Journal of Quality in Maintenance Engineering

Emerald Article: Economic model for maintenance decision: a case study for mill liners Rajiv Dandotiya, Jan Lundberg

Article information:
To cite this document: Rajiv Dandotiya, Jan Lundberg, (2012),"Economic model for maintenance decision: a case study for mill liners", Journal of Quality in Maintenance Engineering, Vol. 18 Iss: 1 pp. 79 - 97 Permanent link to this document: http://dx.doi.org/10.1108/13552511211226201 Downloaded on: 19-07-2012 References: This document contains references to 28 other documents To copy this document: [email protected] This document has been downloaded 85 times since 2012. *

Users who downloaded this Article also downloaded: *

Oscar Fernandez, Ashraf W. Labib, Ralph Walmsley, David J. Petty, (2003),"A decision support maintenance management system: Development and implementation", International Journal of Quality & Reliability Management, Vol. 20 Iss: 8 pp. 965 - 979 http://dx.doi.org/10.1108/02656710310493652 Selim Zaim, Ali Turkyilmaz, Mehmet F. Acar, Umar Al-Turki, Omer F. Demirel, (2012),"Maintenance strategy selection using AHP and ANP algorithms: a case study", Journal of Quality in Maintenance Engineering, Vol. 18 Iss: 1 pp. 16 - 29 http://dx.doi.org/10.1108/13552511211226166 Aditya Parida, Uday Kumar, (2006),"Maintenance performance measurement (MPM): issues and challenges", Journal of Quality in Maintenance Engineering, Vol. 12 Iss: 3 pp. 239 - 251 http://dx.doi.org/10.1108/13552510610685084

Access to this document was granted through an Emerald subscription provided by OXFORD BROOKES UNIVERSITY For Authors: If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service. Information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information. About Emerald www.emeraldinsight.com With over forty years' experience, Emerald Group Publishing is a leading independent publisher of global research with impact in business, society, public policy and education. In total, Emerald publishes over 275 journals and more than 130 book series, as well as an extensive range of online products and services. Emerald is both COUNTER 3 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.

The current issue and full text archive of this journal is available at www.emeraldinsight.com/1355-2511.htm

Economic model for maintenance decision: a case study for mill liners
Rajiv Dandotiya and Jan Lundberg
Department of Operation and Maintenance Engineering, University of Technology, Lulea , Sweden Lulea
Abstract
Purpose Wear life of mill liners is an important parameter concerning maintenance decision for mill liners. Variations in process parameters such as different ore properties due to the use of multiple ore types inuence the wear life of mill liners whereas random order of processing, processing time and monetary value of different ore types leads to variation in mill protability. The purpose of the present paper is to develop an economic decision model considering the variations in process parameters and maintenance parameters for making more cost-effective maintenance decisions. Design/methodology/approach Correlation studies, experimental results and experience of industry experts are used for wear life modeling whereas simulation is used for maximizing mill prot to develop economic decision model. The weighting approach and simulation have been considered to emphasize the contribution of parameters such as ore value and processing time of a specic ore type to a nal result. Findings A model for estimating lifetime of mill liners has been developed based on ore properties. The lifetime model is combined with a replacement interval model to determine the optimum replacement interval for the mill liners which considers process parameters of multiple ore types. The nding of the combined model results leads to a signicant improvement in mill prot. The proposed combined model also shows that an optimum maintenance policy can not only reduce the downtime costs, but also affect the process performance, which leads to signicant improvement in the savings of the ore dressing mill. Practical implications The proposed economic decision model is practically feasible and can be implemented within the ore dressing mill industries. Using the model, the cost-effective maintenance decision can increase the prot of the organization signicantly. Originality/value The novelty is that the new combined model is applicable and useful in replacement decision making for grinding mill liners, in complex environment, e.g. processing multiple ore types, different monetary value of the ore type and random order of ore processing. Keywords Economic models, Replacement decision, Mill liners, Ore properties, Optimization, Process parameters, Replacement control, Optimization techniques Paper type Case study

Economic model for maintenance decision 79

The authors would like to thank VINNOVA Strategic Mine Research Programme (The Swedish Governmental Agency for Innovation Systems) and the supporting companies (Boliden AB and Metso Minerals, Ersmark) for their nancial support, as well as the expert group consisting of personnel from the companies involved in this project. They would like to extend special thanks m (Boliden) and Tage Mo ller (Metso) for sharing their to Jan Burstedt (Boliden), Stig Markstro valuable experiences and knowledge for the improvement of the paper. They would also like to University of Technology) for the valuable offer special thanks to Prof. Uday Kumar (Lulea comments and feedback on the paper.

