This page is under construction 14/02/07
Production control, at first, seems an immense subject with many sub-topics and specialities that are particular to specific, individual industries and companies. This 'course', or reference site, shows that the subject can, in fact, be described in a general way, and that it can be clearly and logically structured around just half a dozen headings. The many topics encountered in production control fall neatly under these headings, and are described below. The paragraphs coloured blue within the Agenda/Index are overall descriptions of the main headings of the subject.
1.1 Definition of the MPS
1.2 Formulation of the MPS
1.2.1 Sales & Operations Planning'
1.2.2 Detailed Formulation of the MPS
1.2.3 Capacity Requirements Planning
1.2.4 Capacity Management
1.2.5 Plan Stability and Time Fences
2.1.1 Gross Requirements
2.1.2 Steps in the Calculation
3.1 Data Feedback and Closed-Loop MRP
3.2 Plan Types in MRP
3.2.1 The 'Open Order'
3.2.2 The Planned Order
3.2.3 The Firm (Fixed) Planned Order
3.3 How the MRP Logic Works
4.1 Leadtime and Queue - Introduction
4.2 The Viscious Circle
4.3 Inpur/Output Control
4.4 Queues at Intermediate Work Centres
4.5 Final Order Release
4.6 Job Despatching Rules (Job Prioritising)
4.6.1 Slack Time
4.6.2 Slack Time per Operation
4.6.3 Critical Ratio
4.6.4 Shortest Processing Time (SPT)
1.1. Definition of The Master Production Schedule
A definition of the 'Master Production Schedule' is "a time-phased set of plans to acquire material that will thus be available for sale or despatch to customers". That is, the master schedule relates to finished products, not to components or raw materials. "Acquiring" the stock could imply simply buying-in, but will usually mean production. If it means productions, then the MPS gives rise to the supporting material plans, but it does not constitute the materials plans themselves.
In addition, the MPS must often be expressed over quite a considerable timespan because of the leadtimes inherent in distribution and manufacture for acquiring the supporting materials and raw materials. As well, because of the ever changing environment in which manufacturing in general operates, the MPS must be checked on a frequent basis to ensure that it remains appropriate - ie that the premises on which it was built remain valid. The frequency of re-checking (and, often, re-formulation) is usually monthly, this being determined by the pace of change of the market in which the MPS products are sold and the liability to change of product forecasts. Companies which review their master schedules weekly should ask themselves whether in fact they are responding to required changes in material and raw material requirements rather than (higher level) master schedule requirements.
1.2 Formulation of the MPS
Classically, formulation of the master production schedule is a two-stage process performed in simulation:
1.2.1 Stage 1: Sales & Operations Planning' (also known as "Business Planning" or "Production Planning")
This term denotes the broad planning done at the product family (ie group) level. For example, senior management will often consider the firm's business in four or five general marketing areas (for example 'electrical kitchen goods', 'washing machines', 'other white goods' etc), and express production targets in money over time. Every firm divides its products into well-recognised groups of some type. A primary input to the Sales & Operations Planning stage of master schedule formulation is therefore the demand forecast over time for each product family, expressed in appropriate units (£, tons, units ...). A mechanism must consequently exist within the company's sales forecasting system for aggregating individual product forecasts over time into their respective families. The forecasts must be reviewed and (usually) revised by senior management to take account of economic and market trends; stock building or order book policies; finance; and capacity capabilities.
1.2.2 Stage 2: Detailed Formulation of the MPS
Because it is, of course, eventually necessary to manufacture individual, real products, a mechanism must exist for translating the family or group production forecasts to manufacturing plans at the product level. Because of the detail involved, individual product master schedules are clearly more demanding of time and analytical effort. The plans finally must lie within the boundaries agreed at the Sales & Operations planning stage outlined above for finance and general volume production. What is potentially far more difficult, however, is to formulate the product-level MPS with on-going checks that the plans being made are viable from the viewpoint of capacity availability. The normal solution is to formulate the MPS, and then simulate the effects of it by calculating the required quantities and schedules of components and raw materials that would be needed to fulfil it, taking into account existing stocks and open (ie on-going) plans. The machine or plant capacity required to support the resulting schedules is then compared to the capacity which exists, in a step known as 'Capacity Requirements Planning' (CRP) - see below.
1.2.3 Capacity Requirements Planning
When a job is released to the shop floor, it is assumed that manufacturing capacity is available for it. This assumption is made valid because of a series of checks, from the broad level to the detailed level, of existing capacity vs. the manufacture which is being planned. These checks constitute a 'Hierarchy of Capacity Planning'. In other words, it is the master scheduling process which ensures the shop does not become overloaded.
A MASTER SCHEDULE MUST NEVER BE RELEASED AS A COMPANY PLAN IF IT IS NOT POSSIBLE TO FULFIL IT FROM THE CAPACITY POINT OF VIEW.
The hierarchy of capacity checking and planning begins with two major master scheduling checks termed (1) Resource Requirements Planning. This typically involves either an analysis of historical data or the use of simple models. It is not detailed, and is carried out by senior managers, then (2) Rough Cut Capacity Planning (RCCP). Typically, RCCP involves the use of simple models or, particularly, the deployment of 'CRP' (see below). Some detail is required at this stage, and it is carried out by middle managers.
There are many methods for carrying out rough cut capacity planning, depending on the inherent complexity of the company's plant and machines. The most sophisticated technique is CRP - capacity requirements planning (below). In this, the Master Production Schedule is exploded over time, using the bill of materials and leadtimes, to produce time-phased schedules of components, sub-components and raw materials needed to manufacture it. After taking into account working stock in the system, these component and other needs are compared to the capacity required to manufacture them. Bottlenecks or under-capacity situations must be resolved at this stage before the MPS can be released to the company and, ultimately, to manufacture as an acceptable, achievable plan. The process of resolution is often by way of capacity management and requires - demands - the participation of the people who will be responsible for the actual achievement by the shop floor of its part in the plan. Only when all potential problems have been resolved should the plan be released as a definite company commitment.
