UNIT-V (Simulation)

Introduction: simulation analysis is a natural and logical extension to the analytical and mathematical techniques used for solving the problems in operation research. Simulation which can appropriately be known as management laboratory, determines the effect of alternate policies without disturbing the real system. Recent advances in simulation methodologies, software availability, and technical developments have made simulation one of the most widely used and popularly accepted tool Operation research and systems analyses. It helps us in deciding the best policy with the prior assurances that its implementation will certainly prove to be beneficial to the organization. Definitions of simulation: Definition1: simulation is a representation of reality through the use of a model or other device which will react in the same manner as reality under a given set of conditions. Definition2: simulation is the use of system model that has designed the characteristics of reality in order to produce the essence of actual operation. Defination3: According to Donald G. Malcolm, a simulated may be defined as one which depicts the working of a large scale system of men, machines, materials and information operating over a period of time in a simulated environment of the actual real world conditions. Types of simulations: simulation is mainly of two types: i. Analogue simulation (or environme ntal simulation): The simple example of simulating the reality in physical form, which we may refer as analogue (or environmental) simulation. ii. Computer simulation (or System simulation): For the complex and intricate problems managerial decision making, the analogue simulation may not be applicable, and the actual experimentation with the system may be uneconomical also. Under these situations, the complex system is formulated into a mathematical model for which a computer program me is developed, and then the problem is solved by using high speed electronic computer. Such type of simulation is called a computer simulation or system simulation. The simulation models can be classified into following four categories: A. Deterministic models: in these models, in put and out put variables are not permitted to be random variables and models are described by exact functional relationship. B. Stochastic models: in these models, at least one of the variables or functional relationship is given by probability functions. C. Static models: these models do not take variable time into consideration;

D. Dynamic models: This model deals with time varying interaction. Why to use simulation: it has been discussed in the chapter on What is operation research? that mainly following techniques are adopted for solving various types of managerial decision making problems in operations research. (i) Scientific method, (ii) Analytical method, and (iii)Iterative method. But each method has its own drawbacks and limitations as discussed below. 1. Drawbacks of scientific method: the steps of scientific method have the following limitations and difficulties: a. It may be either impossible or extremely costly to observe certain processes in the real life situations. b. The observe system may be so complex that it may be impossible to describe it in terms of a set of mathematical equations. c. Even through the mathematical model may be formulated to describe system under study; straight- forward analytical solutions may not be available. For example, such situations may arise in complex queueing problems, job-shop problems, high order difference equatio ns, complicated stochastic models, multi- integral problems.etc. d. It may be either impossible or very costly to perform validating experiments on mathematical models describing the system. Thus on account of these drawbacks the scientific method cannot be used to solve complex managerial decision- making problems.

2.Drawbacks of analytical method: analytical techniques used in dynamic programming, queueing theory, network models, etc., are not sufficient to tackle all the important managerial problems requiring data analysis due to following limitations: a. Dynamic programming models can be used to determine optimal strategies taking into account the uncertainties and can analyses multi period planning problems. But, still it has its own shortcomings. Dynamic programming models can be used to tackle very simple situations involving a very few variables. If the number of state variables becomes larger, the computation work becomes quite complex and difficult. b. Similar limitations also hold good for other mathematical techniques like dynamic stochastic models such as inventory and waiting line situations. Only small scale systems are amenable to these models. But by making a number of assumptions the systems are simplified to such an extent that in many cases results thus are only rough approximations.

3. Drawbacks of ite rative method: a. In linear programming models, we assume that data does not change over the entire planning horizon. This is one time decision processes and assumes average values for the decision variables. If the planning horizon is long, say 15 years, the multi period linear programming model may deal with the yearly averaged data, but will not take into account the variations over the months and weeks. Consequently, month to month and week to week operations are left implicit. b. Other important limitation of linear programming is that it assumes that data should be known with certainty. In many real situations, the uncertainties about the data are such they cannot be ignored. Incase the uncertainty relates to only a few variables, the sensitivity analysis can be used to determine is effect on the decision. But in the situations where uncertainty pervades the entire model, the sensitivity analysis may become too cumbersome and computationally difficult to determine the impact of uncertainty on recommended plan. From above maintained drawbacks, we conclude that whenever the characteristics like uncertainty, complexity, dynamic interaction between the decision and sub sequent event and need to develop detailed procedures and finely divided time intervals, all combined together in one situation, then model becomes too complex to be solved by any of the techniques of mathematical programming and probabilistic models. Then such complex model mus t be analyzed by some other kind of quantitative technique; which may give quite accurate and reliable results. Many new techniques are investigated so far, but among all, the best available is simulation. Generations of random numbers: Random variable: The random variable is a real valued function defined over a sample space associated with the out come of a conceptual chance experiment. Random variables are classified according to their probability density function. Random Variate: It refers to a particular outcome of an experiment, i.e., a numerical or sample value of a random variable. Random number: As generally understood it refers random variable or numerical value assigned to random variable following uniform probability density function (i.e., normal, position, exponential, etc,) in other words, it is a number in a sequence of numbers whose probability of occurrence is the same as that of any other number in that sequence. Pseudo-random numbers: Random numbers are called pseudo random numbers when they are generated by some deterministic process but they qualify the pre-determined statically test for randomness. The sequence of numbers generated by such process is completely determined by the input data (or the first random number) used for the method.

