Knowledge that Transforms

The focus of this paper is on determining appropriate combination rules for idiosyncratic ordinal utility functions in conjoint measurement. An axiomatic diagnosis is used which is based on explanatory criteria rather than goodness-of-fit or predictive criteria. Experimental results are presented comparing the selection of combination rules based on axiomatic tests and empirical fitting procedures for additive (x + y + z) and distributive (x(y + z)) combinations of three factors. A number of conflicting diagnoses result from the simultaneous applications of axiomatic tests and goodness-of-fit criteria, suggesting the importance of testing certain necessary conditions for simple polynomial combination rules, such as additive and distributive rules. Because the axiomatic approach is effective in rejecting inappropriate functional forms, we argue that it should complement rather than replace the empirical fitting diagnostic approaches.

Shift and days-off scheduling problems have received much attention in the literature of integer programming approaches to workforce scheduling. A typical managerial use would be to schedule full-time employees to minimize the number of labor hours while satisfying variable workforce requirements of a service delivery system. We present computational experience to show that an easily implemented application of linear programming frequently produces optimal solutions to these problems. When the context progresses toward a continuous operating environment (service delivery over 24 hours a day, 7 days a week) we stress the need to shed the myopic views of the shift and days-off scheduling formulations in favor of an integrative tour scheduling formulation. For this problem we observe that a simple heuristic initiated by rounding down the associated LP solution consistently produces near optimal solutions. This observation is based on experiments over varying workforce requirement patterns.

The conventional approach to the assembly line balancing problem assumes that the manufacturing methods to be used have been predetermined. However, in practice the design engineer has several alternatives available in the choice of processing, typically involving a trade-off between labor or capital intensive options. The choice of manufacturing method is frequently viewed as an investment or capital budgeting decision, contrasting projected savings in labor cost with the additional fixed cost for the more capital intensive alternatives. The manufacturing tasks (based on the selected processing alternatives) are then assigned to work stations so as to minimize the number of work stations (i.e., labor costs) necessary to achieve a desired production rate. This paper describes a method of simultaneously considering both the choice of manufacturing alternatives and the assignment of tasks to stations so as to minimize total costs (labor and fixed) over the expected life of the production line. The importance of considering these decisions jointly results from the fact that the benefits obtained from specific manufacturing alternatives should not be limited to anticipated labor savings alone. The true measure of achievable labor savings can only be determined after an assignment of tasks to stations has been chosen. For instance, although a processing alternative may reduce the total work content of a set of tasks, if the resulting line balancing assignment does not reduce the number of stations required to achieve a desired production rate, the assumed savings will not be achievable and will serve only to increase the idle time of the line. On the other hand, an apparently trivial reduction in the time to complete a certain task may lead to a more efficiently balanced line, producing a much greater real savings in labor cost than had been anticipated, due to a reduction in both work content and idle time for the line. The combined processing alternative line balancing problem can be formulated as an integer programming problem. Two alternate formulations are provided which differ in the degree of flexibility in selecting a cycle time. A branch and bound procedure is described for the fixed cycle time situation which takes advantage of the special structure of the problem to provide an efficient method capable of solving problems of practical interest. The effectiveness of the proposed procedure is demonstrated by application to an actual redesign of an assembly line for a major auto-industry supplier.

This paper defines a model that covers the multimachine extensions of all existing flow-shop problems with time lags as well as flow-shop problems where setup, processing and release times are separated. Approximate solutions and lower completion time bounds are developed for the restricted and unrestricted cases.

A probabilistic investment model is formulated as a Wiener process with a barrier. A planning horizon, targeted rate of return, discount rate, and the mean and variance rate of return are the important parameters in the model. Sensitivity analyses are studied. Several significant statements can be made: (1) contrary to traditional mean-variance portfolio models, rate of return variance may not always be an undesirable characteristic, since higher variance can increase the chances of achieving certain types of investment goals; (2) one can almost always achieve certain types of investment goals if the time horizon is sufficiently long.

