40 c) Calculation of expected number of failures for tMTW:

41
42 The total expected number of failures in the time interval [Tn, Tn + tMTWn] according
43
44
to the Weibull cumulative intensity function is (Modarres M. et al. 1999):
45
46
47 Λ(Tn , Tn + t MTWn ) =
α
1
β
[(T n + t MTWn ) − (Tn )
β β
]
48 (6)
49
50 Where, as already commented, tMTWn is the minimal time for warranty after the last
51 failure and repair took place, and (Tn + tMTWn) is equivalent to (TMTWn ).
52
53
54 d) Calculation of system reliability for the recommended warranty period (Tn +
55 tMTW n):
56
57 Assuming a Weibull distribution, the reliability function will be according to the
58
59
60
http://mc.manuscriptcentral.com/imaman
Page 5 of 14 Manuscripts submitted to IMA Journal of Management Mathematics

1
2
3 following expression:
4
5
1
6 R (T MTWn ) = exp{ − β
× [( TMTWn ) β − (Tn ) β ]}
7 α (7)
8
9
10
11 Therefore, the failure probability in a system will be consequently:
12
13 F (TMTWn ) = 1 − R (TMTWn )
14 (8)
15
16 e) Calculation of minimal time for the warranty period (tMTW n):
17
18
With all the above calculations, it is possible to obtain now the minimal time for the
Fo
19
20 warranty period after the (n)th repair (see Figure 2) based on the risk until the next
21 failure expressed as a probability derived from the reliability:
22
r
23 tMTW n
24
Pe

25
26
tn ts
27
28
Tn Ts
er

29 Fig. 2. Time line of failures events

30
31
32 Solving tMTW n for 80%, 60% or 40% of reliability:
Re

33

[ ]
34 1
β
35 t MTWn = Tn − α β ⋅ Ln ( R (T MTWn )) β − Tn (9)
36
vi

37
38 As we can see in the above described process, a risk balance has been applied for the
39 calculation of the recommended warranty period, and not an objective function.
ew

40 Basically, that percentage considered as minimal reliability within the warranty

41 period, is a parameter to be estimated during the proposed calculation process,
42 together with the possibility of determining its relationship with costs.
43
44
45
46 2.2 GRP Model proposed for the warranty period assessment
47
48 Now, through the GRP model the flexibility is increased in order to report on the five
49 possible system states after a repair, although the complexity is increased too at the
50 same time:
51
52
53 • As good as new
54 • As bad as old
55 • Better than old, but worse than new
56
• Better than new
57
58 • Worse than old
59
60
http://mc.manuscriptcentral.com/imaman
 Manuscripts submitted to IMA Journal of Management Mathematics Page 6 of 14

1
2
3
4 The NHPP model accounts for the second state, while the rest repair states have not
5 been involved. Generalized Renewal Process (GRP) copes with all the five possible
6 states, modelling the failure behaviour of a specific system and understanding the
7
effects of the repairs on the system’s age (Yañez et al., 2002), introducing the
8
9
concept of virtual age (Vn) and furthermore an indicator of repair quality (q). (Vn)
10 represents the calculated age of the element immediately after the nth repair occurs.
11
12 a) Calculation of parameters:
13
14 n
15
16
Vn = q ⋅ ∑q j =1
n− j
x j = q ⋅ ( xn + vn−1 )
17 (10)
18
The summation (with V0=0 for t0=0) assumes that the nth repair compensates for the
Fo
19
20 damage accumulated during the time between the (n - 1)th and the nth failure, but also
21 damages produced in previous intervals.
22
r
23
The indicator of repair quality (q) represents the state “as good as new” when q = 0,
24
while the assumption of q = 1 corresponds to a non-homogeneous Poisson process
Pe

25
26 “as bad as old”. The values of q that fall in the interval 0 < q < 1 represent the after-
27 repair states in which the condition of the element is “better than old but worse than
28 new”, whereas the cases where q > 1 correspond to a condition “worse than old”.
er

