1. Definition of Intelligent Manufacturing
Manufacturing is the process of transforming raw materials into usable products. It's important to note that manufacturing here is not limited to processing and production. For a manufacturing company, its manufacturing activities encompass all activities related to "transforming raw materials into usable products," such as product development, process design, equipment maintenance, procurement, and sales.
The most common understanding of intelligent manufacturing is simply "applying intelligent technologies to manufacturing." However, what is intelligence? What is artificial intelligence? Although more than half a century has passed since the concept of artificial intelligence was first proposed, its definition remains controversial. It is generally believed that current research in artificial intelligence focuses on six main areas: natural language processing, machine learning, computer vision, automated reasoning, knowledge representation, and robotics. However, it is clear that people do not believe that companies implementing intelligent manufacturing necessarily need to apply all of these technologies.
There are many definitions of intelligent manufacturing.
In their book *Intelligent Manufacturing* (the first monograph in the field of intelligent manufacturing research), Wright and Bourne define intelligent manufacturing as "modeling the skills and expert knowledge of manufacturing technicians through the integration of knowledge engineering, manufacturing software systems, robot vision, and robot control, so that intelligent machines can perform small-batch production without human intervention." Today, the intelligent technologies applicable to manufacturing activities are not limited to those listed in the above definition, and intelligent manufacturing is clearly not limited to small-batch production. However, there is no reason to underestimate its significance because of the limitations of this definition. Proposing the concept of intelligent manufacturing at a time when related technologies were still immature was undoubtedly visionary and pioneering work.
Lu Yongxiang once defined intelligent manufacturing as: "A human-machine integrated intelligent system composed of intelligent machines and human experts, capable of intelligent activities such as analysis, reasoning, judgment, conceptualization, and decision-making during the manufacturing process. Through the cooperation between humans and intelligent machines, it expands, extends, and partially replaces the mental labor of human experts in the manufacturing process. It updates and expands the concept of manufacturing automation to flexibility, intelligence, and high integration." His emphasis on human-machine integration is a profound insight.
In China's "12th Five-Year Plan for the Development of Intelligent Manufacturing Technology," intelligent manufacturing is defined as "information-based manufacturing under ubiquitous sensing conditions, oriented towards the entire product lifecycle. It is based on advanced technologies such as modern sensing technology, network technology, automation technology, and anthropomorphic intelligent technology, and achieves intelligent design processes, intelligent manufacturing processes, and intelligent manufacturing equipment through intelligent sensing, human-computer interaction, decision-making, and execution technologies." This definition of achieving intelligent design, manufacturing processes, and manufacturing equipment is merely a phenomenon of intelligent manufacturing. In other words, intelligent design and equipment are simply means of manufacturing, not the goal itself.
The Ministry of Industry and Information Technology (MIIT) defined intelligent manufacturing in its "Intelligent Manufacturing Development Plan (2016-2020)" released in 2016 as: a new production mode based on the deep integration of next-generation information and communication technologies with advanced manufacturing technologies, permeating all aspects of manufacturing activities such as design, production, management, and service, and possessing functions such as self-sensing, self-learning, self-decision-making, self-execution, and self-adaptation. This definition undoubtedly draws on the wisdom of many scholars and experts, highlighting the technological foundation and application stages of intelligent manufacturing, and revealing its functional manifestations, but it fails to touch upon the essence and connotation of intelligent manufacturing.
In the US, EU, and South Korea, smart manufacturing (SM), which is highly valued, can be seen as a more advanced stage of intelligent manufacturing development. SM is the result of the rapid development of cutting-edge technologies in recent years, such as the Internet of Things (IoT), big data, VR (virtual reality)/AR (augmented reality), smart sensing, cloud technology, and next-generation artificial intelligence. The US National Institute of Standards and Technology (NIST) defines SM as a fully integrated collaborative manufacturing system capable of responding in real time to changes in the needs and conditions of enterprises, supply chains, and customers. This definition is quite simple, not directly specifying the specific functions of the technologies and systems involved, but it more clearly reveals the goal of intelligent manufacturing.