1.2.4 Capacity Management. Capacity requirements planning is a technique used to determine the work load that would result from a master schedule. Capacity management by contrast is the process of attempting to deploy the capacity likely to be involved in achieving the plan. Typically this might be by arranging to switch labour from one site to another, making special use of certain machines and other technical and managerial actions. Capacity management meetings are often weekly get-togethers of senior production staff. The presupposition, of course, is that capacity management is possible. In some companies, capacity is inherently inflexible, especially with regard to labour. It may then be necessary to rearrange the material requirements rather than the capacity capability ... a far trickier undertaking.
1.1.5 Plan Stability & MPS "Time Fences"
The means of calculating the MPS and subsequently for deciding whether or not there is a requirement to reschedule the plans, is the Projected Stock or 'projected available balance'. In each period being assessed, the projected available balance comprises the starting stock quantity plus any master schedule production quantity due in in that period, less the forecast demand in the period. For example, at the end of Period 3, the closing stock may be calculated to be 47 units. So the projected available balance in Period 4 may be "47 units (ie the closing balance from Period 3) plus (say) 50 units, the required MPS schedule of manufacture in that period, less (say) 30 units, the forecast demand for Period 4. Thus the projected available balance in Period 4 is 47 + 50 - 30 = 67 units. If the projected available balance was too low, such as below zero or below the safety stock level associated with the particular product, the system would generate a message to the master scheduler informing him of the fact and suggesting to him that the master schedule should be rescheduled, by scheduling in the next master schedule plan to the date when it is seen to be now required. Clearly, by bringing in the next scheduled production lot to this earlier date, the low stock situation revealed by the projected avaiable balance will be avoided. However, if this alteration to the master schedule were to be made, it would be necessary to recalulate the supporting schedules of materials and raw material plans.
Because of the disruption that a revision of the supporting material and raw material would cause, every effort is made, in fact, to keep the master schedule stable, notwithstanding the outcome of calculating the projected available balance. To achieve this stability, so-called time fences are "placed" on the master schedule. What this means is that over the schedule of plans itself that makes up the master schedule, various points ahead are designated as "time fences" and any recheduling of plans that lie within the fences are given special treatment and are subject to special rules. Thus the first time fence may be set from the current date to one week ahead. This period (of a week) may be designated the frozen zone, and no rescheduling allowed of any plans within that period, regardless of the messages emanating from the projected available balance. The second time fence may be set four weeks out, and the period from one week out to four weeks out referred to as the semi-frozen zone. The rule here might be that there should be no rescheduling in the semi-frozen zone without the permission of the manufacturing manager himself, (or, perhaps, the master scheduler). The designation of time fences and the rules associated with them will vary from company to company and will be dependent on the nature of the production plant and the volatility of product demand.
2.1 Materials Planning and The Bill of Materials
At the centre of materials planning is the notion of independent and dependent demand, and 'stepping down' through the bill of materials. Independent demand is the demand for materials at the top of the bill of materials - ie sales products. The demand is 'independent' because it is not dependent on the demand for any other materials or components lower down in the bill. Dependent demand by contrast is the demand, or requirements, for materials at lower levels of the bill of materials. The demand is dependent because it depends on, or determined by, the demand for the materials immediately above it in the bill. For example, a very simple bill of materials might consist of a bicycle (at the top) and two bicycle wheels immediately below it. So if the demand for bicycles was 13 (say, to meet a sales forecast), the dependent demand for bicycle wheels would be 26 (ie 13 x 2).
The bill of materials can be analysed so as to place all the entries in it at various "levels". At the top level (traditionally referred to as Level 0) there are the sales products (these are the ones having independent demand). The next level down (Level 1) are the products immediately below Level 0. Thus the bicycle just referred to is at Level 0 and the bicycle wheels are at Level 1. And so on, level by level, to the bottom. The final level, or bottom of the bill, is where raw materials (ie bought-in products) are to be found. Traditionally, this is designated Level 99 (there never are 99 levels in a bill, but there is no harm in nominating the last level "99" - it ensures that raw materials are easily distinguished). Before describing the mechanics of materials planning, a number of terms are introduced. Thus:
2.1.1 Gross Requirements
Gross requirements are the total needs from all sources for a particular item at a particular time. Gross requirements are in support of plans for the item, plus any independent demand. For example, a plan for light wheelbarrows (at Level 0) might be 200 barrows; a plan for heavy wheelbarrows might be 150 barrows (but with 2 wheels per barrow), and a plan for spare barrow wheels might be for 50 wheels. The gross requirement for wheels is thus 550 wheels (200 + 150 x 2 + 50). Note, however, that the plans for wheels will be formulated taking into account of the current stock of wheels and the planning rule involved in the manufacture of wheels. In short, the term 'gross requirements' simply means the total requirements for a given product, before any stock on hand and planned manufacture are taken into account.
2.1.2 Steps in the Calculation
Once the data implied above are in place and the techniques understood whereby the actual calculations of plans and gross requirements are made, the calculations of material plans themselves can be undertaken. The steps are straightforward:
1. Calculate plans to meet the gross requirements of each product at one level of the BOM. Next ...
2. Calculate the lower level products needed to support these plans.
3. When one level has been finished, move to the next level down.
4. When the last level has been dealt with, the procedure is complete. This is "Level 99", and the plans generated are purchase plans.