29 Similarly, cases with q < 0 would suggest a component or system restored to a state
30 “better than new”. Therefore, physically speaking, q can be seen as an index for
31 representing the effectiveness and quality of repairs.
32
Re

33
34 b) Calculation of failure probability and system reliability:
35
36 For Vn-1 = y the system has a time to the (n)th failure, xn, which is distributed
vi

37 according to the following cumulative distribution function (cdf):

38
39
F (x + v ) − F (v
ew

)
40 F (t V )= n n −1 n −1 =
41 n −1 1 − F (v )
n −1
42
43
44 − 
(
 x +v
n n − 1
β −v
n − 1
)β 

45  αβ 
46 = 1− e  
(11)
47
48
49
where R(t) = 1 – F(t) is the reliability function.
50
51 There are different formalizations of the GRP process to model equipment evolution
52 over time. In this case, we will try to simplify the implementation of the GRP method
53 based on the study by Mettas and Wenbiao (2005) built on the GRP II by Kijima &
54 Sumita (1986), which uses the maximum likelihood approach (Maximum likelihood
55 Estimation, MLE), to cover complex systems with multiple repairs.
56
57
58
59
60
http://mc.manuscriptcentral.com/imaman
Page 7 of 14 Manuscripts submitted to IMA Journal of Management Mathematics

1
2
3 k
4 LK = log L{ dataα ,⋅β ,⋅q }= ∑ n ⋅ (− lnα + ln β )
l
5 l =1
6 k  Tl − tl ,n + vn β  vn β 
7 − ∑  l l
 −  l  
 α
8 l = 1    α  
9
10 k n  (xl ,i + vl ,i−1 )β − vl ,i−1β 
11 − ∑ ∑  
12 l = 1 i = 1  αβ 
13
k n  xl, i + vl , i −1 
+ (β −1) ⋅ ∑ ∑ln 
14
l = 1 i =1  α 
15
16  (12)
17
18
c) Calculation of minimal time for the warranty period (tMTW n):
Fo
19
20
21 Now, the minimal time for the warranty period after the (n)th repair based on the risk
22 until the next failure expressed as a probability derived from the reliability is:
r
23 Solving tMTW n for 80%, 60% or 40% of reliability:
24
Pe

25
[ ]
1
β
26 t MTWn = v n −1 − α β ⋅ Ln ( R (TMTWn )) β − v n −1 (13)
27
28
er

29
30
3. Case study
31
32 As commented, case studies have been normally used to support and help theoretical
Re

33 subjects in engineering and other research fields. Developing these cases, it is usually
34 found such amount of information that can either trivialize the study or complicate it
35
beyond a reasonable level. Therefore, the intention here is to synthesize a practical
36
example which transmits easily the use of the above described mathematical
vi

37
38 development to calculate an appropriate warranty period for a repair task.
39
ew

40 3.1 Study Scenario

41
42
The case here exposed will deal with a customer (a large entity in the industrial
43
44
sector) who is interested in the acquisition, use and maintenance of specific engines.
45 These engines have been developed and tested by the manufacturing company and
46 they are already in used by the client. That means, there are enough data about the
47 engines behaviour. Therefore, the service company can offer to the client specific
48 warranty periods after maintenance assistances, assuring that the repaired engine will
49 present a determined availability during a specific period of time. The availability is
50 determined by the company, according to the risk that the organization is willing to
51 take. The specific period of time is what here below is going to be determined,
52
applying two different stochastic methods, the NHPP and the GRP. As mentioned,
53
54 the first one is usually applied to simple repair cases, and the second one is usually
55 implemented when repairs are complex. This case refers to a complex repair;
56 therefore, results obtained by GRP will be more appropriate to determine a specific
57 warranty period.
58
59
60
http://mc.manuscriptcentral.com/imaman
 Manuscripts submitted to IMA Journal of Management Mathematics Page 8 of 14