This paper fully discusses methods for ranking a set of alternatives in the fairest possible way according to a minimum violations criterion. New methods, based on finding paths and circuits in graphs, are presented for ranking participants in round-robin and generalized tournaments, and for consensus and group decisionmaking problems. The objective of the paper is to review existing methods for tackling these problems, and compare them with the new methods, according to a “fairness” criterion, and the amount of computing required to reach a solution. It is shown that the new methods often exceed the existing methods in both fairness and reduced computing requirements. In particular, the new methods are generally more versatile than existing methods. This allows organizations to obtain the fairest ranking of a number of alternatives, according to their managers' or employees' wishes. Particular attention is given to incomplete rankings where insufficient exposure of an individual to some alternatives restricts that individual to ranking only the remaining alternatives.

This paper presents a procedure for determining the number of simulation observations required to achieve a preassigned confidence interval for means estimated by simulation. This procedure, which is simple to implement and efficient to use, is compared with two other methods for determining the required sample size in a simulation run. The empirical results show that this procedure gives good results in the precision of estimated means and in sample size requirement.

In variable-base budgeting an organization's resources are partitioned into two categories, those allocated to decision making and those allocated to implementation, and both sets of resources are further subdivided as necessary—between departments, across time stages, etc. The purpose of the budgeting effort is to maximize an expected return, which depends on that part of the organization's resources allocated to implementation, net of risk, which is reduced by allocating resources to decision making. In a previous paper it was shown that zero-base budgeting, in which substantial resources are allocated to decision making, is optimal only under certain conditions, one of which is that management resources must be costless. Since these conditions seldom obtain in the real world, it seems reasonable to examine the sensitivity of the risk/return performance of a zero-base strategy with respect to deviations from the conditions for which that strategy is optimal. The principal conclusion of the research reported here is that the mesa phenomenon—the relative insensitivity of the outcome of resource allocation decisions in the neighborhood of the optimum—applies to the allocation of management resources as well.

A new method for solving the corrugator trim problem is presented. This problem draws its name from the difficult task facing scheduling personnel in corrugator box plants each day: to determine the least-cost method of combining customer orders on the corrugator, where one of the major costs to avoid is waste or excess trim lost from the materials used. Trim loss, however, is only one of several major concerns. Others include corrugator width utilization, cutting pattern changes, avoidance of split orders, and so on. A full discussion of these factors and associated trade-offs is included in the paper. The solution procedure presented is capable of dealing with these and other trade-offs in a systematic manner. In particular, this approach eliminates the spreading of orders over several stock sizes, a characteristic which has plagued the linear programming-based procedures that have been proposed for solving the problem. This improved material handling characteristic of the solution is obtained through the controlled generation of solution elements. These elements then become potential activities to be scheduled by a 0-1 integer programming algorithm. The algorithm used is well suited to computers of moderate size and speed. In addition to solving the corrugator trim problem, the proposed method has sufficient flexibility to make it a useful vehicle for solving similar problems in other industries which manufacture such products as glass, steel, composition board, and furniture.

The effect of decreases in food energy intake on the activity pattern of an individual may be described by an M x /M/1 queueing model. λ represents the rate of activity arrival to the system. E(k) represents the average energy level at which the system performs, μ represents the rate at which the system processes energy units. p 0 = 1 − λE(k)/μ represents the probability of an idle system, of the system at rest. The model was validated by collecting data on food energy intake and activities from ten healthy young men. When food energy intake decreased an average of 500 kcal/day, 1 − p 0 , the probability of a nonidle system, decreased from 0.431 to 0.413, p < 0.005. The greatest parameter change occurred in λ, which decreased from 0.245 to 0.235, p < 0.001. Thus, when food energy intake decreases, the activity pattern of an individual as represented by a queueing system changes: more time is spent at rest; less time is spent in moving activities.