Intelligent decision-making: the difference between intelligent manufacturing and high automation
Academician Li Bacon wrote a sentence in his recommendation for "The Three Body Intelligence Revolution": "As more and more objects around us become intelligent, do not let human intelligence be overwhelmed by those intelligences. The meaning of this sentence is to remind us that the AI flood is changing the world, but compared to human intelligence, it is only a "grain of rice in the ocean". Academician Li once again proposed at a recent meeting: pay attention to the workshop site, intelligent manufacturing is not far away from us. So, the production site is a management issue, a scientific command and decision-making.
The "intelligence" studied in the manufacturing industry and the "artificial intelligence" studied in academia are "two cars" with the same direction but different tracks.
The Position of Human in Intelligent Manufacturing
When we observe the role of humans in an operational system from the perspective of systems science, there are only two roles for humans: decision-makers or executors. People may play different roles in different time and space.
We have marked the location of people in an industrial operation system diagram. Since humans play two roles in this industrial system, the "intelligence" in the industrial field must include decision-making intelligence and execution intelligence. This is a matter of two levels. Intelligent decision-making and intelligent execution are two indispensable and inseparable aspects for achieving intelligent manufacturing. We cannot separate decision-making and execution into independent research at the decision-making or execution level, nor can we mix decision-making and execution together.
In the current implementation of intelligent manufacturing strategy in China, research on intelligence in the decision-making field cannot be lacking or weakened. We need to first define the decision-making, decision-making, execution, and incorporation of "intelligence" in industrial systems before continuing with the following discussion.
Decision making is the allocation of unlimited needs (goals, tasks) and limited resources.
An industrial system is a nested and segmented system. An industrial enterprise system can be divided into two subsystems: resources and tasks (goals).
The resource system includes the decision-making management team of the enterprise's own hierarchical structure, as well as subsystems such as research and development, production, sales, administration, and finance; Including software and hardware equipment, material resources, funds, and energy of the enterprise; This also includes external resources such as suppliers and customers; In addition, it must also include information resources and time resources that we cannot see.
The task (goal) system of an enterprise has long-term, medium-term, and short-term goals, also known as planning, scheduling, and scheduling goals. The goal must also be layered and nested. But no matter how it is divided, the goal of the terminal must be implemented on specific executable entities/or services. We must note that the goals of enterprises are often multi-objective, multi constrained, and dynamically changing. For example, the best service and the lowest cost, such as not being able to work overtime and completing tasks, such as ensuring the delivery time of a few orders even if it affects production capacity this month, and mining production capacity next month.
The operational decision execution system of an industrial system is that high-level decisions are based on the allocation of high-level resources according to the enterprise's higher-level goals; The next level of decision-making is to allocate sub resources based on the corresponding sub objectives; and so on. When the determined goals and resources become a determined configuration relationship that cannot be further separated, the system enters the execution layer. Above this, they all belong to the decision-making level.
Since the invention of stone tools by humans, human intelligence has been solidified on the tools. The history of industrial civilization is the history of human wisdom being continuously solidified through both soft and hard methods in industrial tools, products, and production modes. So, the definition of 'industrial intelligence' is not important.
On the front line of industrial enterprise operations, that is, at the decision-making level, if we use industrial design tools, production tools and equipment, market analysis and marketing networks and technologies that gather artificial intelligence to assist us or proxy workers in materializing decision-making goals, this is called executive intelligence.
The decision intelligence of industrial systems refers to the ability to optimize the allocation of decision-making objectives and limited resources. This is a capability that integrates systems science, management science, and information technology. Intelligent decision-making belongs to the science of the 21st century.
The intelligence of the execution layer belongs to the category of production capacity
At the executive level of industrial enterprises, which is commonly referred to as the front line of design, production, and sales, there are increasingly more intelligent resources available. High end design software, the best CAX system, 3D printing, and perfect virtual reality VR can make design increasingly intelligent and efficient. The machines in the workshop are becoming increasingly intelligent, the equipment is becoming more and more intelligent, and various robots are perfectly integrated with the automation of the production line. Market sales management is increasingly supported by powerful network data and management systems.
However, all of these are fixed assets of the enterprise (soft assets, hard assets), which belong to the category of production capacity. Or rather, this is advanced production capacity. There is no direct causal relationship between these factors and whether a company can gain competitiveness, achieve ideal returns, or sustain long-term development. No matter how intelligent these production resources are, and whether they are willing to admit it or not, this is the definition of production capacity and does not need to be proven. Enterprises with advanced equipment go bankrupt; Enterprises with very average hardware resources are developing normally. We have seen too many cases like this. In other words, the advanced production capacity mentioned earlier can be bought with money. What can be bought with money may not necessarily be core competitiveness.
With the support of correct decision-making, the potential of these advanced production capacities can be fully realized, and enterprises will have enormous development space; Under unwise decisions, these things will become a burden on the enterprise, a high cost, and you will be the first to be eliminated in the future. The simple truth that 'Xiaomi with a rifle can defeat American planes and cannons' is still not fully understood by many companies.
Execute Intelligence
The executive layer also has decision-making problems.
Workers were instructed to open a mold. In the past, workers processed CNC machine tools according to process requirements and experience, what to do first and then what to do, what tools to use, set the speed, etc.
This is intelligent human decision-making. When we give the product to the "intelligent" machine tool, the digital product definition and human knowledge and experience are input to the machine tool, the machine tool system will automatically process according to the instructions, and even the system can optimize the processing path to achieve the purpose of saving time and effort. All decisions at the executive level are made with clear goals and defined resources. In the local scope of the execution layer, the system boundary is clear, the system environment is simple, and the openness is limited, which belongs to the deterministic problem of simple system. It is precisely because of this that a highly "intelligent" device resource can make "autonomous decisions" and "precise execution" in its execution decisions ".
An unmanned workshop for home appliance assembly and production (not to be called an "unmanned factory") is a fully automated production line composed of robots, AGVs, etc., which can be regarded as a device with a high degree of automation. It follows strict procedures and clear rules to implement the established clear production instructions. "Personalized customization" is actually assembled according to instructions, that is, the manipulator grabs different parts from the prepared line-side inventory to assemble products of different specifications. The design theory of this kind of unmanned workshop has matured, but it is not easy to gather the experience and wisdom of many people in the design and implementation of engineering details, and to integrate various technical achievements and information.
However, the unmanned workshop is not a theoretical problem, but a practical problem, and its intelligence belongs to "weak artificial intelligence". Instead of feeling "shocked" by visiting an unmanned workshop, it is better to study how they generate huge cash flow to "support" the production line.
In fact, driverless cars and "alpha dogs" should also be classified into this category, belonging to the deterministic problems of simple systems, "autonomous decision-making" and "precise execution" according to established rules ".
Driverless cars are not likely to hesitate for an obstacle in the middle of the road, nor will they make autonomous decisions between saving time and money.
AlphaGo Go is essentially an executive program. Each execution order was issued by Li Shishi. The goal of execution is clear and the rules of execution are tight. AlphaGo use "value networks" to calculate situations and "strategy networks" to cast their own. The key is that there must be an optimal solution to this problem. In systems science this is classified as a deterministic simple system. But AlphaGo learning ability and computing power is very strong. The AlphaGo can count the chess game, but it can't count the opponent. The act of assigning Li Shishi to play chess is no different from the nature of assigning a work order to workers in a factory.
So far, most of the "intelligent manufacturing" we have seen and talked about are intelligent production capacity. The household appliance unmanned workshop is only a small step forward.
If you use the value chain to analyze such enterprises, you will find that the value addition of their products mainly occurs in the technological innovation and market innovation of the products. Customers will not pay 100 yuan more for products assembled in unmanned workshops than those assembled manually. However, when you customize your online purchase, you will have to pay several hundred yuan more for one more option. The goal of intelligent manufacturing is to enable enterprises to obtain higher profits and stronger competitiveness. We should see the essence from the phenomenon.
At present, a technological revolution with "intelligence" as the key word is booming, unstoppable and the prospect is immeasurable. Technological advances in the field of intelligent design, intelligent production, and intelligent markets at the executive level are obvious. These new technologies have greatly improved the productivity of enterprises. At the same time, we need to realize that the "intelligence" at the execution level is attached to the production resource, and no matter how intelligent it can change the properties of the production resource.
Perhaps some people will suggest that these intelligent production resources will indeed improve the efficiency and quality of product production and reduce the cost of products. Standing at the level of enterprise system, the "efficiency, quality and cost" of the implementation layer is neutral, and the overall efficiency and sustainable development of the enterprise is the fundamental goal pursued.
Information source: Intelligent Manufacturing
