Sergey Peterkin, Reitstep.
MRP does not work ... What is the alternative?
Synchronous Planning and Optimization
Introduction
The long-term efforts of the ERP (Enterprise resource planning) class information system suppliers in Russia have finally been successful: the leadership of a significant share of domestic enterprises are confident that they need such systems, since they can (as stated) solve the problems of production planning and supply. One by one, the leading Russian developers announce the availability of planning and production management functions in their systems that meet the MRP-II "standard" (Manufacturing Resource Planning -, planning of production resources). Enterprises that have introduced ERP systems and are trying to manage production with their help are even starting to appear in various industries. However, planning with their help is for some reason more difficult than before. Why? Experienced consultants state that the reason is that "... the wrong production culture, inaccurate specifications and technical processes, constant design changes, non-operational inventory level tracking ...", and, oh, horror, "... there is no aggregated strategic and / or planning master” .
All this is true, but how many have wondered: are all Western management methods (especially 30 years old) as good as presented? It’s one thing to plan the work of a Western enterprise of 50-100 people, working long and successfully in a stable environment, it’s another thing to “spin” in Russian conditions, managing an industrial giant standing up on its feet, or a young developing plant, many of whom would envy the reaction speed to changing demand Western competitors. Perhaps the roots of many unsuccessful ERP system implementations at Russian industrial enterprises lie precisely in an attempt to use an imperfect algorithm for planning, which is effective only when “everything is good”?
Nowadays, few people will be surprised or interested in a detailed description of algorithms for calculating needs and planning. Nevertheless, the author took the liberty of once again briefly describing the MRP planning algorithm and analyzing why it is not effective in many of our real enterprises in practical application. Moreover, this is not just a rejection of what is well known to all, and was proposed, in particular, once by the author. The article proposes another scheduling algorithm devoid of the traditional flaws of MRP II. This method can be applied to any enterprise with discrete production or continuous production, which can be reduced to discrete.
Resource Planning for MRP-II. What's wrong?
Let us briefly recall the principles of calculating needs according to the MRP-II algorithm.
Initial data:
- Production plan / forecast / customer orders.
- specifications and operational processes of manufactured products,
- level of stocks at warehouses and places of storage,
- expected revenues from suppliers and from production, expected shipments to consumers,
- work sites with a work schedule.
Planning
Based on information about the demand forecast or production plan, product specifications, gross needs for materials, components and intermediate assemblies are determined.
If there is information about work in progress, stocks of materials and components, materials on the way, the net needs for produced and purchased parts are determined in quantity and time, by the planning method back without taking into account the actual load of production.
If there are technological routes indicating the operational complexity and production rates, as well as taking into account information about the equipment operating schedule and production tasks performed, the production load is calculated, and, in the event of an overload, re-planning is performed.
In the event of any anomalies, for example, as a result of the automatic planning procedure, the start dates for production / purchase are in the past) manual rescheduling for those positions for which the system generates messages on exceptions.
The algorithm looks quite holistic and logical. And in theory this is true. If we take it apart using the example of production of a real product (flashlight, shovel, table, bicycle, computer, etc.), there is no doubt whatsoever: it suits us. But wait a moment! In order to produce a flashlight or a shovel, assemble a computer from the main units, an automated calculation of needs can be performed using Excel tables. Such enterprises do not need a complex automated planning system. MRP, on the other hand, needs calculation functions tend to use Russian enterprises producing fairly complex products.
And in this case, in practice, the MRP-II calculation cannot help them in any way. Why? Because…
1. MRP "knows nothing about today . " The calculation of the start dates is always made back from the date of the planned need. In the event that the period to the planned need (the time from “today” to the date of shipment) is less than the total lead time of the planned product, the system plans the start of work according to the plan in the past.
