IE434 Lecture Notes SPRING 2007 SCHEDULING III. CONTRIBUTION OF KEY TOPICS TO SCHEDULING Objective of this section is to emphasise the importance of some key topics which become very important in real life applications in which a schedular is involved or affected in some way or another. 3.1. Development of “Standard Operation Times” The terminology of “standard time” is believed to be self explanatory. In this case, what we are talking about is the “standard operation time per piece” which have to be spent for a shop order of a specified part number when it is processed at a work center. Thus, the sum of all “standard operation times per piece” in the routing of a shop order multiplied by the number of pieces on it will be equivalent to its “standard work content” in terms of man or machine hours (Figure 32). Figure 32 A schedular must have information or at least an idea of “standard work content”s of shop orders to be loaded to a work center in order to come up with respective “remaining work content” values associated with each respective shop order. Thus, it will be possible to use specific techniques to determine the sequence of loading. Such techniques will be discussed in later sections of the lecture notes. One last thing we should say about “standard operation times” is that, in majority of real life cases, it is usually difficult to find them especially in small and medium scale companies. In large scale companies, they are usually obtained by treating the “actual operation time”s with some factors which are found to be appropriate. Sometimes, the operation times which are standardized in this way, are assigned to the work center itself. In such cases, it is assumed that all the part numbers passing through that specific work center will have exactly the same “standard operation time”. This approach may arise some questions. Forexample (Figure 33), is it acceptable to assign the same standard operation time to two different part numbers at a drilling work center, when one of them requires one drilling of one hole while the T. Cengizhan PAMİR Page 1 / 13 IE434 Lecture Notes SPRING 2007 Figure 33 other one needs drilling of three holes? Even the differences in the size or shape of the holes or the type of the material itself (aliminum or steel) cause differences in operation times. For the above stated reasons, development of reliable “standard operation time”s is a very important subject for realistic scheduling especially in job shops. 3.2. The concept of “Earned Hours” In the previous section, the “remaining work content” of a shop order was mentioned. In this section, we are going to tell some specific facts about the “earned hours”, because, inorder to determine the “remaining work content”, one has to know the “earned hour” value of a specific job at any point in time(Figure 34). “Earned Hours” is the measure of how much of the work is completed in standard terms. Number of actual hours spent is rarely equal to earned hours. In practice there is a wide range of factors causing this fact. T. Cengizhan PAMİR Page 2 / 13 IE434 Lecture Notes SPRING 2007 Figure 34 Determination of earned hour values in flow shops (either flow assembly work or flow fabrication work) is somewhat straight forward. On the contrary it is difficult to make exact determinations in job shop environments. Such environments may require additional tasks of mapping the routing of each part number in terms of manufacturing progress steps. An example of such a mapping of routing can be seen in Figure 35. Figure 35 T. Cengizhan PAMİR Page 3 / 13 IE434 Lecture Notes SPRING 2007 Obviously, the approach given in Figure 35 requires extensive computations by a specifically designed software in the existince of hundreds of shop orders floating in job shops of a manufacturing company. 3.3. The concept of “Group Technology” (GT) The central theory of group technology states that; various situations that require decisions can be grouped together based on preselected, commonly shared criteria, and that the decision which applies to one situation in the group will apply to all of them in that group. Thus, the definition of group technology tells that; it is a technique for identifying and bringing together related or similar parts/components or machines in a production process, in order to take advantage of their similarities by making use of the benefits of flow production methods. Although both MRP and Group Technology are usefull in planning an control of multiproduct, small lot-sized job shops, they appear to be in conflict with respect to their individual operational characteristics. When group technology is applied to the manufacturing of parts, the emphasis is on grouping them in terms of their common manufacturing characteristics regardless of the level of product structure they are in or the timing of when they should be manufactured. On the other hand, MRP strictly considers the position of respective part within the product structure hierarcy, and the timing of fabrication to meet the required end item due date. In other words, group technology wants similar parts manufactured at the same time regardless of the