1
2
3 3.2 NHPP application to a practical case
4
5 Exercise 1:
6
7
8 As above mentioned, the manufacturing company has at its disposal historical data
9 about the engine behaviour. As table 1 can illustrate, failure and repair events are
10 presented according to ti. With this ti (i = 1,..., 6), the values already defined are:
11
12 Ti (hours); Tn / Ti; Ln (Tn / Ti)
13
14
15
With these values and applying the above developed formulas we will obtain: αn, βn,
16 TNFn, tMTWn, Λ, R(Ts) and F(Ts). However, in order to see how impacts the NHPP in
17 the calculation process; we will apply the mentioned formulas considering different
18 numbers of known events. That means, we will analyse the system behaviour under n
Fo
19 = 4, n = 6 and n = 8.
20
21 TABLE 1. NHPP Engine Historical Data related to time per warranty event (n = 8)
22
r
23
24 Event ti (hours) Ti (h) Tn / Ti Ln (Tn / Ti)
(i)
Pe

25
26 1 13.460,00 13.460,00 0,09 1,63
27
28 2 9.287,00 22.747,00 0,01 1,10
er

29 3 3.128,00 25.875,00 2,65 0,97

30 4 10.148,00 36.023,00 1,90 0,64
31 5 9.037,00 45.060,00 1,52 0,42
32
6 8.771,00 53.831,00 1,27 0,24
Re

33
34 7 7.459,00 61.290,00 1,12 0,11
35 8 7.252,00 68.542,00 1,00 -
36
vi

37 Applying the process for n = 6 and n = 4, we obtain the results shown in tables 2 and
38 3:
39
ew

40
41 TABLE 2. NHPP Engine Historical Data related to time per warranty event (n = 6)
42
43
44 Event ti (hours) Ti (h) Tn / Ti Ln (Tn / Ti)
45 (i)
46
47 1 13.460,00 13.460,00 4,00 1,39
48 2 9.287,00 22.747,00 2,37 0,86
49 3 3.128,00 25.875,00 2,08 0,73
50 4 10.148,00 36.023,00 1,49 0,40
51
5 9.037,00 45.060,00 1,19 0,18
52
53 6 8.771,00 53.831,00 1,00 -
54
55
56
57
58
59
60
http://mc.manuscriptcentral.com/imaman
Page 9 of 14 Manuscripts submitted to IMA Journal of Management Mathematics

1
2
3 TABLE 3. NHPP Engine Historical Data related to time per warranty event (n = 4)
4
5 Event ti (hours) Ti (h) Tn / Ti Ln (Tn / Ti)
6 (i)
7
8 1 13.460,00 13.460,00 2,68 0,98
9 2 9.287,00 22.747,00 1,58 0,46
10 3 3.128,00 25.875,00 1,39 0,33
11 4 10.148,00 36.023,00 1,00 -
12
13
14 Applying different percentages (80%, 60% and 40%) for reliability, we can obtain
15 different times as recommended warranty period based on the time till next failure.
16 The use of one tMTWn or another is a decision which can be more easily taken if we
17 know the expected amount of failure Λ, as well as the system reliability R(Ts). For
18 that calculation, it has been here considered one year of function after the repair (ts =
Fo
19 8.760 h).
20
21
22 TABLE 4. NHPP Results obtained with n = 8, n = 6 and n = 4
r
23
24 n=8 n=6 n=4
Pe

25 αn 18107,47 18594,02 19471,96

26 βn 1,562172 1,685553 2,253470
27
28 TNFn 5.367,67 5.155,20 3.749,66
er

29 tMTWn (20 %) 1.217,77 1.178,91 878,32

30 tMTWn (40 %) 2.770,34 2.673,73 1.973,40
31 tMTWn (60 %) 4.926,88 4.735,69 3.452,79
32
Λ 1,653580 1,736069 2,532629
Re