True, in this case, the system will generate messages for exceptions . However, in practice, there may be so many of them that there simply may not be enough for their "analysis" of the real time allotted for planning. Consider the average Russian industrial enterprise with a total number of approximately 500 people. Suppose it takes on average several dozen customer orders per day and produces products “more complex than bicycles,” that is, having a specification consisting of several tens of lines. These conditions are met by most Russian industrial enterprises in almost all industries. For him deviations from ideality, the number of exceptions after each MRP run will be in the hundreds of records. “Happy” users of systems with the MRP-II planning algorithm are well acquainted with this fact. The bottom line: the plan is unrealistic .
2. MRP plans without real resource utilization . When planning back, a standard fixed lead time is used, i.e. total non-operational production time. The fact that any work area may be busy at this point in time is not taken into account by the algorithm. However, there is a way out: launching the CRP ( Capacity Requirement Planning ) function, with the help of which it is most often possible to "expand" bottlenecks. In this case, as a rule, the production terms of all parts planned by the system also change. Those. after starting the CRP function, you must restart the MRP, which again will form a production plan without taking into account today and limited resources. And so - in a circle ...
In theory, such consistent approximations in the formation of plans are quite feasible. However, as a rule, in the practice of Russian enterprises, when we solve planning problems for dozens or more products having, as a rule, more than five levels of nesting and consisting of tens - hundreds - of thousands of units, parts, components, as a result of such planning, anyway, it turns out to be unrealistic plan. Not to mention the fact that when calculating capacity utilization (CRP), automated planning is not possible taking into account alternative routes, and the calculation itself (CRP or MRP calculation) at final computing capacities will last more than one hour. Of course, the calculation speed depends on the software and computing power. But on the basis of practical experience, the author can conclude that for a medium-sized machine-building enterprise with a database of 30 thousand products (which is not so much for our machine-builders), components, parts, materials and components, manufacturing products having 5-7 levels nesting, MRP and CRP calculation on average will take about 4 hours each.
3. MRP cannot solve the issue of changing the timing of the production of a particular head product in case of problems with any component of the lower level, because knows nothing about the root source of the need. This drawback is inherent in the calculation algorithm itself. MRP calculates requirements from top to bottom, level by level for all database products (see Figure 1). Because of this, in most cases it is not possible to correctly determine the source of need.
As a result, for MRP all orders are “gray” (see below). Those. in case of problems in the production of a component included in several products, it is not possible to determine, for example, which client should be informed of the postponement of the order.
In connection with the above figure, consider an example. Under the plan of shipments for customer orders, it is planned to purchase components. When planning according to the MRP algorithm, it turns out that even taking into account the safety stock, the purchase of several batches of the necessary components had to be carried out 5 days ago. Due to the fact that the needs calculated by the MRP algorithm are depersonalized, it is not possible to determine the date of which of the orders received must be changed.
What is the alternative?
An alternative exists in the form of a relatively recently appeared and already adopted by many ERP providers scheduling algorithm used in APS and MES systems that were once positioned completely apart (APS - Advanced Planning and Scheduling System - synchronous planning and optimization - STR, MES - Manufacturing Execution System) . Now the APS algorithm is considered a breakthrough in the practice of managing industrial enterprises, comparable to the development of the MRP-II algorithm more than three decades ago. “... obliged to a large extent by the development of computer technology, APS offers not only the fastest, but also the best answers to the eternal questions of an industrial enterprise:
- what can we produce;
- when we can ship;
- how can we better use the available resources to meet demand ... "
But this is all advertising. So say the representatives of almost any company operating in the field of information systems. However, few people can reasonably explain how the APS algorithm can solve the planning problem and how it should be used. Most often, when selling a “black box” with a magic button is offered. But, according to the author’s deep conviction, not understanding the algorithm of the system’s operation, it cannot be used to solve the planning problems of a modern Russian enterprise.
