Simulating coolers 2014

50 %
50 %
Information about Simulating coolers 2014
Business

Published on April 21, 2014

Author: KhaledMabrouk

Source: slideshare.net

Description

Training seminar on simulating Produce Coolers

1 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 1 Simulating Produce Coolers (Operational Flow) April 11, 2014 Grower-Shipper Association Large Conference Room 512 Pajaro Street, Salinas, California Facilitated by: Khaled Mabrouk Operational Engineering Leader Sustainable PRODUCTIVITY Solutions kmabrouk@ReduceOR.com (831) 331-6259 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 2 Facilitator Background • B.S. Industrial Engineering - Purdue University • Experience: - Fortune 500 Corporations (Transportation, Mfg) - Consulting Industry - Software Industry - IE Program Instructor for WSU, EMU, & SJSU • Additional Activities Include: - Active in various professional organizations - Member of Six Sigma Conference Board - Published 20+ papers - Frequent presenter at professional conferences Consumable Goods Warehousing Specialty Coffee Transportation Postal Service Office Products Auto Appliance Call Centers High Tech HW Steel Dirt Racing Aerospace Forest Products Software Insurance Banking Placement Agency Waste Management Agriculture Build CPI Culture Simulation Modeling & Analysis

2 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 3 When Do We Use Analytical Tools? • Uncertainty of outcome • Amount of dollars at risk © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 4 Where Do Analytical Tools Fit? Your interpretation of reality Model used to answer "What ifs?" Reality Analytical Tools

3 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 5 Compare Decision Support Tools • Discrete Event Simulation • Spreadsheet • Slide rule/Calculator • Pencil & Paper • Intuition Effort Required Quality of Answer © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 6 Types of Models • Simulation models may be further classified as being static or dynamic, deterministic or random behavior, and discrete or continuous processes. – A static simulation model, sometimes called Monte Carlo simulation, represent a system at a particular point in time (single video frame) – Dynamic simulation models represent system as they change over time (the whole video)

4 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 7 Training Schedule 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 8

5 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 9 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 10 Types of Stochastic Models - Monte Carlo Simulation • Passage of time plays no substantive role. • “static” in general. • Used often in sales forecasting, financial analysis, and geometric tolerance studies. “Basically, a spreadsheet with random value cells”

6 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 11 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 12 The other line always moves faster ..… -Murphy

7 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 13 Queuing Theory - Queuing Model Basics Customers Arriving Serviced Customers Leaving Discouraged Customers Reneging Customers Balking Service Facility © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 14 Queuing Theory - Defining Customer Arrival • Probability (inter-arrival time). • Do customers arrive simultaneously (batch arrivals)? If so, probability distribution describing the size of the batch. • Reaction of customer entering system; are they patient or impatient (wait, balk, renege, jockeying). • Does arrival pattern change with time? Customers Arriving Serviced Customers Leaving Discouraged Customers Reneging Customers Balking Service Facility

8 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 15 Queuing Theory - Defining Service Patterns • Probability distribution to describe sequence of customer service times. • Is service single or batch? • Is service state dependent (i.e. does the length of the queue affect the service time)? • Is learning a factor? Customers Arriving Serviced Customers Leaving Discouraged Customers Reneging Customers Balking Service Facility © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 16 Queuing Theory - Queue Structure • How are customers selected for service (FCFS, LCFS, random, etc.)? • Is either priority or pre-emption a factor? • Is there a physical limit to the queue? • Single vs. multiple queues. Customers Arriving Serviced Customers Leaving Discouraged Customers Reneging Customers Balking Service Facility

9 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 17 Queuing Theory - # of Service Channels • Multi-server with 1 queue. • Multi-server with separate queues (operating independently). Service Facility Service Facility Service Facility Service Facility Service Facility Service Facility © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 18 Queuing Theory - Stage of Service • Single vs. multi-stage service. • Recycling/feedback may occur. Service Facility Service Facility Service Facility

