Simulation of a Tomato Processing Plant by Abasiano Udofa An Engineering Project Submitted to the Graduate Faculty of Rensselaer Polytechnic Institute in Partial Fulfillment of the Requirements for the degree of MASTERS OF ENGINEERING IN MECHANICAL ENGINEERING Approved: _________________________________________ Ernesto Gutierrez-Miravete, Engineering Project Adviser Rensselaer Polytechnic Institute Hartford, CT May 2014 © Copyright 2014 by Abasiano Udofa All Rights Reserved ii CONTENTS Simulation of a Tomato Processing Plant ........................................................................... i LIST OF FIGURES ........................................................................................................... v ACKNOWLEDGEMENT ................................................................................................ vi ABSTRACT .................................................................................................................... vii 1. Introduction.................................................................................................................. 1 1.1 Manufacturing Systems ...................................................................................... 1 1.2 Modeling Manufacturing Systems ..................................................................... 1 1.3 Discrete Event Simulation.................................................................................. 2 1.4 Tomato Processing Plant .................................................................................... 2 2. Methodology ................................................................................................................ 5 2.1 Mass Balance ..................................................................................................... 5 2.2 Deterministic vs. Stochastic Simulation ............................................................ 7 2.3 Hand Calculation and Queuing Model ............................................................... 8 3. Discrete Model Event ................................................................................................ 12 3.1 Assumptions ..................................................................................................... 12 3.2 Locations .......................................................................................................... 12 3.3 Entities and Arrivals ......................................................................................... 13 3.4 Processing and Routing .................................................................................... 13 4. Results........................................................................................................................ 15 4.1 Deterministic Runs ........................................................................................... 15 4.2 Stochastic Runs ................................................................................................ 18 4.3 Recommendations ............................................................................................ 24 4.4 Simulation vs. Hand Calculations .................................................................... 24 5. Conclusions................................................................................................................ 26 6. References.................................................................................................................. 27 7. Appendix.................................................................................................................... 33 iii LIST OF TABLES Table 2.1: Distribution Functions ...................................................................................... 8 Table 2.2: Deterministic hand calculation ......................................................................... 8 Table 2.3: Spreadsheet of Hand Calculations .................................................................. 11 Table 4.1: Location Statistics for deterministic0 ............................................................. 16 Table 4.2: Location Statistics for deterministic1 ............................................................. 16 Table 4.3: Location Statistics for stochastic1 .................................................................. 19 Table 4.4: Location Statistics for stochastic .................................................................... 19 Table 4.5: Location Statistics for stochastic10 ................................................................ 22 Table 4.6: Throughput times............................................................................................ 24 iv LIST OF FIGURES Figure 1.1: Layout of the Tomato Processing Plant .......................................................... 4 Figure 2.1: Mass Balance for Tomato Processing Plant .................................................... 6 Figure 3.1: Locations ....................................................................................................... 13 Figure 3.2: Processing and Routing ................................................................................. 14 Figure 4.1: Location Utilization for deterministic0 ......................................................... 17 Figure 4.2: Location Utilization for deterministic1 ......................................................... 17 Figure 4.3: Time Plot for Deterministic runs................................................................... 18 Figure 4.4: Location Utilization for Stochastic1 run ....................................................... 20 Figure 4.5: Location Utilization for Stochastic run ......................................................... 20 Figure 4.6: Time Plot for Stochastic1 run ....................................................................... 21 Figure 4.7: Time Plot for Stochastic run ......................................................................... 21 Figure 4.8: Location Utilization for Stochastic10 run ..................................................... 23 Figure 4.9: Time Plot for Stochastic10 run ..................................................................... 