Process and straight-line flow organization: how to combine?
Individual projects of technical re-equipment, which enterprises could afford, and the transition to more advanced technologies cannot fully compensate these inconsistencies. Therefore, in the distribution of orders by available capacity, that is, in the formation of operational production plans, enterprises have to solve problems in the conditions of existing restrictions: the productivity of equipment, its number or insufficient number of production personnel for the organization of works in two or three shifts. Simultaneously, require the solution of economic problems of the use of unclaimed production capacity, retention of the volumes of unfinished production reserves, as in many cases they just grow nowhere else.
All these are immediately made and as a result, there is simply no enough time and resources to carry out systemic changes in the organization of production processes. This stalemate has hardly changed for the third decade. Is it possible to move it from the dead point by providing enterprises with the necessary resources? It is quite obvious that in this case the technical equipment would radically change both qualitatively and quantitatively. However, the vast majority of enterprises could hardly refuse the functional principle of production organization.
Much has been written and said about the functional organization of the production structure and its preferred organization of a continuous unit flow. But we do not compare these two principles in this article. We propose to consider steps for the transition from the functional principle of organization of production to the organization of a single flow one on the fact that for many enterprises, of course, such a transition will bring a serious positive economic effect. Refusing here from the enumeration of well-known arguments in favor of Lean Manufacturing, we have tried to solve the problems of transition to a single stream from the point of view of the tasks of operational management (drawing up production schedules).
As an object of analysis and recommendations, we consider a machine-building enterprise, the manufacturing of which consists of several production divisions (workshops), they include several dozen units of equipment and workplaces, there are several hundred types of products (parts and assembly units) produced monthly. The production organization of divisions is executed on the functional sign forcing to work with rather high values of production parties and, as a result, with the high level of incomplete production.
For a comprehensive description of the production structure, organization and management, the term "production system" is widely used.
Advantaged and disadvantages of production system
The principal appearance of the production system (PS) is determined by the method of combining operating elements (machines, work centers) in the organizational unit (area, shop), the organization of the production route, methods of managing a wide range of nomenclature. The two alternatives and defining options are considered to be as follows:
- Process (or functional) organization, which are grouped shops of the same type of operations (turning, grinding) to work with the entire range of enterprises on all routes, which is more consistent with a single and small-scale type of production.
- A product (straight-line flow) organization in which machines are grouped by the unity (proximity) of processing routes, which usually corresponds to one item group. Corresponds to mass and large-scale type of production.
In practice, there is a mixed production system with the flow organization of assembly processes and the process organization of the manufacture of parts more often.
Let's compare the advantages of product (straight-line flow) and process organizations with comparable loading and performance of equipment (see figure).
Commercial right motivated practicing managers understand that the straight-line flow production system is preferable. However, in the context of a wide variety of nomenclature and, accordingly, a small serial production, it is not always possible to organize an "ideal" flow in practice, since the nomenclature of machining production at our enterprises is quite large.
An example of a radical solution to the problem of diversity of the nomenclature — the automotive industry: in fact, in the production structure of the automaker is the assembly and a very limited number of units and components (engine, front panels with small variations of the model range) with a PS of strictly flow. The overwhelming volume of the "filling" of the car was transferred to an independent manufacturer. Moreover, together with the problems of balancing between the process and product (straight-line flow) production systems: the terms of delivery of components are formulated on the principle of "just in time", they force the supplier to either "continue" a single flow at his enterprise, or work with a sufficiently high level of stocks with a wide range of nomenclature.
In this article, we analyze the management problems faced by enterprises of discrete type of production (mechanical engineering, instrumentation as typical industries) with a complex, multi-component product consisting of parts, most of which require the passage of a multi-stage route (sequence of operations of processing the workpiece). And we tried to focus on the most difficult tasks.
We will adhere to the common classification of management tasks according to planning horizons: long-term plan (balancing the manufacturing volume, capacity and investment), medium-term, conditionally annual (calendar distribution of planned volumes of output, taking into account restrictions on capacity and other critical resources) and operational (detailed calendar planning of individual operations).
Almost everywhere, the first two tasks are assessed as satisfactorily solved, that is, on time, reasonably and are not criticized by the methods used. The tasks of operational planning (scheduling operating plans) descend to the level of the main dispatcher at best, and more often — the shop dispatcher and even the master. Is it because they are not important? Or is it because they are objectively complex — both because of the mathematical complexity, and because of the susceptibility of the created calendar to inevitable and various destructive factors in the form of constantly emerging operational changes?