Journal of Quality in Maintenance Engineering Vol. 18 No. 1, 2012 pp. 79-97 q Emerald Group Publishing Limited 1355-2511 DOI 10.1108/13552511211226201

JQME 18,1

80

1. Introduction Autogenous mills are used in mineral processing for particle size reduction. Mill liners are important spare parts of grinding mill in terms of reducing particle size and mill shell protection. Replacement decisions of mill liners are critical concerning mill prot due to inuence of maintenance activities on process efciency and production losses. Therefore, it is important to combine maintenance parameters with process parameters for the decision making of replacement of mill liners. Based on correlation studies, liner wear causes an increase in energy consumption and reduction in process efciency, which leads to signicant monetary losses. Mill liners also need to be replaced at regular intervals once they worn out which causes heavy monetary losses due to production loses during the mill stoppage for liner installation. The study of the present paper is the part of the same case study, which has been discussed by Dandotiya and Lundberg (2011a). A detailed explanation on economic inuence of maintenance activities on mill liners, uses of mill liners, variation in energy consumption, metal recovery, process efciency and correlation studies can be found in the same reference. Many researchers have performed research in the eld of liners performance and replacement strategy. In the study Santarisi and Almomany (2005), have proposed the liner replacement strategies based on mathematical modeling of wear rate. The inuence of load behavior on liner wear using DEM simulation was discussed by Kalala et al. (2008). Dandotiya and Lundberg (2011a) have briey described different characteristics of mill liners. They presented a replacement model based on lifecycle prot, which combines process parameters with maintenance parameters. However, their model has not considered the inuence of ore properties on liner wear and their model is limited to the mill, processes single ore type. The purpose of present paper is to develop a economic decision support model for mill liner replacement which considers inuence of variation in process parameters and its affect on maintenance decisions. The present paper also suggests approaches for combining various process parameters with the maintenance parameters. The proposed model in the present study extends the lifecycle prot model presented by Dandotiya and Lundberg (2011a) by considering the characteristics of various types of ore from different mines, together with the cost associated with maintenance activities performed on the mill liners.

2. Case study University of This paper presents a part of collaborative research conducted by Lulea Technology for a mining company (Boliden Mineral AB) and a liner manufacturer (Metso Minerals), all located in Sweden. Within the research framework, the comminution process and current replacement decision policies for the liners were analyzed as whole. The research focuses on assessing liner replacement policy using the correlation between throughput, grinding performance, power consumption, ore properties and maintenance cost. The standard grinding mill used here is 5.5 m in diameter and 5.7 m long with a power consumption of approximately 1800 kW (for more details see Dandotiya and Lundberg, 2011a).

3. Data collection and analysis The data for the process parameters and the replacement of mill liners, such as the replacement schedule, the inspection data and data for other maintenance activities, were obtained and analyzed. The process parameters, e.g. power consumption, process efciency, torque, throughput, mill speed, mill load etc., were correlated over various lifecycle periods of the mill liners. By performing a correlation study, redundant parameters were removed from the investigation. A process ow chart was made, as shown in Figure 1, which briey explains the data collection and analysis process and its uses during the various stages of the investigation. As shown in Figure 1, an hourly-based process and maintenance data collected from the industry. For this study, periodic discussions with the reference group were carried out at different phases of the research work. For the details of mill liners and expertise and experience of the reference group of the present case study (see Dandotiya and Lundberg, 2011b). Data collection The mill in the case study processed ore types that came from different mines and possessed different physical characteristics, such as different grade values, ore densities etc. Besides the process data, the maintenance department of the mining company also provided the data for all the maintenance activities concerning the mill studied in the present case study, such as data on mill stops due to mill liner inspection and replacement, motor repair, overhauling etc. This data is important for making decisions on replacements, making it possible to synchronize the replacement activity with other scheduled mill stops, so that the overall downtime of the mill can be minimized. Data analysis Correlation studies were carried out to determine the correlation between the liner wear and the process parameters over the life span of the liners. The other objective was to determine and remove the redundant parameters to minimize the model complexity. The trends of the data and the correlation of the data sets to each other were used as model inputs and this has a signicant inuence on the model outcomes. Hence, the outliers from the process data needed to be removed in order to achieve an appropriate correlation and appropriate trends. Data analysis for the single ore type. In the context of the present study, the term single ore type is used to denote a given ore type which comes from a specic mine and possesses a specied range of physical properties, such as density, grade value (percent of metal content), ore hardness, rock size etc. In the present case study the mill grinds various ore types which come from different mines and exhibit variations in their material properties and characteristics, such as density, grade value, rock size etc. The order of processing for the different ore types in the mill, over the life span of the mill liners, also varies, as the ore milling schedule also depends on the availability of ore in the mines. Therefore, in order to develop a generalized approach to investigating the economic efciency of the mill and the optimum replacement interval, the process data for each ore type is segregated (see Figures 2 and 3). Using the methodology (see Dandotiya and Lundberg, 2011a) requires continuous process data for a single ore type over the life span of the mill liners. Therefore, the