33
34 R(Ts) 0,191364 0,176212 0,079450
35 F(Ts) 0,808636 0,823788 0,920550
36
vi

37 With this exercise, we observe how decreases the time till next failure as well as the
38 system reliability, when the amount of failures and repairs events decrease. If we
39
analyse separately the development with the time of these cases, the mentioned
ew

40
41 differences in the system reliability are even more remarkable. The calculated
42 parameters not only depend on the correct number of failures but also the order of
43 them. A small number of divergent failures can distort the parameters, as in the first
44 four events (see event 3, ti = 3.128,00 Hr), so it is desirable to consider convergent
45 times of failures avoiding the mixture of different failure modes.
46
47 Consequently, the different risks assumed for warranty (in other words, the different
48
choices for the percentage taken in the warranty time calculation) yields to different
49
50 results in the values for the minimal time for the warranty period. Figures 3 and 4,
51 show the evolution of R(Ts) along 2 years after the events n = 4, …, 8 have taken
52 place for Ts = [2190, 4380, 6570, 8760, 10950, 13140, 15330, 17520].
53
54
55
56
57
58
59
60
http://mc.manuscriptcentral.com/imaman
 Manuscripts submitted to IMA Journal of Management Mathematics Page 10 of 14

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Fo
19
20
21 Fig. 3. NHPP Reliability evolution depending on time and the amount of events
22
r
23
24
Pe

25
26
27
28
er

29
30
31
32
Re

33
34
35
36
vi

37
38
39
ew

40
41 Fig. 4. NHPP Surface representation of R (t, n)
42
43
44
45 3.2 GRP application to a practical case
46
47 Exercise 2:
48
49
50 In addition to the described exercise, which has helped to illustrate the calculation of
51 warranty periods, it is possible also to explore now considering GRP model in order
52 to obtain the different warranty parameters. For that purpose and as a second
53 example, we are going now to consider the same repair events, obtaining the results
54 of Table 5.
55
56
57
58
59
60
http://mc.manuscriptcentral.com/imaman
Page 11 of 14 Manuscripts submitted to IMA Journal of Management Mathematics

1
2
3 TABLE 5. GRP Results obtained with n = 8, n = 6 and n = 4
4
5 n=8 n=6 n=4
6
7
αn 10901,91 11551,98 11971,63
8 βn 3,464468 3,355038 2,663281
9 TNFn 9.016,30 8.850,22 8.087,37
10 tMTWn (20 %) 5.199,15 4.735,58 3.301,54
11 tMTWn (40 %) 7.099,28 6.770,40 5.548,63
12
13 tMTWn (60 %) 8.745,11 8.554,81 7.715,04
14 Λ 0,920753 0,973892 1,162304
15 R(Ts) 0,398219 0,377610 0,312765
16 F(Ts) 0,601781 0,622390 0,687235
17
18
In addition, now for the GRP model, Figures 5 and 6, show the evolution of R(Ts)
Fo
19
20 along 2 years after the events n = 4, …, 8 have taken place for Ts = [2190, 4380,
21 6570, 8760, 10950, 13140, 15330, 17520].
22
r
23
24
Pe

25
26
27
28
er

29
30
31
32
Re

33
34
35
36
vi

37
38
39
ew

40
41
Fig. 5. GRP Reliability evolution depending on time and the amount of events
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
http://mc.manuscriptcentral.com/imaman
 Manuscripts submitted to IMA Journal of Management Mathematics Page 12 of 14

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Fo
19 Fig. 6. GRP Surface representation of R (t, n)
20
21
22 Now as with NHPP, through GRP model the higher is the failure probability, the
r
23 higher the slope of the warranty period curve. In other words, if the company
24 assumes higher risks in its warranty policies, the increase in the warranty times shall
Pe

25
26
be more significant.
27
28 The adjustment of the reliability obtained from the GRP vs. time and number of
er

29 expected failures confirm its better fit thanks to the inclusion of the parameter q, see
30 Table 6 where a comparison of both models, NHPP and GRP are realized using the
31 likelihood value as a representation of the adjustment.
32
Re