Generation of random numbers: Monte-Carlo simulation needs the generation of a sequence of random numbers which constitute an integral part of the simulation model and also help in determining random observations from the probability distribution. Random numbers are assigned in such a manner that their proportion is exactly to the probability distribution. Random numbers may be found by computer, by random tables, or manually. Generally, it is not practical to obtain random numbers manually because it requires much labour. Usually such devices as: roulette wheels, dice rolling, card shuffling, etc., are used in these method. But the most common method to obtain random numbers is to generate them by a computer program. These numbers lie between 0 and 1 (0 and 100 %) in conjunction with the cumulative probability distribution of a random variable, including 0, but not 1. Many tables of random numbers are available in the literature. These numbers are considered to be truly random numbers because these were generated using some random physical process. While choosing the random numbers from the table, the starting point on the table is immaterial. We may start with any number in any column or row, and proceed in the same column or row to the next number, but a consistent, unvaried pattern should be used in drawing random numbers. We should not jump from one number to another indiscriminately. In the case of choosing random numbers for more then one concerned variable, different list of numbers for each variable should be used; other wise same random numbers would imply dependence among different variables. Monte-Carlo simulation: Monte-Carlo technique has become so much important part of simulation models that the terms are often assumed to be synonymous. However, it is only a special technique of simulation. The technique of Monte-Carlo involves the selection of random observations with in the simulation model. The sequence is restricted for application involving random numbers to solve deterministic and stochastic problems. The principle of this technique is replacement of actual statically universe by another universe described by some assumed probability distribution and then sampling from this theoretical population by means of random numbers. In fact this process is the generation of simulated statistics (random variables) that can be explained in sample terms as choosing a random number and substituting this value in standard probability density function to obtain random variable or simulated statistics. Applications of Simulations: Simulation has a large number of applications. It can be used for learning about the operating characteristics of a new airplane by simulating a. flight conditions in a wind tunnel, b. on electronic or hydrolic analog models of production processes or c. economic systems, or

d. on mathematical models of such real- life systems as inventory control, e. production scheduling, f. Network analysis, and so on. It can also be used for a. b. c. d. e. f. planning military strategy, traffic control, management games and role playing, medical diagnosis, hospital warehouses, Evaluation of industrial and commercial policies.

Application to inventory control: For providing efficient to the customers, it is necessary to choose to reorder point with proper consideration of demand during lead time. If the lead time and of inventory per unit time both are random variables, then the simulation technique can be applied to determine the effect of alternate inventory policies on a stochastic inventory system, e.g. different combinations of order quantity and order point. The basic approach would be to find the probability distribution of the input and output functions of the past data. Then, we run inventory system artificially by generation the future observation on the assumptions of the same distributions. The method involves a good amount of computation. But in simple problems it is possible to generate artificial samples for future with the help of random numbers and then the entire computations are done with the help of desk calculator. Of course, it becomes necessary to use electronic computers for solving more complex problems. Application to queueing problems: In fact, some queueing problems cannot be solved explicitly by analytical methods, in such cases, the only possible method of solution is to simulate the experiment. Simulation Languages: So far, we have discussed the material which provides a background for the study of the systems that need simulation. Now the next important thing is the formulation of the computer mode deciding how many variables can be included in the model. A simulation computer program can be written in one of the conventional computer languages such as FORTRAN, COBOL, ALGOL, or PL/1 with out restoring to a special purpose simulation language. However, these languages require extensive programming experience. It can be observed that even in a simple queueing problem, various complicated details are involved in the simulation model. However, the development of special simulation languages has considerably reduced the program, preparation time and cost with features especially for different types of models and systems. The special purpose simulation languages which have been developed so for are: GPSS, SIMSCRIPT, GASP, SIMPAC, DYNAMO, SIMULATE, CSL. The aim of these languages is to speed up the conversion of a simulation model to computer program. The several simulation

languages are made so that they can be applied to different types of problems in order to make their simulation procedure more easy and automatic for the user. The most popular among these languages are GPSS, SIMSCRIPT, and DYNAMO. These three languages give automatic instructions for time keeping and other common simulation operations. The GPSS (General Purpose Systems Simulator) can be used to a wider class of systems in order to maintain a fixed set of procedures for performing the simulation automatically. There are 4 entities that constitute the foundations of GPPS: (i) transactions (ii) facilities (iii) storage, and (iv) blocks. The basis of SIMSCRIPT language is a description of the system in terms of concepts of entity, attribute, set and event, etc. This language is more flexible then GPSS but require knowledge of FORTRAN. Both of these languages are well-suited to queueing type problems. The DYNAMO language is similar to SIMSCRIPT except that it simulates continuous type models instead of models involving individual events. This language is not particularly suitable for queueing type problems. Thus it seems to be more applicable to Macro simulation which involve small-scale models of pieces of a firm, such as queueing, production scheduling and inventory, etc. Scope of simulation technique: Simulation is an approach rather then the application of a specific technique. Thus there is a wide range of applications of computer based simulations models. The major use of computer based Monte-Carlo simulation has been to solve the complex queueing problems. The GPSS program has been developed has a queueing simulation tool. As a large number of job-shop simulation program have been developed which involve deterministic times for the individual operation of prescribed order. Because of different processing times for similar type operations and different order operation seq uences, it becomes difficult to predict the waiting time for specific order at any prescribed service station. For better scheduling, orders must be scheduled with a provision of waiting at the various service stations they will pass through. Simulation helps us in accurately estimating such waiting times. A good amount of work has been done for the development of inventory simulation models. There is fairly wide applicability and specialized use of such models as determination of optimal order level and lot size under probabilistic demand and lead time, optimal review period and ordering policy for continuous review inventory models. A large number of network simulation models have also been developed so for. For example, we can determine the critical path even if we are given randomly selected activity-time for each activity. Repeating this process large number of times, the probability distribution for project completion time can be determined. Also, we can determine the probability that each given

activity is on the critical path. Many other applications of simulations models include the following (i) Financial studies involving risky investments. (ii) Military studies of logistics, support planning and weapon system effectiveness. (iii)Testing of decision rules for hospital admission and operating policies. (iv) Studies of individual and group behavior.