Economic model for maintenance decision 81

JQME 18,1

82

Figure 1. Data collection and analysis process

process data for a single ore type was segregated from the mixed ore data over one replacement cycle of the mill liners. In the present case study, multiple types of ores were processed between two consecutive replacements of liners shown in Figure 2. Figure 2 shows various types of ore processed over the period of one lifecycle of the mill liners, with t ij representing the milling time of ore type j for time sample i of a lifecycle period of the mill liners. The data for a single ore type was segregated from the lifecycle of the mill liners and is shown in Figure 3. The process data set for a single ore type 1 was collected for the time intervals, 2 3 i i.e. for t 1 1 , t 1 , t 1 ..t 1 (see Figure 3). A given ore type is processed in discrete intervals of time over a lifecycle period of the mill liners; i.e. no continuous process data is available for the ore type over one life span of the mill liners. Based on the segregated data, the process data was generated for one whole life span of the mill liners. For example, as described in Figure 2 and Figure 3, data segregation was performed for the process parameter power consumption (see Figures 4 and 5). Figure 4 shows the segregated data for a given ore type and the blank space between the data points shows that ore was not processed in the mill during that time interval. Furthermore, using a trend test, simulation, interpolation and extrapolation, data was generated for the whole lifecycle period of the mill liners (see Figure 5). Figure 6 shows the various activities performed in the case study for data generation for each process parameter for a specic ore type over the life span of the

Economic model for maintenance decision 83

Figure 2. Different ore types processed over a replacement cycle of the mill liners Figure 3. Segregated data for a single ore type for one replacement cycle of the mill liners

Figure 4. Segregated power data

JQME 18,1

84

Figure 5. Generated power data

mill liners. Apart from trend analysis, distribution analysis and simulation were also performed in order to generate a continuous data set over the life span of the mill liners. The Easy-Fit, Visual Basic and Matlab software tools were used to perform the analysis of the process data. The time series method is helpful in nding a best t line in a type of data when it has both trend and randomness (Box and Jenkins, 1976; Kendall, 1984; Shumway, 1988). If a trend in the data was found, then linear interpolation and extrapolation were used to generate the data for the periods when the specic ore was not actually being processed in the mill (see Figure 4 and Figure 5). Independent and identical distribution (IID) has been veried before tting the the data in any distribution (Ascher and , 1992). Feingold, 1984; Kumar and Klefsjo 4. Model formulation The approach of the present paper improves the replacement model for single ore type proposed by Dandotiya and Lundberg (2011a). The present model also included the ore density parameter in the replacement model which incorporates the inuence of ore properties on wear life of mill liners which makes the optimization approach applicable for multiple ore types. Literature on maintenance optimization models can also be referred in the references, e.g. Dekker (1996); Roll and Naor (1968); Dogramaci and Fraiman (2004); Christer and Scarf (1994); Sethi and Chand (1979); Jardine et al. (1998); Wang (2002); Yun and Choi (2000); Duffuaa and Al Sultan (1999); Hsu (1988). Model assumptions The following assumptions have been made: . Liner prole is not considered in the model. . Minor replacement of mill liners does not inuences the mill performance. . For the detailed explanation of rst two assumptions see Dandotiya and Lundberg (2011a).

Economic model for maintenance decision 85

Figure 6. Data generation for the whole life span of the mill liner from a mixed ore type to a single ore type

Due to lack of process data, it is assumed that the decreasing pattern of process efciency for the mine A ore type was the same for the ore types from mine B and C (see Figure 7). Throughput inuences the wear life of mill liner i.e. an increase in throughput leads to decrease in wear life. However, the present study random throughput but the variation in wear life, due to randomness in throughput assumed to be constant.

JQME 18,1

86

Figure 7. Process efciency trends for ores from mine A, B and C

.

The mill prot is linearly dependent on the milling time; i.e. the total prot is proportional to the milling time of a given ore type,P i at i . The proportionality constant is the average prot per unit of time pavgi of the given ore type. A weighted arithmetic mean is considered for calculating the overall or effective replacement interval of the mill liners for the multiple ore types. Relative importance is given to a specic ore type based on its value and quantity processed over a life span of the mill liners. Simulation is used to merge different distributions of replacement intervals for corresponding ore types.