23 v ACKNOWLEDGEMENT I would like to first thank God for his grace and strength to study. I also want to thank all of my teachers and professors I have had throughout my educational journey. A particular note of thanks is given to my mother who had the idea of studying tomato processing plants. I would also like to thank all of those who contributions to the field of simulation I have used in this paper. The books Modeling and Analysis of Manufacturing Systems and Simulation Using ProModel provided an understanding of manufacturing systems and the ProModel computer program. I would like to thank Dr. Ernesto for his guidance through this process and his helpful insights on manufacturing systems and ProModel. Lastly I would like to thank Dave Hoeppner for his assistance in finishing this project. vi ABSTRACT Tomato processing plants are large complex manufacturing systems that can often be difficult to anticipate, plan, and react to the variability. Simulation is the imitation of a dynamic system using a computer model. Simulation began to be used in commercial applications in the 1960s. Now simulation is used by businesses to design, implement, and optimize complex manufacturing and service systems. Simulation tools can be used to gain further understanding of interdependencies and effects of variability on the production of a tomato processing plant. This paper will discuss the development of a simulation of a model of a tomato processing plant using Pro Model, a discrete event simulation software. The results of deterministic vs. stochastic simulation will be discussed. The benefits of simulation vs. hand calculation method will be presented. Recommendations for the most beneficial use of simulation software in the tomato processing industry are vii presented. 1. Introduction The purpose of this report is to demonstrate and discuss the modeling and analysis of a tomato processing plant using the discrete event simulation program Pro Model. A brief overview of manufacturing systems and how they are modeled computationally and virtually will be given. A description of the tomato processing sequence of events will be given for a better understanding of the system under study. The results of the computer simulation will be compared with those obtained using simple hand calculations. 1.1 Manufacturing Systems Manufacturing systems are processing systems where raw materials are transformed into finished products through a series of operations performed at workstations. These systems consist of entities, activities, resources, and controls that define the parameters for processing. Entities are the items being processed through the system. Activities are the tasks being performed in the system. Resources are what is being used to perform the activities. Controls dictate when, where, and how the activities are performed. The resulting interactions of these elements are what make manufacturing systems complex and difficult to evaluate. Interdependencies and variability are the two factors that make up the complexity of manufacturing systems and make the behavior difficult to analyze and predict. The manufacturing system under study in this paper is a tomato processing plant. 1.2 Modeling Manufacturing Systems Hand computation and simulation are methods used to model manufacturing systems to help understand the complexities and make responsible decisions. Hand computation methods are extremely useful, but can be limited and inefficient with larger, more complex systems. Simulation is a modeling and analysis technique used to evaluate and improve dynamic systems of all types. Simulation is typically performed on a computer utilizing various computer programs designed for capturing the behavior of specific systems. Accurately predicting the performance of complex systems and having the ability to test various scenarios before making major decisions that affect the system is 1 why simulation is important. Proper simulation accounts for interdependencies of a system that cannot be obtained using other analysis techniques. This allows for risk-free trial with no disruption to the current system and provides objective evidence to substantiate changes to the system or guide in building a new system. 1.3 Discrete Event Simulation How simulation works is dependent upon the method of simulation chosen. The common ways of characterizing simulation methods are static vs. dynamic, stochastic vs. deterministic, and discrete event vs. continuous. Static simulation is one that is not timedependent while dynamic simulation is dependent on time. Dynamic simulations are well suited for service and manufacturing systems. Deterministic vs. stochastic simulation has to do with the nature of the inputs and outputs. Deterministic simulation has fixed inputs and outputs, while stochastic simulation has random inputs and outputs. Deterministic simulation should always produce the same outcome no matter the number of run times. Stochastic simulation requires several runs to get an accurate performance estimate due to variations of outputs for a given run. Discrete-event simulation is based on the tracking of events as they occur at distinct times during the simulation, while continuous simulation is based on the tracking of events as they change continuously with respect to time. Discrete-event simulations typically reflect many manufacturing and service systems. The methods chosen for the simulations used for this project are dynamic, discrete-event, both stochastic and deterministic. These parameters best represent tomato processing plants due to the change of state from solid tomatoes to tomato paste. As shown in the following section tomato processing plants are processoriented systems, which are represented well by discrete-event simulation. Pro Model was the software chosen to complete this simulation. Pro Model is a powerful commercial simulation tool designed to effectively model any discrete-event simulation processing system. 1.4 Tomato Processing Plant A tomato processing plant takes fresh tomatoes and turns them into paste by chopping, heating, and removing the water from the tomatoes. Tomato processing plants are 2 complex manufacturing systems. The tomato paste manufacturing system is a process layout known as a flow line. All the tomatoes move along the same sequence shown in Figure 1.1. The tomatoes are dumped into the receiving stations from the trucks transporting them. The receiving station also includes 3 to 5 greater times the amount of water then the weight of the tomatoes for cleaning purposes. The tomatoes are routed to the washing/sorting station next. Having been washed under a clean water spray the unfit tomatoes are picked out and sent to scrap while the rest of the tomatoes are sent to the chopping station. At the chopping station the tomatoes are chopped by a hammer chopper or a special mono-pump where they are broken and pulped. The pulp is preheated to 85-98°C for Hot Break processing. The pulp is then sent to the hot break station consisting of a de-juicing unit (or juice extractors) composed of two centrifugal stations: a pulper and a refiner, equipped with two sieves having different meshes. The first sieve treats solid pieces up to 1 mm (or more), while the refiner processes solid pieces up to 0.6 mm (or more). Two products therefore come out of the de-juicing unit: refined juice ready for concentration and waste which is sent to scrap. The entire concentration process (evaporation) takes place under vacuum conditions and at low temperatures, significantly below 100°C. Product circulation inside the various concentric tubular exchangers is carried out by special stainless steel pumps designed to ensure that the product is conveyed inside the exchanger tubes at enough speed to avoid “flash evaporation” (burning the tomato paste). The concentrate is sent from the evaporator directly inside the aseptic system tank where it is packaged. 