10 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 19 Queuing Theory - Performance Measures • Some measure of waiting time that a typical customer might have to endure. • An indication of the manner in which customers may accumulate. • A measure of the idle time of the server. Customers Arriving Serviced Customers Leaving Discouraged Customers Reneging Customers Balking Service Facility © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 20 Queuing Theory - As An Analyst • Determine value of appropriate measure of effectiveness for a given policy. - Must relate waiting delays, queue lengths, to the input stream and the service procedure. • Design an “optimal” system. - Want to balance customer waiting time against idle time of servers according to some inherent cost structure. • Trade off between better customer service and expense of providing more service capability. Customers Arriving Serviced Customers Leaving Discouraged Customers Reneging Customers Balking Service Facility

11 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 21 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 22 Why Simulate? • Reduce risk • Increase number of options available • Quantify effect of operational changes • Avoid disrupting operations • Represent randomness • Unravel system complexity

12 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 23 Simulation Allows us to Imitate our Operation • A simulation is the imitation of the operation of a real- world process or system over time • It requires the: 1) generation of an artificial history of a system 2) the observation of that artificial history 3) to draw inference concerning the operating characteristics of the real system © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 24 Simulation Requires Assumptions • Simulation modeling starts with a set of assumptions concerning the operation of a system • These assumptions are expressed in mathematical, logical, and symbolic relationship between the entities, or objects of interest of the system • Once developed and validated, a model can be used to investigate a wide variety of what-if questions about the real-world system

13 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 25 Components of a Simulation System • An entity is an object of interest in the system • An attribute is a property of an entity • An activity represents a time period of specified length • In a produce cooler system, pallet of produce might be one of the entities, the cooling process type is an example of an attribute, and going through the cooling process is an activity © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 26 Simulation Maps Inputs to Outputs Cycle time TBF & TTR Arrival rate etc.. Simulation Model Throughput Bottlenecks Utilization etc.. Simulation is the art and science of building models to evaluate alternatives for the purpose of making decisions

14 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 27 Simulation Provide a More Accurate Answer! Simulation = Random Behavior + Dynamic Interaction Reflected Over Time © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 28 Simulation Project Steps 1) Specify an Objective 2) Understand System Logic 3) Develop Model Strategy 4) Organize Data for Model 5) Build & Debug Model 6) Verify & Validate Model 7) Experiment with Model 8) Analyze Results 9) Implement Solutions Steps in a Simulation Project © Sustainable PRODUCTIVITY Solutions, 2013 2.2 Build Code Modules 1.0 Design, Review, & Approve Functional Spec: 1) Map Process 2) Input Data Framework 3) Prioritized list What-Ifs? 4) Interface Requirements 5) Select Simulation Software 2.3 Debug Code Modules 2.4 Verify/ Validate Model 3.1 Collect Data 5.2 User Test Interface 4.1 Design Initial Set of Experiments 3.2 Fit Distributions to Data 6.0 Purchase SW directly from SW Vendor 2.1 Design Code Modules 7.0 Deliver Recommendations & Training 4.2 Run Experiments Buy Software? Interface Required? No No Yes Yes 4.3 Interpret Results 4.4 Design Next Set of Experiments 5.1 Design & Code Interface

15 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 29 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 30 Factors Impacting Whether We Should Simulate • Time span before decision makers need information. • Amount of risk in situation. • Level of predictability in new/modified process. • Ability to ensure realistic expectations of simulation. • Ability to secure effective simulation engineers.