This section provides a very simplified definition of intelligent manufacturing and systems. This is precisely because intelligent manufacturing is still under development. A simple definition may encompass a broader range of functions and technological elements, both existing and future; a simple definition may also have deeper meanings, whether superficial or intrinsic; whether explicit or implicit.
Intelligent manufacturing: Integrating machine intelligence into various manufacturing activities to meet the corresponding goals of enterprises.
Key terms in the definition: machine intelligence, integration, manufacturing activities, and goals.
Machine intelligence encompasses computation, perception, recognition, storage, memory, presentation, simulation, learning, and reasoning. It includes both traditional intelligence technologies (such as sensing and knowledge-based systems like KBS) and next-generation artificial intelligence technologies (such as deep learning based on big data). Generally, artificial intelligence is divided into three stages: computational intelligence, perceptual intelligence, and cognitive intelligence. The first stage is computational intelligence, which refers to rapid computation and memory storage capabilities. The second stage is perceptual intelligence, which refers to sensory abilities such as vision, hearing, and touch. The third stage is cognitive intelligence, which refers to the ability to understand and think. Cognitive intelligence is currently the area where the gap between machines and humans is greatest; teaching machines to reason and make decisions is exceptionally difficult.
Although machine intelligence was developed by humans, many of its functionalities (such as computation and memory) far surpass human capabilities. Integrating machine intelligence into various manufacturing activities to achieve intelligent manufacturing typically offers the following advantages:
(1) Intelligent machines possess computational intelligence superior to that of humans. In areas with fixed mathematical optimization models, requiring extensive computation but not knowledge reasoning—such as engineering analysis of design results, advanced production scheduling, and pattern recognition—machines can provide better solutions more quickly than humans rely on experience. Therefore, intelligent optimization technology helps improve design and production efficiency, reduce costs, and increase energy utilization.
(2) Intelligent machines possess a higher capacity for proactive perception and automatic control of manufacturing conditions than humans. Taking CNC machining as an example, vibration and temperature changes in the "machine tool/workpiece/tool" system have a significant impact on product quality, requiring adaptive adjustments to process parameters. However, humans clearly struggle to perceive and analyze these changes in a timely manner. Therefore, applying intelligent sensing and control technologies to achieve closed-loop control of "perception-analysis-decision-execution" can significantly improve manufacturing quality. Similarly, in a company's manufacturing process, many dynamic and changing environments exist. Certain elements in the manufacturing system (equipment, testing mechanisms, material conveying and storage systems, etc.) must be able to dynamically and automatically respond to system changes, which also relies on the autonomous intelligent decision-making of the manufacturing system.
(3) Manufacturing enterprises may possess massive amounts of product lifecycle data. The development of technologies such as the Industrial Internet and big data analytics brings enterprises faster response times, higher efficiency, and deeper insights. This is incomparable to traditional methods that rely on human experience and intuition.
Machine intelligence is the condensation, extension and expansion of human wisdom. Overall, it has not surpassed human wisdom, but the intelligence of some units far exceeds human capabilities.
A company's manufacturing activities include R&D, design, processing, assembly, equipment operation and maintenance, procurement, sales, and finance. Integration means not completely overturning previous manufacturing methods, but rather further improving manufacturing efficiency by incorporating machine intelligence. The definition states that the purpose of intelligent manufacturing is to meet the company's corresponding goals. Although specific goals are not explicitly stated, it is easy for readers to understand that improved efficiency, reduced costs, and environmental sustainability are all implicitly included.
Intelligent manufacturing system: Integrating machine intelligence into a system that includes people and resources, enabling manufacturing activities to dynamically adapt to changes in demand and the manufacturing environment, thereby meeting the system's optimization goals.
In addition to the key terms in intelligent manufacturing, the key terms here also include: system, people, resources, demand, environmental change, dynamic adaptation, and optimization goals. Resources include raw materials, energy, equipment, tools, data, etc.; demand can be external (considering not only customers but also society) or internal to the enterprise; environment includes equipment operating environment, workshop environment, market environment, etc.; in this definition, "system" is a relative concept, as shown in the figure. That is, a system can be a processing unit or production line, a workshop, an enterprise, or an enterprise ecosystem composed of the enterprise, its suppliers, and customers; dynamic adaptation means being able to respond in real time to environmental changes (such as temperature changes, tool wear, market fluctuations); optimization goals involve the goals of enterprise operation, such as efficiency, cost, energy saving, and consumption reduction. The various means required by the system are all implicitly included.