3 Figure 1.1: Layout of the Tomato Processing Plant 4 2. Methodology This section explains the methodology used for the hand calculations. A conservation of mass for the system is completed first. Given the arrival time of the tomatoes and the assumed service rates of the locations the utilization rate, effective arrival time, throughput time, and entities in the system are calculated. This information will help to guide the building of the simulation. The hand calculations will be used as a comparison to the outputs of the simulation. 2.1 Mass Balance When building simulation models the first step is model conceptualization. In conceptualizing a tomato processing plant a mass balance is necessary to know how entities are flowing through the plant. Although tomato processing plants change the state of tomatoes to paste there still must be a conservation of mass. This model was based on a plant that processes 1000 kg of tomatoes/day. This means 41.660 kg/hr are being discharged into the system. One can assume that the tomatoes being discharged were handpicked so we can expect ~2% waste between the receiving station and the sorting station. The tomatoes next go to the chopping station then from the chopping station to the hot break station. We can expect ~3% losses between the tomatoes going from the hot break station to the evaporator. The evaporator station will evaporate ~84% water at this station and the resulting paste will be sent to the aseptic filler. Figure 2.1 below shows a breakdown of the 5 mass balance described above. Figure 2.1: Mass Balance for Tomato Processing Plant 6 2.2 Deterministic vs. Stochastic Simulation Both deterministic and stochastic simulations have been used for the modeling of the processing plant. Deterministic simulation contains no input components which are random while stochastic simulation contains one or more input components which are random. Due to this fact stochastic simulation produce outputs that are random and may not be repeatable and deterministic simulation all future states are determined based on the input data. Stochastic systems involve probability distributions to generate random variables. A random variable is a function that assigns a value to every outcome of a random experiment. The random variables for our purposes are continuous. The continuous random variable X is defined by its probability density function f(x). The probability that the continuous random variable X takes on a value in the interval [a,b] is defined by equation 1 below. π π(π ≤ π₯ ≤ π) = ∫π π(π₯)ππ₯ [1] The exponential distribution is used for the arrival time of the tomatoes. The probability density function is defined by equation 2 below. π(π₯) = ππ −ππ₯ [2] The exponential distribution is commonly used for inter-arrival times because the preceding event does not affect the following event. It contains the “memoryless” property. The normal distribution is used for the processing times the probability density function is defined by equation 3 below. 1 1 π₯−π 2 ) π π(π₯) = π√2π π 2( [3] Normal distributions are typical used for time waiting to perform a task or when many different factors affect the outcome. The simulations scenarios ran will highlight further the advantages and disadvantages of both types of simulation as it pertains to tomato manufacturing plants. The distributions functions are built into Pro Model and generate random variables based on the specified distribution. distributions and Pro Model expressions. 7 Table 2.1 below shows the Table 2.1: Distribution Functions Distribution Exponential Normal Pro Model Expression E(mean) N(mean, std. dev.) 2.3 Hand Calculation and Queuing Model A hand calculation of a deterministic tomato processing plant with no waste is shown in table 2.2. This was run to understand how the expected throughput time and the total number of batches being processes in a 24 hour time period. The table shows that 1440 batches and a throughput time of 45 minutes. Table 2.2: Deterministic hand calculation Arrival quantity (batches) Inter-arrival time (min) Total time (hrs) Total arrivals (batches) Process times (min) Receive WashingSorting Chopping Hot Break Evaporation Total 10 10 24 1440 9 9 9 9 9 45 A queuing network consists of one or more locations which provide some service or make arriving entities wait for service at a queue when locations are busy. Queuing network theory is the basis for the calculations utilized to model the system. Queuing networks provide good estimates for the characteristics addressed in simulation. “Queuing theory is the science of waiting lines” (Harrell, 41). Understanding and correctly utilizing the queuing networks was key to correctly building the hand 8 calculation and computer models for the tomato factory. In general the tomatoes arrival and service rate are stochastic (as mentioned in section 1.3 stochastic arrivals involve random inputs and outputs). Therefore to appropriately model the system random numbers were used. Poisson arrivals, exponential distributions, and first come first serve (FCFS) service provides the proper parameters for the hand calculations. The distributions of interarrival times are exponential. FCFS service has been assumed for this type of system. The commonly used abbreviation for queuing systems has the form A/B/s where “A” is the type of inter-arrival distribution, “B” is the type of service time distribution, and “s” is the number of servers. The tomato processing plant uses M/M/1 queuing system for each workstation or process unit. The arrival rate is represented by (ο¬ο©ο¬ο the service rate is represented byο ο¨οο©. The utilization factor (ο²) equals the arrival rate divided by the service rate. π = π/π [4] Calculations were also performed based on Little’s Law. Little’s law states the expected number of entities in the system (L) is equal to the arrival rate (ο¬ο©ο times the throughput time (W). πΏ = ππ [5] where: π = 1/π(1 − π) [6] Open networks are systems consisting of multiple interconnected workstations typically having jobs moving between pairs of stations according to some routing scheme. The tomato processing plants consists of network workstations and was solved as an open network. Open networks admit jobs from the outside world, which are then routed along the network. The following properties of stochastic systems are applicable to open networks. 1. The sum of independent Poisson RV is Poisson 2. If rates are Poisson inter-arrival times are Exponential 3. Inter-departure time from an infinite capacity M/M/c system is exponential. 