16 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 31 Factors Impacting Level Of Detail • Time span before decision makers need information. • Scope of model (tend to have inverse relationship to level of detail). • Level of experience of simulation engineer. • How much we know about the process. • What we are willing to invest, to achieve the desired level of accuracy . © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 32 Simulation Project - Specify An Objective • Increase throughput • Reduce in process inventory • Increase worker/machine utilization • Improve on time delivery performance • Reduce capital requirements or operating expenses • Verify a proposed design

17 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 33 Simulation Project - Understand System Logic • Investigate how the system works • Determine the boundaries & restrictions to be used in defining the system • Determine the system configurations to be studied © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 34 Simulation Project - Develop Model strategy • Organize an "assumptions" document • Determine the level of detail • Select the measures of performance • Select the factors to be varied – Work-in-process (queue size) – Cycle time – New methods and processes – New equipment – Staffing – Output demand – New products

18 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 35 Simulation Project - Organize Data To Be Used By Model • Identify input data needed by the model • Collect information on system logic • Collect data for input distributions • Collect data for validation efforts © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 36

19 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 37 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 38 There are Two major Data Categories Numeric: • Process times • Time between failures • Time to repair • Percentage of time field is hit by a destructive insect/disease • Percentage of time weather is not favorable for growth Logic: • Procedures & Processes • Logic around decision points, or branch points in the process • Types of equipment breakdown • Planning logic/sequence • Prep activity required based on product we plant

20 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 39 Collect Causal Data • Most data collection mechanisms in place today are “Effect” oriented. • Proper data analysis will increase the payback from simulation. Gamma(shape 3, scale 1) 0.0 0.2 0.4 0.0 2.2 4.5 6.7 9.0 11.2 Expon (mean 1) 0.00 0.51 1.03 0 2 4 6 8 10 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 40 Sampling & Descriptive Statistics Basics • Sample Size: - 20 to 40 data points is sufficient for most processes • Collect data from different sample populations. For example, for a multi-site (or team) operation, you want to collect data from each of the sites/teams. Also, you will want data for various days of the week (for most situations). • Measures of Central Tendency: - Mean = Average - Mode = Value that occurs the most - Median = Value that is at the mid-point of all data points • Measures of Dispersion: - Range = Difference between highest and lowest value - Variance = Measure of how far each data point is from the mean - Standard Deviation = Measure of how far each data point is from the mean (square root of Variance)

21 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 41 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 42 Sources of Randomness • Almost all real systems contain one or more sources of randomness (input random variables) – Inter-arrival times of trucks – Processing times for loading/unloading – Number of arriving trucks – Number of pallets that need to be cooled – Equipment repair times – Processing requirements of different products – Failure rate of equipment

22 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 43 Methods of Representing Randomness in a Model • Use observed system data directly in simulation (avoid doing this) • Use observed system data to define an empirical distribution (do if you have to) • Fit a standard distribution to the observed system data (preferred option) © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 44 Review of Important Continuous Distributions Exponential Applications: Time between arrivals. Expon (mean 1) 0.00 0.51 1.03 0 2 4 6 8 10 Expon (mean 2) 0.00 0.51 1.03 0 2 4 6 8 10 Weibull Applications: Time to complete task, time between failures. Excellent for reliability since tail can veer to the left. Weibull (shape 2, scale 1) 0.0 1.3 2.5 0.0 0.5 1.0 1.6 2.1 2.6 Weibull (shape 6, scale 1) 0.0 1.3 2.5 0.0 0.5 1.0 1.6 2.1 2.6

23 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 45 Review of Important Continuous Distributions Normal Applications: Variables that are the sum of a large number of quantities. Excellent as a measure of error. Normal (mean 0, variance 1) 0.0 0.2 0.4 0.00 0.62 1.24 1.85 2.47 3.09-0.62-1.24-1.85-2.47-3.09 Normal (mean 1, variance 2) 0.0 0.2 0.4 0.0 1.4 2.9 4.3 5.7 7.2-1.4-2.9-4.3-5.7-7.2 Lognormal Applications: Time to perform task. Provides a spike close to x=0 that is often reflective of a tight process Lognorm (scale 1, shape 2) 0.0 0.3 0.6 0 6 12 18 24 30 Lognorm (scale 0, shape 1) 0.0 0.3 0.6 0 6 12 18 24 30 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 46 Review of Important Continuous Distributions Gamma Applications: Time to complete task. Used often for processing time. Gamma (shape 2, scale 1) 0.0 0.2 0.4 0.0 2.2 4.5 6.7 9.0 11.2 Gamma(shape 3, scale 1) 0.0 0.2 0.4 0.0 2.2 4.5 6.7 9.0 11.2 Triangular Applications: As a rough model, for processing times, where there is little data available.

