Chapter 3 Branched Paths, Rate Tables, and Entity Dependencies: Airport Revisited mn an airport there is normally more than one type of check-in, as well as more than one type of passenger (Ge. a curbside check-in or an no check-in traveler). Our goal is to embellish the aiport model from Chapter 2, to handle routing to different check-in stations, varying arrival rates of passengers, and varying types of passengers with rate tables, and data tables. Part 3.1; Choice of Paths As in the previous chapter, our primary concern is with the amount time it takes the passengers to reach the security checking station in determining the number of check in stations necessary Step 1: Delete the Excess? object in the prior chapter model by right clicking on the Experiment in the [Navigation] Panel.! We will embellish the previous airport model to allow a choice of three Gifferent routes for passengers arriving at the airport. Passengers who need to check bags and get tickets can either check in at the curbside station or check in at the main station inside the airport while some can proceed directly to the security check point. Refer to Figure 3.1 which depicts the modified airline check-in process to be modeled. One Aine Statt 20% ‘Two Check in Stations Uniform(2,8) Minutes NHPP Rate: Table 125 yards Figure 3.1: Modified Airline Check-in Process Routel: Passengers who use the curbside check-in need to walk 10 yards to reach the station where a single airline staff member takes between one and five minutes, with a most likely time of two minutes, to check-in a passenger. Once they have checked in, they walk 125 yards to the security ‘check line Route2: Passengers who have checked in online and do not need to check any luggage, only need 10 walk the 125 yards to the security check line once they arrive to the airport. "5 The [Navigation Panel] in the upper right corner contins all ofthe models, objects, and experiments inthe project. You can access the properties, rename the item, or delet the item from the profes. rash Paths, Rite Tables, sn Entity Dpendesies Aor Revisited 19 ‘conga to: Gils Coier-Takhassussi Streot Riyadh 11533 - goorier@abssal. odLUconsed to: Gils Cormier -Takhassussi Street Riyadh 11533 - gcoumier@ate Route3: Passengers who plan to use the inside check-in first need to walk 50 yards to reach the station where there are currently two airline personnel (instead of four previously) checking, passengers into the airport. It takes between three and ten minutes uniformly to process each passenger. Once they have checked in, they then walk 65 yards to reach the security check line, Step 2: Modify the model from Chapter 2 to include the new curbside check-in (SesvEs) as well as the direct path (i.e., Route 2) from the arrivals to the security gate as seen in Figure 3.2. Specify the properties according to description and Figure 3.1. Routel igure 3.2: Creating Different Routing Paths Step 3: Ithas been observed that passengers follow the probabilities in Table 3.1 in determ route they will follow. ‘Table 3.1: Passenger Type Percentages Check-in Ty Percentage ‘Curbside Check-in 10% Tnside Check: 70%. ‘No Check-in (Carry on) 20% Step 4: To model the decision of a passenger to choose their check-in route the Selection Weight property of Path objects will be utilized. The probability of choosing a particular path is that path’s individual selection weight divided by the sum of all path selection weights from that node.16 Place the proper weights to each different check-in route (see Figure 3.3). "© SIMIO automaticaly normalizes the selection weights to ensure the probabil sald up to 1. Therefore one can specify 70% oF 7. 