9 The inter-arrival time (π′) is defined by the arrival time multiplied by the probability of the entity transferring to the next station (p). π′ = ππ [7] The three steps procedure used for analyzing this open queuing network are: 1. Determine effective arrival rates 2. Analyze each station as if it were alone 3. Aggregate the results over the network. The expected throughput time is found by aggregating the results over the network. π = ππ ππ [8] where: ππ = π′ [9] π A spreadsheet was created with the solutions of these steps and can be seen in table 1 below. Using open network theory for this tomato processing plant the throughput time is 30.19 minutes. The calculations also give the utilization rate for each location below. 10 Table 2.3: Spreadsheet of Hand Calculations Sorting Chopping Hot Break Evaporating Packaging 66.67 66.67 66.67 66.67 66.67 66.67 1.00 0.98 0.02 1.00 0.97 0.03 0.16 0.84 1.00 60.00 58.80 58.80 57.04 9.13 9.13 0.90 0.88 0.88 0.86 0.14 0.14 1.00 0.98 0.98 0.95 0.15 0.15 0.15 0.13 0.13 0.10 0.02 0.02 9.00 7.47 7.47 5.92 0.16 0.16 Receiving Arrival Quatinty (batches) Interarrival time(min) Service times (min) 10.00 10.00 9.00 Arrival Rate (batches/hr) Service Rate (batches/hr) Probability Scrap Eff. Arrival Rate Utilization Rate Aggregate Throughput Rate Expected Throughput Time (min) 60.00 *Please note 10 batches contains 600 tomatoes 11 30.19 3. Discrete Event Model 3.1 Assumptions Several assumptions were made in the development of the simulation. The entities arriving to the system are assumed to be in batches. The arrival rate is 10 batches per minute. 1 batch contains 60 tomatoes, which means that there are 600 tomatoes/min. In reality, there are multiple sorting stations and people monitoring and adjusting the conveyor speeds for entrance into the plant. For this simulation there was only one sorting station and the conveyors were given a speed of zero. These assumptions are considered good assumptions because instead of a conveyor with speed queues were utilized where you could hold the tomatoes until the next location is open. The service times of the machines were assumed based on the desired per hour production output. In reality there is more variability with the service times of the machines and the assumed service times would be validated by data. Tomato processing plants use water and steam as resources for processing. These resources were not added to this model to simplify the model. There is an added location called scrap where all the waste is routed within the system. In reality there are flow lines that will route the scrap out of the system. Pro Model version 8 was used to model the system. 3.2 Locations Figure 3.1 below shows the layout with all the locations. The queues, the receiving, and the scrap locations have infinite capacity. The washing sorting, chopping, hot break, evaporation, and packaging locations have capacities of 10 batches each. 12 Figure 3.1: Locations 3.3 Entities and Arrivals The only entities in the system are tomatoes. The tomatoes arrival rate for the deterministic runs is 10 batches/min. The inter-arrival time for the deterministic runs is 10 minutes and for the stochastic run is an exponential distribution with a mean of 10 minutes. Each batch contains 60 tomatoes. 3.4 Processing and Routing The processing times for the deterministic runs are 9 minutes for every location except for the queues, scrap, and packaging. The queues do not have a processing time and entities are routed to exit from the scrap and packaging location. For the stochastic runs with random service times, the processing times for every location is a normal distribution with a mean of 9 minutes and a standard deviation of 1. The routing is arranged for a serial flow through the system with a queue between every part of contiguous locations. It is important to note that the routings account for the scrap during processing through the probability command. From the receiving station to the washing/sorting station, 100% of the entities were successfully processed. From the washing/sorting to chopping station, 98% were successfully routed through the station 13 while the remaining 2% sent to scrap. From the chopping station the hot break station, 100% of the tomatoes were successfully routed through the system. From the hot break to evaporation station 97% of the entities entering were routed while 3% were sent to scrap. From evaporation to packaging station 16% are routed to be packaged while 84% were sent to scrap. The 84% consists of all the water that is removed for the paste to be made. Figure 3.2 shows the routing scheme in Pro Model. Figure 3.2: Processing and Routing 14 4. Results This section discusses the results obtained from the five simulations consisting of two deterministic runs and three stochastic runs. For the deterministic runs the first one had no scrap (deterministic0) and the other run had scrap (deterministic1). The first stochastic run uses an exponential distribution for the arrival rate and no variability in the processing times (stochastic1). The second stochastic run used an exponential distribution for the arrival time and a normal distribution for the processing time of all the locations except the queues (stochastic). The third stochastic run provided the average of 10 replicated runs (stochastic10). The data for the locations utilization and time plot tracking the arrival of the tomatoes will be discussed, as well as charts with statistics. The section also includes recommendations on improvements in the simulation of tomato processing plants. The discussion ends with a comparison of the simulated resulted with the hand calculations. All of the simulation runs were done based on a twenty four hour day. 4.1 Deterministic Runs The scrap location in the simulation depicts where the waste is sent. The first deterministic run (deterministic0) was simulated without any waste, knowing that this is not how the plant functions in reality, to verify that the simulation was working properly. The second deterministic run (deterministic1) has the correct waste allocations. Table 4.1 and 4.2 shows the location statistics for both runs. The results show that the simulation is behaving as expected. In table 4.1 the total entries for scrap are 0 batches, but in table 4.2 the total entries are 1195 batches. The total entries for all the other locations and the percent utilization of the locations are affected between the two runs because of changing the tomatoes being sent to scrap. Figure 4.1 and 4.2 show the same effects of nothing being sent to scrap through the fluctuations in the percentage utilization at each location for the deterministic0 and deterministic1 runs. The main difference seen is that the chopping, hot break, and evaporation locations are occupied approximately 2% and full 87% of the time in the deterministic0 run. This is contrasted with being occupied 19% and full 70% of the time for chopping and hot break locations, and occupied 36% and full 52% of the time at the evaporation location in the 15 determinstic1 