3.1 Data Feedback and Closed-Loop MRP
The mechanics of feedback can be summarised as follows:
1. Input each day to the system of data relating to 'events' (ie the submission of data transactions notifying the system of various important facts, such as plans (identified by their number) which have been started, or "released", or plans which have been completed etc.);
2. Recalculate all plans to meet the master schedule, in the light of the revised data which has just been fed back;
3. Take action relating to the differences between the newly revised plans just calculated and the previous plans.
A revised materials plan for an individual product may either directly take the place of the previous equivalent materials plan for that product (ie one previously issued prior to closing the loop) or instead the system may recommend to the planner and shop supervisor, by way of a message, that they themselves should change the previous plan. Which of these two alternative actions are to be taken is governed by the rules of the MRP manufacturing logic. In order to understand this logic, it is necessary to know and remember the character/rules of the various plan types within the MRP system, as explained next.
3.2 Plan Types in MRP
When plans are created, or formulated, through the materials planning logic indicated above, they seem at first glance to be all of the same character and format. That is, each one specifies a particular product to be manufactured (or purchased), a particular quantity to be made, a start date when the plan is to be commenced, and a finish date when the plan must be completed. However, within the MRP material planning system, plans in the system are, in fact, of different types, and understanding and remembering what these types are is absolutely essential to an understanding of the system. There are three types:
3.2.1 'Open' Orders
These are plans currently being worked on in the factory (or raw materials actually under despatch). That is, they are plans which have reached or passed their start dates and have been "released" to actual manufacture or purchase (as notified through a feedback transaction). The rule is that open orders are under the planner's (or buyer's) control - ie they cannot be rescheduled to new dates by the system;
3.2.2 'Planned' Orders
These are works orders required in the future and which are entirely under the control of the MRP planning system. That is, they are planned entirely by the materials planning system and are exclusively under the system's control with regard to standard manufacturing lot quantities and start/finish dates determined exclusively by the materials planning logic as described above with respect to the projected available balance, (or, for raw materials, they are standard raw material order quantieies needed in the future, with dates again determined by the projected available balance planning logic under MRP's control);
3.2.3 'Firm (or Fixed)' Planned Orders
These are works orders (or purchase orders) within the system required in the future but which are under the (human) planner's control. (Since they are under the planner's control, the plan quantities and plan manufacturing leadtimes may be non-standard, as specified by the planner.) When the MRP planning system is re-run - ie when the loop is closed - although the firm plans are taken into account, they are never changed with regard to quantities or timing. However, if the quantity or timing of a firm plan is at variance with the ideal quantity or timing of the plan as determined by the strict logic of the MRP planning system, the MRP planning system leaves them in the overall plan, but generates a message to the planner recommending that he/she reschedules the plan by so-many days so as to conform with the strict planning logic. (The message reads, perhaps, "Reschedule Plan X from Date 1 to Date 2") (firm raw material plans may also be fixed, and therefore may also give rise to rescheduling messages for buyers or suppliers)
3.3 How the MRP Logic Works
As stated above, starting at the top of the Bill of Materials (ie Level 0), time-phased plans are worked out to meet the requirements of the MPS. Then, plans in support of these are worked out to manufacture materials at successively lower levels of the bill of materials, finally down to raw materials. The process stops when the lowest level of the bill has been dealt with (ie the raw materials level, traditionally designated "Level 99"). Plans generated here, of course, are raw material purchase plans.
After a time (*), transactions (ie data records) relating to relevant activities that have occurred in reality are fed back into the system in order to update the data held there. The most important transactions will clearly relate to plans in the planning schedule which have been been "released" (ie started) or plans which have been completed. These transactions will state that the relevant plan has indeed been started or has been completed and will state also the quantities of material manufactured. The latest stock records are also available. (*The frequency of feedback these days is usually daily. In the past, feedback might have been weekly, with all the problems this brought. Feedback is almost never more frequent than daily, since the subsequent more frequent rescheduling of physical manufacture that this intimates is rarely practical.)
When the feedback of data has taken place, new (ie revised) material plans are recalculated. The new material plans are next compared to the current, or 'previous', material plans.
The action MRP now takes depends on whether the old, previous plans were ordinary "planned orders" (as defined above) or whether they were either open or firm/fixed planned orders.
In reality, old or previous planned orders are deleted by the system and replaced by the new plans just calculated. As far as the system user is concerned, however, it appears merely as if the old, previous plans have been left as they were, except probably for being slightly amended as to their required dates.
Open and firm planned orders cannot be deleted or even amended by the MRP system, however, and they appear in the new/revised schedule of plans as now calculated. However, the MRP replanning logic takes note of when these plans are due and the quantities of material associated with them, and also calculates the ideal, correct dates when and by how much the plans should be scheduled for, to meet the precisely calculated requirements. If the date/quantity of an open or firm plan differs from the ideal requirement, the system produces a message for the user requesting that he/she should reschedule it. (Clearly, if the date/quantity of an open or firm plan coincides exactly with the ideal requirement, no message is issued.)
3.4 Problems in using Closed-Loop MRP.
Three major problems centre round ...
Information Feedback;
Technical Infeasibilty; and
Capacity Infeasibility.
These drawbacks to closed-loop MRP, any one of which might prevent the system being successful, together, explain why use of the system is very widely regarded as having been disappointing since its inception in 1972 and its introduction to the UK in about 1976.
4.1 Leadtime and Queue - Introduction
Manufacturing leadtime is the time which elapses between a job's being released to the shop floor and the subsequent availability of the manufactured items at the place they are required. As we have seen above, product leadtime is a central item of data in planning. Hitherto, however, it has simply been assumed that leadtime is "known". In production planning, it is now necessary to take a closer look at it.
The principle components of leadtime are:
Queue Time;
Set-Up Time;
Running Time;
Wait Time; and
Move Time.
In a great many shops, the chief element of these five is queue. Queues, however, are frequently engineered by foremen to prevent machines and men falling idle.