20 ‘Simulation Modeling with SIMIO: A WorkbookProperties: Path Path) || Show Commonly Used Properties Only [2 [reaveliagic Type Usieectonl Init Traveler Capac” Iniity EntyRavkingRe Fest InFestOut Drawn To Scale False Selection Weight Logie Length 50 sits Yards Alon Passing True (Speedin Infrty © Routing Logic Selecton Weight ” Figure 3.3: Weighting Paths Step 5: Remember to select the Outbound Link Rule to “By Link Weight” on the TxawsresNone. Output @SrePassengersArrive (see Figure 3.4).17 Properties: Output@srcarival (Traneferode) [[) Show Commorly Used Properties Only 5 Inial Traveler Capacity Infirity Entry Rankng Rule Fist in rst Out Link Rule Routing Logic OutboundtinkPreference Any Outbound ink Rule By Link Weight Entity Destnaton Type Contue Figure 3.4: Outbound Link Rule Step 6: Run the model for one, twenty-four hour replication (no “Experiment” needed). You might want to improve on our animation Question 1: What is the average number of passengers waiting at Curbside and the Inside Check-in? Question 2: What is the average wait time at both Check-ins? Question 3: What is the average time it takes for a passenger to check-in? Part 3.2: Changing Arrival Rate Passengers do not arrive at a constant rate (j.e., homogeneous process) throughout the entire day in a real airport (even though they may arrive randomly), so we should expand our model (© handle such changes. (Qur approach is to allow the arrival rates to change throughout the day. To model this changing rate, we will use a data object in SIMIO called a Rate Table and reference the Rate Table we create in the arrival logic of ‘our Source object. It is important to understand that the interarrival times for this rate will be assumed to be "Although “Shortest Pa isthe dealt Qtabound Link Rue, “By Link Weight” rule wil sil be used unless the entity destination has been assigned toa particular node. Therefore, we could have used the default rule Branched Pas, Rate Tbs, and Entity Dependencies: lpn Revised 21 yah 11583 - gconmior@ataical odsExponentially distributed, but the mean ofthis distribution changes with time. Ian interarival process has, an Exponential distribution, then itis known statistically thatthe numberof arrivals per unit time is Poisson distributed, Since the parameter (its mean) changes with time, it is formally referred to as a “non- homogeneous Poisson” arrival process, also called a NHPP (Non-Homogeneous Poisson Process). Therefore when you use a SIMIO Rate Table you are assuming the arrival process is a NHPP. Error! eference source not found. Table 3.1 shows the hourly rate over the 24 hour day where more passengers on average arrive during the morning and dinner time hours. ‘The zero rates imply no arrivals during that time period ‘Table 3.2: Hourly Arrival Rate Begin T End Time ‘Hourly Rate ‘Midnight (time 0) Tam (ime 1) o Lam (time 1) am (ime 2) oO 2am (time 2) Sam (time 3) 0 3am (time 3) damm (time 4) 0 ‘am (time 4) Sam (time 5) 0 ‘6am (ime 6) 30 Tam (time 7) 90 Bam (time 8) 100 ‘8am (time 8) Sam (time 9) 15 ‘Sam (time 9) Tam (time 10) 60 TOam (time 10) Tam (time 11) 60 Ham (time 11) Noon (time 12) 30 ‘Noon (time 12) Ipm (time 13) 30. Ipm (time 13) 2pm (time 14) 30 2pm (time 14) Spm (time 15) 60 3pm (time 15) 4pm (time 16) 60 4pm (time 16) ‘Spm (time 17) 75 Spm (time 17) ‘6pm (time 18) 100 ‘6pm (time 18) “pm (time 19) 50 “pin (time 19) ‘8pm (time 20) 30. 8pm (ime 20) ‘Spm (time 21) 0 ‘Spin (time 21) 10pm (time 22) 0 10pm (time 22) Lipm (time 23) 0 pm (ime 23) ‘Midnight (time 24) 0 Step 1: Arrivals are only present from Sam to 8pm for & total of 15 hours. A simulation run length of 24 hours would cause the time-based statistics like number in station, number in queue, and scheduled utilization to be computed when there were no arrivals, causing lower than expected values for these statistics. Therefore we will set the Starting Time within the “Run Setup” tab to 5:00 am and the Ending ‘Type to Specific Ending Time at 8:00 pm as shown in Figure 3.5 Starting Time:[ 1/24/2011 5:00:00 [+ Ending Type: Figure 3.5: Replication Start and Stop Time Step 2: To create a Rave Tans, select the “Data” tab and then click Create —>Rate Table, Name the lable PassengerArrivalRate, Set the fterval Size to 1 hour and the Number of Intervals to. 15, corresponding to the run length, While you can modify the size (e.g, half hour increments) and number 2 ‘Simulation Modeling with SIMIO: A Workbook Llconsed to: Gils Cormier. Takhascussi Street Riyadh 11583. geormior@atalca odyof time intervals during which the arrival rate is constant, you cannot modify the rate, which is fixed at events per hour. Step 