run. This means that in the deterministic0 run more tomatoes are being processed at these locations than in the deterministic1 run. This is consistent with expectations as these are the locations with waste. Figure 4.3 shows a graph of the time plot at the receiving location which verifies that the model is deterministic since there is no change in the amount of time. The graph looks the same for both runs so only one graph is show. Table 4.1: Location Statistics for model deterministic0 Location Receive Washing Sorting Scrap Chopping Hot Break Scheduled Total Average Time Average Maximum Current % Time (Hr) Capacity Entries Per Entry (Min) Contents Contents Contents Utilization 24.00 999999.00 1450.00 8.94 9.00 10.00 10.00 0.00 Evaporation Packaging Queue1 Queue2 Queue3 Queue4 Queue5 24.00 24.00 24.00 24.00 10.00 10.00 10.00 10.00 1440.00 0.00 1430.00 1420.00 8.99 0.00 8.95 8.95 8.99 0.00 8.89 8.82 10.00 0.00 10.00 10.00 10.00 0.00 10.00 10.00 89.94 0.00 88.86 88.24 24.00 24.00 24.00 24.00 24.00 24.00 24.00 10.00 10.00 999999.00 999999.00 999999.00 999999.00 999999.00 1410.00 1400.00 1440.00 1430.00 1420.00 1410.00 1400.00 8.95 0.00 0.00 0.42 0.81 1.22 1.15 8.76 0.00 0.00 0.42 0.80 1.19 1.12 10.00 1.00 10.00 10.00 10.00 10.00 10.00 10.00 0.00 0.00 0.00 0.00 0.00 0.00 87.60 0.00 0.00 1.13 1.13 1.13 1.12 Table 4.2: Location Statistics for model deterministic1 Location Receive Washing Sorting Scrap Chopping Hot Break Evaporation Packaging Queue1 Queue2 Queue3 Queue4 Queue5 Scheduled Time (Hr) 24.00 24.00 24.00 24.00 24.00 24.00 24.00 24.00 24.00 24.00 24.00 24.00 Capacity 999999.00 10.00 10.00 10.00 10.00 10.00 10.00 999999.00 999999.00 999999.00 999999.00 999999.00 Average Time Total Per Entry Average Maximum Current % Entries (Min) Contents Contents Contents Utilization 1450.00 8.94 9.00 10.00 10.00 0.00 1440.00 8.99 8.99 10.00 10.00 89.91 1195.00 0.00 0.00 1.00 0.00 0.00 1399.00 8.95 8.70 10.00 9.00 86.97 1390.00 8.95 8.64 10.00 10.00 86.37 1341.00 8.94 8.33 10.00 10.00 83.29 206.00 0.00 0.00 1.00 0.00 0.00 1440.00 0.00 0.00 10.00 0.00 0.00 1399.00 0.42 0.41 10.00 0.00 1.10 1390.00 0.81 0.79 10.00 0.00 1.11 1341.00 1.22 1.14 10.00 0.00 1.08 206.00 1.15 0.16 5.00 0.00 0.17 16 Figure 4.1: Location Utilization for deterministic0 Figure 4.2: Location Utilization for deterministic1 17 Figure 4.3: Time Plot for Deterministic runs 4.2 Stochastic Runs Stochasticity simulated the randomness that many manufacturing plants experience. The first stochastic run (stochastic1) uses an exponential distribution with mean of 10 minutes for an inter-arrival time and the same deterministic processing times as the two deterministic runs. The second stochastic run (stochastic) uses the same distribution for the inter-arrival time and normal distributions for the processing times. The location statistics for both of the scenarios are in table 4.3 and 4.4 respectively. Adding randomness to the inter-arrival time reduced the total batches of tomatoes at the receiving location to 1290 from the 1450 seen in table 4.2 for the deterministic1 run. However, table 4.4 for the stochastic run has the most entries to the receiving location of 1630 batches. This difference is due to the exponential distribution. The tables also show an increase in the percent utilization for the stochastic vs. stochastic1 run. The average time per entry is about the same for both runs except for at queue1. It is 10.74 minutes for the stochastic1 run vs. 60.53 minutes for the stochastic run. The difference in the amount of batches the plant is seeing accounts for the change to the average time per entry for queue1. Figure 4.4 and 4.5 shows a breakdown of what is happening at each 18 location in terms of percent of the time empty, occupied, or full for the stochastic1 and stochastic run respectively. The washing-sorting, chopping, hot break and evaporation locations are full approximately 30% more in the stochastic run vs. stochastic1 run. This can be attributed to the fact that the stochastic run has more batches in the system. Hourly time plots for the receiving locations for the stochastic1 and stochastic runs are shown in figure 4.6 and 4.7 respectively. The graph shows there are more batches being process for the stochastic run because the range slightly greater than 18, while for the stochastic1 run it is 12. Unlike the deterministic run in figure 4.3 these figures are not a constant value. Table 4.3: Location Statistics for stochastic1 Scheduled Locations Time (Hr) Receive 24.00 Washing Sorting 24.00 Scrap 24.00 Chopping 24.00 Hot Break 24.00 Evaporation 24.00 Packaging 24.00 Queue1 24.00 Queue2 24.00 Queue3 24.00 Queue4 24.00 Queue5 24.00 Total Capacity Entries 999999.00 1290.00 10.00 1290.00 10.00 1100.00 10.00 1249.00 10.00 1249.00 10.00 1208.00 10.00 168.00 999999.00 1290.00 999999.00 1249.00 999999.00 1249.00 999999.00 1208.00 999999.00 170.00 Average Time Per Entry Average Maximum Current % (Min) Contents Contents Contents Utilization 9.00 8.06 50.00 0.00 0.00 8.99 8.05 10.00 10.00 80.51 0.00 0.00 1.00 0.00 0.00 9.00 7.81 10.00 0.00 78.06 8.95 7.76 10.00 10.00 77.62 9.00 7.55 10.00 0.00 75.50 0.00 0.00 1.00 0.00 0.00 10.74 9.62 40.00 0.00 0.00 0.42 0.37 10.00 0.00 0.98 0.81 0.71 10.00 0.00 1.00 1.22 1.02 10.00 0.00 0.97 1.14 0.13 5.00 2.00 0.14 Table 4.4: Location Statistics for stochastic Scheduled Locations Time (Hr) Receive 24.00 Washing Sorting 24.00 Scrap 24.00 Chopping 24.00 Hot Break 24.00 Evaporation 24.00 Packaging 24.00 Queue1 24.00 Queue2 24.00 Queue3 24.00 Queue4 24.00 Queue5 24.00 Total Capacity Entries 999999.00 1630.00 10.00 1572.00 10.00 1291.00 10.00 1534.00 10.00 1521.00 10.00 1470.00 10.00 235.00 999999.00 1620.00 999999.00 1534.00 999999.00 1524.00 999999.00 1474.00 999999.00 235.00 Average Time Per Entry Average Maximum Current % (Min) Contents Contents Contents Utilization 8.94 10.12 60.00 10.00 0.00 8.93 9.75 10.00 10.00 97.46 0.00 0.00 1.00 0.00 0.00 8.96 9.54 10.00 10.00 95.40 8.97 9.47 10.00 9.00 94.70 9.04 9.23 10.00 10.00 92.26 0.00 0.00 1.00 0.00 0.00 60.53 68.09 147.00 48.00 0.01 1.57 1.68 8.00 0.00 4.49 1.99 2.11 8.00 3.00 2.98 1.84 1.88 7.00 4.00 1.78 1.15 0.19 3.00 0.00 0.19 19 Figure 4.4: Location Utilization for Stochastic1 run Figure 4.5: Location Utilization for Stochastic run 20 Figure 4.6: Time Plot for Stochastic1 run Figure 4.7: Time Plot for Stochastic run The final stochastic run (stochastic10) is the average results of 10 runs with the same distribution functions as the stochastic run. In reality the processing times and interarrival times will be random so this is very useful data since every stochastic run produces different random numbers. Table 4.5 shows the location statistics which reports the total entries to the receiving location at 1485 batches. This is lower than the stochastic run but higher than the stochastic1 run. The 1485 batches are slightly greater 21 than the average of the prior runs. This number would probably be closer