The lifetime model considers the ore density parameter in order to incorporate the inuence of processing multiple ore types on lifetime of mill liners. The other unknown factors (see Equation 1), which may also have inuence on liner wear, are assumed to be constant in the study. The main reason is due to lack of experimental data, which is needed for all individual ore types, for establishing the relationship between lifetime and ore properties. Hence, as per the experience of reference group, the ore density is assumed to be a signicant factor. Radziszewski et al. (2005) have developed an advance wear model for lifter bar which also suggests a linear relationship between density and wear rate of lifter bar, if other parameters are assumed to be constant (see Equation 1): Wearrate r tanbF mF w P Hr 1

where r is the ore density b is the abrasion angle, F is the loading force, Hr is the ore hardness, mis the friction angle and w is the sliding velocity. However, an improved lifetime model considering more number of parameters of the ore properties will lead to more effective results for replacement interval of mill liners.

Abbreviations i: j: k: tin : Number of sampling intervals in one liner lifecycle. Ore type. Number of days in one time sample. Duration of the sampling intervals (days).

Economic model for maintenance decision 87

T Cycle j: Threshold life of mill liners for ore type j when risk of mill shell damage arises (days). T rep : M i: N: Ei: Time taken during the mill liners installation including preparation time (days). Mass ow from the mill during the ith time interval (tones/hour). Number of replacement cycles during the period T Cycle j. Energy consumed by the mill during the ith time interval (kWh). Process efciency. Revenue generated by processing ore ($/tone). Downtime cost per replacement cycle ($/hour). Total cost incurred during liner replacement (i.e. liner cost, labour cost, startup and synchronization cost) ($/replacement). Energy cost ($/kWh). Total inspection ($/inspection/day). cost per day (including downtime cost)

hp :
V: C DT : C rep : C energy : C ins : P gross :

Gross prot over a Tmax ($). Average density of group of different ore types. Density of ore type j.

ravg : rj :

A detailed discussion on model inputs parameters can be found in reference (Dandotiya and Lundberg, 2011a). Relation between ore properties and wear life. According to the reference group, the processed ore type, depends on its internal properties, will signicantly affect the wear rate of the liners up to approximately 25-50 percent, in the present particular case study. The amount of pyrite in the ore by some reasons affects the wear properties. The high amount of pyrite has been seen leading to high amount of wear on the liners. On the other hand, increased amount of pyrite increases the overall density of the ore, this increased density of the processed ore will directly increase the wear forces and thus increase the wear of the liners. The densities of the ores in the present case study are also varying between approximately 3.0-3.8 kg/dm3 because of the amount of pyrite. In order to incorporate the effect of physical properties of ore from different mines on the lifecycle of mill liners, a relation between density ore density and lifecycle of mill liners has been considered. Let the lifecycle of mill liners TCycle( j) (when only ore type

JQME 18,1

j is processed) and the ore density is r j therefore, based on information provided by reference group an inverse relation has been considered between lifecycle of mill liners and ore density: T Cycle ja 1 k1 ) T Cycle j rj rj 2

88

where, k1 is the proportionality constant for ore density. Therefore, from the above explanation, a relation between the lifecycle of mill liners TCycle( j) (when only ore type j is processed) and the average lifecycle for mill liners for multiple ore types Tavg has been established and represented by Equation 3:   ravg :T avg T Cycle j 3 rj The relation for gross prot over a replacement cycle and optimization approach is adopted from , which is represented by Equation 4. In the proposed approach of the present paper the maximum lifecycle period for each ore type will be different as wear life of mill liners depends on ore properties, therefore in the present model the maximum wear lives will vary from one ore type to other, hence in the present case T max Dandotiya and Lundberg (2011a)will be T max ( j). Where, T max ( j) TCycle ( j) Trep and the no. of replacement cycle is:   T Cycle j T rep N T Cycle T rep The gross prot is calculated at different lifecycle period. Where, TCycle varies from 1 to T max ( j) with the sampling interval of DT 1 day: ! T T T Cycle Cycle Cycle X X X M i hip V 2 E i C energy 2 C iinsp 2 C DT 2 C rep N 4 P gross
i 1 i1 i1

The replacement model for the single ore type considers the corresponding process parameters, various cost parameters and ore density. Since each ore type possesses a different grade value and density and different process parameters, the cost and revenue parameters also vary from one ore type to another. Hence, the processing time and monetary value of a specic ore in the mill become important key parameters concerning optimum replacement decisions for grinding mill liners. An approach is proposed in the present study by considering an example (see Figure 8). In this example three ore types from three different mines are considered, because, in the present case study, these three ore types were processed over the investigated life span of mill liners. After using the process and maintenance data in the model developed, the optimum replacement intervals for individual ore types are determined. The total milling or processing time and prot per unit of a specic ore type in the mill are used as an inputs for the mathematical formulation for determining the overall or effective replacement interval, which is represented as T eff . The term overall or effective replacement interval for multiple ore types is used to denote the optimum replacement