requirement dates, and MRP wants parts to be manufactured at the scheduled times regardless of the manufacturing impact. At this point, one may ask the question of whether it is possible to integrate MRP and GT. The answer will be a carefull “yes”. Especially at workstations where both the number of unique part numbers and the volume of production is high, respective parts, reardless of their position in the product structure, can be grouped by their manufacturing characteristics and by the requirement quantities for fixed intervals of time. For such a group, most probably, the production schedule (qty and timing) will be the production schedule of the part which has the earliest due date. The rest is to define the respective part numbers by a specific code or group number within the MRP data base and assign ordering parameters (e.g. order policies) accordingly if MRP software is considered to be modified for handling such parts. But, since trying to modify the engine of MRP is a dangerious task and costly, required replenisments can be downloaded to a PC as raw schedules for further processing and group scheduling. The most important aspect to be remembered of this case is that, “dynamic grouping of parts” must be performed. By “dynamic grouping” it is meant that, each time when the machine is being loaded, the group content may be different from the previous case. That is, if a specific load is composed of Part A, Part B, and Part C; the following load may contain Part A, Part C, Part E and Part F. If one tries to utilize “static grouping” (e.g. Part A-Part B-Part C always togerher), after the first release of a specified group of shop orders, things will get much more complicated. T. Cengizhan PAMİR Page 4 / 13 IE434 Lecture Notes SPRING 2007 Deficiencies of the “static grouping case” can be observed in the example given by Figure 36. Figure 36 3.4. The concept of “Flow efficiency” Measuring flow efficiency in job shop manufacturing environment will give important clues to the scheduler about the shop behaviour in terms of the speed of production flow. This is important especially where jobs (shop orders) are flowing through multiple stages (work stations). In order to provide a clear vision of flow efficiency, the following has been summarized from “Factory Flow Benchmarking Report, Lean Aircraft Initiative, 1996”. “Flow Efficiency is defined as the ratio of the fabrication time (touch labour time) to the cycle time (also called throughput or flow time). Factory Flow Benchmarking Report (#RP96-06-61) by the Lean Aircraft Initiative defines major components of the cycle time as; fabrication time, lot process delay, storage delay, and transportation delay. Subject report says that; in the airframe sector, proportion of cycle time that the product was waiting was 96 %. Also it is indicated that, the airframe sector had flow efficiencies from 0.02 percent to 0.8 percent (Note: These values belong to 1996). Actual labour hours per part per crew member Flow Efficiency = ( )*100 Total cycle time the part is in the system – router queing T. Cengizhan PAMİR Page 5 / 13 IE434 Lecture Notes SPRING 2007 Figure 37 Correlation analyses carried out had shown that, there would be higher flow efficiencies with “lower lot sizes” and with “shorter distance travelled”. Other findings can be listed as in below; The flow efficiency, at its best, will vary inversely with the lot size (Max attainable flow efficiency = 1 / lot size). It has been detected that, there were no flow efficiencies above 2 % if the parts travelled more than 2000 feet (609.6 m). It has been found that, the process type layout had a great deal to do with the flow efficiency. Job shop layouts did not achieve above 0.1 % flow efficiencies while flow shops, cells or dedicated lines were able to achieve values as high as 18.7 % flow efficiencies. The median values for routing queuing (mnfg approval time) ranged from 11% to 32% of the cycle time. Maximum value was 83 %. The distances travelled by the parts found to be averaging from 2416 feet (736 mt) to 5023 feet (1531 mt). Process control steps ranged from a low of 9.3 % to a high of 81.4 % and averaged from 44 % to 55 %. The predominant process control method (at about 80 %) was process verification consisting mostly of manual inspection.” An analysis had been carried out at Turkish Aerospace Industries in 1999, about fabricated flat sheet metal parts. This study was covering a total of 599 shop orders (4867 pieces) for which “flow efficiency” & “cycle times” were calculated. T. Cengizhan PAMİR Page 6 / 13 IE434 Lecture Notes SPRING 2007 Obtained results were as following; Scheduled Flow Efficiency = 0.36 % Actual Flow Efficiency = 0.60 % Attainable Max Actual Flow Efficiency = 30 % Scheduled Cycle Time = 66 M-days. Actual Cycle Time = 70 M-days. However, there is a trap in dealing with flow efficiency values, from which a schedular must avoid. This can be better explained by the following example; Let’s assume that, shop order F200200 is released for manufacturing of 6 pieces of part number 353535-35. Again, let’s assume that, o Only one operator worked on the shop order at each work station it