As stated, in a very great number of manufacturing concerns, the dominant and often troublesome component of leadtime is queue time. It is not unknown in certain engineering job shops for queue to comprise 80% or more of total leadtime. And a justification put forward by the manager or production controller for doing nothing about this state of affairs and making no attempt to reduce them, is that the release of work to manufacture from planning is uneven, or that it is liable to unevenness. Efficient manufacture requires a steady flow of work, they say, if an alternation of idle machines and temporary bottlenecks is to be avoided. If queues of jobs are held in the shop, the manager can himself guarantee that his workload will be steady and that idle machines and unemployed labour will be avoided.
4.2 The Vicious Circle
First, a definition: a gateway work centre is a work centre where a starting operation is performed, immediately following job release - say, a turret machine or press. The action required to eliminate unevenness in the flow of work to gateway work centres and with it the reason for maintaining queues there is for the production controller to manage the rate of release of jobs in the most careful manner, sometimes holding them up, sometimes bringing them forward. The mechanism for overseeing this process is input/output control. Before illustrating it through the input/output control table, a cautionary tale must be told of leadtime inflation in companies which did not understand the fundamental link between leadtime and queue and which therefore failed to adopt the simple corrective procedures necessary. Thus:
The Viscious Circle
* A flurry of orders is received and all of them are released at once to the shop. Q. What happens? A. The published leadtime increases and complaints are received.
* The company responds by increasing the 'official', quoted, product manufacturing leadtime so that complaints about not meeting promises will no longer be true. Q. What happens? A. Because the leadtime is now so much longer, customers' future jobs in their own planning systems, which were to have been released later, now become due (and in some cases are past due, so are marked 'urgent'). These are duly sent into the company, and this rush of new work is again released to the shop,
* With this rush of more new work, the shop becomes seriously overloaded, and even the newly promised longer leadtimes are not met. Q. What happens? A. Following more complaints, the company escalates its promised leadtimes even more, and again customers find that future jobs which were to have been released (much) later have now become due or past due. These jobs are again duly sent into the company creating yet a further rush of work, and once more this new work is released to the shop. Once more, the leadtime grows and even the previous extended leadtimes are missed ...
The damage includes WIP, the need for progress chasers, inflexibility and uncertainty.
The Virtuous Circle is simply to reverse matters! .. decrease the official leadtime.When the official leadtime is decreased, customers will desist in sending in their orders for a period - ie orders will dry up for a time, enabling the manufacturer to complete the work he has on hand.
4.3 Input/Output Control The purpose of input/output control in the job shop is to control queues, and thereby simultaneously control manufacturing leadtime, work-in-progress and the utilisation of machines. The relationship between queue, WIP and machine utilisation for intermediate (ie non-gateway) work centres is described in 4.3 below. The question of target queues at intermediate work centres is also discussed there. However, the assumption made there about the rate of arrival of jobs at the work centres is not valid for gateway work centres, since the release of jobs to the gateway work centres is directly controlled by the production control manager through the input/output table. Consequently, queues serve no purpose at starting operations on the shop floor and should therefore be non-existent or, at least, very small.
In describing the terminology of jobs at work centres, the analogy can be used of an ordinary funnel into which is being poured a liquid. The input (to a work centre) is the liquid being poured into the funnel at the top. The output (from the work centre) is the liquid actually discharging from the funnel at the bottom. Meanwhile, the queue is the measurable depth of liquid standing within the funnel. The load is the total quantity of liquid in the funnel itself. The capacity (of the work centre) is a measure of the funnel's mouth at the bottom, which is enabling it to discharge the liquid. Continuing with the analogy, to maintain a level of queue (denoted by the level of liquid in the funnel), input and output must match. Input can be increased only if output is increased correspondingly. The usual format of the input/output table is a simple Table for a given work centre. The figures within such a table are standard hours of work and the time periods are weeks. Typically, four past weeks (-4 to -1) and the current week (0) are shown, and, perhaps, four future weeks (+1 to +4). It may be supposed that the planned inputs for Weeks 0 to +4 have been obtained by averaging the total materials planning load for these five weeks. Actual inputs (Weeks -4 to -1) will have been obtained from transactions relating to completed work at this work centre, submitted to the shop floor data collection system. The planned outputs for Weeks 0 to +4 are the week by week totals of standard hours the shop floor supervisor or foreman has specifically agreed to produce. Every possible effort must be made in input/output control to avoid using an output rate based on wishfulness rather than on cold calculation. A maximum output rate often used is an average of recent output achievement, referred to as the demonstrated output, or demonstrated capacity. The actual outputs will again have been obtained through the shop floor data collection system. (If the shop undertakes relatively lengthy jobs, data should be collected and included for partial job completions.) The queue in standard hours at the end of a period is calculated as shown in the following equation (set out vertically):
Queue at the end of this period =
Queue at the end of the previous period +
Input over the period -
Output over the period
For example, suppose the following data pertained: Queue at the end of Week -1 = 8 hrs, Planned input in Week 0 = 106 hrs; Planned output in Week 0 = 104 hrs; Planned queue at the end of Week 0 = 8 + 106 - 104 = 10 hrs.
It is also normal to associate with an input/output table figures for a cumulative input variance and a cumulative output variance. The cumulative variances are a quick indication of whether a work centre is running consistently ahead of plan or consistently behind. Cumulative variances are typically also used as "control limits", associated with each work centre. If a limit is exceeded (say, ±20 hours), the computer system draws the fact to the production controller's attention. Control limits should be tightly set so that potential problems can be addressed early on, before they become crises. The normal way of scrutinising the input/output table is, first, to examine output - is the work centre executing its capacity plans? If input is as planned but output is down, there may be a problem.
4.4 Queues at Intermediate Work Centres
Intermediate work centres are work centres other than the gateway work centres. That is, they are the work centres receiving work sent on from another work centre on the shop floor.