Economic model for maintenance decision 89

Figure 8. Approach used for determining the overall replacement interval

interval for mixed ore types that combines the inuence of all the ore types which have been processed over the investigated life span of the mill liners. Figure 8 shows the investigated ore types from mine A, B and C, the process data, operational and maintenance cost data which were used in the proposed combined model for single ore type to determine the optimum replacement interval for given ore types. The three inputs of the model for replacement intervals for the multiple ore type are: (1) the optimum replacement interval; (2) the ore processing time; and (3) the monetary value for each ore type. The evaluation criteria are named weighting factors and must be taken into consideration during the subsequent evaluation to determine the optimum solution (Pahl, 1996). A weighting factor is a real, positive number. It indicates the relative importance of a particular evaluation criterion. The present model assumed the average prot per day ($/day) for a specic ore type to be a relative importance of the given ore type. Hence, for a given ore type, the average prot per day is determined and is later used to determine the overall prot during the corresponding process time of the given ore type. An example is considered (see Figure 2) where t ij denotes the processing time for the j th ore type for the i th time interval. Therefore, the total processing time of ore type j in one replacement cycle of the mill liners will be: tj
j X x1

t ix

JQME 18,1

Let the average prot per day for ore type j be pavg( j) ($/day) and let the corresponding process (milling) time be tj (days) during the lifecycle period of the mill liners. Then the total prot over one lifecycle period of the mill liner for a given ore type, j, is: 6 P j pavg j tj $ Therefore, the gross prot of the mill for all three ore types is: P total
3 X i 1

90

P j

A term prot share (wj ) is introduced to consider the worth of the ore together with the processing time inside the mill over the life span of the mill liners. The prot share can also be dened in another way, i.e. as the ratio of the individual prot of a given ore type, j, to the gross prot from all the ore types over the life span of the mill liners, which can be determined by Equation 8. If wj is the prot share of the given ore type j, then, using Equation 1 and 2: wj P j P total 8

Later, the value ofwi is used as a weight for a given ore type to estimate the overall replacement interval, determined using Equation 9: T eff
j X x1

wx T x

where T1, T2. . . Tj are the optimum replacement intervals for ore 1, 2. . .j, combining Equation 8 and 9:
j X

px T x 10 px

T eff

x1 j X x1

Due to the randomness in throughput, the replacement model generates a specic distribution for the optimum replacement interval rather than a discrete value. Therefore, in order to determine the overall replacement interval, these distributions need to be combined. Hence, the overall replacement interval formulation is simulated for merging these distributions together. After simulation a convergence in the result for the overall replacement interval was obtained (see Figure 9). Results The real data set for each ore type (consisting of the process, maintenance and operational (energy) data) was collected from the ore dressing plant and the liner manufacturing company. The cost and revenue data are classied information and were used in the model to obtain the result.

Economic model for maintenance decision 91

Figure 9. Effective replacement interval for the combination of different ore types

A decreasing trend in the process efciency was found when the liner was worn out, which means that, at the end phase of the liners life, the grinding performance decreases (see Figure 7). Based on the process efciency data (at the same level of input metal (Cu) content), a trend curve was set up for ore from mine A with a 95 percent condence interval. The same comparison of the process efciency should also be carried out for mine B and C. However, due to a lack of desired process efciency data at similar levels of metal content for mine B and C and discussions with the reference group, the process efciency for the ore types from mine B and C follow the trend for process efciency of ore from mine A; i.e. that the decreasing trend was the same for all three ore types, as shown in Figure 7. However, the values of the average level of process efciency for the different ore types were considered, i.e. Mine A ore: 0.83, Mine B ore: 0.71 and Mine C ore: 0.84. Density and ore processing time data The density and process time data for each ore type was collected from the company and is given in Table I.
Mine A (Ore type j 1) Processing time over the miller liners life span (days) t j Design kg =dm 3 pj Prot share (a) wj Mine B (Ore type j 2) Mine C (Ore type j 3) 72 3.25-3.34 0.46

115 132 2.99-3.77 3.55-3.66 0.28 0.25

Note: aThe prot share or weights were calculated based on the methodology given in Section 3.