passed through. o The scheduled cycle time for shop order is 10 M-days. o The total standard work content (std touch labour) for shop order is 150 minutes. o The standard router queue time is 5 M-days. o The actual router queue time is realized as 7 M-days. o The actual cycle time for shop order is realized as 15 M-days. o The total actual touch labour for shop order is realized as 270 minutes. Therefore the calculation for realized flow efficiency will yield following result; FE = {[(270/6)/1]/[[(15-7)*7,5]*60]}*100 = (45/3600)*100 = 1,25% and Max FE = 1/6 = 16,7 % Where the scheduled flow efficiency for the case was as follows; FE = {[(150/6)/1]/[[(10-5)*7,5]*60]}*100 = (25/2250)*100 = 1,11 % , and Max FE has the same value of 16,7 %. Above example suggests us, “the longer touch labour hours on the shop order” tends to show “higher flow efficiencies”. So, especially during the setups of machinery or where man power is utilized to perform the required operation, if the job progress is slower than some target value, flow efficiency value might increase artificially. We can conclude that, in the above case, either something is wrong with the demonstrated performance of the labour, machinery, equipment or with the standard values we are using. A good schedular would have good guess of what was happening at the shop floor. T. Cengizhan PAMİR Page 7 / 13 IE434 Lecture Notes SPRING 2007 3.5. Activity Grouping in large Assembly Work Stations In large assembly lines where the end item moves through a series of work stations, each station corresponds to a specific “work breakdown structure” (as discussed in section 2.3.2). In this way, the whole end item is represented by a series of “work breakdown structure”s each of which is a subject of master scheduling. That is, there is a separate master schedule for each “work breakdown structure” where each master schedule is sequentially dependent on the previous one in the series (Figure 38). For this reason they are called “scheduled work breakdown structures” (SWBS). Figure 38 Within each SWBS, there are a number of assembly operations some of which are also sequentially dependent to others. Usually, the time required to complete the whole tasks for an SWBS may take several weeks. Naturally, the respective assembly operations require detail and subassembly parts together with some materials which are fabricated in the same facility or procured from outside. Figure 38 shows an example assembly line for such an end item and its breakdown into SWBSs. It also shows breakdown of the work content of sample SWBS which is named SWBSC2 in terms of the sequential operations to be completed. Let’s assume that the make span assigned to SWBSC2 is 45 M-days and the part requirements of the 14 operations are as in Figure 39. T. Cengizhan PAMİR Page 8 / 13 IE434 Lecture Notes SPRING 2007 Part Requirements for Operations of SWBSC2 # of required OPERATION required qty make items 1 2 3 4 5 6 7 8 9 10 11 12 13 14 220 170 125 130 450 210 180 280 190 175 375 230 195 345 480 512 200 150 970 230 190 420 225 210 420 520 380 460 TOTALS 3275 5367 Figure 39 In MRP database, the independent requirement for SWBSC2 will normally have due date as M-Day 4545, and the operation part numbers will have process time of 45 M-Days and, lets say, 2 M-Days as stock time which is going to be used as a part of move time from SWBSC2 to SWBSAC1. Exploding MRP only with this information will result in a necessity to complete all 5367 pieces just before M-Day 4500 which is the scheduled start date of SWBSC2. All of the Rp schedules of respective details and subassemblies will be calculated accordingly. At the time of ordering these parts to fabrication areas, there will be a sudden jump in the work load and most probably temporary bottlenecks will be created. Activity grouping in such assembly lines may prove to be useful in the operation of job shops by distributing the work load as evenly as possible through out the time. This grouping is about the grouping of the operations within an SWBS by considering their interdepencies. Thus, it becomes possible to distribute the requirement dates of detail parts over the make span of the respective SWBS. In the example of Figure 38, 14 operations of SWBSC2 are divided into three groups for which the respective requirement dates are M-Day 4500 (for 2532 pieces), M-Day 4515 (for 1365 pieces) and M-Day 4530 (for 1470 pieces) respectively. The next step to do is to treat the process times of 14 operations as follows; First Group (6 operations) ; Process time : 45 M-days (no change). Second Group (4 operations); Change Process times from 45 to 30. Third Group (4 operations); Change Process times from 45 to 15. The rest will be performed by MRP software yielding to a smoother shop order release activity and in this way, a smoother loading of shops. T. Cengizhan PAMİR Page 9 / 13 IE434 Lecture Notes SPRING 2007 3.6. Takt Time and Cycle Time concepsts : What do they mean to a Schedular? Both “takt time” and “cycle time” are the powerfull tools in understanding the shop floor. Continuous observation of these values gives a good idea about the behaviour pattern of a work center or of a specific shop itself. Variances in the respective values may occur from period to period due to some expected or unexpected reasons such as “over loading”, “machine breakdown”, “unavailability of a tool”, “inexperienced worker”, “seasonal high or low temperatures”, or even “the announcement of raise % in monthly salaries”. In addition to understanding the shop floor, sometimes, decision making process may state “which time value is to be used as a basis for scheduling purposes : takt time or cycle time or standard time for a process”. 