New solutionsfor productiondilemmasRisk-based planning and schedulingis the wave of the futureBY DAVID STURROCK December 2012 47
new solutions for production dilemmasIf you ask a group of plant between predicted schedules and actual Possible solutions to these problemsmanagers to list their top problems, performance. Without a clear picture of include using some well-establishedmany will mention meeting customer the factory floor, it is difficult to gener- technologies in new ways. But first, let’scommitments, or, more specifi- ate an accurate production schedule. explore the problem in a bit more depth.cally, creating a plan or schedule and To protect against delays, the schedulerperforming to that schedule. Why? You must buffer with some combination Why variability mattersmight expect that the companies still of extra time, inventory or capacity, Variability is the often unpredictablescheduling with magnetic boards or all adding cost and inefficiency to the deviation of processing that occursspreadsheets would have problems as system. And to do this effectively while in every real system. This variabilitycomplexity and volume increase. But minimizing waste requires the sched- often is accounted for ineffectively, andmany have made significant investments uler to exercise significant judgment sometimes ignored totally, in commonin enterprise resource planning (ERP) based on years of experience. planning/scheduling systems. Thinksystems and still have found them to fall about your drive to work. If someoneshort in detailed production scheduling The human-based solution asks how long it takes, you can probablytasks. Many of those same companies Companies now cope with software give an answer based on a typical day,have gone on to implement advanced limitations by relying on experienced perhaps 30 minutes. But does it requireplanning and scheduling (APS) solu- people to make up for shortcomings exactly 30 minutes every day? In fact, iftions to integrate detailed production of their planning tools. A person with you travel in the late evening, it mightscheduling into their ERP, and they still years or perhaps decades of experience only require 20 minutes. However,have problems. What is missing? Why might know a good way to work around during a heavy rush hour it mightdo they still not have enough forward common problem situations and essen- require 40 minutes. And when therevisibility to support meeting customer tially say, “Ignore what the plan says. Do is an accident or construction it mightdelivery commitments? this instead.” But that experience level require more than an hour. For the most part, the ERP system varies from person to person, and this Now let’s think about the simplestand day-to-day production remain approach puts daily production effi- production system possible – a singledisconnected. APS solutions have some ciency in the hands of just one or two machine producing a single part typewidely recognized shortcomings. A crit- key individuals. Further, the schedules with a single arrival rate. And let’s sayical problem with the traditional APS manually generated with this approach this machine works 24 hours, sevenapproach is that it requires that all the might not be as good as they could be days a week with no breaks or down-data be fully known and deterministic – because people cannot process all the time. Anyone could predict its operation,all processing times must be fixed. APS information necessary to create a good right? If that machine had parts arrivingsolutions typically integrate the data production schedule. at exactly 60 minutes apart and it tookfrom the ERP system but rarely deal with Perhaps more importantly, this exactly 55 minutes to process each part,current data from their factory floor. For human-based solution might not be you could predict the average waitingexample, equipment downtimes, work viable in the long run. Studies have time of each part to be zero minutesin progress details, and shift informa- shown that many industrialized nations because each part is finished before thetion often are ignored. And when an are approaching a major staffing crisis next part arrives. But if each part arrivesemployee calls in sick, the APS solution triggered by a growing number of an average of 60 minutes apart (usingtypically is not updated to reflect the employees reaching retirement age a random exponential distribution)change in resource availability. and a shrinking number of qualified and takes an average of 55 minutes to Hence, the resulting APS-generated employees to replace them. Production process (again using a random expo-schedule is by nature optimistic and schedules often are created by some of nential distribution), can you stilldifferent from what occurs in the real the more experienced workers. When predict the results? Except for queuingfacility. It is common that what starts these highly experienced people leave, theory experts, it is the rare person whooff as a feasible schedule turns infea- their critical knowledge is lost, and it can correctly predict the average waitingsible over a short time as variation and is challenging to find people with the time of about 10 hours.unplanned events degrade performance. experience and judgment needed to The impact of variability is hard toIt is normal to have large discrepancies create good schedules. predict and often is not intuitive in even48 Industrial Engineer