Table I. Ore type specic inputs for mine A, B and C

JQME 18,1

92

Results for the optimum replacement interval for ore from mine A. The process data were segregated for the whole lifecycle period (see Figure 6). The lifecycle of mill liners depends on the density of the specic ore and, therefore, the lifecycle of the mill liners for ore from mine A is determined using Equation 2. In the present case the average life of the mill liners is 319 days. As the wear is assumed to be directly dependent on the ore density, it is also important to consider the total processing time of the given ore in the mill. Therefore, the weighted mean of all the densities is assumed to be the average density of the mixed ore types, and the processing times of the different ore types are assumed to be weights:
3 X

ravg

rx x1 3 X
x 1

tx 11

tx

Using the data from Table I, the value of the average density of the mixed ore type is calculated as ravg 3.48 kg/dm3, where the average values of the densities of individual ores are considered. Using Equation 3, the average lifecycle of the mill liners when ore from mine A is processed will be T Cycle A 330 days. However, if the variation in the density (see Table I) within the same ore is considered, the variation in the lifecycle of the mill liners for the case of ore from mine A will be 294 # T Cycle A # 371days. The objective of determining the average lifecycle for individual ore types is to determine the feasibility of the model results; i.e. in the case of ore from mine A, in the worst case scenario, if ore with a high level of density, rA 3:77 kg/dm3, is processed, then the threshold limit for the replacement of mill liners will be 294 days. In order to generate the throughput data for simulation, the EasyFit software was used to detect the distribution within real throughput data and its corresponding parameters (see Figure 10).

Figure 10. Distribution of throughput data

The optimum replacement interval was determined using the proposed replacement model of the present study. Due to the large variation in the throughput data (1,500-2,500 tonnes/day) (see Figure 10), a wide range of optimum replacement intervals (250 to 310 days) was obtained (see Figure 11). Ore from mine B. Similarly, the value of the average lifecycle of the mill liners in the case of ore from mine C was found to be T Cycle B 308 days and the average life of the mill liners varies between 303 and 312 days and the optimum replacement interval was found to be within the range of 245 to 290 days. Ore from mine C. Similarly, the value of the average lifecycle of the mill liners in the case of ore from mine C was found to be T Cycle C 334 days and the average life of the mill liners varies between 324 and 338 days and the optimum replacement interval was found to be within the range of 240 to 275 days. Estimation of the overall optimum replacement interval for the case of mixed or combined ore types. Clemen and Winkler (1999) and Winkler and Makridakis (1983) have discussed methods of combining subjective probability distributions in decision making, and these researchers have provided an excellent summary of the current state of information regarding the uncertainty of interest. They stated that decision and risk analysis applications involving the combination of probability distributions or other quantities have often used simple combination rules (e.g. a simple average) and tend to perform quite well because of this usage. More complex rules sometimes outperform in some instances. The present study uses a weighted mean approach in order to combine three distributions for the optimum replacement interval for the ore from mine A, B and C. The weights are dened based on a monetary value (prot/day). The processing time of the ore type in the mill is also an important parameter in terms of determining the overall prot over the life span of the mill liners. Equation 5 was used and simulated to determine the optimum replacement interval for multiple ore types. Since the replacement intervals for individual ore types are used as input and since the replacement intervals also follow different distribution types, in order to combine

Economic model for maintenance decision 93

Figure 11. Distribution of optimum replacement interval

JQME 18,1

94

the three distribution inputs, the results for the overall replacement intervals were simulated until the results started to converge. After running the simulation, the optimal replacement interval converged between 258 and 286 days with a 95 percent condence interval (see Figure 9). In addition, the process time for each individual ore was obtained from the case study and used as input in the model in order to determine the annual prot. Hence, in order to obtain the optimum replacement interval, the processing time needs to be known in advance. Economic inuence of the suggested maintenance policy. An economic analysis of the suggested maintenance policy was performed. In order to determine the nancial benets for the mill over the lifecycle period of the mill liners, the average prot per day is calculated. The average prot per day M avg i for a given ore, j, is calculated from the following relation: M avg i P gross i $=day T Cycle j 12

P gross j is the gross prot for ore type j for a time period of T Cycle j. Let M avg 1, M avg 2 and M avg 3 be the average prot per day for the ore types from mine A, B and C, respectively, when the maintenance management follows the present replacement policy of 310 days. Similarly, letM avg N 1,M avg N 2and M avg N 3 be the new average prot per day generated when the replacement interval is changed to a suggested 265 days (determined through the mathematical model), and let t1, t2 and t3 be the processing time in the mill for the ore from mine A, B and C, respectively. The percentage change in the prot is calculated as follows: P present M avg 1 t1 M avg 2 t2 M avg 3 t 3 P suggested M avg N 1 t1 M avg N 2 t2 M avg N 3 t3 P suggested 2 P present % change in total profit 100 P present

