Risk-Based Multi-criteria Decision Management for a Railway System in Thailand

Infrastructures Management includes a wide range of activities that are essential to efficient working of country utilities and services. The nature of infrastructure management involved a variety of operations such as planning, design, construction, condition evaluation, maintenance, improvement and so on. Infrastructure Maintenance Management System has gradually become an important field for not only developed countries but also for developing countries particularly Thailand. This is because of nature of the infrastructures deteriorate over time and it is crucial to maintain them in order to preserve their condition or performance to be at the acceptable level which satisfies the user’s needs. If the maintenance is neglected, it can lead to disruptions in services for example, traffic accidents, water leaks, or power cuts.

Among different facilities, railway is one of the important infrastructures. A railway system combines several objectives, such as provision of an adequate level of service or safety, preservation of the facility of a number of physical facilities that include railways, bridges, terminals, and traffic control devices. Once these facilities deteriorate, the suitable maintenance action needs to be taken. Most of maintenance policies have been developed in order to reach the most beneficial plan under the budget initiation. However, in the long-term planning, most projects or programs are subjected to risk and uncertainty. The expected maintenance costs of facilities increase as the measurement of uncertainty increases (Madanat and Ben-Akiva, 1994).

Maintenance practice and strategies have a strong impact on the management cost of a facility. Deferred maintenance results in increased life-cycle costs rather than a cost saving (NCHRP 58, 1979). The total ownership cost of infrastructure system is greatly influenced throughout different its lifecycle from the beginning to disposal includes maintenance costs, which are affected by the types of maintenance selected and implementation schedule for these policies. Maintenance costs are also affected by the risk associated with unplanned or unscheduled repairs arising from contingencies such as accidental damages (Ayyub and Popescu, 2003). The decision makers have to establish an optimum maintenance management strategy for an infrastructure system which can reduce or minimize the aforementioned cost effectively and efficiently, while ensuring that the system goals are conformed. The main objective of a decision-making process is to maximize the benefits to its customers and users, based on well-defined goals and with available resources (NYSDOT, 1998).

The current maintenance practice of State Railway of Thailand (SRT) is based on Cyclic Maintenance. Cyclic maintenance is the maintenance program which indicates the exact time and the duration of maintenance. The decision-making of selection of maintenance alternatives is based on track condition inspection i.e. Track Quality Index (Q.I.) (SRT, 1995). However, the development of maintenance management system for individual objective is not effective because it does not take into account all the needs of the system. Without the effective decision-making process, the agency cannot claim that each maintenance alternative meets the obligation to maintain the railway network effectively. Because of the variety of system objective of railway maintenance management, the decision-making process takes into consideration of a multicriteria decision-making process. In addition, the benefits under different measurement units make the decision-making process more complicated. Therefore, the decision makers have to evaluate properly among competitive maintenance alternatives (Li and Sinha, 2004). In order to mitigate this complication, multiobjective decision-making approaches would be useful. It provides a convenient set of mathematical tools to identify an optimum alternative given a set of competing objectives (Clemens, 1997). In this study, one of multiobjective decision-making technique called Analytic Hierarchy Process (AHP), is employed to identify an important of objectives to use in a multicriteria decision-making process. With AHP, weighting factors of all objectives can be identified so that the decision makers can be easily used in a calculation of decision-making.

Mr. Arthit Krachang’s study proposed to develop risk-based multiobjective functions that influence the railway maintenance management. In addition, weighting factors of individual objective provides a non-dimensional output so that planners or decision makers can use those functions for decision-making about a suitable action in railway maintenance management system.

The main objective of his study is to develop a risk-based multicriteria decision framework for the railway system in Thailand. The basis of the decision-making is to maximize the overall benefits as well as to minimize the maintenance costs and risks regarding to the budget constraint. In order to achieve the main objective, secondary objectives were defined as follows: (1) determine objectives of railway maintenance management system and develop the objective functions regarding to influencing parameters as well as their weighting factors; (2) respond to risk through an effective and efficient decision-making process by using quantitative analysis of the effect of influencing parameters that are subjected to risk; and (3) develop an overall risk-based decision-making process for the railway maintenance management system.