And, before proceeding to the description of the open source software algorithm, we note that APS is not:
- only by means of “quick and subtle planning”,
- fast version of MRP-II,
- planning optimizer or the second stage of implementation of the MRP-II / ERP system,
- scheduling algorithm only for perfectly perfect data,
- methodology only for advanced enterprises and enterprises working on order,
- APICS standard.
APS Resource Planning. How it works
Currently, there are several commercially available APS algorithms: network models, simulation models, mathematical modeling models (modeling using neural networks, using linear programming, etc.), network optimization, etc. More details about the advantages and disadvantages of these algorithms and the limits of their application are described in the author’s book “Just In Time for Russia”. Below, in summary, an algorithm for calculating plans is presented using the most widely used optimization network algorithm.
Initial data:
- production plan / forecast / customer orders.
- specifications and operational processes of manufactured products,
- level of stocks at warehouses and places of storage,
- expected revenues from suppliers and from production, expected shipments to consumers,
- resources and resource groups.
The APS algorithm uses a real enterprise model that is fundamentally different from MRP, i.e. not abstract and rigidly defined work centers (sections) with identical machines in it, which, as a rule, must correspond to work sections and cannot be “spread out” across several workshops or the entire enterprise, but the real resources of the plant, i.e. people, machine tools or equipment groups, areas, tools and equipment, etc. (fig. 3). In the framework of MRP II, an adequate description of a team of workers simultaneously performing work on several work sites leads to the fact that for work centers in the ERP system the number of workers is determined as 1.5; 3.17 people, etc.
Unlike classical MRP, the resources of the APS algorithm (network optimization model) can have additional qualitative attributes, such as qualifications of workers or their rank, characteristics of equipment or tools, for example, “old”, “new”, where “old” equipment determined for example by a real accuracy group or processing speed. Also, each work schedule can be defined for each resource, and resources can be combined into resource groups, which is an approximate analogue of the work sites (work centers) of the MRP model. Moreover, resources can be combined into resource groups that also have a very real name (“turners of the 2nd category”, “turners of the 6th category” - see above. If there is no GOST-standard discharge grid for the enterprise personnel, its own , which really exists in every enterprise. One example is work experience.
An example of a process definition in a new enterprise model is as follows.
Product A:
- operation 050 "precision metalworking"; necessary resource groups: precision processing and a turner of the 6th category , necessary resources: a worker of the 6th category, machine No. 1 .
- operation 055 "normal metalworking"; necessary groups of resources: normal processing, a turner of the 2nd category , necessary resources: a worker of the 2nd category OR a worker of the 3rd category OR a worker of the 6th category, machine No. 1. OR machine number 2 .
With such a definition of resources, there is no need to determine alternative routes, it is enough to determine the possible groups of resources, the components of which will be sequentially sorted during planning. At the same time, their simultaneous availability in time will be checked, for example, for day or evening shift
Planning
Formation of a production or shipping plan for customer orders using the APS algorithm will be as follows.
Step 1. Calculation of material requirements. Determination of the earliest start date, procurement date and shipment date / finished product release date
The calculation is performed according to the MRP algorithm, but with two significant differences.
1. The calculation is not for all objects of all database products of the same nesting level, but for all components of each “head” product, the need for which at a certain date can be determined by the production plan, customer order or demand forecast. Those. first, the first product is taken (in accordance with a certain selection rule), and calculated “from top to bottom, from the future to the past,” then the second, etc. until for all products the requirements for materials will not be calculated (see. Fig. 4). Looking ahead, we note that it is due to this that it is possible to calculate the expected completion dates of each specific customer order or production task.
2. In the event that the dates of the start of production / purchase of some parts / materials are in the past (the system “runs into” today - see Fig. 5), the planning method changes for. Elements are sequentially taken with the start dates of the work, earlier today , and, using the planning method forward from today , a new date for the completion of production or a delivery date from the supplier is determined.
Step 2. Determining the start / end dates taking into account the loading of existing resources.
“Push”, i.e. bottom-up planning, from the past forward.