13

Based on the case study, if the present maintenance policy of applying a replacement interval of 310 days is changed and a suggested replacement interval of 265 days is applied, the percent change in the total prot of the mill during one lifecycle of the mill liners is 0.5 percent. This is a signicant amount of savings when the prot amount is of a high order. For the studied case in this paper, if the suggested maintenance policy is considered with a 95 percent condence interval, i.e. if the replacement interval varies from 258 days to 286 days, then the corresponding savings vary from 0.3 percent to 0.5 percent of the annual prot of the mill, respectively. This means that, if the maintenance management schedule the replacement action with other maintenance activities within a range of one month, the management can still have an approximate signicant saving of 0.3 percent of the annual prot of the mill. However, the savings also vary with the types of ore and the duration of the processing time. Therefore, the savings mentioned (0.3 percent to 0.5 percent) are only valid for the given case provided in the paper. Based on the inuence of individual ore types on the nancial benets, it was also found that the nancial benets could go up

to 1 percent. Therefore, the processing time and the ore types play a critical role in terms of nancial savings. Model usability Data is available for the ore types which are going to be processed in the future. For prediction of the replacement interval of mill liners, in the case where data such as the power consumption of the mill, throughput (tonnes/day), process efciency, downtime cost (per hour), liner cost, energy cost and inspection cost are available for the ore types which are going to be used in the future, the present model can be used simply according to the methodology described in both parts of the present paper. Data is not available for the ore types which are going to be processed in the future. Let us consider the case where a mining company with a completely different setup concerning the mill dimensions, liner properties, ore properties etc. wants to use the proposed model to predict the optimum replacement interval and its corresponding savings. In this type of case, the prediction can be accomplished by comparing the physical properties of the ore types, which are going to be processed with those of the ore type for which data is available. As far as the present model is concerned, the variation of the throughput, process efciency and energy consumption needs to be considered. In order to determine the approximate range of the values of model parameters, the mill should be run for each new ore type for 10-15 days, so that the average level of throughput, process efciency and energy consumption can be obtained for the new types of ore that are to be processed. The next step is to incorporate the trends of process efciency and energy consumption over the life span of the mill liners for the ores that have been processed in the past. Based on this information, the presented model can provide a rough estimate of the replacement interval of the mill liners for the given case. Model risk. After performing a sensitivity analysis, it was found that the model sensitivity depends on the following parameters: throughput, process efciency and downtime cost. The process efciency and the throughput level may change in the future due to signicant changes in the ore properties and the availability of ore in the mines, which might lead to a variation in the results in reality compared to the results predicted by the developed model. Hence, there can be a technical risk (the hazard of damaging the mill shell) and an economic risk (nancial loss) associated with the model, which is explained as follows: Technical risk. No technical risk is associated, because, for the worst case scenario, the model considers the average lifecycle of the mill liners as a constraint and the model results for the optimum replacement interval will always be less than the threshold limit or the average lifecycle of the mill liners (the maximum delay in the replacement of mill liners). Economic risk. The present model considers a slight decrease in the process efciency parameter, which means that the grinding efciency decreases when the liner is worn out. Therefore nancial losses can occur only in the case where the grinding efciency of the liners is considered to be constant at each phase of the mill liners life, which is only theoretically possible. Based on the case studied in the paper, the maximum loss that could occur is 0.15 percent. The present model can also be used for replacement decisions for other similar systems consisting of non-repairable equipment by considering the periodic