Conclusion

In this study, the risk-based multicriteria decision-making process for the railway maintenance selection is carried out by using risk simulation technique. The weighting factors and objective functions are used to setup an overall objective value optimization framework. The developed cost effectiveness and overall objective value optimization model are used to identify the optimal maintenance plan. Finally, the cost effectiveness method is carried out to present the results comparison. The Northern line and the Eastern line are selected to present a case study.

Questionnaire surveys were carried out among the SRT experts who involved in the railway maintenance. The outcome of the questionnaire surveys present the objectives, influencing factors, and their weighting factors that involved with the railway maintenance selection. The obtained weighting factors are determined and then, validated by consistency ratio in the AHP methodology. According to the SRT expert opinions and the obtained weighting factors, all of objectives are the important objectives for the railway maintenance while some of influencing parameters have less important but not the negligible factors. The most three important objectives are safety, costs, condition.

The deterioration model is developed from the historical railway condition and integrated into cost effectiveness optimization model and overall objective value optimization model for the Track Quality Index (Q.I.) prediction. The developed deterioration model is based on the regression analysis of historical Q.I. value and then, the coefficient R2 is used for the model validation. The processes of deterioration model development are discussed in section 4.4. The predicted track condition is used to calculate the cost effectiveness ratio and the objective value.

Development of objective functions of different railway maintenance objectives is the important part of this study. Interview and discussion with SRT experts were carried out to develop the objective functions and validate the results. The objective function refers to a function that quantifies the quality of a solution in an optimization model. The objective function returns an unknown quantity representing the quality measurement of a decision that needs to be made. In this study, there are two optimization models were created for results comparison that are the deterministic model and the probabilistic model.

The development of the optimization model consists of decision variables, objective function, and constraints. The first step when formulating a model is to identify and assign names to the decision variables. The decision variables are the variables that can be directly controlled by the decision makers and those values determine the solution of the model. In this study, the decision variable is the maintenance type. In the optimization model, the maintenance plan which gives either the maximum or the minimum value of the objective function needs to be determined. The second step is to develop the objective function in terms of the decision variables. The objective function is the formula which specifies the goal that is trying to achieve. The goal can either be to maximize or to minimize the value of the objective function.

However, the deterministic cost effectiveness optimization model is the used traditionally in the policy analysis by treating the input variables as exact value. The probabilistic cost effectiveness optimization model is a modified optimization model which takes risk and uncertainty into account. In the probabilistic model, the variables that are subjected to risk are treated probability density function (PDF). In this study the track initial condition is the parameter which is subjected to risk. The historical data is fitted with the best fitted probability density function (PDF) based on chi-square test. The developed PDF of track condition is used in the simulation process. It was randomized treated in the cost effectiveness function until the cost effectiveness ratio is statistically converged.

The optimization results which obtained from both model show that the most effective maintenance plan is 4-year maintenance cycle and in middle of 4 years are do nothing. There are some difference is that the maintenance type in some first year are different. The optimization results were introduced to the SRT experts to validate in term of acceptance. Both of the obtained optimal maintenance plan yield very close maintenance costs and effectiveness ratio. The maintenance costs for both of the optimal maintenance plan are under the budget limitation which is specified by SRT policy. The predicted conditions of the railway are above the minimum acceptable level.

An effective railway maintenance selection is the multiobjective decision-making process. Since the cost effectiveness optimization is solely based on single objective optimization, this situation leads this study to modify the cost effectiveness optimization by applying the concept of multiobjective optimization. According to the results of the questionnaire survey, there are 6 objectives and 19 influencing parameters in railway maintenance selection. In order to accumulate all of objectives in railway selection process, creation of 19 objective functions must be carried out. Due to the time and historical data limitation, the safety and cost objective are selected the present the case study.