the simplest systems. Think how much unable to represent material handlingharder it would be in any real system. or unusual machine configurationsYet even though such variability can accurately. Much greater flexibility ishave a huge impact on production, it available using an off-the-shelf simula-is virtually ignored by most common tion package such as ExtendSim, Arenascheduling approaches. and Simio (among others). These pack- ages often allow creation of the facilityAlgorithmic versus model more intuitively by graphicallysimulation-based solutions describing the workflow. The facility willWhile many planning and schedul- be modeled by dropping objects onto aing approaches are available, they can workspace and connecting objects tobe classified broadly as algorithmic or other objects. The objects will representsimulation-based. Algorithmic-based resources, and the links will representscheduling solutions are based on either the various routes that are needed togeneric or custom-coded algorithms. produce a product. The objects can beThe custom-coded algorithms tend extended to model the exact behaviorto produce better results but are more of the corresponding resources. Today’sexpensive to implement and maintain. simulation packages typically offer 3-DUnfortunately, efficient generic algo- animation, so you can see your schedulerithms are not available for many of the results in all four dimensions, includingdifficult problems in production sched- time.uling. Algorithmic-based solutions The data schema needed to run theare best suited for long-term supply schedule is also customizable. Simula-chain planning applications where tion packages designed to work acrosslarge computation times are less of an numerous industries (e.g., healthcare,issue, the environment is less dynamic, mining, packaging, complex assembly)and the constraints are less complex to generally require that the data elementsrepresent. needed to drive the simulations also be Simulation-based scheduling solu- flexible. Data can be modeled withintions are based on a generic or custom- the simulation package to match thebuilt facility model. While some data needs to schedule the facility. Thesoftware tools feature a simulation data needed can be maintained withincomponent based on a generic fixed the model or interfaced to an exter-facility model, the greatest accuracy and nal system. Simulation packages haveflexibility results from use of a custom standard methods for importing andfacility model. Simulation-based sched- exporting data from text files, spread-uling solutions are best suited for highly sheets and databases. They also havedynamic factory scheduling applica- solid application program interfaces – ations where a fast response is required set of routines, protocols and tools forand a detailed, accurate representation building software – that enable integra-of complex constraints on equipment tion to almost any external data source,and operators must be represented to including an existing SAP or APSgenerate a good schedule. system. The full flexibility of simulation The ability to create and analyzegenerally is not available in most sched- experiments in simulation packagesuling packages as their facility models drastically exceeds the capabilitiesare based on a predetermined data that exist in traditional APS solutions.structure. For example, they often are Typical APS solutions might have the December 2012 49
new solutions for production dilemmasability to compare the performance oftwo schedules using a number of prede- examining variables Figure 1. This Gantt chart has incorporated risk analysis by includingtermined factors (due date, lead time, the expected variability.WIP). With simulation-based systems,the use of experiments is enhanced. Thenumber of factors and number of sched-ules that can be compared is endless. Inaddition, add-ins can be used withinthe experiments to help determineoptimum scenarios. Simulation-basedscheduling can deliver much moredetailed analysis to help determine thebest production schedule to run givenyour constraints and key performanceindicators. When selecting a simulation prod-uct for design use, you should considerhow appropriate it is for what you want up flexibility and accuracy, you often system behavior, usually with much lessto model, as well as its ease of learn- implement a solution much faster. You expertise required than with customing, ease of use, and the time required can get similar rapid implementation in algorithms.to create a solution. When looking for a a simulation-based approach by start- The base simulation technology usedsimulation product for use with sched- ing with a somewhat simplified facility is similar to what you may be using foruling, you should consider the features model, while retaining the advantage of evaluating design and process changes.it has that specifically make scheduling later enhancing the model to solve your Depending on the simulation tool youeasier. The more basic products allow unique problems. select, you may have a design model thatsufficient data import and export to There is an old saying that often is could be used as a starting point. If yousupport scheduling activities, but these applied to scheduling: “When you are don’t, you might want to take advan-would require more custom work and up to your waist in alligators, it is hard tage of the opportunity to model yourpossibly programming to prepare them to remember that your objective is to facility with the objective of identifyingfor effective use by a scheduler. Some drain the swamp.” A simulation solu- bottlenecks and other problem areas toproducts have features like the ability tion based on a custom facility model improve efficiency. Then take advantageto generate Gantt charts automatically, can be implemented at any desired of your simulation product features towhich helps support scheduling activi- level of detail, allowing the tool to start extend that facility model for schedul-ties. The top tier products (from a generating useful schedules quickly. ing purposes.scheduling perspective) have schedul- You can deal with the biggest alligators While custom simulation models caning extensions that allow you to build while simultaneously starting to drain be as complex as needed and consumea fully customized scheduling environ- the swamp. The model can be enhanced and produce a great deal of data, it isment within a single product with no as needed to improve the results by not necessary (or even wise) to startprogramming required. They provide a incorporating the best practices of the with such a complicated model. Simu-fast and flexible route to a custom solu- experienced schedulers. For example, lation is a perfect example of the Paretotion. after you have a working schedule, you Principle, as about 80 percent of the can pick an area with the greatest poten- benefits typically come from about 20How simulation works tial payback and improve that next, thus percent of the effort. The key is to imple-When you have a critical scheduling continuing to drain the swamp. You can ment the most important 20 percentproblem, speed of implementation is keep repeating this step as necessary first. For example, if process planningimportant. One advantage of using until your solution is as good as you by part family provides good enoughthe fixed model available in traditional need. As your system changes, often you results, then there may be no need toscheduling systems is that by giving can update the model to match the new provide any more detail than that in50 Industrial Engineer