24 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 47 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 48 Fitting a Distribution 1) Use a histogram and summary statistics to determine appropriate types of distribution 2) Estimate the parameters for these distributions from the observed data 3) Determine how well the distributions "fit" the observed data Can use Data Fitting Software, or Minitab

25 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 49 Fitting a Distribution – Analyze Data Collected Step One - Use a histogram and summary statistics to determine appropriate types of distributions • Collected 60 data points for service time with the following characteristics: We limit our analysis to the lognormal, normal, exponential, gamma, Poisson, and weibull distributions Minimum Value 23.3 Maximum Value 95 Mean 53.9 Mode 55.6 Varience 245 skewness 0.5 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 50 Fitting a Distribution – Compare Histograms (skinny cells) Step One

26 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 51 Fitting a Distribution – Compare Histograms (fat cells) Step One © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 52 Fitting a Distribution – Compare Histograms (just right) Step One

27 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 53 Fitting a Distribution – Conclusion from Histograms Step One • Observations – Data are skewed to right and has no negative values - Normal would not be appropriate – Data are continuous - Poisson would not be appropriate – Probability decreases as value goes to zero - Exponential would not be appropriate We can now hypothesize that it must be a gamma, lognormal, or weibull distribution © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 54 Fitting a Distribution – Estimate Distribution Parameters Step Two - Estimate parameters for these distributions. • For each distribution hypothesized, we must estimate its parameters from the observed data. • We used a curve fitting software to calculate the following parameters.

28 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 55 Fitting a Distribution – Estimate Distribution Parameters Step Two Gamma Distribution: Scale Parameter 11.94 Shape Parameter 4.52 Lognormal Distribution: Scale Parameter 3.95 Shape Parameter 0.3 Weibull Distribution: Scale Parameter 3.44 Shape Parameter 59.19 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 56 Fitting a Distribution – Select Distribution Step Three - Determine how well the distributions "fit" the observed data. • Graphical comparisons of the fitted distributions and the sample data • Goodness-of-Fit tests

29 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 57 Fitting a Distribution – Comparison to Gamma Step Three - Graphical Comparison Comparison of Input Distribution and Gamma(11.92,4.53) Values in 10^1 0.00 0.02 0.03 2.3 3.8 5.2 6.6 8.1 9.5 Input Gamma © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 58 Fitting a Distribution – Comparison to Lognormal Step Three - Graphical Comparison Comparison of Input Distribution and Lognorm2(3.95,0.30) Values in 10^1 0.00 0.02 0.03 2.3 3.8 5.2 6.6 8.1 9.5 Input Lognorm2

30 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 59 Fitting a Distribution – Comparison to Weibull Step Three - Graphical Comparison Comparison of Input Distribution and Weibull(3.44,59.22) Values in 10^1 0.00 0.02 0.03 2.3 3.8 5.2 6.6 8.1 9.5 Input Weibull © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 60 Fitting a Distribution – Statistical Tests Step Three - Goodness of Fit Tests • Chi-square Test – Measures how well the sample data fit a hypothesized probability density function. • Kolmogrov-Smirnov Test – Compares an empirical distribution function with the distribution of the hypothesized function. • Anderson-Darling Test – Detects discrepancies in the tails of distributions. • In general, the lower the test value, the better the fit

31 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 61 Fitting a Distribution – Statistical Test Results Step Three - Goodness of Fit Tests Hypothesized Distribution Chi-Square Test Kolmogrov- Smirnov Test Anderson- Darling Test Weibull (2.87,593.7) 8.99 0.099 0.791 Gamma (9.69,5.61) 6.65 0.067 0.266 Lognorm (3.94,.33) 6.76 0.069 0.331 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 62

32 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 63 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 64 Guidelines for Designing Model • Break process down into components – Use process map to drive your thinking – Use Pseudo-Code for larger models Components include key processing steps, path networks, and critical infrastructure requirements (for example selecting a storage location, creating an arrival pattern). • Scale Layout • Think, before you act: – Define Infrastructure – Worksheet w/ list of components • Build code in sections

33 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 65 Simulation Project - Build and Debug Model • Perform a "structured walk-through" of conceptual model – assumptions – level of detail – scope – review game plan • Construct & debug computer model © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 66

34 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 67 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 68 Simulation Project - Verify & Validate Model • Confirm that model operates as expected • Confirm that the conceptual model is an accurate representation of the real world • Have system "experts" review results for reasonableness • Use animation to find invalid model assumptions

35 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 69 Verification & Validation • Verification - Does the simulation model accurately represent the conceptual model? - Does it operate as intended? • Validation - Does the model accurately represent the real system for the purpose of the study? Conceptual Model Simulation Model Validation Verification Real System © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 70 Validating the Model - Basic Inspection Approach • Compare model results to system results under similar settings • Consider the model valid if – Key performance measures are close in value – The model and the system exhibit the same type of randomness

36 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 71 Validating the Model - Correlated Inspection approach • The model is "driven" with the same input data experienced by the real system • Consider the model valid if: – Key performance measures are close in value – The model and the real system exhibit the same type of randomness © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 72 Validating the Model - Turing Test • Organize system output data and model output data on standard format sheets • Ask system "experts" to determine which of these sheets reflects the real system and which of these sheets reflect the model • Use their explanation of how they were able to identify the real system output data from the simulation model output data to improve the model

37 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 73 Credibility ….. As you go! • The model and its results are accepted by management and line personnel as being valid • The model is used as an aid in making decisions • Develop credibility as you go! – Include representatives from management and line personnel in simulation team – Ask for input from team members – Keep team members involved and informed= © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 74

38 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 75 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 76 Simulation Project - Experiment with the Model • Execute original "what ifs?" • Utilize results to determine additional "what ifs?"

39 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 77 Output Analysis - Purpose • Estimating a simulation's true measure of performance • Determining whether different scenarios give statistically different results • Interpreting simulation results to answer questions of interest © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 78 Output Analysis - Requirements • Make 5 or more independent replications • Determine the warm up period - transient state • Estimate the mean of the performance measures • Construct a confidence interval for each performance measure

40 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 79 Transient & Steady State Behavior 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 Transient State Steady State © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 80 Output Analysis - Confidence Intervals • The mean will show the average of these values but actual output will probably be different from that value. • A confidence interval will give a range of values where the actual output will most likely be. • Mean = 951.9 • 95 % confidence Interval ( 935, 969 )

41 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 81 Output Analysis - Confidence Intervals This means that we are 95% confident that the actual value will be between 935 and 969 (in common sense terms) 1 2 3 4 5 6 7 8 9 10 900 920 940 960 980 1000 Total Throughput Upper Limit Mean Lower Limit © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 82

42 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 83 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 84 Case Studies Review • Why did they simulate? • What was involved in completing the project? • What were the results.

43 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 85 Produce Cooler Example WET Box Dry Box L o A D I N G D O C K S < 3B 233 10 Bays 2D, 3W, 3H 10 Bays, 3D, 3W, 3H 7 Bays 2D, 3W, 3H 11 Bays, 3D, 3W, 3H 3B 533 5 AFC 2x3 3B 733 1B 233 1B 533 1B 333 < 3 AFC 3x3 3 AFC 2x3 1B 121 1B 131 10 Bays 7D, 3W, 3H 5 Bays 7D, 3W, 3H 2 Bays 2D, 3W, 3H 3B 233 < 2B 33 3 Tube 04 Tube 07Tube 09 10 Bays 7D (deep), 5W (wide), 1H (height) 2 Bays 2D, 5W, 1H Ice Injctr Ice Injctr I/B Gate Ice Injctr © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 86 Produce Cooler Processes Queue Truck for Receiving I/B Truck QC checks product Forklifts unload pallets Receiving updates inventory Receiving verifies product, and print stickers Forklifts place pallets under shade bays Forklifts Move pallets to Vacuum Tube 01 or 02 Forklifts Move pallets to Ice Injection 01 or 02 Forklifts Move pallets to Wet Box doors Forklifts Move pallets to Dry Box doors Loaders Place Pallet in racks Forklifts Move pallets to Hydra-Vac Loaders move pallet to Forced Air Cooling Tunnel Pallets wait for cooling process to complete Forklifts Move pallets to Dry Box doors Accumulate appropriate number of pallets for cooling process Execute Cooling Process Accumulate appropriate number of pallets for cooling process Execute Cooling Process Accumulate appropriate number of pallets for cooling process Dry or Wet Cooling? Wet Cooling Type Dry Cooling Type Dry Wet AFC Vac Tube IceInjctr HydraVac

44 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 87 Arrival Rates Berry 1 13 8 Broccoli 2 12 5 Anise 1 12 2 Artichoke 1 13 2 Broccoli 2 14 2 Cauliflower 2 12 2 Celery 2 14 8 Lettuce 2 12 6 Mixed Lettuce 2 11 4 Romaine 2 12 7 Spinach 1 12 3 Company A Company B ProductOwner Pallets per Trailer Trucks per Day # Trailers per Truck © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 88 Cooling Approach By Product Hydra- Vac Vacuu m Tube Ice Inject Forced Air Cool Anise 100 Artichoke 100 Berry 100 Broccoli 35 65 Cauliflower 100 Celery 100 Lettuce 100 Mixed Lettuce 100 Romaine Heads 40 60 Spinach 75 25 Cooling Approach (by %) Product

45 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 89 Produce Cooler Challenges • Produce degrades quickly • Must be in cooler within 2.5 hours of harvesting, takes 30-60 minutes for product to arrive • Each product has it’s own cooling requirements; process and time in process • Different forklifts for inside vs. outside cooler; concern about emission affect on product © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 90 Produce Cooler – Simulation Challenges • Must keep track of every pallet • Arrival pattern varies by time of day • Should wait for sufficient number of pallets of a product type before starting cooling process • Use shade bay for staging • 6 pallet forklift

46 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 91 Produce Cooler - Sample Model • There are many excellent simulation software products. • The use of a specific product on this project is not an endorsement for any one specific simulation software. • Let’s RUN THE MODEL © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 92 Lessons Learned • Tube capacity should be a multiple of 15 • Extra receiving personnel needed during busy period to reduce in-bound trucks queuing • Room available in Shade Bay Area to convert to other use • Both Dry & Wet Boxes are sized appropriately • Single move for AFC products has significant impact on Average Time to Complete Cooling Process

47 © Sustainable PRODUCTIVITY Solutions - 2014 “Processes + People Drive Performance” ReduceOR.com Page 93 Rubber Band Effect 1) Monte Carlo Simulation 2) Queuing Theory 3) Discrete Event Simulation A. Developing Model Strategy B. Data Collection C. Representing Randomness in Model D. Fitting a Distribution E. Build & Debug Model F. Verify & Validate Model G. Experimentation with Model H. Simulation Case Studies 4) Key Learning © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 94 Benefits of Discrete Event Simulation • Avoid optimizing locally at no benefit to operations, or worse yet, to detriment of operations. • During simulation experience, found that most capital expenditure projects provided little benefit. DES allows you to identify these wasteful projects. • Improve understanding of system behavior “pattern recognition”. Developing conceptualization skills. Much of tweaking of operations has zero, or negative impact on performance. • Data requirements for BUILDING the model leads to useful insight into real problem(s).

48 © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 95 Simulation Project Steps 1) Specify an Objective 2) Understand System Logic 3) Develop Model Strategy 4) Organize Data for Model 5) Build & Debug Model 6) Verify & Validate Model 7) Experiment with Model 8) Analyze Results 9) Implement Solutions Steps in a Simulation Project © Sustainable PRODUCTIVITY Solutions, 2013 2.2 Build Code Modules 1.0 Design, Review, & Approve Functional Spec: 1) Map Process 2) Input Data Framework 3) Prioritized list What-Ifs? 4) Interface Requirements 5) Select Simulation Software 2.3 Debug Code Modules 2.4 Verify/ Validate Model 3.1 Collect Data 5.2 User Test Interface 4.1 Design Initial Set of Experiments 3.2 Fit Distributions to Data 6.0 Purchase SW directly from SW Vendor 2.1 Design Code Modules 7.0 Deliver Recommendations & Training 4.2 Run Experiments Buy Software? Interface Required? No No Yes Yes 4.3 Interpret Results 4.4 Design Next Set of Experiments 5.1 Design & Code Interface © Sustainable PRODUCTIVITY Solutions “Processes + People Drive Performance” ReduceOR.com Page 96

Add a comment

Related presentations

Boat chandlery

Boat chandlery

October 26, 2014

http://ecommerce-for-business.com/shackles-boats-theboatonlinestore/ We offer a ...

http://vendere-su-internet.com/shackles-for-boats-from-theboatonlinestore/ Moori...

Silver bar!

Silver bar!

October 21, 2014

Pretty similar to gold bars are these silver slabs. Silver is considered as the mo...

Gold coin prices!

Gold coin prices!

October 21, 2014

If you are an investor of gold bars and coins, one of the major things that you ou...

CyberSecurity's social media stats for one week as of Oct 21st 2014

CyberSecurity's social media stats for one week as of Oct 28th 2014

Related pages

Simulating Produce Coolers Operational Flow

“Simulating Produce Coolers Operational Flow” Description In this seminar, we teach how to use computer modeling to mimic Produce Cooler’s operation ...
Read more

Efficiently Simulating Thermoelectric Devices and Heat ...

Efficiently Simulating Thermoelectric Devices and Heat Pipes. ... April 10, 2014. ... (also known as thermoelectric coolers or Peltier devices) ...
Read more

Simulating the invisible - Phys.org - News and Articles on ...

Simulating the invisible July 28, 2014 by Poncie Rutsch ... all nanoparticles melt at cooler temperatures than they would normally, ...
Read more

Grower-Shipper Association

"Simulating Produce Coolers Operational Flow" Workshop: 2014-10-10 - 2014-10-10 - 8:30 a.m. - 12:00 noon: GSA Large Conference Room: Hosted by GSA and ...
Read more

Simulating a Passive Downdraft Evaporative Cool Tower

Simulating a Passive Downdraft Evaporative Cool Tower Liam Buckley1 1 IES Ltd, Boston, MA, USA Abstract This paper is focussed on the concept, design and ...
Read more

39 CPU coolers review: what is the best choice? | Hardware ...

A CPU cooler can be very cheap, very expensive and can even cost you up to 100 euros. For this article we tested 39 current CPU coolers and try to find out ...
Read more

POSSIBILITIES FOR SIMULATING THE SMOKE ROLLBACK EFFECT IN ...

... Volume LX (2014), ... POSSIBILITIES FOR SIMULATING THE SMOKE ROLLBACK ... and the cooler lower layer in a burning compartment.
Read more

McCreesh, N; Booth, M (2014) The effect of simulating ...

McCreesh, N; Booth, M (2014) The effect of simulating different in-termediate host snail species on the link between water temperature and ...
Read more

Rigorous Simulation of Heat Exchanger Networks: PRO/II - HTRI

Heat Exchanger Network Simulation – 2014 . 3 ... Simulating Heat Exchanger Networks ... (Air Coolers) PRO/II ® LINK HTRI ...
Read more

Condensation-Fouling Interaction in Low-Temperature EGR ...

Condensation-Fouling Interaction in Low-Temperature EGR-Coolers 4.2 Condensation rate Condensation of the water - sulfuric acid vapor occurs in
Read more