In many manufacturing environments with a regular pattern of work - ie not in engineering job shops - the queue element of job leadtimes may be readily controllable through careful work scheduling. In the job shop, with manual scheduling, synchronisation of job arrival times and departure times at intermediate rather than gateway work centres may be difficult to accomplish. Adherence to such a schedule even if formulated may be impossible. Nonetheless, at intermediate work centres, even though there is some loss of control regarding work centre input because of the seemingly erratic arrival of work from other the work centres, input/output reports and plans should still be prepared just as they are for the gateway work centres. The main values of the I/O table for an intermediate work centre are as a monitor of the target queue there (see below) and as a monitor of the general equilibrium between input and output. In this last regard, if the table indicates the work centre is starting to run out of work, the shop floor supervisor may be able to expedite jobs due into it from other work centres. If it seems that it is beginning to fall behind, it may be possible to increase its capacity by some means such as by the authorisation of overtime or by the transfer into it of additional labour.
Although the simple formula from above still, clearly, applies (ie previous queue + expected input over the week - expected output over the week = new queue), it may be preferable to tabulate jobs rather than hours of work.
Although a degree of input and output management is thus possible, the pattern of arrival of jobs into the intermediate work centre remains liable to unevenness, so justifying the maintenance there of a queue to ensure a steady rate of work and the satisfactory utilisation of physical resources. The question therefore is: how much queue should there be?
From theory, if the average size of the queue at a machine were to be plotted on the y axis and the percentage machine utilisation were to be plotted on the x axis, then (a) when machine utilisation is low (say, 60%), it is found that queue size is also small (say, 1 job); (b) when machine utilisation is high (say, 90%), queue size is found to be very high (say, 15 jobs). (c) the ideal compromise is when machine utilisation is about 50 - 60%, corresponding queue size is about 3 or 4 jobs.
It will normally take a considerable period before the ideal queue is established. (The production controller is respinsible for the smooth loading of the shop; the foreman is responsible for achieving plan. As well, remember the existence of suddenly small queues must be explained to the workforce - they may well suspect that the company is running out of business, and might start slowing down.)
4.5 Final Order Release (and Despatching Rules)
There has been very extensive research into despatching rules over the past 30 years employing research through computer simulation. (Nowadays, the focus of academic and manufacturing research has switched to Advanced Planning Systems - see below.)
Candidate jobs for release to the shop floor and actual manufacture are clearly those in the planning schedule which have reached their start dates, commensurate with the total workload being in accordance with planned input as above. Even though the company may not be employing closed-loop MRP, the same steps are followed as those described for "open orders", or "scheduled receipts", ... that is, the availability of the required components is checked on the stock records and the stock quantities involved are transferred on the stock record to 'allocated' status. Picking lists for the physical marshalling and transfer of the parts to the shop floor are prepared for use by stores staff. Tool availability may also be checked at this stage by reference to a tools database. Advance notice of the requirement for job set-up may be conveyed to the shop floor. Shop documentation includes a route card and an id card. A release priority system should be operated when I/O control prevents all jobs from being released on their start dates. (Often, the priority system is based on commercial principles - that is, jobs for more important customers are released ahead of less important ones.)
Note that to sustain an even flow of work into the shop according to the I/O plan, it will also be necessary to release jobs earlier than their start dates as well as to delay them. If si, the same scheme as above is used for prioritising such jobs for early release, although, before prioritisation of them takes place, clearly the earlier availability of the components needed for the manufacture must be checked.
4.6 Job Despatching Rules (Job Prioritisation Rules)
As stated above, the first point to make is that there has been very extensive research into 'despatching rules', or job prioritisation rules, over the past 30 years, using computer simulation. However, it is a universally acknowledged fact that the key to superior shop floor performance is the elimination of causes of variance in performance by the use of Rough Cut Capacity Planning (RCCP), Fast Set Up times (Through SMED), Total Productive Maintenance (TPM) and the complete elimination of defective output through Spatistical Process Control (SPC).
The Despatching Rules below are divided into (1) Dynamic Despatching Rules, which require data feedback pertaining to the status of the jobs being prioritised, (rules 4.6.1 to 4.6.3) and (2) Static Despatching Rules, which rely purely on information about the jobs themselves, with no data required from feedback (Rule 4.6.4).
4.6.1 The Slack Time Rule (dynamic)
Slack-time is defined as follows:
(time due minus time now) minus (processing time still remaining)
Note that the first part of the expression ((the time the job is due minus the time now)) refers simply to elapsed time, and the second part ((processing time remaining)) refers to machine set-up time and machine running time. Importantly, note further that job queue time, job wait time and job move time are not directly incorporated in the rule.
Suppose that the time now is Day 1 and that a job is due on Day 10, and that there are 5 days of work on it remaining to be done. Then the job's slack-time is:
(10 - 1) - ( 5 ) days
= 4 days.
In other words, there are 4 days in hand with regard to this job. Greatest priority rests with jobs with the smallest slack-time. If there are three jobs A, B and C with slack-times of 1 day, - 3 days and 4 days respectively, they would be ranked in priority order B, A, C. Slack-time is a dynamic rule.
4.6.2 Slack Time per Operation (dynamic)
If two jobs each had slack-time of 4 days, but the first had only 1 operation remaining and the second had 2 operations remaining, it would seem that the second job should have a higher priority. Each operation to be gone through brings its own delay, especially if there is a queue at the machine. Slack-time per operation is therefore defined as:
slack-time / number of operations remaining
Greatest priority rests with jobs with the smallest slack-time per operation. For the two jobs with slack time of 4 days but with 1 and 2 operations remaining respectively, the slack-time per operation quotients are 4.0 days per operation and 2.0 days per operation. The second job therefore has the higher priority. Slack-time per operation is clearly also a dynamic rule.