3: Enter values in the rate table as shows [einen [eco [eae sera] bor, 0:00:09 Doy 4, 01:00:90 | [ay 0:00:20 doy, oz:00 oar 1 2:00:09 bay 1, 03:00:00 ber 3 03.00: Day 4, 0400:00, [oar 00:20 Dey 050000 Figure 3.8Figure 3.6. [Jos se: 0:00 Days, 07:00:00 [ony 7:00:29 Dey e800 | Jos, c:20:00 Days, 03:00:00 [Jos 08:00 Days, 10:00:00 |_ |pay 4, 10:00:00 Day 4, 11:00:00 | Joay 1:90:90 Days, 0000 bers, 12:00:09 Day 1, 13:0:00 [ony 1200290 bays, 00:00 | Jooys, 00:00 Day, 15:00:00 es8ausseessessus of igure 3.6: The Rate Table Step 4: Modify the SrePassengersArrive source accontingly (see Figure 3.7 for details) [Esto Corset Used Properties Only EntPassenger ‘Time Varying Arrival Rate PassengerArrivalliate RateScaleFecter 0 EntesPer Aral 1 © ‘Stopping Coneitions Figure 3.7: Specifying a ‘Time Varying Arrival Rate Step 5: Run the model again and observe the results. Question 4: What are the average Wait Times at each check-in location (Curbside and Inside)? What is the average check-in time for a passenger (Curbside and Inside)? Question 5 Step 6: If the length of the simulation (replication/run length) exceeds the length of the rate table, then the rate table simply repeats until the simulation replication is terminated. So if the run length should be longer and there should be no arrivals during that time, then the rate table should contain zeros accordingly. Branched Pots, Rale Tables, and Entity Dependencies: Aiport Revisited 23, LUcontedt: Giles Cormier -Takhasausei Street Riyadh 11530 - gcormier@alfaloaLedyPart 3. istinguishing Entities: Data Tables In an actual airport more than one type of passenger arrives. Some are continental travelers while others are intercontinental travelers, Still there are some passengers which are handicapped and require special autention o are families. The different types of passengers result in different processing times. To model these three cases the use of Data Tables is extremely helpful. There is a 33% chance for each type of passenger to arrive 1 our airport. Step 1: To create a Oxta ‘TABLE click “Tables” in the “Data” tab, then “Add Data Table.” Call it ‘TablePassenger. In our Data. Tans®, the processing times for each of the three types of passengers for each check-in location will be specified. Select the properties from the “Standard Property” option in the Properties section. PassengerPriority will be an Integer property whereas the check-in times will be Expression properties named CurbsideCheckInTime and InsideCheckInTime."" Priorities 1, 2, and 3 represent Continental, Inter-Continental, and Handicapped passengers respectively as in Figure 389 ; Random. Uforn(2,5) 2 2 Random-Tianguar(2,36) Random. form) 3 Random Tisngur(3,4,7) Random Uniform(4,6) Figure 38: The Data Table Step 2: Since the check-in times are in minutes, we want to specify the “Unit Type” for these properties as seen in Figure 3.9. Defautveue 0.0 Specity Unit Type Candidate Refer... False ‘Time (Minutes) Unt Type Time Defeutunts ‘Minutes ‘Appearance DévloyNene _nsdecheckanTime ‘operational Planning General pe InsidecheckaoTime Desertin ReqiredVaue Tue Figure 3.9: Inside Check-In Properties ° Note that if you fist specify the “Genera” name, it wil become the default forthe “DigplayName” propery. Also, by using proper ease, the table labels as seen will have spaces between the words ° Even though the check-in times are expression properties, the familiar drop down expression editor isnot available in Dac» ‘TAMLES, Therefore, the easiest way i to putin all thre priorities and type inte Fist value ofthe first ow. Then copy ito the other rows an just modifying the values ofthe parameters, Tt may seem obvious the Unit Type forthe time properties sbould be specified. However, one can leave the units as unspecified then the numbers wll ake on the units of the property When they are used (i, default value of hours). 