1. Definition of production capacity under mass customization production mode

APS planning and scheduling is the medium - and long-term planning and short-term planning of manufacturing enterprises. Capacity is the most important basis for production planning. Production planning decisions are based on defined goals, so that the specified tasks and capacity to achieve the best configuration. A correct assessment of one's own resource production capacity is a prerequisite for optimal allocation.

Definition of capacity. Production capacity refers to all the fixed assets involved in the production of the enterprise in the planned period, under the established organizational and technical conditions, the amount of products that can be produced. It can also be expressed as: capacity = unit working time/beat time. This standard traditional definition of productivity should have existed for more than 100 years. In the seller's market, planned economy environment and mass production mode of commodity shortage, industrial enterprises have several characteristics: production capacity is determined and static, which is itself the decision goal; According to the production capacity, enterprises can easily develop medium and long-term production plans, and can be divided into weekly plans or daily plans; The capacity of an enterprise's equipment resources (fixed assets) determines the production capacity, and there are usually bottleneck resources. In this case, factory capacity = workshop capacity = equipment/person capacity.

When we enter the mass customization mode of modern industry, we continue to use the past capacity definition to make production planning, which has a great problem. First, the rhythm of small-batch, multi-variety, mixed-line production is dynamic, and the production capacity is dynamic. Second, production capacity is not only a decision of enterprise production resources, but may be more determined by supply chain capacity transactions. The production capacity of factories, workshops and equipment is not equal due to the outsourcing of orders, work orders and processes. The supply chain is not only the raw materials and parts purchased, but also includes tooling, information resources, logistics resources, etc. Third, the biggest difference is that the goals of production planning decisions are different from production tasks, and the goals of decision-making at different levels are inconsistent with tasks and resources at different levels. The goal of long-term planning decisions is the on-time delivery of orders, the goal of short-term planning decisions is the shortest production cycle and minimum WIP, and the goal of real-time production planning is critical resource OEE and process quality. In the mass customization model, the main goal of making production plans has been not only to complete the number of products, but to achieve the economic goals of the enterprise. The capacity of a factory is defined as the number of products it can produce. The capacity of a plant is the on-time delivery capacity of a plant to meet the product market.

With the increasing degree of small-batch, multi-variety, mixed-line production in manufacturing enterprises, enterprises will obviously feel that there are a lot of rich production capacity (equipment resource capacity), but it cannot be played out. We analyze this phenomenon from two aspects.

From the perspective of system science, we can think that the whole factory in the mode of mass production is a linear system. When we get an order, we know immediately when it will be delivered. Because we can calculate the "beat" of each workshop from parts production to total assembly production. All production activities are orderly from end to end. When we entered the mass customization mode, the process of each order was very different, and the production beat of each workshop could not be found, or it was a dynamic beat. Another phenomenon of the mass customization model is that industrial production is becoming more and more specialized. The production of a factory is seriously constrained by the supply chain, which not only includes raw materials, parts and components, but also includes various substances, tools, information, energy and so on in the production of products. Therefore, a factory of a linear system is divided into a system of three levels of equipment, workshop and factory, and presents nonlinear characteristics. The operation management structure of the factory is formed by the system of internesting, intercollaboration and interrelation between the three different levels. The business environment has changed, the production model has changed, and we continue to use the original perspective can not see the essence of the problem.

Analysis from the economic level. Many enterprises tend to think of two ways when the production rhythm cannot be controlled and the order cannot be met. Investment in fixed assets to expand production capacity; Or to build dedicated lines to control the pace (often better called process reengineering). The essence of this is to resolve the non-linearity of the production process into linearity. Any complex problem can be simplified if you are willing to pay the money and time. The problem is that they don't know how much of their existing resource capacity is still available. Experience has shown that most companies have 10% to 30% of their capacity to tap potential, or even more. Unfortunately, they just threw away their most profitable business. Depreciation of fixed assets is an important cost for enterprises. After capacity reaches a certain level, the depreciation cost of fixed assets is equalized. The capacity to be dug up again is very large.

APS has been born for more than half a century and is the product of large-scale production mode. Its basic theory is: factory capacity = workshop capacity = equipment/person capacity. On such a theoretical basis, under the same production capacity model, in the mode of mass customization, it is obviously very difficult to make (cut into) three different levels of production plans with different decision goals, and it will inevitably cause the plan to be unenforceable.

2. Correctly handle the relationship between the whole and the part of the enterprise

To make a production plan, you first need to use MRP to break each order into a work order and then decompose it into each process. The BOM structure constraints of the product determine the production and processing sequence of each component. According to this sequence constraint, as well as the constraints of the limited production capacity of the production resources of each process, APS processes the production and processing in each workshop of all factories during all available working hours. There is no problem with this logically. But the uniform ordering of the whole factory is "affecting the whole body". Any disturbance that needs to be rescheduled is adjusted by each workshop at the same time. There are at least two aspects of this rigid scheduling that conflict with the factory's operating mechanism.

The organizational structure of an enterprise is the result of long-term operation optimization and adjustment. Decision-making structure layer by layer, responsibility and obligation layer by layer implementation, for the enterprise is the most efficient structure. The management structure of APS breaks the original management authority and responsibility, and realizes "flat" management. However, this mechanism of "military commander directly commands soldiers" has produced contradictions with the original management mechanism. At the very least, this contradiction is difficult to reconcile under the existing APS management structure. Let me give you a simple example. The factory has five workshops, among which a process of the parts machining workshop is suspended for some reason, requiring APS rearrangement. The rearrangement involves all the workshops. So the other four workshops question, why the machining workshop so a little problem let us follow the toss?

The internal relationship between a whole factory and a part of the workshop is interdependent and exists each other. The whole plays a dominant role in the decision of the local commander, and coordinates the development of the local unity to the overall strategic goal of the factory. But there are also local interests of the workshop. Local changes also affect the overall change. A plant addresses the problem in the pursuit of the optimal effect of the system as a whole. However, if we do not pay enough attention to the positive role of some parts, the overall interests will be damaged. A factory must pursue customer satisfaction. Under this premise, some workshops may need to sacrifice production capacity to ensure the construction period, and some workshops may need to maximize production capacity. Similarly, the digital management system corresponding to the enterprise entity needs to reflect the overall system target state, and at the same time, each part needs to have its own management control mechanism. If it is a rigid integrated management system, it is certainly not possible to achieve such a goal.

3. APS modeling problem

We need to establish a production process mechanism model for a production operation system, which is also a white box model. This model should not only reflect the input-output relationship, but also reflect the state of the production process, becoming an accurate mathematical model that truly reflects the physical/chemical mechanism of the production process and the material flow mechanism.

The process of industrial production and its operation management is extremely complex. This level of complexity is unmatched by the consumer Internet industry in the field of circulation. Order a product on Jingdong, even if the complex circulation link, this order information will never disappear, can be found at any time. In the field of industrial manufacturing, orders cannot be found once they enter the workshop. Only after the final assembly and packaging is marked with the customer's label can the order information be found. This is called "order information lost." The design process of an equipment includes the deconstruction process from the assembly to the bottom parts, and the reconstruction process from the bottom to the assembly needs to go through several virtual and real verification and testing processes. This process involves hundreds of thousands of individual task nodes that management must involve. The production process is more complicated. A wide variety of production processes and a wide variety of production equipment are related to management.

The flow of materials in the production process should be the key to management control. The simplest flow is a single piece/batch flow (serial) similar to turning, milling, planing and grinding. There are also multi-capability machines that simultaneously process parts of various sizes, such as heat treatment, spraying, surface treatment (parallel). Some equipment may process different numbers and varieties of parts at the same time. After the completion of a process, it is possible to produce several different parts at the same time. An assembly line or an automatic CNC machine may complete multiple processes at once. The equipment of some processes is space. Material residence time and ambient temperature are all involved in the management factors. The device has primary resources, as well as secondary and tertiary resources, which are also very complex. These complications are numerous. For the mathematical modeling of the production process, these phenomena, characteristics and laws need to be described truthfully. This is undoubtedly a very difficult and necessary thing to do. Of course, it can be simplified in some cases.

Building mathematical models with tables is a very common method in science and engineering research. APS is a mathematical model of factory production planning and scheduling system based on Excel. The problem is that the production and operation management of manufacturing enterprises and the control of manufacturing processes are a very complex system. Even if we can design a table that reflects the basic operation law of enterprises, if the table is too complicated for users to fill in correctly, then the data model will lose its practical role.

In order to enable manufacturing companies to fill out the form, the APS form is not too complicated. This leads to the disadvantage that the model cannot reflect the objective reality. In order to make up for the defects of the model, APS usually designs a lot of configuration items in the software interface. In order to compensate for the shortcomings of the APS model, many user configuration interfaces become a powerful feature of APS.

APS's modeling is flawed, and I don't think that's the core issue. Rough modeling is not necessarily unusable. Modeling too accurately does not necessarily work. The core problem is that the basic theory and management structure of APS is no longer suitable for mass customization production mode.

4. APS algorithm problem

The model and the algorithm are bound together. The effectiveness of the model is the basis of algorithm optimization. We assume that the model is available. This section deals only with algorithms.

Many APS are describing their various optimization algorithms, the vast majority of which are for commercial marketing. More than a decade ago I talked about the difficulty of scheduling optimization, and gave a simple sorting example: Assuming that the computer can handle 1.000.000 sequences per second, we want to build an optimal scheduling system. Nine jobs in less than a second. 11 takes a minute. Given 20 jobs, it takes 77,147 years to find the optimal solution! Actual shop scheduling problems involve hundreds or thousands of jobs. Asking APS to find the optimal solution is basically nonsense. This is a very easy question to prove. I thought you said you could optimize it. Well, let's arrange it before the end of the day and save it. Then let Solver compute overnight and see how it improves on yesterday's results. I guess there is no APS for you to verify this, because the APS scheduling result is one of countless theoretically feasible solutions.

Is the optimization capability of APS important? The truth is not important at all. The meaning of APS is not scheduling optimization, but scheduling automation. If we can get a scheduling result quickly, it is much simpler than manual operation, and the scheduling efficiency is greatly improved. On the other hand, APS workshop scheduling is a short-term production plan, and people still need to go through the actual implementation of the plan! APS scheduling down to the second makes no practical sense. Many uncertain problems can be buffered and adjusted by people.

5. What can APS do and what can't APS do

APS is defective, but not unavailable. In theory, if the degree of standardization is good, mass production is based, and various industrial enterprises with small disturbances should be able to be applied.

If you do not arrange orders, but for the scheduling of work orders in each workshop, it should be able to be applied. For example, as MES each workshop scheduling management of an APS module. But it is best that the processing process of the workshop is a single piece flow or batch flow. For example, the roughing or finishing workshop that only includes the turning milling and grinding equipment, or the workshop is a simple process environment such as stamping and casting.

In the occasions where various products have exact production beats, such as home appliance assembly, electronic product assembly, clothing and other production lines should be able to be applied.

The current APS architecture is difficult to implement multi-workshop collaborative production, and it can not handle the relationship between local interests and overall collaboration. The production management of large-scale equipment can not use APS, because the system environment involved in large-scale equipment is very complex, the long production cycle has great uncertainty, and the supply chain and multi-workshop synergistic contradictions APS is difficult to deal with. APS cannot be applied to production management in the field of R&D and design. The data management mechanism of APS is not good at dealing with dynamic data, so it is difficult for smes with poor data environment to implement. In addition, APS models and algorithms are not suitable for the introduction of domain knowledge, so the application of artificial intelligence in machine learning is also difficult.

There are other reasons why APS projects fail, but they are not the core factors. For example: scheduling results do not conform to the habits of workshop teams or employees; A few emergency work orders "cardiopulmonary bypass"; There are some missing items in production feedback that can not form an effective closed-loop management and so on. Standard working hours are the key to implementing APS, but in many cases, the so-called standard working hours simply do not exist, and APS cannot be managed. As for inverted, forward, drag and pull and other requirements are the production management as a game, not worth mentioning.

6. Two recommendations for implementing APS

In smes, APS does short-term planning of workshops, one workshop at a time. It is better to do it together with the implementation of MES. Medium - and long-term planning is still done manually. If there is the APS output data of the workshop as a basis, the experience of the staff making medium and long-term plans is still sufficient. Don't aim too high. The role of scheduling automation in the workshop is far more important than scheduling optimization, and APS really can't be too good. For large equipment manufacturing enterprises, personalization is too strong, APS can not help too much. APS cannot be counted on for highly automated production workshops, such as chips, display panels, printed circuit boards, and automated automotive and home appliance production workshops, which require accurate real-time planning, and this field is where models and algorithms come in.

APS projects are best implemented by the production lead of the enterprise, with the IT team as a support. IT professionals can easily fall into the trap of "managing your business with software."