4.6.3 Critical Ratio (dynamic)
Critical ratio is the basis of a further dynamic rule, and is defined as follows:
(time due minus time now) / leadtime remaining
As with the two slack-time rules, (time due minus time now) in this expression is simply elapsed time. But what is different is that the denominator, the "leadtime remaining", is the sum of all five of the elements which go to make up a job's remaining duration, given previously, not just two of them. In applying the rule, jobs with the lowest critical ratios have the highest priorities.
The incorporation of queue time in the critical ratio rule gives it an advantage over slack-time per operation. The reason for this is that the standard, or target, length of queue decided for each job, as implied from the target queues discussed previously, will contribute to the determination of the job's start time and due time (along with set-up, run, wait and move). It is because of queue's contribution to the establishment of the start time and due time that the critical ratio rule is claimed to be superior to slack-time per operation. In operation of the rule, if the job is on schedule to meet its due time, so that, among other things, its target queue is being maintained, the job's critical ratio will be 1.0. If the job falls behind (critical ratio less than 1.0), the rule will cause its priority to increase until it catches up. When it has caught up, the balance of the elements set-up, run, wait, move and, of course, queue will have been restored. If the job's progress is too fast (critical ratio greater than 1.0), the rule will cause its priority to decrease until, eventually, the critical ratio is back to 1.0, thus again restoring the queue element of the job's remaining duration.
4.6.4 Shortest Processing Time (SPT) (static)
Under the SPT rule, jobs are prioritised in order of their total processing times (ie set-up time plus running time), jobs with the shortest times having the highest priorities. Thus consider three jobs A, B and C with processing times of 10, 5 and 1 hour. The SPT priority order is C, B, A. The processing and queuing times are as follows
Job C : (1 hr) queue 0 hrs (first job) ... Job B : (5 hrs) queue 1 hr (must wait for Job C) ... Job A (10 hrs) queue 6 hrs (must wait for Job C and Job B) ....... total time 16 hrs processing + 7 hours queuing. Total 23 hrs.
This should be compared to the queuing times in which the jobs are despatched in the order A, B, C, the reverse of SPT.
Job A : (10 hrs) queue 0 hrs (first job) ... Job B : (5 hrs) queue 10 hrs (must wait for Job A) ... Job C (1 hr) queue 15 hrs (must wait for Job A and Job B) ....... total time 16 hrs processing + 25 hours queuing. Total 41 hrs.
Not surprisingly, SPT consistently outperforms other despatching rules in simulation studies in which jobs are considered to arrive in no particular sequence, when measured by the yardstick of minimum job lateness. However, there are two warnings. The first is that SPT is liable to give poorer performance in a controlled environment in which adherence to the schedule is important. The second relates to long-running jobs. In the standard treatment of queues in queuing theory, the assumption is made that "the calling population", or members of the queue, are dealt with on a first come, first served basis. Because this is not the case when jobs are being despatched by SPT, long-running ones will be pushed to the bottom of the priority list in every queue they join. To overcome this problem, if SPT is used by the foreman as a way of determining the despatch priority of jobs, some means must be found of "rescuing" the lengthy jobs to prevent them getting stuck in the system. There needs to be an additional subsidiary rule, such as if a job has been waiting longer than so-many days, it is automatically released regardless of SPT. One company devised a rule that from Monday to Thursday, they would release jobs using SPT, and on Friday they would use FIFO (First in First out). FIFO would automatically ensure that jobs which had become 'stuck' would be released.
5.1 The Use of Siumulation in APS Scheduling Engines
A common way in which computers might be employed to solve a problem such as finding a 'best' schedule for several jobs to be scheduled through a number of factory work centres is through the use of simulation. Simulation involves the creation of a 'model', or representation, of the schedule or environment in a so-called 'simulation language', and allows the user to track progress through time. That is, the computer program might systematically examine one possible schedule after another, and, after all possibilities have been examined, indicate to the user which one of those examined was overall the 'best' (defining 'best' in some way, such as shortness of time). However, the notion that the computer can examine every possible schedule, and simply pick out the best, in thwarted when we learn that the general expression to derive the number of possible ways of scheduling n jobs through m machines is (n!) **m - that is, in words, factorial n to the power of m. Such a potentially very, very large number of possible ways presents insurmountable practical difficulties for even the most powerful of computers in solving the problem of scheduling jobs through machines by straightforward mathematical means . For example, the "10 ×10" problem, - ie the scheduling of 10 jobs through 10 machines - yields some 10 to the power of 70 possible solutions, an astronomical number (ie several billion billions) .
5.2 Discrete Job APS Systems
To illustrate a scheduling procedure, consider three jobs, Job 1, Job 2 and Job 3 below. All three jobs utilise work centres A, B and C. The representation below illustrates the sequence in which the jobs require each work centre, and the time the job variously occupies the work centre. (Every repetition of a capital letter denotes one hour's occupancy.)
Job 1 : A A B B B B B C C C C C C D D
Job 2: B B B C C C C C C D D D D A A A
Job 3: A A A A B B B C C C C D D D
To simulate a schedule for the three jobs through the four work centres, two rules are established. The first is that the 'events' which constitute each job (ie the loading/occupancy of a particular job on a particular work centre) are dealt with separately, on an individual and system wide basis. For example, although Job 1 must go on four work centres, the job is considered as consisting of four distinct, separate parts (Job 1 on WC A, Job 1 on WC B, Job 1 on WC D ....), so that altogether the completion of the scheduling consists of 12 separate events when the demands of all three jobs are taken into account. Taking into account all 12 events is done on a system wide basis - that is, all 12 requirements are considered together. We do not simply deal with Job 1 first, only considering the needs of Job 2 when Job 1 has been dealt. The second rule is that the various jobs are despatched according to rules relating to the situation at the time. The term 'despatched' means scheduled through a work centre. The rules relate to the overall position at the time. For example, there is a rule that Job 1 must go to WC A first. When this has been done, the next time Job 1 is considered clearly it does not have to go through A, because by that time this requirement has already been fulfilled.