24 Simulation Modeling with SIMIO: A Workbook Ucensed to: Gilles Cormier - Takhassuss Sret Riyadh 11533 - gconnior@alalsaLedyStep 3: We will directly reference these process times from the DxcA "Taste by using the Initial Priority property and the Priority state variable of the EntPassenger object. In order to obtain different ial priorities to represent each of the passenger types (i.e, 33% chance for each type), we must use the Discrete Distribution (entered in cumulative distribution form). In Figure 3.10, each passenger type is assumed to be equally likely Random. Discrete(1, 0.33, 2, 0.66, 3, 1.00) © Travel tog © Initial Desired Speed Random.Uniform(2,4) nts Hes per Hour Discrete isatnete bat Distribution NetweeTurarcund Method Ex Re-enter where each © Routing Logic passenger type is ntl Prenty Randombiscrete3}033;20.66,34.0 equally likely Seopa (one-third) © rnancals © Advanced options Desay tame CanTeanafer In ROu OF. Tre Figure 3.10: Passenger Propet Step 4: A “property” of an object is initialized at the time an object is created but cannot be changed. Furthermore, it shares its values will all other objects. A “state variable” is a characteristic of an object that can be changed. Also, state variables can be associated with individual objects and thus each object has its own (changeable) value. ‘The Initial Priority property is defined by SIMIO for all model entities, bbut is immediately assigned by SIMIO on creation to the Priority state variable (also defined by SIMIO) Question 6: — Where do you find the MopeLEwsr2y defi tion of the Priority state variable? Step 5: Now we must specify the processing times at the check-in stations to depend on the passenger type as seen in Figure 3.11. ‘The specifications needed for the new processing times use the Priority state variable as: ‘© ArSrvCurbSideCheekIn:”? TablePassenger (Passengers . Priority) .CurbSideCheckInTime © At SrvCheckin: TablePassenger [Passengers . Priority) .TnsideCheckInTime Notice that the content of the [123 are used to designate the specific row in the table, similar to arrays in programming languages. Unfortunately, the expression editor does not know about the Data Table © If you receive an enor when using Passengers-Priority you probably did not rename the Moost Br: (Navigation) panel and therefore, should use Nodelentity. Prioricy instead object in the ® To enter this specification s0 as to employ the expression editor during enty is 10 put in the tle name and property as ‘TablePassenger CurbsideCheckInTine andthen inset [Passengers.Priority). Chapter 7, we will demonstrate the ability o assign a paricular row of table 0 the entity 30 the row will not need (0 be cxplicily set with the (]. This will make things easier bur there ae instaees when you need to specify a row expicily. Brahe Pa ables and Enity Dependencies AispotRevished 25 LUcensed to: Giles Cormier -Takhassussi Sireet Riyadh 11533 -gcomniar@alaisal ecproperties, so you must type in the property into the expression. Also be sure that the processing time units in the check-in stations are specified in minutes. Ranking Rule FestinFrstout DyramcSelectonRue None Braterintne 0.0 [rrocessng Tine _TablePassenger(Passengers.Pority).nsideChecktaTime Figure 3.11: Utilizing the Table to Specify the Processing ‘mes Step 6: — Run the model, Remember no arrival occurs before Sam, so no animation appears until then. Question 7: Wat is the average wait times at each check-in location (Curbside and Inside)? Question 8: What is the average check-in time for a passenger? Part 3.4: More on Branching At times, the waiting lines at the inside Check-in station seem overcrowded. We decided to ald a second Inside Check-in station that passengers may use during rush hours. Thus, passengers will be routed to whichever Check-in station has the smallest number in queue. The distance from this second Check-it station tothe Security Check is 75 yards as seen in Figure 3.12. ne Airline Statt (4.2.5) Minutes E rr 126 yards 425 yards {in Process with Two Inside Cheek In Figure 3.12: Airline Cheet Step 1: Add the second check-in station, Call it SrvCheckIn2, Assume it has a capacity of one and it's processing times are the same as those for the regular check-in station. Add the path of 75 yards from this second station to SnkSecurity. Step 2: We need to model the choice made by “inside” check-in passengers to go to the check-in station with the smallest