to actual figures. The percent utilization is down from the stochastic run also. The same trend continues when looking at the location utilization chart figure 4.8. The numbers are in the middle of the previous two runs as far as percent full and occupied. The hourly time plot figure 4.9 shows the randomness at the receiving location and is different than both the prior runs. Table 4.6 shows the total number of batches that exited the plant in the 24 hour period that it was run and the throughput time. The two deterministic runs differ by 1 batch, while with the stochastic runs there is greater variability. The throughput times for the deterministic and stochastic runs are between 46-49 minutes. This shows that even with the randomness the throughput times are generally consistent. Introducing stochasticity is closer to the reality of how a real plant would function. The different simulated scenarios and statistics provide valuable insight on how to modify a plant for optimum output. Table 4.5: Location Statistics for stochastic10 Scheduled Name Time (Hr) Receive 24.00 Washing Sorting 24.00 Scrap 24.00 Chopping 24.00 Hot Break 24.00 Evaporation 24.00 Packaging 24.00 Queue1 24.00 Queue2 24.00 Queue3 24.00 Queue4 24.00 Queue5 24.00 Total Capacity Entries 999999.00 1485.00 10.00 1441.10 10.00 1182.70 10.00 1403.60 10.00 1393.20 10.00 1340.00 10.00 216.20 999999.00 1478.00 999999.00 1404.20 999999.00 1394.90 999999.00 1342.00 999999.00 216.30 Average Time Per Entry Average Maximum Current % (Min) Contents Contents Contents Utilization 8.96 9.24 51.90 7.00 0.00 8.95 8.96 10.00 9.80 89.58 0.00 0.00 1.00 0.00 0.00 8.97 8.75 10.00 8.70 87.46 8.97 8.68 10.00 9.90 86.84 8.97 8.35 10.00 9.40 83.46 0.00 0.00 1.00 0.00 0.00 34.87 36.89 109.80 36.90 0.00 1.34 1.33 7.10 0.60 3.55 1.66 1.62 7.20 1.70 2.29 1.59 1.49 6.80 2.00 1.41 1.15 0.17 3.20 0.10 0.17 22 Figure 4.8: Location Utilization for Stochastic10 run Figure 4.9: Time Plot for Stochastic10 run 23 Table 4.6: Throughput times Run Deterministic0 Deterministic1 Stochastic1 Stochastic Stochastic10 Average Time Total Exits In Operation (Min) 1400.00 1401.00 1268.00 1526.00 1398.90 48.61 46.69 46.64 48.30 47.68 4.3 Recommendations The simulation of a tomato processing plant can be further refined to provide the maximum benefit to the industry. Any tomato processing plant looking to use simulation should take statistics of their specific plant to minimize uncertainty from assumptions. Statistics should be taken for the arrival rate of tomatoes to the plant, the processing times at each location, and the inter-arrival times of the tomatoes. The statistics can then be fitted to the most accurate distribution function and whether deterministic or stochastic simulation should be done. The importance of this has been seen in section 4.1 and 4.2 as the results differ based on assumptions and distribution functions chosen. Information about the amount and flow of resources can also be added to the simulation to determine the effects on production. Resources needed for a tomato processing plant are water, steam, and power. 4.4 Simulation vs. Hand Calculations The deterministic and open network queuing theory hand calculations performed in section 2.3 provided some helpful insight but the Pro Model simulation would be the preferred method due to the complexity of this system. Pro Model simulation allows for good decision in the shortest time possible. The hand calculations predicted the utilization rate for each location but the Pro Model simulation showed the rate of utilization as well as how it was being utilized empty, full, or occupied. The built in distribution functions allowed for quick changes and evaluations on the effects which would not be possible with the hand calculations. The ability to model queues in the simulation is insight that the hand calculations would not be able to depict accurately. The throughput time and entities in the system for the deterministic simulations and the 24 deterministic hand calculations are very close. The hand calculations shows entities in the system at 1440 batches while the Pro Model runs show at the receiving location 1450 batches entered the system. The throughput time for the deterministic hand calculation was 45 minutes and for the Pro Model deterministic it was 48.61 and 46.69 minutes. The queuing theory had the throughput time at 30.19 minutes which was off from what Pro Model shows. The percent utilization for the washing-sorting, chopping, and hot break for the queuing theory is within 1% of the Pro Model numbers of the deterministic runs. The hand calculations are good in providing rough solutions but would fail to provide the accurate solutions needed for a dynamic system such as the tomato processing plant. 25 5. Conclusions Discrete event simulation is valuable and should be utilized in the tomato processing industry. Discrete event simulation accurately converted the activities performed in a tomato processing plant to time triggered events and consequent reactions which were chronologically processed. With the many uncertainties that arise from variability and interdependencies of a tomato processing plant, simulation can be used for production planning, scheduling, and guide decision to increase productivity. Discrete event simulation shows the effects on process parameters when modeling manufacturing systems. These parameters are static vs. dynamic, stochastic vs. deterministic, and discrete event vs. continuous. Tables 4.1-4.2 along with figures 4.1-4.3 showed the results for deterministic Pro Model runs and the effect of adding waste to the model. The effects of adding stochasticity can be seen in table 4.3-4.4 and figures 4.4-4.6. Table 4.5 and figures 4.7-4.9 show the benefits of replications on predicting results. Collecting data statistics about the processing plant and converting that into the proper distribution functions is important to accurately capture the system dynamics. Simulation is the preferred method rather than hand calculation for complex systems such as tomato processing plants. Although performing hand calculations may be initially cheaper the potential benefits of simulation can have more financial pay offs for companies than hand calculations. 