According to SRT expert discussion and the obtained weighting factors, railway safety is the most important objective for the railway maintenance. Therefore, the agency maximizes the railway safety by applying the cyclic maintenance policy. However, the maintenance costs are required to be minimized due to the budget limitation. In this study, safety objective function and costs objective function are developed to present the multiobjective optimization. In addition, each objective has more than one influencing parameter. Under this circumstance, weighting factors of different influencing parameters of an objective are used. A product of objective function and weighting factor of influencing factor represents an objective value of an individual influencing factor. The summation of these values represents a total objective value of an objective. The summation of a product of weight and total objective value of an objective represents the overall objective value of each maintenance plan.

The result of the overall objective value optimization reveals that the most effective maintenance plan is 4-year maintenance cycle. It should be noted that the obtained optimal maintenance plan is similar with the optimal plan which obtained from the cost effectiveness optimization. However, the type of maintenance in first year is different. In addition, for the sections that their rails have reached the service life the heavy or medium maintenance are applied in first years.

The overall risk-based decision-making process is completed by accumulating the weighting factors and objective functions for the railway maintenance. The optimal maintenance plans which are obtained from the cost effectiveness and the overall objective value optimization, and current SRT maintenance policy are simulated to identify the PDF of cost effectiveness ratio. Among three maintenance plan the optimal maintenance plan which obtained from the cost effectiveness optimization yields the highest cost effectiveness ratio, the optimal maintenance plan which obtained from the overall objective optimization is the follower, and the SRT current maintenance policy yields least one. In comparison, the maintenance plans which yield the higher cost effectiveness are subject to higher variation of the cost effectiveness ratio.

The mean of cost effectiveness ratio or the standard deviation can be used for prioritize the prefer maintenance alternative in the decision-making process. In this study, if the decision maker prefers only the maximum benefit of the maintenance plan, the optimal plan which obtained from the cost effectiveness optimization is the final decision. In comparison, if the decision maker prefers the less variation of the benefit, the current SRT maintenance policy or the optimal maintenance plan which the overall objective value optimization are the better decision. The developed risk-based multicriteria decision framework is recommended to apply in the network-level management for planning purposes.

An overall risk-based multiobjective optimization reveals the other view point of railway maintenance selection that should be incorporated into the decision making process. The risk-based simulation does not intend to show that the result yields the better maintenance plan or higher accuracy than the deterministic one or the current SRT maintenance practice, but it demonstrates that the risk-based simulation reveals the possible cost effectiveness ratio that are neglected in the deterministic one. Moreover, the risk-based simulation results encourage decision makers to be concerned not only with the outcome value but also with the amount of risk each decision carries.

His thesis abstract is copied and posted.

ABTRACT
State Railway of Thailand (SRT) is a public agency which has the responsibility for planning, constructing, managing, and maintaining railway network in Thailand. Due to the maintenance budget limitation, SRT has established the Railway Maintenance Management System to manage their railway network. SRT has conducted the cyclic maintenance and collected their railway inventory data and its condition data. However, the determination of railway maintenance management system is solely based on the inspected track condition data.
However, an effective railway maintenance selection is the multiobjective decision-making process which is very complex and it is subjected to risk and uncertainty. This situation leads this study to modify the existing determination by applying the concept of objective function. The overall objective value is the summation of objective values for different objectives, influencing parameters, and their weighting factors which obtained from the Analytical Hierarchy Process (AHP). Moreover, some influencing parameter can be changed over time or beyond the control of agency. Therefore, the decision-making process may still expose to risk by invisible risky information.

To address this problem, a framework which takes risk into account is developed for selecting optimal maintenance alternatives. Incorporation of risk analysis by considering the inputs variables in the form of probability density function (PDF) then simulated by using Monte Carlo simulation technique. The result of a risk-based decision-making process can help SRT determine the optimum maintenance alternatives those are targeted towards maximizing benefits and minimizing cost under risk control. Finally, Northern and Eastern routes are selected to present as a case study of an overall risk-based decision-making process.