number in queue. There are several ways to create this model, Perhaps the most obvious way is to add a TravsreaNove. that branches passengers to the shortest queue. ‘The distance from the SrePassengersArrive to this RawsreRNone is 50 yards (remember that the selection weight is 70 for this path and you will have to delete the original path to the SrvCheckIn object) 26 Simulation Modeling with SIMIO: A Workbook LUponsed to: Gills Cormior-Takhassussi Street Riyadh 11583. gconmior@alfsicaLedu‘+ The TRaNSrERNoDE should specify the Outbound Link Rule as “By Link Weigh © Set the Selection Weight properties of the two paths from the ‘TRaNsrERNoDe (the “value” of @ “true” expression is Land 0 if itis “false”) (1) To SrvCheekIn: srvCheckIn. InputBuffer Contents .NumberWiait ing SrvCheckIn2. InputBuf fer. Contents .NumberWaiting (2) To SevCheckIn2: SrvCheckIn2, InputBuffer.Contents.NumberWaiting < SrvCheckIn. InputBuffer Contents .NumberWaiting Question 9 Which check-in station is chosen if there are “ties” in the size of the InputBufier (For example, suppose both are empty)? Question 10: — What is the utilization of SevCheckIn2 and does it make sense to have both check Part 3.5: Work Schedules Suppose you decide only to staff the second inside check-in station from 10am to 6pm. This kind of concern requires a work schedule which alters the capacity of a server over time, Work schedules will be built from the “Schedules” component in the “Data” tab. Step 1: Click on “Work Schedule” button under in the “Schedules” igon in the “Data” tab to add a new ‘work schedule. By default, the StandardWeek schedule has already been defined as seen in Figure 3.13. Rename this schedule SecondCheckInStation. Change the Days to “I” ~ this schedule will repeat daily 24 For each day of the work schedule, you need to specify the work pattern for that day. Therefore, you can have different patterns for different days (i.e, weekends, etc.). The StandardDay pattem was Specified for Day 1. The Start Date property can be left as the default since we have a daily pattern.25 ‘Work Schedules | Day Pa Rane surtoste [Sesion oye [cateony [oar | Your “default” date will be different Creating Second CheckInStation WorkScHEDUzE Step 2: In the “Day Patterns” tab, you can add Day ParzeRws by clicking the Day Pattern button in the ‘Create” section, SIMIO defines the StandardDay pattern of Sam (o Spm with an hour lunch at 12pm, Since we start the day at 10am and end the day at 6pm, we only need one work period in the pattern. Select the second work period (Ipm to Spm) and delete it, Then specify the Start Time to be 10am and tf here are seven days inthe pater, then the days ofthe week will appear, Inoue cae there is only the Day I pater. SIMO will repeat the paltem based on the number of days in the Wax Sceatuts, For each day inthe work schedule you need to specify a paltera, Also, we have the ability to specify exceptions to the day pattern or work peviods for particular dates. "The Start Dae property represents the particular day the ftt day ofthe pattern will tars F his date i nthe future compared (0 the start of the simulation, i yal repeat backwards tothe curent date based on the pattern, For example, if we have a three day paltern named Dayl, Day2, and Day and the work schedule start date is set to 0/17/13, schedule will follow the Day pattern for this dae, If the simulation stats on 09/13/13, then the patra for 09/13 is Day3, 08/14 is Day, 09/15 is Day2, 08/16 Day which ‘makes Dayl be on 09/17 and Day? on 09/18, ec, Brahe Pats, Rate Tables and Entity Dependence: Aton Revised 27 LUcensed 1: Gils Cormier - Takhassuse! Soot R¥yadh 11533 -geormlor@atalsal ecuFigure 3.14. Specify a | for the the Duration to be eight hours or the “End Time” to be 6pm as see Value which will specify the capacity of the second inside check to be 1. If work period has not been efined for a period in the day patter, itis assumed to be off shift (i.e, zero capacity). edie | bay Patterns Nome [Description | Category > i standsreoay Baton enn |v eo ar [eso hows com 1 Figure 3.14: Setting a Day Pattern with Different Work Periods Change the Capacity Type to