33
34
TABLE 6. Comparison of NHPP and GRP Likelihood values obtained with n = 8, n
35 = 6 and n = 4
36
vi

37 LK n=8 n=6 n=4

38 NHPP -79,756 -59,919 -39,398
39
GRP -73,907 -55,900 -37,981
ew

40
41
42 GRP is a general model, providing a way to describe the rate of occurrence of events
43 over time understanding the effects of the repairs on the age of that system, and
44 including the Renewal Process or the NHPP representation. For example, consider a
45 system that is repaired after a failure, where the repair does not bring the system to
46 an as-good-as-new or an as-bad-as old condition. In other words, the system is
47 partially rejuvenated after the repair.
48
49
50 3.3 Results discussion
51
52 The obtained result for the warranty period is, as mentioned, a minimal value
53 considering different percentages (20%, 40% and 60%) of the reliability till the
54 expected time next failure (TNF), calculated with both models. The company should
55
now decide as a business issue, if this period is convenient or not to be enlarged with
56
57
the consequent increase of warranty costs. Therefore, the assessment of such
58 percentage is one of the steps to take into account in this process, where it is
59
60
http://mc.manuscriptcentral.com/imaman
Page 13 of 14 Manuscripts submitted to IMA Journal of Management Mathematics

1
2
3 necessary to know as an input, the risk that the company is disposed to assume for
4 the warranty period. For that reason, two exercises have been here also represented.
5 In case that the warranty period is a mandatory topic included in the repair order or
6 the maintenance contract (or even enforced by the current legislation, (González-
7 Prida V. et al. 2012b)), the manufacturing firm (or the outsourced company which
8
offers maintenance services with their corresponding warranty) can foresee with the
9
10 above described calculations, the time which will be possibly needed till next
11 warranty assistance, just in order to save part of the general budget for this after sales
12 service and to prepare its technicians team ready for such events in the future
13 (González-Prida et al. 2012). For that purpose, the NHPP model has been proved to
14 provide good results for realistic situations with minimal repairs (Asiedu Y. and Gu
15 P. 1998). Based on this, and given its conservative nature and manageable
16 mathematical expressions, the NHPP has to be selected for the particular case of
17 minimal repair. However, when the repairs are no minimal the best adjustment can
18
be obtained with GRP model through its adaptability with three dimensions (α, β, q),
Fo
19
20 at the expense of more complexity. Of course, the GRP model has advantages and
21 limitations. In general, the more realistic is the model, the more complex are the
22 mathematical expressions involved. The main strengths and weakness of this model
r
23 are summarized next:
24
Pe

25 Strengths:
26
o It is a useful model to represent equipment under aging (deterioration).
27
28 o It is a better approach for the assessment of the warranty period for a given
assumption of risk.
er

29
30
31 Weakness:
32
Re

33
o Involves relatively complex mathematical expressions.
34
35 o It is not adequate to simulate repair actions when the number of events is
36 small.
vi

37
38
39 4. Conclusions
ew

40
41 Warranty period is frequently used by companies as an effective marketing tool in
42 front of other competitors. Throughout this document, we have applied two specific
43 tools already developed in the area of LCCA, to a particular case of the after-sales
44 management process. With that intention, it has been analysed the Life Cycle Cost
45 and particularly the Non-Homogeneous Poisson Process (NHPP) and General
46
47
Renewal Process (GRP). These methodologies for the LCCA can make easier and
48 also improve the decision-making process in cases as the after-sales service. The
49 application of the models can for instance enable the forecast of eventual failures in
50 the product and to anticipate the consequent repair costs that such events will bring
51 to the management of the product post-launch. In other words, a proper management
52 of warranty assistances helps to reduce costs, enables to take suitable decisions, and
53 improves the image of the company in front of the client. NHPP is a good and simple
54 method to estimate the cost of the warranty period, although due to the fact that, the
55 failure rate will obviously change with the operative time of the system and,
56
57
consequently, it will not be homogeneous. On the other hand, GRP is the recommend
58 method for a better adjustment, considering complex repairs of the item. As future
59
60
http://mc.manuscriptcentral.com/imaman
 Manuscripts submitted to IMA Journal of Management Mathematics Page 14 of 14