1. After determining the start date of the production / purchase of the very first material / part in the production chain, using the planning method forward from the earliest production / purchase date, sequentially, starting from the first, the start and end dates and times are determined, each operation taking into account the limited loading a resource from a given resource group. If there is not enough free power in a given time range for this resource (all workers / units of equipment are loaded, or tools are unavailable):
1) according to the rules set in the system, from the group of resources defined for this operation, the available resource is selected, the operation is redefined for it,
2) if there is no such resource, all operations of this part are sequentially transferred to the nearest free temporary “window” of the first free resource. Thus, the start / end dates of each operation for the production of a given component or finished product are determined (Fig. 6).
2. Actions according to claim 1. performed for all parts by levels, starting from the lowest. Thus, the start date of the production of the next one is determined by the product technological chain (the latest end date is taken).
3. Actions in paragraphs 1 - 2 are carried out throughout the structure of the product from bottom to top, until the final date of readiness of the final product is determined.
Step 3. Reduce total production time.
“Pulling”, i.e. top-down planning from the present to the past.
By the date of the final completion of work defined above, the dates of production / procurement of all parts / materials of the product in question are “pulled” taking into account the load of resources, i.e. for parts / materials of the lower level, the start / end dates are redefined to later (in the general case).
After completing the planning cycle for one item, the same planning is performed for the next item from the production database. At the same time, the requirements for materials reserved for the first product are consumed only if there is an excess (production taking into account the minimum lot), and when calculating the time taking into account the load of resources, the planned load of the new calculation is “superimposed” on the one created by the previous calculation. Next, planning is carried out for the next product, etc.
Depending on the software implementation of the algorithm, the calculation can be performed both for one product (for a specific production task or for a customer’s order) and for all products in the database, either with a general zeroing of the load and preservation of the reserved materials.
In the "single" planning mode (necessary for the enterprise working on order), each order is calculated top-down, superimposing its needs on previously planned ones.
In the general re-planning mode, for all the planning horizon, all previously planned needs are reset (except for the tasks put into production), a sequence of objects for planning is selected (plan, production tasks, insurance reserve replenishment needs, demand forecast, customer orders, etc.). Further, according to the algorithm described above, one by one the selected products are planned sequentially.
With such a rescheduling, the completion dates for tasks previously planned but not put into production can change (“move forward”). However, the overall plan, however, will be real.
Optimization
The second part of the method is production optimization. Production optimization is carried out for planned and / or already launched production tasks, taking into account all the bottlenecks. Moreover, in most systems, it is possible to impose a number of restrictions on the process of operational production management. Unfortunately, the description of the most common optimization algorithms is beyond the scope of this article. We only note the most common ones, such as: combining production batches from the calculation of the optimal collection for launching into production, drawing up the sequence of production tasks, and calculating the optimization of equipment preparation (for example, when painting a product with different colors, the sequence of tasks will be: from light tones to more dark, to reduce the intermediate cleaning of equipment), etc.
Conclusion
This is not to say that the APS algorithm is the magic button that perfectly plans everything and everything at the enterprise. A corporate information system with the APS planning algorithm is only a necessary but not sufficient condition for effective planning of the industrial enterprise. However, using old and ineffective planning algorithms in a corporate system can significantly limit a company's chances of success.
ERP , – , , . , , , . , APS.
… ERP APS . APS + CAD – PDM . , ( ) . - , , , .. ERP , APS . — .
… !
MRP II . .
Peterkin S.V.
« MRP »
1 . ()
, , .
2 . - . At this stage, the required quantities are calculated taking into account all receipts and expenses of materials. If at this step the system detects a decrease in the level of the material below a certain level, then the quantity that needs to be purchased or produced to meet the need is determined.
It is also possible to calculate net requirements taking into account the lot rule (minimum order lot, lot multiplicity, order frequency).
3rd step. Determination of terms of purchase and manufacturing. ( ) -. MRP «» , , , () , .
, , .