The Figure below shows the horizontal loading of the 3 jobs, using so-called "discrete event simulation". That is, the four operations relating to Job J1 are not simply loaded one after another onto the four work stations. Instead, here, Operation 1 of J1 is already loaded on W/C A, and is completed at the end of Period 2 and Operation 1 of Job 2 is already loaded on W/C B. At the end of the second time period, Operation 1 of Job 1 has finished on W/C A. This completion constitutes the first "event" following the commencement of the schedule. Consequently, the simulation model now intervenes and all loading opportunities are examined on a system-wide basis at this point, to discover what event should proceed next. The next event to commence is the loading of Job 3 on W/C A, since W/C A has now become free. The next event is the completion of Job 2 on W/C B. When this occurs, the system finds that Job 1 is waiting and requires W/C B. Also, that Job 2 requires W/C C, and the work centre is available. And so on, with the system examining the situation at the completion of each event. 4. RULE-BASED APS SYSTEMS 1. Lateness Rule: if Days Cover 2.0, Score = Days Cover Else Score = Days Cover × 100; 2. 1st Set-Up Rule : If the Previous Job = same product filling (ie rice or noodles) Score = Score minus 50; 3. 2nd Set-Up Rule : If the Previous Job = same flavour (beef or curry) Score = Score minus 100. The points shown in Table 1 relate to the initial points allocation, made immediately after considering the number of days cover of each on the six products. (Prior to this starting situation, it is assumed the production lines are clean.) Job Customer Days Cover Filling Flavour Initial Points Allocation J1 A 3.6 Noodle Beef 360 J2 B 2.4 Rice Beef 240 J3 A 3.3 Noodle Curry 330 J4 C 3.9 Rice Beef 390 J5 C 3.5 Noodle Beef 350 J6 B 1.9 Noodle Curry 1.9 Table 1 : Rule-based Scheduling : Scores assigned after Initial Assessment Job J6 is to be despatched first, having the lowest score. If J6 is first, it is possible to apply the second and third rules on product filling and product flavour to evaluate job scores to determine the job to be despatched second. Thus after this second round of scoring, the position is as shown in Table 2 overleaf. Job Customer Days Cover Filling Flavour Revised Points Allocation J6 B 1.9 Noodle Curry n/a J1 A 3.6 Noodle Beef 310 J2 B 2.4 Rice Beef 240 J3 A 3.3 Noodle Curry 180 J4 C 3.9 Rice Beef 390 J5 C 3.5 Noodle Beef 300 Table 2 : Rule-based Scheduling : Scores assigned after Second Assessment The second job despatched in the Schedule is thus J3 with 180 points. Because of the specifications of the four jobs remaining, whatever the next job is to be, there must be a change of filling or a change of flavour, or both. This being so, a work centre rule is invoked whereby a plant cleandown is scheduled. Consequently, the despatching priority of the third job reverts purely to considerations of lateness and the points in Table 3. Of the four remaining jobs (ie not despatched) in Table 1, the one with the lowest score is J2. This is therefore despatched third, and, on that basis, the points of the three remaining jobs are evaluated and as given in Table 3. Job Customer Days Cover Filling Flavour Revised Points Allocation J2 B 2.4 Rice Beef n/a J1 A 3.6 Noodle Beef 260 J4 C 3.9 Rice Beef 240 J5 C 3.5 Noodle Beef 250 Table 3 : Despatching of the Fourth Job in the Snack Foods Example From Table 3, the fourth job to be despatched is J4. Since there is now bound to be a change in product filling in proceeding to the next job, the cleandown rule is again invoked and the point scoring position of the two remaining jobs J1 and J5 reverts once more to Table 1. The final despatch order of the six jobs is J6, J3, J2, J4, J5 and J1. Note that the use of a weighting scheme such as above may be unnecessary if sequence is governed by simple technical factors, such as, say, colour. To give effect to the rule, a colour code of each batch to be manufactured must be carried with the job data so that the APS can correctly sequence them. In some instances, best sequences require the application of combinatorial optimisation to find an optimum sequence. 5. DISPLAY OF THE DISCRETE JOB SCHEDULE The centrepiece of discrete job APS systems after the creation of a schedule of work is the output on the VDU of a display of the schedule in "planning board" format on the VDU. Because the VDU screen cannot accommodate the entire schedule, the display allows scrolling, panning and zooming. To give the attendee an idea of the format and style of this output, such a display is reproduced on the coloured printed page on the next page of the notes. The factory's individual work centres to which jobs are assigned are shown on the left of the screen. Because space is limited, it is necessary to scroll round the bottom of the screen to see all of them. The scheduled jobs are denoted as coloured horizontal bars over time opposite the work centre to which they have been assigned. (Red denotes a job which is late, green ahead of schedule, blue on time.) Since only a limited time span can be viewed, provision is made for panning across the screen to see further into the future. In addition, it is typically possible to click on a job so as to break out more detail - ie zoom in. The detail from a zoom is seen at the bottom of the screen overleaf. e d i t i n g t h e s c h e d u l e A reason for wishing to edit (ie change) a schedule just produced automatically by the APS system might be that it fails to take account of the "fine detail" of reality. In turn, this could be because it is not felt worthwhile to get a precise result through the software alone. Alternatively, editing and re-running the system may be simply a matter of "what if?" - exploring alternative possibilities in order to make a good schedule better). One way of editing is through the work centre rules (see below). That is, provided the planner is familiar with the computer code used, the logic of a rule can be changed pro tem in order to force a particular action. Another method is to substitute an alternative rule / program in place of the standard one. Yet again, rules may operate by taking into account "weightings" placed on various features such as stock holding or timeliness as explained above, so that these weightings can be varied to explore new possibilities. Editing, finally, can be used to evaluate real production differences, such as the possibility of manufacturing non-standard batch quantities or following non-standard production routes (see below). Besides re-running with different rules, the APS will also allow changes to be made direct via the VDU screen. Reasons for doing so might include: * forcing a job to be done at a particular time; * holding a work centre, or a machine in a work centre, open for a particular job; * moving jobs to a different time or from one work centre to another graphically on the VDU screen with a mouse. 