3.6.1. What is “Takt Time” ? "Takt" is the German word for the baton that an orchestra conductor waves at the musicians, to regulate the speed at which they play. In manufacturing terms, takt time is a calculated value. Takt time is the speed at which parts must be manufactured in order to satisfy demand. It is simple enough to calculate: Determine the daily demand (order) volume. Lets say we have orders for 215 arm chairs per day. Determine the number of working minutes in a day. Let's say that we have an eight-hour day, with 30 minutes for lunch and two 10-minute breaks. This means we have: (8 x 60)-30-10-10=430 minutes, or 430 minutes in a working day. Divide the number of minutes by the number of products needed. In our current example, the calculation would be 430÷215, which equals 2. This means that one unit must be manufactured every other minute in order to meet demand. Takt time is 2 minutes. Here what we have to clearly indicate is that, “takt time is the goal”. It must be reached to satisfy demand. In the mean time we also need to mention that there are certain points which one has to keep in mind when talking about “takt time”. These points can be clarified as follows; “Takt time” does not consider the terms “capacity” or “hours per piece”. It simply tells “the lenth of time at which one unit of product is to be completed” to meet the demand volume for the specified period. If takt time is less than the process time of one unit, the ultimate goal should be to achieve some process improvements which will shorten the lead time. In practice, usually the first thing which is considered in such situations is to provide additional T. Cengizhan PAMİR Page 10 / 13 IE434 Lecture Notes SPRING 2007 capacity by using overtime, which most probably will not yield to desired results. Work centers for which takt time will be the measure of performance, should be arranged in a way to process “part families”. This means that, the parts to be processed in respective work center should be similar with respect to operational characteristics such as, process requirements, material type, material thickness, dimensions etc. In a production environment which is going to be monitored by takt time, the work content or product mix and part routings, should remain same as much as possible. Again in such environments, all of the shop orders must be “workable”. 3.6.2. What is “Cycle Time” ? The second type of time is "cycle" time. This is a measured value, not a calculated value as takt time is. In other words, you must go out to the floor, and measure the time it actually does take to manufacture the product. When making time observations, it is important to measure both the total cycle time for each operator (how long the job takes from beginning to end), and the time of each of the component tasks that make up the cycle. The cycle cannot be improved without a detailed understanding of what makes it up, and often it is possible to reassign component tasks to rebalance the operation. After making the observations, one can draw out an Operator Loading Bar Chart to graphically express what is going on in previous example (Figure 40). The horizontal axis on this bar chart represents operators, whose times are indicated by the use of stacked bars along the vertical axis, which is the time axis. Draw a thick line across the chart to represent the takt time (two minutes, in this case); this makes it easy to tell at a glance whether any individual operator is working within takt time, or exceeding it. Let us suppose that the six operators are producing one arm chair in two minutes, but that their cycle times look like as in Figure 40. Figure 40 T. Cengizhan PAMİR Page 11 / 13 IE434 Lecture Notes SPRING 2007 Arm chairs cannot be produced any faster than the slowest operator works, so in this case, we get one arm chair every two minutes, but the six operators are working for a total of 8 minutes to do it. Again assume that, we observe enough waste in the work cycle to set a team goal for reducing the total cycle time from 8 to 6 minutes. To determine the staffing of this line, we divide the new total cycle by the takt time: 6 ÷ 2 = 3 operators. Three operators would be sufficient, and the new bar chart would look like as in Figure 41. Figure 41 Thus we can assign three freed up operators to some other value added tasks. A similar “Load Chart” is given in Figure 42 which is from a real life example of what has been achieved in assembly of flight deck panel production. T. Cengizhan PAMİR Page 12 / 13 IE434 Lecture Notes SPRING 2007 Figure 42 T. Cengizhan PAMİR Page 13 / 13