26 6. References 1. Askin, Ronald and Charles Standridge. Modeling and Analysis of Manufacturing Systems. New York: John Wiley & Sons, Inc., 1993. Print 2. Harrell, Charles, Biman K. Ghosh, and Royce O. Bowden Jr. Simulation Using Promodel. New York: McGraw-Hill, 2004. Print 3. Gutierrez-Miravete, Ernesto. [Online] 18, January, 2001. [Cited: 5, November, 2013] http://www.ewp.rpi.edu/hartford/~ernesto/C_S2001/mams/notes/mams02.html 4. Gutierrez-Miravete, Ernesto. [Online] 26, September, 2002. [Cited: 23, October 2013] http://www.ewp.rpi.edu/hartford/~ernesto/C_F2002/DES/Notes/s04/s04.pdf 5. Fenco Food Machinery. [Online] [Cited: 25, April, 2013] http://www.fenco.it/eng/tomato-paste-processing.asp 6. Whitney, Daniel E. [Online] 20, November, 2002. [Cited: 16, December 2013] http://ocw.mit.edu/courses/mechanical-engineering/2-875-mechanical-assemblyand-its-role-in-product-development-fall-2004/lecture-notes/cls20_smltion04.pdf 27 ******************************************************************************** * * * * * * * * Formatted Listing of Model: C:\Users\udofaa\Downloads\deterministic0.MOD ******************************************************************************** Time Units: Distance Units: Minutes Feet ******************************************************************************** * Locations * ******************************************************************************** Name Cap Units Stats Rules Cost --------------- -------- ----- ----------- -------------- -----------Receive Washing_Sorting Scrap Chopping Infinite 10 10 10 1 1 1 1 Hot_Break 10 1 Time Series Oldest, FIFO, Evaporation Packaging Queue1 10 1 10 1 INFINITE 1 Time Series Oldest, FIFO, Time Series Oldest, FIFO, Time Series Oldest, FIFO, Queue2 Queue3 Queue4 Queue5 INFINITE INFINITE INFINITE INFINITE Time Time Time Time 1 1 1 1 Time Time Time Time Series Series Series Series Series Series Series Series Oldest, Oldest, Oldest, Oldest, FIFO, , , FIFO, Oldest, Oldest, Oldest, Oldest, FIFO, FIFO, FIFO, FIFO, ******************************************************************************** * Entities * ******************************************************************************** Name Speed (fpm) Stats Cost ---------- ------------ ----------- -----------tomatoes 150 Time Series ******************************************************************************** * Processing * ******************************************************************************** Process Routing Entity Location Operation -------- --------------- ------------------ Blk Output Destination Rule ---- -------- --------------- ---------- tomatoes tomatoes tomatoes tomatoes 1 1 1 tomatoes Queue1 FIRST 1 tomatoes Washing_Sorting FIRST 1 tomatoes EXIT FIRST 1 tomatoes tomatoes tomatoes tomatoes tomatoes tomatoes Queue3 Queue4 Scrap Queue5 Scrap EXIT FIRST 1 1.000000 1 0.000000 1.000000 1 0.000000 FIRST 1 Queue2 Scrap Hot_Break Chopping 1.000000 1 0.000000 FIRST 1 FIRST 1 Receive Queue1 Scrap Chopping Wait 9 Wait 9 tomatoes Hot_Break Wait 9 1 1 tomatoes Evaporation Wait 9 1 tomatoes Packaging 1 tomatoes Washing_Sorting Wait 9 1 tomatoes Queue3 tomatoes Queue2 1 1 tomatoes tomatoes tomatoes tomatoes tomatoes Queue4 tomatoes Queue5 1 1 tomatoes Evaporation tomatoes Packaging FIRST 1 FIRST 1 ******************************************************************************** * Arrivals * ******************************************************************************** Entity Location Qty Each First Time Occurrences Frequency Logic -------- -------- ---------- ---------- ----------- ---------- -----------tomatoes Receive 10 0 INF 10 min 28 Move Logic ------------ ******************************************************************************** * * * * * * * * Formatted Listing of Model: C:\Users\udofaa\Downloads\deterministic1.MOD ******************************************************************************** Time Units: Distance Units: Minutes Feet ******************************************************************************** * Locations * ******************************************************************************** Name Cap Units Stats Rules Cost --------------- -------- ----- ----------- -------------- -----------Receive Washing_Sorting Scrap Chopping Infinite 10 10 10 1 1 1 1 Time Time Time Time Hot_Break 10 1 Time Series Oldest, FIFO, Evaporation Packaging Queue1 10 1 10 1 INFINITE 1 Time Series Oldest, FIFO, Time Series Oldest, FIFO, Time Series Oldest, FIFO, Queue2 Queue3 Queue4 Queue5 INFINITE INFINITE INFINITE INFINITE Time Time Time Time 1 1 1 1 Series Series Series Series Series Series Series Series Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, FIFO, , , FIFO, FIFO, FIFO, FIFO, FIFO, ******************************************************************************** * Entities * ******************************************************************************** Name Speed (fpm) Stats Cost ---------- ------------ ----------- -----------tomatoes 150 Time Series ******************************************************************************** * Processing * ******************************************************************************** Process Routing Entity Location Operation -------- --------------- ------------------ Blk Output Destination Rule ---- -------- --------------- ---------- tomatoes tomatoes tomatoes tomatoes 1 1 1 tomatoes Queue1 FIRST 1 tomatoes Washing_Sorting FIRST 1 tomatoes EXIT FIRST 1 tomatoes tomatoes tomatoes tomatoes tomatoes tomatoes Queue3 Queue4 Scrap Queue5 Scrap EXIT FIRST 1 0.970000 1 0.030000 0.160000 1 0.840000 FIRST 1 Queue2 Scrap Hot_Break Chopping 0.980000 1 0.020000 FIRST 1 FIRST 1 Receive Queue1 Scrap Chopping Wait 9 Wait 9 tomatoes Hot_Break Wait 9 1 1 tomatoes Evaporation Wait 9 1 tomatoes Packaging 1 tomatoes Washing_Sorting Wait 9 1 tomatoes Queue3 tomatoes Queue2 1 1 tomatoes tomatoes tomatoes tomatoes tomatoes Queue4 tomatoes Queue5 1 1 tomatoes Evaporation tomatoes Packaging FIRST 1 FIRST 1 ******************************************************************************** * Arrivals * ******************************************************************************** Entity Location Qty Each First Time Occurrences Frequency Logic -------- -------- ---------- ---------- ----------- ---------- -----------tomatoes Receive 10 0 INF 10 min 29 Move Logic ------------ ******************************************************************************** * * * * * * * * Formatted Listing of Model: C:\Users\udofaa\Downloads\stochastic1.MOD ******************************************************************************** Time Units: Distance Units: Minutes Feet ******************************************************************************** * Locations * ******************************************************************************** Name Cap Units Stats Rules Cost --------------- -------- ----- ----------- -------------- -----------Receive Washing_Sorting Scrap Chopping Infinite 10 10 10 1 1 1 1 Hot_Break 10 1 Time Series Oldest, FIFO, Evaporation Packaging Queue1 10 1 10 1 INFINITE 1 Time Series Oldest, FIFO, Time Series Oldest, FIFO, Time Series Oldest, FIFO, Queue2 Queue3 Queue4 Queue5 INFINITE INFINITE INFINITE INFINITE Time Time Time Time 1 1 1 1 Time Time Time Time Series Series Series Series Series Series Series Series Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, FIFO, , , FIFO, FIFO, FIFO, FIFO, FIFO, ******************************************************************************** * Entities * ******************************************************************************** Name Speed (fpm) Stats Cost ---------- ------------ ----------- -----------tomatoes 