1
2
3 research lines, it is proposed to develop more elaborated models with, for example,
4 the implementation of e-technologies in order to achieve higher levels of quality
5 effectiveness in the technical assistances.
6
7
8
9 REFERENCES
10
11 Ahmed, N. U. (1995). A design and implementation model for life cycle cost
12 management system, Information and Management, 28, pp. 261-269.
13 Asiedu, Y. and Gu, P. (1998). Product lifecycle cost analysis: state of art review,
14 International Journal of Production Research, Vol. 36 No. 4, pp. 883-908.
15
16 Costantino F., De Minicis M., González-Prida V., Crespo A. (2012). On the use of
17 Quality Function Deployment (QFD) for the identification of risks associated to
18 warranty programs. ESREL Conference 2012, Helsinki, Finland.
Fo
19 Crespo A. and Iung B. (2007). A Estructured Approach for the Assessment of
20 System Availability and Reliability Using Montecarlo Simulation, Journal of
21 Quality in Maintenance Engineering. Vol. 13. Núm. 2. Pag. 125-136.
22 González-Prida V., Crespo A., Moreu P., Gómez J., Parra C. (2009). Availability and
r
23
reliability assessment of industrial complex systems: A practical view applied on
24
a bioethanol plant simulation, Safety and Reliability for Managing Risk. London,
Pe

25
26 UK. Taylor & Francis Group. ISBN 978-0-415-48513-5. Pag. 687-695.
27 González-Prida V., Barberá L., Crespo A., Parra C. (2012). Warranty period
28 calculation after the repair of a complex industrial asset applying Non-
er

29 Homogeneous Poisson Process. DYNA Ingeniería e Industria 87, 655-662 2012

30 González-Prida V., Barberá L., Gómez J., Crespo A. (2012). Contractual and quality
31 aspects on warranty: best practices for the warranty management and its maturity
32
assessment, International Journal of Quality and Reliability Management,
Re

33
34 IJQRM. Vol. 29 No. 3, pp. 320-348.
35 González-Prida V., Crespo A., A framework for warranty management in industrial
36 assets. Computers in Industry 63 (9), 960–971 2012
vi

37 Hurtado, J.L., Joglar, F.and Modarres, M. (2005). Generalized Renewal Process:

38 Models, Parameter Estimation and Applications to Maintenance Problems,
39 International Journal on Performability Engineering, Vol. 1, No. 1, paper 3, pp.
ew

40 37-50.
41
42
Levy, H. and Sarnat, M. (1990). Capital Investment and Financial Decisions, 4th
43 Edition, Prentice Hall, New York.
44 Modarres, M., Kaminskiy, M. and Krivtsov V. (1999). Reliability engineering and
45 risk analysis. Marcel Dekker Inc., New York.
46 Moreu P., Gonzalez-Prida V., Barbera L., Crespo A. (2012). A practical method for
47 the maintainability assessment in industrial devices using indicators and specific
48 attributes. Reliability Engineering and System Safety 100 (2012) 84–92.
49
50
Parra C., Crespo A., Moreu P., Gómez J., González-Prida V. (2009). Non-
51 homogeneous Poisson Process (NHPP), stochastic model applied to evaluate the
52 economic impact of the failure in the Life Cycle Cost Analysis (LCCA). Safety,
53 reliability and risk analysis: theory, methods and applications. Taylor & Francis
54 Group, London, ISBN 978-0-415-48513-5. Pag. 929-939.
55 Yañez, M., Joglar, F., Mohammad, M. (2002). Generalized renewal process for
56 analysis of repairable systems with limited failure experience, Reliability
57 Engineering & System Safety, Vol. 77, pp. 167-180.
58
59
60
http://mc.manuscriptcentral.com/imaman