analyzing the metricsFigure 2. RPS records statistics based on any targets by which your system can be measured. This chart shows an analysis of targetbudgets and ship dates.your process plans. On the other hand, manner. A risk-based planning tool with the potential for much greaterif late supplier orders are a major source now becomes the intersection between facility detail and real-time facility data.of disruption, then you might want to what the business needs and what the However, RPS then uses the same simu-build in extra detail around supplier system is capable of achieving given the lation model to replicate the scheduledelivery performance variability into constraints and variation of the factory. generation multiple times, all whileyour model. And by all means, you never While your base plan must be created including variability such as down-want to re-enter data. You should be deterministically, technology can add times and late material problems. Inable to tie your simulation model to take stochastic analysis for more advanced Figure 1, you see a typical Gantt chartits input directly or indirectly from your scheduling. Risk-based planning and generated without variability, but it hasexisting data systems, such as a link to scheduling (RPS) is one name for this incorporated risk analysis that resultsoutput from SAP. technology. RPS extends traditional from the subsequent inclusion of the APS to account fully for the variation expected variability. RPS records statis-Accounting for risk present in nearly any production system tics on the schedule performance acrossProcess variability introduces risk into and provides the necessary information replications, including targets likeschedule execution. Manufacturers have to the scheduler to allow the upfront the likelihood of meeting a due date,made strides to capture the variability of mitigation of risk and uncertainty. You production budget, expected milestonetheir systems using history and overall can leverage your existing investment completion date, or any other targetsequipment effectiveness calculations. in planning systems, such as SAP’s by which your system can be measuredThis data provides a historical view of APO, to close the gap between master (Figure 2).how their systems perform that demon- planning and detailed production RPS allows for flexible schedulingstrates the inconsistency and variability scheduling, thereby driving more reve- strategies to support your key produc-in their manufacturing process. Unfor- nues and greater customer satisfaction tion objectives and lets you quicklytunately, as discussed above, most APS at reduced cost. reschedule in response to unplannedsystems cannot handle that data and RPS begins with a deterministic events. You can model your complexso cannot generate schedules using schedule generated by executing the production processes to capture allthis valuable information. Simulation simulation model with all randomness critical constraints so that the result-technology provides the opportunity to turned off. Note that this is roughly ing schedules reflect the reality of youraccount for the variability in an adequate equivalent to the APS solution, albeit systems. You can display schedules in a December 2012 51
new solutions for production dilemmascut thisFigure 3. This root cause analysis details how the “cut” resource is a major contributor to non-value-added waiting time.wide range of outputs, including interac-tive Gantt charts that display individual a room many industries and application areas. Risk-based planning and scheduling iswaiting times at critical resources, as with a view a new technology that permits broaderwell as the root causes for non-value- Figure 4. Three-dimensional animation understanding of performance risksadded time in the system. Figure 3 can let you view how the shop floor would and their causes and provides a valuableillustrates an analysis that identifies look at any future time you specify. management tool for evaluating strate-the “cut” resource as a major source of gic business decisions. dnon-value-added waiting time. SinceRPS is an extension of your simulation David Sturrock is vice president of opera-model, you might be able to integrate a tions for Simio LLC. He graduated from The3-D animation of your planned sched- Pennsylvania State University with a bach-ule to provide a unique and insightful elors degree in industrial engineering. He haspreview of your facility operations. more than 30 years of experience in simula-Figure 4 illustrates how you could view tion and has applied simulation techniquesthe anticipated shop floor situation at in the areas of production scheduling, trans-any future time specified. custom-built solutions. Today’s tech- portation systems, plant layout, call centers, nology takes you well beyond those early capacity analysis, process design, healthcare,Real solutions now solutions as well as providing a tool to packaging systems and real-time control. HeWhile the concept of simulation-based help reduce your dependence on manual has co-authored two simulation textbooksscheduling has been around for decades, human judgment and reduce the impact and teaches simulation at the University ofto date it has been used mainly with of knowledge lost from employee turn- Pittsburgh.inflexible standard facility models or over. The flexible models extend use to52 Industrial Engineer
Copyright of Industrial Engineer: IE is the property of Institute of Industrial Engineers and its content may notbe copied or emailed to multiple sites or posted to a listserv without the copyright holders express writtenpermission. However, users may print, download, or email articles for individual use.