6. WORK CENTRE RULES 6.1 BASIC RULES A predominant practical issue in using an APS system is the system's ability to take account of the different rules which relate to each factory work centre at which jobs are to be processed. Rules may cover such factors as choice of operation date, cycle time, set-up time, priority, resources, waiting time, operating float and more besides. The rules relating to many of the work centres in a factory may well be straightforward. For the rest, however, as we have seen, there can be great diversity and great complexity. 6.2 SPECIAL RULES The provision of a programming facility in this way will be in the form of a type of computer programming language. The languages provided by vendors are typically C, BASIC and FORTRAN look-alikes. The user writes the program, or sub-routine, which is then incorporated within the software and is "executed" whenever a job is under consideration for that work centre. In order to use the facility, it will be necessary for the persons who are to be in charge of the APS System to attend the vendor's training course. 6.3 DOCUMENTING WORK CENTRE RULES 7. OTHER DATA FILES BILL OF MATERIALS STOCK RECORDS THE ROUTINGS FILE The task of setting up and maintaining the routings file usually falls to the production control manager. To support an APS system, the accuracy of the data held must be 100%. Considerable difficulty can be experienced in reaching and maintaining this level, however. Manufacturing operations staff usually know informally "what happens", but cannot always be relied on to record it accurately or may be reluctant to notify changes that have occurred. 8. SCHEDULE ADHERENCE It can be guaranteed that the APS system will fail if the schedules produced for the execution of work in the plant are invalid in even the smallest respect. Schedule validity is the absolutely essential sine qua non of APS. Reasons for non-adherence can be any one or any combination of the following: (i) Modelling inaccuracy (technique and rules); (ii) Modelling inaccuracy (basic data); (iii) Non-adherence elsewhere; (iv) Not trying ! 9. DATA FEEDBACK AND APS NAPS135 The question of feedback under APS is quite different from MRP and a normal shop floor environment, especially a job shop. The fineness and detail of manual planning in such circumstances is so broad, that progress must be reported back to keep track of job and machine status, and the progress of the schedule. In a Closed-Loop MRP system, feedback is an essential part of the system. Under an APS, however, feedback is not required since the progress of work is in line with the APS plans for it ... feedback would merely confirm what was already anticipated. 10. DISCRETE JOB APS AND OPERATIONS MANAGEMENT Overleaf - example of the use of APS and simulation. The manager of the shop floor concerned wished to verify the timeliness of his current workload (shown in the middle horizontal bar ... ie 29 jobs will be late) , and then examine the effect on schedule of (a) putting on a Saturday shift (top horizontal bar ... 4 jobs will be late) and (b) putting on additional operators on the day shift (bottom bar ... 15 jobs late). Note the buttons Operation, Cost and WIP on the top of the screen. Clicking on the Cost button will get the three alternative costs of the action simulated. 11. THE APS SOFTWARE SCENE The Internet would seem a more useful source than literature, even including The Software Users' Yearbook. Specify "APS" and "Manufacturing Scheduling" at the Google Search Engine (http:// www.google.com).The picture of software development is that APS products have emerged from a number of directions, as below. Category (i) seems most prevalent: (i) ERP/MRPII vendors, who have bought out or merged with an independent company selling a finite scheduling software package; (ii) ERP/MRPII vendors who are developing their own finite scheduling capability in-house; (iii) Established Finite Scheduling package vendors who are developing their own ERP/MRPII functionality. It should be noted that the view of ERP/MRPII vendors is that they will not survive unless they acquire a finite scheduling/APS facility ... in other words, that closed-loop MRP is dead. The main difficulty that companies from Category (i) above have experienced in merging their products is that of achieving data integration. There have been substantial underestimations of the time needed to integrate, although the benefits to the user are considerable on its completion ... one set of files and one means of updating/accessing data regardless of the APS module involved. General advice is difficult to formulate. The finite scheduling package companies from the late 1990s spent very many years developing their algorithms. One has to wonder, therefore, whether ERP companies in Category (ii), which include the mighty SAP, are likely to come up with products as good as Category (i). Chesapeake Decision Sciences MIMI (originally written for the oil industry); The Baan Company, Baan SCS Planner (acquired Berclain and CAPS Logistics); J.D.Edwards Schedulex (buy out of Numetrix; Schedulex is a continuous flow system); Logility (Logility Value); PeopleSoft (Red Pepper) Integration began in 1997; Synquest; Made2Manage Systems (acquired Bridgeware in 1998); MAPICS - joint development of MAPICS XA Wisdom with Symix Systems; SAP R/4 SCOPE I2, RHYTHM (Rhythm also licensed to ORACLE) Manugistics (Rhythm) Provisa, from Lanner Group in Redditch (contact Mike Ellerby, 01527-403400)Preactor; Preactor International, Chippenham, Wiltshire (ph 01249-650316) For Official Use: NSFC034F, NSFC512F, NSFC501F, NAPS181F, NSFC553F, 1st Display Screen, NAPS201F, NAPS211F, NSFC042F, NSFC047F, NAPS216F, NSFC555F, NSFC049F,2nd Display Screen,NSFC042F
![[Homepage]](butt1on.gif) |
![[About GMCS]](butt3on.gif) |
|
 |