Here, I think, will be to remember the quote of Shigeo Shingo, one of the creators of the Toyota Motors production system:
«Why do we have to make a schedule? Don't plan because the world is dynamic, everything is constantly changing. Just forget about the schedule.…»
It is clear that this statement that can be consider as provocative, in fact, means something completely different — critically review your production system (including) in order to minimize efforts to comply with critical deadlines, changeovers, unproductive movements, etc, strengthening the reliability of the production process in all senses at the same time.
Fully agreeing with the need for a critical look at the current enterprise production system lets focus on practical ways of such transformation. But first — let's fix a common understanding of the problem.
The enterprise (or part of it, for example, a shop) has a structure that can be described as “n” machines producing “k” types of parts (nomenclature items). In this case, each type of part passes through “m” operations (machines). The task of operational planning is to determine the start-up batches and the sequence of operations on each machine, taking into account the route for each part and so that the total time of the actual production cycle is sufficient for the timely execution of the order, subject to restrictions on the available capacity of the equipment and personnel.
The main problem with which the planner, who received a valid (satisfying the specified restrictions) version of the schedule, is struggling - a fairly long cumulative (for all machines) waiting time for the submission for processing. The approximate graphical ratio of the "conditions" of the machine at the process production system is shown in the figure below. Obviously, with production (straight-line flow) production system waiting time will be a multiple of a smaller and acceptable. But while in a similar situation, a typical small-scale, multi-operational production, we consider only the problems of the process production system. The task of scheduling the launch and release date with fixed duration of movements, changeovers and operations is to minimize the waiting time (downtime), which leads to a reduction in the production cycle.
For the level of shop (shops) machining machine-building enterprise usually takes the dimension of 80 machines and 300 types of parts.
For our case — discrete production "quantum" of solution is the sequence of running parts on this machine, and the actual solution — all sequences of launches on all machines. Satisfying the constraints (capacity, number of shifts, policy terms) is a valid solution.
Such problems are combinatorial and consist in considering of all or a part of the possible combinations and finding the best (for example, by the criterion of minimum downtime) solution of the available permissible.
In General, it is believed that the problem can be "manually", but not easy, solved for the matrix 2x2, almost not solved exactly in the case of "three machines" and "three products". For higher dimensions machine computation are required.
The number of combinations for one machine which is n! in our case is 80! for all machines — 80!300. The computational complexity of this problem is such that it cannot be strictly solved in a reasonable time, regardless of the power of computers. In other words, a palliative approach is needed. Practice has developed many assumptions that significantly reduce the dimension of the problem and even make the computational complexity acceptable. These assumptions are called in professional slang "heuristics" that is, intuitive insight, logical in its own, but with no evidence. It is generally accepted that heuristics work, as they began to use long before the computer was invented. However, the authors of these assumptions cannot guarantee that the solutions obtained will be the best. It's more dramatic in real. After all, usually the management of enterprises is not directly engaged in solving shop tasks — this is not his level. The exception is a situation when problem-specific workshop led to the failure of the plan. But it will be a situational response, nothing more. In addition, one or two specialists really form the schedule, gaining a very great influence in the absence of any practical criticism.
Like any complex task, scheduling has long attracted the attention of researchers — both theorists and developers of algorithms. The modern market offers a wide range of application software for operational scheduling, "packed" in specialized standalone it applications and integrated enterprise-class systems (ERP). The last, of course, is preferable because it allows you to prepare data for calculations with an acceptable percentage of data errors and for an acceptable time. But all (familiar to us) IT solutions have common features: they do not disclose the applied basic algorithms and assumptions ("heuristic rules"). The arguments are different — "it is very difficult to understand the user" or "trade secret". When accepted the tool remains a "black box", which does not add to the user confidence in its benefits.
At the other pole is the humanitarian aspect of the application of machine algorithms. First of all, the algorithm uses a different model than the one that the scheduler works with in practice. The algorithm is aimed at scheduling, while the planner — solving production problems. Here are the different criteria for the effectiveness of the solution, different "technologies": (quasi) optimization if the machines are against negotiation, compromise — humans have. But, ultimately, a machine solution is undoubtedly useful if it is applied to support the scheduler solution (for example, as a simulation model). Since, unlike the algorithm, the planner clarifies, discusses and jointly solves problems, uses informal power (access to information and networks, the right to organize, expert power and even personal power).
Practice shows that in real application "machine" and "traditional" production schedules (operational schedules of production) do not differ essentially that is shown on an example further.
Unfortunately, we could not find examples in the Russian practice and we have to refer to the study conducted in the company Nederlandse Spoorwegen — the main passenger rail carrier of the Kingdom (the example is considered quite relevant). As can be seen from the table, the best formal criteria solution was only a promise of the algorithm developer, but the real example of the algorithm did not show the declared parameters. Besides used the "space" in the formulation — actively used expensive alternative - taxi. The restriction on the alternative was not significant in the formal setting, but was actually perceived by the planner as important.
As an interim summary: the human factor is necessary in complex planning, so both the problem statement and the applied it applications should be focused on human decision-making. And IT - tools should be applied directly by the planner, not by IT service of the enterprise, what we usually meet in our practice.
If the human factor is neglected, then the problem is transformed from solving the problem to correcting errors, the discrepancy between the mental mathematical models is fixed and the possibility of really comparing the solutions disappears.
Focused production system
So why should we plan? The obvious answer is: "Because we can affect machine downtime, part waiting time and changeover time by changing the start sequence."
Why, in this case, Taiichi Ono, President of Toyota Motors and a recognized guru in the organization of production, says:
"Planning is a waste; it adds no value. Can't we get rid of this necessity?»
For our case of small-scale diversified and complex production route — cannot. But we must ultimately move towards reducing the dimension of the problem.
The process production system with the grouping of machines by type of operation differs from the product production system with the grouping by the family of the item. In both cases, the same composition of the machines is used, but with different grouping criteria.
The task of transformation, thus, is represented in the form of reconfiguration of the available equipment fleet.
As an image of the original structure the matrix representation of "machines-parts" is suitable, where the color shows the use of the machine for the production of this part.
The ultimate formulation of this task seems overwhelming, including computational restrictions. But it will become much more real if you approach it systematically, breaking into a number of quite acceptable for the complexity subtasks. It should be noted that the following sequence of actions does not include measures to reduce the changeover time to a level at which the number of changeovers does not increase the production time of the entire pool of planned orders. This is a separate task that is beyond the article`s frames.
Rare sheets allocation, cells search
There is fixed chain of machines for each item unit in accordance with the technological route. Obviously, the details of one item group will have identical or very close routes, and grouping them into a single flow is natural. But since our standard range of nomenclature is several times wider than the machine fleet, it is advisable to include in the formed flow and other families of parts with similar sets of types of operations (and machines).
The group of machines, implementing similar (but not strictly identical) sets of operations and, accordingly, routes, is called "cell". The cell does not necessarily provide the execution of all operations of the technological route, but most of it. The cell in our case provides processing of several families of parts. But, most likely, not all details will get to such cells, that is will be processed, as well as earlier, in the scheme of process production system. Nevertheless, the main utilitarian task will be achieved — reducing the number of different options under consideration to create a better schedule.
As a result of the transformation, we will get original matrix with a block structure where the monotone block is a separate cell.
For small dimension of tasks, a significant array of cells can be formed manually by visual analysis of the matrix. With a larger dimension of the matrix "number of types of parts x number of types of machines" will be effective simple method of sequential sorting of rows ("machines") and columns (nomenclature"). At an even larger dimension, one can apply variations of algorithms of many linear algebra problems, the so-called simplex method, which basically has the same sorting of rows and columns.
What should we do with the rest units?
As a rule, when carrying out such groups, special cases of deviations from clearly identified patterns of production routes will appear: with the exclusion of some details from the considered nomenclature, the organization of production cells will be explicit and technologically justified. In such cases, it is important to keep the system and try to solve some problems by other methods: revision of technology, purchase of additional simple equipment, etc. The feasibility of this approach is estimated to gain in the total load (reduction of downtime) of the input and new cells.
There will be too many options of selecting cells, among them should be selected a few (5-7), the best on the integral criterion of loading equipment and the duration of the production cycle.
The next task is to check the applicability of the new production system on the real scale of production (statistical data on the volume of output in the context of the nomenclature). You should determine and consider the behavior of production system at a statistically significant interval of variations of the volumes and nomenclature. Already at this step cannot do without machine calculation on simulation models. Modeling capabilities are typically available in it applications that support scheduling tasks1.
The model should reflect the following production aspects:
- The location of the elements (work centers) production cells and the organization of movements between the elements,
- The scheme of maintenance of equipment units by production personnel, including requirements for professional skills and due technological competence.
The following actions are performed and the data are determined by the constructed model:
- Analysis of causes of downtime
- Identify the resources with the lowest productivity that limit the rate of production flow.
Changes are made to the model in accordance with the decisions taken in order to expand the "bottlenecks".
Since the unit time of execution of operations of different parts and assembly units on the same type of equipment can be different values, in the model of the production cell should be distributed relatively: short operations – on one unit of equipment, the performance of the average — on the other, the performance of long — on the third. We are not talking about the technological "linking" operations of the production route of the product to a specific unit of equipment. The actual execution of the operation can and should be performed on the first available production resource. Here we are talking only about the model.
Statistics on the number of quick, medium and long-term transactions should be taken into account while the distribution.
- Comparison of different options, the choice of one of the options;
- Define the work cycle of production cells. In the model shown in the figure, the cycle is one unit of time;
- Reformatting production routes into routes that are performed in production cells with a standard run time expressed in integer values of the work cell stroke. In the model shown in the figure, the stroke is one unit of time, the piece time of fast operations is equal to one stroke, the piece time of the average duration — to three strokes, long — to seven strokes of the production cell;
- Determine the required levels of work in progress and inventory between organized work cells. In some cases, it is necessary to move away from the organization of the unit approach, determining the unit flow of particular consignment of parts and assembly units, the value of which is more than one;
- Development of methodology for the cell production planning and the rules of the alternation of positions of the nomenclature in terms of launch.
When unsuccessful simulation results are obtained, the technological operations that were the cause are distinguished, and the technologists of the enterprise should divide the operation into more elementary, closer to the standard ones.
As a practical example, the long-term operation of turning the overall product (the wheel axis of the truck of the railway car) was divided into roughing on one machine and consistently — finishing on the next machine. As an alternative, the variant with parallelization of two machines was considered, but it turned out to be worse, including the wear of the tool and the number of inconsistencies. Another example of stroke alignment is the acceleration of the operation to move the product between machines. The replacement of the manual manipulator with the robotic one proved to be effective due to the reduction of the waiting time in general, despite the significant investment expenses.
According to the simulation results, the final decision on the technical and organizational measures for the reorganization of the production infrastructure is made.
Reorganization of manufacturing structure
This stage, of course, is labor-intensive and expensive, also it does not contribute to the stability of the current production process. But it means a transition to a fundamentally new paradigm of organization and management of production. The transition to a new production system will fundamentally change the competitive position of the enterprise, allowing it to become a reliable supplier, which also benefits in terms of order fulfillment and operating costs.
Technical reorganization can be carried out routinely, consistently, in many respects — own forces.
Operational management in the conditions of new production system
The complexity of management, that is, first of all, the production schedule in the new production system will be significantly reduced. We list the requirements for the production system that supports the management processes of a set of single continuous production flows:
- Formation of nomenclature plans of production taking into account maintenance of the established levels of stocks and work in progress between the organized production cells;
- Ensuring the timely launch of production in the cells to perform production on schedule;
- To ensure the continuity of the flow: the production planning algorithm must form the alternation of different parts and assembly units in the production cell`s flow
- Formation of printed forms of plans, performance reports and accompanying production documentation: labels, accompanying statements, etc, including technical means and methods of bringing to the executors of normative reference data necessary for the performance of production operations.
IT-application used for modeling cells in some cases can be used in the operational planning mode. But this is not the main thing. It is important to understand the possibilities and limitations of the proposed information systems. But it is even more important to find the right balance between machine calculations and decisions made by human. This balance is the key not only to the effective use of modern it tools, but also to a real increase in the efficiency and competitiveness of the enterprise.
As a summary…
The material of this article is based on the real experience of solving problems on the development of production system and control systems in machine-building enterprises, including the problems of implementing information systems for managing production resources that support the MRP-II methodology.
The enterprises which production processes were built on a stream method: cable production, printing houses, production of packing, etc — statistically always gave much better results, than the enterprises which production process is organized with functional division and with large consignment work, and it, first of all the enterprises of mechanical engineering and difficult instrument-making. It is for companies with a significant range of manufactured components and having a negative experience of various projects to improve their efficiency, the authors have formulated the following conclusions:
- For machine-building enterprises and complex instrument-making projects of information systems should be considered only as part of the reorganization activities aimed at the organization of a continuous flow;
- One hundred percent realization of effectiveness increase, which potential is inherented in the functionality of information systems for engineering enterprises and complex instrumentation is not sufficient for a significant increase in the results of the enterprise itself, as the process (functional) organization of production is a significant and insurmountable limitation of productivity;
- The choice of information systems should be carried out not only and not so much in order to support the processes of operational activities management, but also in order to provide the processes of reorganization with a tool for modeling new production and technological processes.
 Here, according to the authors, the use of IT-applications, necessarily supporting the lot-splitting mode, becomes a key factor, since it is on the calculations of the systems that conclusions will be drawn and decisions will be made. Lot-splitting is a simulation mode of production plans up to the level of transfer batches. This mode will be simulated by a single (or limited enough) flow.