Step 3: Go to the “Facility” view and click on the SrvCheckin? object. “WorkSchedule” and the Work Schedule property to SecondCheckInStation as in Figure 3.1 Properties: SryCheckin2 (Server) show Commonly Used Properties Only 5 [Process Logie! Capaaty Type WorkSchedule work sehede Secondcheckinstation Ranking Rule Fst In First Out Figure 3.15: Specifying a Work Schedule Step 4: Now we need to be sure that people who choose the inside check-in only go to the SrvCheekIn2 station if it is open (itis open only between hours 10 and 18). This requires another TxawseERNoDe as ‘gure 3.16, Connect the original TraNsFERNODE to the new one via a Connector and connect the "TRANSEERNODE to choose shortest Queue Input@SryCheckIn2 Figure 3.16: Adding another TransferNode to Handle the Logie ‘The new TRaNseERNODE should determine routing By Link Weight as well length and Selection Weight property new TRANSFERNODE {o the inputs of both check-in stations, From SrePassengersArrrive New TransFERNoDE 10 choose open SrvCheckIn, Path to Input@SrvCheckIn has zero TimeNow <=10 || TimeNow>=18% © Note, SIMIO utilizes || asthe Or logical operator and £5 38 the And logical operstor. Als, tis expression works oly hee ince If we were simulating more than one day, then we would need to specify we are simulating. just one day. Yath.Renainder (Timetiow, 24) *24 which converts the runing tne into a number between O and 24 28 Simulation Modeling with SIMIO: A Workbook: Uconeedt: Giles Cormier Taknassussi Street Ryash 11532 -goormior@atasaLedu* Path to TRansreRNoDE to choose shortest queue has zero length and Selection Weight property of TimeNow > 10 && TimeNow <18? Step 5: Note that when there is no context, the ut Now run the model and obtain these basie statistics. of time are hours (the units of distance is meters). Question 11: Whats the average Check-In time for a Passenger? Question 12: What is the average wait time for Curbside Check-In? Question 13: What are the average wait times for Inside Check-In? Question 14: — What is the utilization for an employee at the Curbside Check-i Question 15: What are the utilizations for an employee at the Inside Check Step 6: You may have noticed that in the simulation, there are entities in the queue of the SrvCheckIn2 that are in fine at the time the station closes. ‘These are marooned in this model. We will consider how to handle this kind of problem in a later chapter. Step 7: Ifthe length of the simulation replication exceeds the work schedule time, then the work schedule repeated until the replication ends. Part 3.6: Commentary '* The specification of the time-varying arrival pattern in SIMIO is almost identical to the ones in other, even including the fixed units forthe arrival rate (arrivals per hour). ‘+ Pay special attention to the SIMIO specification of cumulative probability and its value ‘+ While work schedules in SIMIO offers considerable flexibility in terms of specifying exceptions te, they are somewhat complicated to input. One has to be careful in specitying the correct times and duration, Also be careful in specifying the time to begin and end the simutation run. In SIMIO, when a server's capacity goes to zero while there is an entity (or entities) being processed, SIMIO uses, by default, a form of “Ignore” so that the items “in process” will be processed while atthe same time the server capacity goes to zero. If another behavior is desired, then you will need to implement it using more advanced concepts, which will be described later. 2 We could have modified the original link weights and avoided having 1 add the new paths and transfer node Specify the expression o be (TieNow <=10 | TineNow >= 18) | SevCheckIa[aputButfer Contents Nunberiiaiting <= Srvchackta2. InputButfer Contents. tiumbertaiting for the link weight for the path to SrvCheekln and for the path Tink Weight to SrvCheckin2 to (TimeNow >10 && TineNow < 18) && CevCnackInd. InputBudter .Contents.Sunbervaiting < SrvcheckIa. InputBuffer contents. unberwaiting. Branch Pats, Rate Tables, and Entity Dependencies: Apr Revisited 29 LUconsed to: Glos Cormier. Takhassussi Sroat Riyadh 11538 - gcormiar@atalsatecs