The hierarchy of intelligent manufacturing systems
It is particularly important to note that the above definition implies:
Intelligent manufacturing systems do not require machines to completely replace humans; even highly intelligent manufacturing systems in the future will require human-machine symbiosis.
South Korean scholars Kang et al. point out that intelligent manufacturing (SM) should not only focus on economic indicators such as increasing efficiency and reducing costs, but should also be able to create new value for society in a sustainable way. A lack of consideration for human and social issues may lead to problems. Intelligent manufacturing should not be simply viewed as the application of cutting-edge IT technologies; it should be a manufacturing engine based on the philosophy of "sustainable development" oriented towards people and society, capable of leading to continuous growth.
2. The Basic Concepts of Intelligent Manufacturing
The introduction outlines the inevitable development from automation to digitalization, networking, and then to intelligence. Automation technology has matured considerably after more than a century of development. Let's observe and reflect somewhat abstractly on the problems that automation technology is well-suited to solve.
The problems that automation technology can solve are mostly deterministic. All automated lines and machines have deterministic processes, definite motion trajectories, and definite controlled objects. Of course, the actual movement of the machine may contain errors, which will be reflected in the quality of the manufactured goods; in other words, uncertainty is not entirely absent. However, from the perspective of designing an automated system, its inputs, outputs, operating methods, paths, objectives, etc., are all deterministic. It is only necessary to ensure that the resulting errors are within acceptable limits.
Classical automation technologies primarily address structured problems. Problems that can be described using classical control theory are structured, such as automatic regulation problems and PID (Proportional-Integral-Derivative) control. The development of electronics and computer technology has accelerated the application of program control and logic control in automation systems, which also deal with structured problems. In modern control systems, knowledge-based systems, similar to IF-THEN, are used in certain situations; they are themselves structured and deal with structured problems.
Traditional automation technologies deal with problems in a fixed pattern, such as automated processing, assembly line production, and automated material conveying.
Traditional automation technologies address relatively localized problems, rarely tackling enterprise-level systemic issues such as supply chain problems, customer relationships, and strategic responses.
Let's observe and consider the real-world problems faced by businesses. Businesses face numerous uncertainties, such as the quality issues that every company must address. For some known, deterministic problems that may lead to quality defects, solutions can be found through appropriate processes and automation—this is the capability of traditional automation technology. Many random factors affect quality, such as temperature and vibration. While it's known in advance that these factors will impact quality, they are only qualitative concepts and cannot be controlled in advance. This necessitates real-time monitoring of changes in relevant factors during manufacturing and applying corresponding controls, such as adjusting ambient temperature or automatically compensating for processing errors. This is the initial stage of intelligent control. These random factors that cause quality problems, while uncertain, are explicit and easily perceived. However, there is another type of uncertainty: implicit uncertainties that engineers and managers may find difficult to perceive. For example, how many related and combined factors influence the performance of an advanced, complex engine system, and to what extent? Similarly, what parameters might non-explicitly affect the performance of a new process, and to what extent? For engineers, these may be uncertain. In fact, some factors and their related influences have a certain aspect, but people lack an understanding of their objective laws, leading to subjective uncertainty. Furthermore, some originally deterministic issues become uncertain due to a lack of digitization. For example, the arrangement of various activities and processes within a company is inherently deterministic. However, because it involves too many people and occurs at different times, it becomes chaotic to human understanding without special methods. This is also known as subjective uncertainty or cognitive uncertainty. Why is subjective uncertainty also considered uncertainty in the manufacturing system? Because the manufacturing system should inherently include relevant people. There is also a category of implicit influencing factors that are inherently uncertain. For example, subtle inconsistencies in the properties of raw materials during precision manufacturing, energy instability, and sudden environmental factors (such as sudden external vibrations) can lead to quality instability; temporary changes in personnel positions in the workshop can cause quality problems; changes in the work schedules of certain employees due to major social events (such as the FIFA World Cup) can cause quality problems; and the specific impact of major public health emergencies on the company is related to various specificities (which vary from company to company) such as the company's supply chain, location, population flow, and the infection status of employees. Currently, people can only have an abstract and qualitative understanding of such problems, making it difficult to develop relatively precise solutions based on the specific degree of impact. Classical automatic control technologies are naturally relegated to the sidelines for problems like these, and even modern control technologies with certain intelligent features are powerless to address them.
