May. 06, 2020 Source： 21ic
There are two types of industrial manufacturing: discrete manufacturing and process manufacturing. The two types of manufacturing basically see no difference on the production line, in fact, the underlying industrial equipment and control systems are different. Generally speaking, the control system guarantees the automation of the product. However, in some cases the product cannot be automated, such as personalized customization.
Personalized customization requires that the underlying industrial equipment can process different products. This requires that the equipment must be intelligent, and the control system must become an intelligent system to perceive different equipment and processes. However, the current production line cannot solve the high efficiency of personalized customization-this is also one of the important goals proposed by Industry 4.0.
Industrial manufacturing still relies on knowledge workers
The current industrial process is to determine the parameters and turn the production line into a "black lamp factory". But if the raw material or product variety changes, the process will have to be redone. This requires the decision-making department to adjust the indicators, and then engineers set it in the control system.
This is the status quo of the automation system. The structure of the automation system is actually a system in which people and information physics systems merge, that is, information physics systems in which people participate—the information obtained by information systems and the information obtained by human perception and cognition are used for analysis and decision-making.
And such a system has constraints. Because it is difficult for people to perceive dynamically changing operating conditions, it is also difficult to process heterogeneous information in a timely manner. In addition, people's decisions are subjective, and different people's decisions are different. This cannot guarantee that the entire production line is efficient and excellent.
Three Challenges to Realize Intelligent Manufacturing
In order to realize the efficiency of individual customization and the global optimization of the process industry, it is necessary to turn the current human and control systems and equipment into autonomous systems, and turn the system management system into a decision-making system for human-machine cooperation. This is different from the original system in that it has the functions of perception, cognition, and decision-making, and its ultimate goal is the direction of efficiency and optimization, so that the production structure and efficiency of the enterprise will undergo fundamental changes.
The first challenge is that the typical representative of artificial intelligence is deep learning based on big data, but deep learning has not been fully applied to the manufacturing process until now. To achieve intelligent manufacturing, we must solve the three problems of multi-scale, how far information and dynamic perception.
The second challenge is that if artificial intelligence is better than people in manufacturing, it is necessary to predict and trace the product quality, energy consumption and material consumption, including the operating status. The so-called tracing means that after a problem occurs, you can perceive which process or action caused it.
The third challenge is to integrate and optimize decision-making and control.
"Small Data Big Task"
There are two types of artificial intelligence today: strong artificial intelligence and weak artificial intelligence. Strong artificial intelligence means that it has comprehensive intelligence compared to humans, but a considerable number of scientists believe that this cannot be achieved; weak artificial intelligence is better than humans in certain scenarios, just like today's autonomous driving and robots can play chess very well Good, but it cannot do both. The third wave of artificial intelligence originated from big data, powerful computing and deep learning algorithms. I think that in the future artificial intelligence must move towards intelligent systems.
Why can't AlphaGo be used in industry? The reason is that it is in a completely determined rule, and the decision of the industrial process is in an open environment, the rules are uncertain, and it is difficult for the industrial process to establish a model of decision-making and trial and error. It can be said that the current artificial intelligence technology and game technology belong to "big data and small tasks", and the decision-making problem encountered by industry in the future is "small data and big tasks"-industrial big data is small data for computers.
What is "industrial artificial intelligence"?
Why develop industrial artificial intelligence?
Industrial artificial intelligence has begun to be proposed internationally, including the industrial artificial intelligence proposed by the United States and the "combination with the economy to promote intelligence" proposed by Germany. In China, the development plan for the new generation of artificial intelligence prepared by the Chinese Academy of Engineering also mentioned how to use artificial intelligence to solve the problem of intelligent manufacturing. In summary, industrial artificial intelligence currently completes three major tasks in the manufacturing process: the perception and cognition of multiple information in operating conditions, the collaborative decision-making at the work management layer, production layer, and operation layer, with the goal of optimizing the comprehensive production index of the enterprise Control system of automatic cooperative control equipment.
There are several key technologies to solve here: the first is the key technology, the intelligent sensing and recognition technology of multi-scale and multi-information operating conditions in complex industrial environments, and the second is the fast and reliable transmission technology based on 5G multi-information in complex industrial environments The third is the technology of intelligent modeling, dynamic simulation and visualization combining system identification and deep learning; the fourth is the prediction and traceability technology of key process parameters and production indicators; the fifth is the intelligent optimization decision-making technology of human-machine cooperation, In particular, the result end, edge, and cloud cooperate to realize the intelligent algorithm technology. Only by conquering these technologies will it be possible to revolutionize industry.
This article is reproduced from the network, if there is infringement, please contact the website to delete, thank you!