150 Time Series ******************************************************************************** * Processing * ******************************************************************************** Process Routing Entity Location Operation -------- --------------- ------------------ Blk Output Destination Rule ---- -------- --------------- ---------- tomatoes tomatoes tomatoes tomatoes 1 1 1 tomatoes Queue1 FIRST 1 tomatoes Washing_Sorting FIRST 1 tomatoes EXIT FIRST 1 tomatoes tomatoes tomatoes tomatoes tomatoes tomatoes Queue3 Queue4 Scrap Queue5 Scrap EXIT FIRST 1 0.970000 1 0.030000 0.160000 1 0.840000 FIRST 1 Queue2 Scrap Hot_Break Chopping 0.980000 1 0.020000 FIRST 1 FIRST 1 Receive Queue1 Scrap Chopping Wait 9 Wait 9 tomatoes Hot_Break Wait 9 1 1 tomatoes Evaporation Wait 9 1 tomatoes Packaging 1 tomatoes Washing_Sorting Wait 9 1 tomatoes Queue3 tomatoes Queue2 1 1 tomatoes tomatoes tomatoes tomatoes tomatoes Queue4 tomatoes Queue5 1 1 tomatoes Evaporation tomatoes Packaging FIRST 1 FIRST 1 ******************************************************************************** * Arrivals * ******************************************************************************** Entity Location Qty Each First Time Occurrences Frequency Logic -------- -------- ---------- ---------- ----------- ---------- -----------tomatoes Receive 10 0 INF E(10) min 30 Move Logic ------------ ******************************************************************************** * * * * * * * * Formatted Listing of Model: C:\Users\udofaa\Downloads\stochastic.MOD ******************************************************************************** Time Units: Distance Units: Minutes Feet ******************************************************************************** * Locations * ******************************************************************************** Name Cap Units Stats Rules Cost --------------- -------- ----- ----------- -------------- -----------Receive Washing_Sorting Scrap Chopping Infinite 10 10 10 1 1 1 1 Time Time Time Time Hot_Break 10 1 Time Series Oldest, FIFO, Evaporation Packaging Queue1 10 1 10 1 INFINITE 1 Time Series Oldest, FIFO, Time Series Oldest, FIFO, Time Series Oldest, FIFO, Queue2 Queue3 Queue4 Queue5 INFINITE INFINITE INFINITE INFINITE Time Time Time Time 1 1 1 1 Series Series Series Series Series Series Series Series Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, FIFO, , , FIFO, FIFO, FIFO, FIFO, FIFO, ******************************************************************************** * Entities * ******************************************************************************** Name Speed (fpm) Stats Cost ---------- ------------ ----------- -----------tomatoes 150 Time Series ******************************************************************************** * Processing * ******************************************************************************** Process Routing Entity Location Operation -------- --------------- ------------------ Blk Output Destination Rule ---- -------- --------------- ---------- tomatoes tomatoes tomatoes tomatoes 1 1 1 tomatoes Queue1 FIRST 1 tomatoes Washing_Sorting FIRST 1 tomatoes EXIT FIRST 1 tomatoes tomatoes tomatoes tomatoes tomatoes tomatoes Queue3 Queue4 Scrap Queue5 Scrap EXIT FIRST 1 0.970000 1 0.030000 0.160000 1 0.840000 FIRST 1 Queue2 Scrap Hot_Break Chopping 0.980000 1 0.020000 FIRST 1 FIRST 1 Receive Queue1 Scrap Chopping Wait N(9,1) Wait N(9, 1) tomatoes Hot_Break Wait N(9,1) 1 1 tomatoes Evaporation Wait N(9,1) 1 tomatoes Packaging 1 tomatoes Washing_Sorting Wait N(9,1) 1 tomatoes Queue3 tomatoes Queue2 1 1 tomatoes tomatoes tomatoes tomatoes tomatoes Queue4 tomatoes Queue5 1 1 tomatoes Evaporation tomatoes Packaging FIRST 1 FIRST 1 ******************************************************************************** * Arrivals * ******************************************************************************** Entity Location Qty Each First Time Occurrences Frequency Logic -------- -------- ---------- ---------- ----------- ---------- -----------tomatoes Receive 10 0 INF E(10) min 31 Move Logic ------------ ******************************************************************************** * * * * * * * * Formatted Listing of Model: C:\Users\udofaa\Downloads\stochastic10.MOD ******************************************************************************** Time Units: Distance Units: Minutes Feet ******************************************************************************** * Locations * ******************************************************************************** Name Cap Units Stats Rules Cost --------------- -------- ----- ----------- -------------- -----------Receive Washing_Sorting Scrap Chopping Infinite 10 10 10 1 1 1 1 Time Time Time Time Hot_Break 10 1 Time Series Oldest, FIFO, Evaporation Packaging Queue1 10 1 10 1 INFINITE 1 Time Series Oldest, FIFO, Time Series Oldest, FIFO, Time Series Oldest, FIFO, Queue2 Queue3 Queue4 Queue5 INFINITE INFINITE INFINITE INFINITE Time Time Time Time 1 1 1 1 Series Series Series Series Series Series Series Series Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, Oldest, FIFO, , , FIFO, FIFO, FIFO, FIFO, FIFO, ******************************************************************************** * Entities * ******************************************************************************** Name Speed (fpm) Stats Cost ---------- ------------ ----------- -----------tomatoes 150 Time Series ******************************************************************************** * Processing * ******************************************************************************** Process Routing Entity Location Operation -------- --------------- ------------------ Blk Output Destination Rule ---- -------- --------------- ---------- tomatoes tomatoes tomatoes tomatoes 1 1 1 tomatoes Queue1 FIRST 1 tomatoes Washing_Sorting FIRST 1 tomatoes EXIT FIRST 1 tomatoes tomatoes tomatoes tomatoes tomatoes tomatoes Queue3 Queue4 Scrap Queue5 Scrap EXIT FIRST 1 0.970000 1 0.030000 0.160000 1 0.840000 FIRST 1 Queue2 Scrap Hot_Break Chopping 0.980000 1 0.020000 FIRST 1 FIRST 1 Receive Queue1 Scrap Chopping Wait N(9,1) Wait N(9, 1) tomatoes Hot_Break Wait N(9,1) 1 1 tomatoes Evaporation Wait N(9,1) 1 tomatoes Packaging 1 tomatoes Washing_Sorting Wait N(9,1) 1 tomatoes Queue3 tomatoes Queue2 1 1 tomatoes tomatoes tomatoes tomatoes tomatoes Queue4 tomatoes Queue5 1 1 tomatoes Evaporation tomatoes Packaging FIRST 1 FIRST 1 ******************************************************************************** * Arrivals * ******************************************************************************** Entity Location Qty Each First Time Occurrences Frequency Logic -------- -------- ---------- ---------- ----------- ---------- -----------tomatoes Receive 10 0 INF E(10) min 32 Move Logic ------------ 7. Appendix Below is the mass flow chart that the simulation was based on. It is from Fenco Food Machinery. 33