Intelligent manufacturing systems do not require machines to completely replace humans; even highly intelligent manufacturing systems in the future will require human-machine symbiosis.
Note: Both explicit and implicit uncertainties!
A significant number of problems within enterprises are unstructured. When striving to improve quality, identifying the factors influencing quality becomes difficult; similarly, the specific impact of major public health emergencies on businesses is hard to quantify, let alone address; this is because the environment and the problems themselves are unstructured. Enterprises possess a vast amount of information that is neither conventional numerical data nor structured data stored in databases and logically expressed using two-dimensional tables. This includes full-text, images, audio, and multimedia information—this is unstructured data. This unstructured data contains valuable information for businesses, such as reports from R&D personnel and collected external materials (text, images, etc.). Traditional automation technologies have failed to effectively utilize this information and have only scratched the surface.
How can we leverage unstructured data to make sound judgments and decisions?
Many problems within businesses are not fixed patterns. Today, many companies implement personalized customization to better meet customer needs. Different types of companies will certainly implement personalized customization in different ways. Even for the same company, different products and different types of customers may require different models. The methods of data collection and processing, and data-driven personalized design and production, are all different. Similarly, energy conservation in factories or workshops may take different approaches depending on the type of company. Even companies producing similar products will have different energy-saving models due to differences in equipment, regional environments, and factory structures. Those working in traditional automatic control technology naturally wouldn't concern themselves with these non-fixed pattern issues.
Our ancestors had a valuable cultural tradition: emphasizing holistic interconnectedness. Ancient Chinese materialism, with its concept of the five elements (metal, wood, water, fire, and earth) and their interrelationships of mutual generation and restraint, while not scientifically rigorous, contained a reasonable element in its emphasis on overall interconnectedness. Traditional Chinese medicine views the human body as a whole, as exemplified by the meridian theory, which emphasizes the interconnectedness of the entire body. Although its scientific limitations are undeniable, the effectiveness of certain practices (such as acupuncture) still demonstrates the validity of its underlying principles.
A company is a large system containing many subsystems and components, various activities (design, processing, assembly, etc.), various resources (raw materials, tools, parts, equipment, manpower, etc.), suppliers, customers, etc. Are all these factors interconnected and influential within this large system? Certainly, they are—though this is based on imagination and intuition. But what is the specific extent of their impact on the overall efficiency of the large system? Senior managers and engineers may not be clear. Even within an equipment system, the interactions between its components, subsystems, operating parameters, environment, and other elements are often only qualitatively understood; the full extent of their influence is difficult to discern. In short, our understanding of the overall connections between a company's large system and its subsystems is very limited. This is not only due to the system's size and complexity but also because it is filled with the aforementioned uncertainties, unstructured processes, and non-fixed patterns.
A clearer understanding of the overall connections can help further improve the overall effectiveness of an enterprise.
It's not that people were unaware of the existence of interconnectedness and uncertainty in the past; rather, they lacked the tools and intellectual capacity to address these issues. Humanity has never ceased its pursuit of "supernatural" tools. Driven by the desire to more clearly understand and even more precisely manage interconnectedness, uncertainty, unstructured nature, and non-fixed patterns, humanity has finally created suitable tools: the Internet of Things, big data analytics, and artificial intelligence (especially the next generation). It is precisely because of these tools and methods that we can no longer allow interconnectedness and uncertainty to continue to trouble us, and the manufacturing field is no exception. Thus, we can more deeply understand the essence of intelligent manufacturing:
The essence and true meaning of intelligent manufacturing is to use advanced technologies such as the Internet of Things, big data, and artificial intelligence to understand the overall connection of the manufacturing system and control and manage the uncertainties, unstructured and non-fixed pattern problems in the system in order to achieve higher goals.