Economic model for maintenance decision 95

JQME 18,1

operational cost data and performance index (e.g. process efciency) and the revenue together with the downtime and inspection costs. Conclusions An economic decision model for mill liner replacement has been developed considering maintenance policies and the process performance for the mill liners. The results obtained from the model show that an optimum maintenance policy can not only reduce the maintenance cost, but also affect the process performance, which leads to signicant savings. The results obtained for the optimum replacement interval show an increase of 0.3 percent to 0.5 percent, with a 95 percent condence interval, in the gross prot of the mill per year, by changing the current replacement policy to a proposed policy based on the given case study. A methodology is suggested that combines the different optimum decisions for individual ore types into a single optimum decision while taking the economic and technical characteristics of each ore type into consideration. The study presents an alternative method of decision making that does not use periodic wear measurements, due to the unavailability of sufcient mill liner wear data for analyses. The present approach also deals with the problems related to variation in the process data due to technical problems in the plant.
References Ascher, H. and Feingold, H. (1984), Repairable Systems Reliability: Modeling, Inference, Misconceptions and Their Causes, Marcel Dekker, New York, NY. Box, G. and Jenkins, G. (1976), Time Series Analysis: Forecasting and Control, Holden-Day, Oakland, CA. Christer, A.H. and Scarf, P.A. (1994), A robust replacement model with applications to medical equipment, Journal of the Operational Research Society, Vol. 45 No. 3, pp. 261-75. Clemen, R.T. and Winkler, R.L. (1999), Combining probability distribution from experts in risk analysis, Journal of Risk Analysis, Vol. 19, pp. 187-203. Dandotiya, R. and Lundberg, J. (2011a), Replacement decision model for mill liners, Journal of Quality in Maintenance Engineering, (submitted for publication). Dandotiya, R. and Lundberg, J. (2011b), Evaluation of abrasive wear measurement devices of mill liners, International Journal of COMADEM, Vol. 14 No. 3, pp. 3-17. Dekker, R. (1996), Applications of maintenance optimization models: a review and analysis, Reliability Engineering & System Safety, Vol. 51 No. 3, pp. 229-40. Dogramaci, A. and Fraiman, N.M. (2004), Replacement decisions with maintenance under uncertainty: an imbedded optimal control model, Journal of Operations Research, Vol. 52 No. 5, pp. 785-94. Duffuaa, S.O. and Al Sultan, K.S. (1999), A stochastic programming model for scheduling maintenance personnel, Applied Maintenance Modelling, Vol. 25, pp. 385-97. Hsu, J.S. (1988), Equipment replacement policy a survey, Production and Inventory Management Journal, Vol. 29, pp. 23-7. Jardine, A.K.S., Makis, V., Banjevic, D. and Ennis, M. (1998), A decision optimization model for condition based maintenance, Journal of Quality in Maintenance Engineering, Vol. 4 No. 2, pp. 115-21.

96

Kalala, J.T., Breetzke, M. and Moys, M.H. (2008), Study of the inuence of liner wear on the load behaviour of an industrial dry tumbling mill using the discrete element method (DEM), International Journal of Mineral Processing, Vol. 86, pp. 33-9. Kendall, M.G. (1984), Time Series, Oxford University Press, New York, NY. , B. (1992), Reliability analysis of hydraulic system of LHD machines Kumar, U. and Klefsjo using power law process model, Reliability Engineering and System Safety, Vol. 35, pp. 217-24. Pahl, G. and Beitz, W. (1996), Engineering Design: A Systematic Approach, 2nd ed., Springer, London. Radziszewski, P., Varadi, R., Chenje, T., Santella, L. and Sciannamblo, A. (2005), Tumbling mill steel media abrasion wear test development, Minerals Engineering, Vol. 18, pp. 333-41. Roll, Y. and Naor, P. (1968), Preventive maintenance of equipment subject to continuous deterioration and stochastic failure, Journal of the Operational Research Society, Vol. 19, pp. 61-71. Santarisi, N.S. and Almomany, R.M. (2005), Mathematical modelling of wear rate of non-repairable parts and their replacement strategies: cement mill liners as a case study, Journal of Quality in Maintenance Engineering, Vol. 11 No. 1, pp. 68-81. Sethi, S. and Chand, S. (1979), Planning horizon procedures for machine replacement models, Journal of Management Science, Vol. 25 No. 2, pp. 140-51. Shumway, R.H. (1988), Applied Statistical Time Series Analysis, Prentice Hall, Englewood Cliffs, NJ. Wang, H.Z. (2002), A survey of maintenance policies of deteriorating systems, European Journal of Operation Research, Vol. 139 No. 3, pp. 469-89. Winkler, R.L. and Makridakis, S. (1983), The combination of forecast, Journal of the Royal Statistical Society, Vol. 146, pp. 150-7. Yun, Y.W. and Choi, C.H. (2000), Optimum replacement intervals with random time horizon, Journal of Quality in Maintenance Engineering, Vol. 6 No. 4, pp. 269-74. Further reading Jardine, A.K.S. and Tsang, A.H. (2005), Maintenance Replacement and Reliability: Theory and Applications, CRC Press, Boca Raton, FL. Law, A.M. and Kelton, W.D. (1991), Simulation Modeling and Analysis, 2nd ed., McGraw-Hill, New York, NY. Metso Minerals (2002), Product Handbook: Basics in Minerals Processing, 2nd ed., Metso Minerals, Ersmark. Park, K.S. (1988), Optimal continuous-wear limit replacement under periodic inspections, IEEE Transactions on Reliability, Vol. 37 No. 1, pp. 97-102. Scarf, P.A. (1997), On the application of mathematical models in maintenance, European Journal of Operational Research, Vol. 99 No. 3, pp. 493-506. Corresponding author Rajiv Dandotiya can be contacted at: [email protected]

Economic model for maintenance decision 97

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints