Week-by-week what we did

Mon 8/1. Explain course description, the Canvas homepage, demo an Arena model. Quantitative (counters, time delays) vs. qualitative metrics. What is lead time? Show a previous exam. Introduce the Banks paper. What is an event? What is discrete (time)? Simulation state (view it in Runtime Variables). Explain the numbers (NumberIn, NumberOut, WIP). What is WIP (Out-In, i.e. left inside). Discrete uniform distribution. Hand simulation of some events, draw n(t). Descriptive statistics on the collected samples (of waiting time).
Mon 15/1. 1) Beamon paper. Qual. vs. quant. metrics. Supply chain specific metrics (output performance) vs. decision variables (our input, that we can control). Stages/echelons. 2) Arena Users guide. Mortgage application case (that includes Create, Process, Decide (probabilistic) and Dispose. The common EXP(a) distribution, the TRI(a,b,c). Animation (entities moving, in queue). A Clerk resource. How to interpret the Output report. 3) The Little paper, wine cellar case.
Mon 22/1. Explained oblig 1. Expression editor (how to use). Did the Data analytics paper (machining), how they validated the model (the power usage plot). Arena Plots (how to). Revisited the Banks paper (the project flow figure). Started the "er" (Emergency patient, priority attribute (1 or 2) using DISC, priority queue.
Mon 29/1. Explained oblig 2 (search for an optimum). Runtime error due to "insufficient array space" (i.e. Arena runs out of memory). Continued "er" (into "er2", see Files). Added autosave. Stability problem since 4 Doctors doing 1 per hour, gave a utilization of 100 % (as we have 4 arrive per hour). Fix instability by either a) reduce demand or b) increase capacity or processing speed. We did the latter (reduce time per patient to 0.98, which gave utilization < 1. Showed that changing queue scheduling (FIFO, LIFO, PQ) does not affect the WIP, i.e. the n(t) curve in the Little paper. But it affects waiting time variations. Made some patients go back to revisit the Doctor (which adds to the resource demand, possibly making it unstable). Did an N-way probabilistic split/branch, checked split probabilities (using Out-counters). Added X-ray (Process requiring 1 Xraymachine and 1 Radiologist) and Test (Process requiring 1 Radiologist). Note that Radiologist is used several places. Distributions used/discussed: UNI(a,b), Tri(a,b,c), Norm(a,b), Constant, Exp(a). Discussed output related to resources, i.e. instantaneous vs scheduled utilization, nbr. busy. Tried to tune the 3 resources, discussed manual optimization (as we are doing) vs. automated (using software such as Optquest). Showed what a Schedule is (capacity schedule, arrival schedule). Discussed final exam, it will be on paper, so you have to explain your model by drawing the model (and indicate your settings), explain the model when necessary using words (story). Pseudo code is also useful.
Mon 5/2. First two hours: Did a model "on paper", using the blackboard. The "supply base" model of 2016 exam II is quite similar to the Assignment 3 (hospital). It has priority, priority queueing, revisiting (counter) and needs to remember service times. Third hour: Introduced Arena modules SEPARATE and BATCH, RECORD and READWRITE. These will be used in later exercises.
Mon 12/2. No lecture, but oblig 4 has a video about RECORDers.
Mon 19/2. Discuss oblig 4 (recording pretimes). Started on input analysis (Law paper). Looked at a HiMolde masters thesis that used Arena (by Dominik Schittenhelm, 2014).
Mon 26/2. Discuss exercise 5 (counting deadlines), identified the impossibility for P1. Look at syllabus: For programming, we have two kinds of data: System (defined by Arena for our use) or user-defined (we create/use them). Data is either variables (global, belongs to "everybody") or attributes (local, belongs to each entity). Datatypes (number, string (=text), boolean (=True/False). Data structures: Scalar (=single value), 1D array (need row index), 2D array (need row and column index). Truth table (for boolean logic) is that "A and B" is False except when both A and B are True; "A or B" is True except when both A and B are False. Arena computes expressions left to right, but will do innermost parantheses first. Then (for arithmetics), multiply/divide is done before plus/minus. HOUR 2: Functions, Arena has plenty of functions (EXPO, EP, LN, MAX,...) which have to return a value when used in expressions. We can (usually, but not in Arena) make user-defined functions for common tasks. Files are for reading and writing input and output that Arena does not offer, such as writing pretimes samples to "pretimes.txt", which is output. HOUR 3: Used pretimes.txt as input to Rstudio, showed a<-read.table(filename), names(a), pretimes <- a$V3, mean(pretimes), max(pretimes), gives same result as Arena RECORDers. Showed plot(a$V1,pretimes). Info supply chains, info "lead time". Finally, added an input part: Reading 9 inter-arrival times from "interarrivals.txt", and a Patient generator loop (instead of the usual CREATE). The resulting file is in "Files/hospital-readwrite-260218.doe" (which is dependent on the input file).
Mon 5/3. Discussed input analysis assignment. Used Excel filters to find distribution of priorities. Group work: Exam 2017, questions 3 and 1. Used NQ(Queue name) to find number in a particular queue. Used MOD(Entity.SerialNumber, N) to alternate among N machines in question 1c), since e.g. MOD(13,3) gives 1 as remainder.
Mon 12/3. Inventory. How to represent demand, inventory, backorders, lost sales, inventory review and replenishment etc (some notes). HOUR2: Made Arena model vInv (scalar, for current inventory), myDemand=quantity ordered (constant or DISC, POISSON for integer demands, UNIF/NORM/EXPO ... for decimal demands). HOUR 3: Added vIO and vBO tracking, backorder hold queue. See Files/invModel-120318.doe.
Mon 19/3. HOUR 1: Changing the periodic review to continuous (and looked at question 2 of 2017 exam, you should do this yourself before easter). HOUR 2: Made a lead time estimater based on exponential smoothing. After receiving replenishment: vLeadtimeEst = 0.6*vLeadtimeEst + 0.4*(TNOW - myStartWaitReplenishment). Used this in setting vReorderPoint to 15*24*vLeadtimeEst/24. HOUR 3: Multiproduct: Used arrays for vInv(), myDemand(), ... The first DECIDE (fulfill order?) is all or none (no partial fulfillment): If vInv(1)<=myDemand(1) && vInv(2) <= myDemand(2) ... (the && means logical AND). The second DECIDE (to replenish) uses a logical OR: We order replenishment if one (or more) of the products are below the reorder point.
26/3 and 2/4. No class (Easter).
Mon 9/4.
Mon 16/4
Mon 23/4.
Mo 14/5. FINAL EXAM (4 hour, 0900–1300). Written (pen and paper), no-book (closed book). This is not a PC/Digital exam.

And that is it.