| Source: Auto Parts Circle
| Submission/Business Cooperation: Please add WeChat juyou18611756963
Expert systems have reached a consensus on their benefits to the technological progress of a globalized society, and efforts must be made to realize the evolutionary process of systems, forming an advanced paradigm aimed at breakthroughs in social applications.
Author | Chen Guangzu
Image | Network
△ Senior Expert in Automotive Industry Chen Guangzu
Automotive intelligence is a rising trend in the current automotive industry, representing a “brain revolution” that embodies signs of modernization in technology and knowledge revolution. Expert systems can be seen as an emerging branch in the field of automotive intelligence, being widely, actively, and enthusiastically researched, manufactured, and applied, potentially forming a new type of product in the automotive industry that embodies wisdom, advancement, ecology, civilization, and wealth. It will also become one of the most active and practical areas in artificial intelligence, facilitating significant breakthroughs in knowledge from theoretical discussions to practical applications. Expert systems (EC-expert control) are a knowledge engineering of modern intelligent systems. They contain a vast amount of knowledge and expertise from specialists in a specific field, processed by computers, forming human expert knowledge and problem-solving methodologies for real-world issues in that field. Building an expert system, or knowledge engineering program, typically begins with the application of experts known as knowledge engineers or through special forms of collaboration among multiple experts in a specific field to query, solve, and analyze relevant problems, gathering strategies and experiences. Furthermore, it aims to compute the imaginative and efficient high-level knowledge of professionals in a limited manner, forming a specialized computer language and tools to serve end-users, conceptually expressed through the three major elements of knowledge acquisition, expression, and application. Currently, the structure of expert systems mainly consists of: databases, which continuously store facts, evidence, hypotheses, and goals, primarily including parameters and thresholds from expert experience and summarized experimental methods. Rule bases: contain rules provided by experts, typically describing methods for knowledge usage, which process data in complex and rapidly changing environments using natural methods, as well as rules for system monitoring and diagnostics during application. Inference engines: determine the order of rule application, focusing on whether the predictive control scheme is suitable for the problem domain. Human-machine interfaces: generally come in two forms: one is the user interface during operation, which includes interpretation tools to assist user inquiries, and the other is for updating the knowledge base through rule modifications and edits to further enhance the overall application level of the system.
Currently, modern expert systems possess characteristics such as high performance, heuristic capabilities, accuracy, comprehensiveness, practicality, and transparency. Therefore, there are numerous conditions and challenges in building such expert systems. We believe that the following key tasks must be accomplished.
Building a highly specialized team of experts: Expert systems require knowledge to be narrow and targeted toward a specific field, but must possess a specialized, high-level team of experts to form a high-efficiency application capability.
Possessing good symbolic processing capabilities: The ability to quickly and accurately use computer symbols to express information and knowledge related to their field, and to perform strong processing and reasoning functions, focusing on knowledge rather than just data.
Possessing high-level problem-solving capabilities: To enable expert systems to reason correctly and align with problem-solving rules, specialized domains and various problems and difficulties should be established, indicating that the reasoning of expert systems must be deep and adaptive, capable of addressing problems amidst constant change and uncertainty.
Having excellent explanatory functions: Expert systems should be able to answer questions like “Why is this done?” “What benefits does this approach have?” and “What goals can it achieve?” This explanatory function incorporates many independent information modules from various expert domains, helping expert systems better integrate artificial intelligence functions, enhancing application levels and effects.
Further developing intermediate databases: Also known as blackboard systems, the focus of the blackboard is to better control the effective working process of expert systems in dynamic states, which is also referred to as the construction of a “dynamic knowledge base.” The middleware development of expert systems engineering software involves considerable intellectual labor and talent, requiring efforts to keep pace with the continuous upgrades of expert systems, addressing the potential “software crisis” and its associated risks.
The Evolution of Expert Systems
Expert systems represent a new type of soft science, and the discovery, invention, and experimental application of science do not progress smoothly like a stream; they often experience individual actions, decentralized processing, continuous exploration, controversies, and developmental pauses. It is precisely this arduous evolution, that has led to a clear consensus on the role of expert systems in promoting technological progress in a globalized society. Efforts must be made to realize the evolutionary process of systems, forming an advanced paradigm aimed at breakthroughs in social applications.
According to this pattern, in 1965, a team led by Professor Feigenbaum at Stanford University created the first expert system for reasoning about chemical compound molecular formulas and inferring mass spectrometry data, which is considered the world’s first invented expert system.
Currently, developed countries such as the United States, Japan, Germany, and Israel are actively researching and implementing expert systems, and China is also a key country in the development and application of expert systems. For example, the Hefei Institute of Intelligent Machinery at the Chinese Academy of Sciences is vigorously researching expert system applications, and Peking University has established a major in sociology and statistics for expert systems, focusing on better developing services in the financial industry. Since the 1990s, steel companies such as WISCO, Benxi Steel, Kunming Steel, and Baosteel have introduced Finnish blast furnace expert systems to develop “artificial intelligence blast furnace smelting” expert systems, thereby providing conditions for the application of expert systems in the steel industry. Thus, it can be seen that over the past seventy years, the evolution of expert systems globally has roughly gone through three eras, and is currently moving towards the fourth era.
Initial Stage: As mentioned above, in the 1950s, the E.A. Feigenbaum team discovered the world’s first dendral chemical molecular structure. During that era, logic and simulated psychology were primarily used, along with some universal disciplines, to establish expert system solving procedures, marking the entry of expert systems into an early development stage, which represented a significant advancement in human technological discovery. However, as a system, expert systems at that time were merely individual scientific inventions, laying a good foundation for future development, which must be acknowledged. But the structure and system of expert systems still showed weak problem-solving capabilities, which is a natural sign of early development.
Mature Stage: This occurred around the early 1980s, where expert systems developed over nearly twenty years achieved significant progress across various professional fields, but the degree of advancement was inconsistent, with some areas progressing more than others, generally remaining in a heuristic application era. Theoretically, for certain single disciplines, considerable progress was made in forming relatively complete and transferable structures, and effective experiences and results were achieved in human-machine interfaces, explanation mechanisms, knowledge acquisition techniques, handling uncertainty reasoning, and enhancing expert systems’ knowledge expression and processing methods, thus allowing expert systems to penetrate social related platforms and achieve good benefits.
For example, advancements in physics, chemistry, mathematics, medical treatments, agriculture, meteorology, military, engineering technology, space technology, and certain manufacturing industries have shown significant results. As a result, the international scientific community recognizes that expert systems have reached a relatively mature development stage, but progress in the automotive industry has not been very ideal.
Collaborative Stage:Entering the 21st century, with the massive upgrades in global technology, information, networks, energy, and computer hardware and software, especially the magical role of software, the crystallization of intellectual labor, has propelled the rapid development of computer data technology, pushing human society into a digital development era. Consequently, the trend of social intelligence development has become increasingly evident, and expert systems have also advanced, becoming one of the most active disciplines in artificial intelligence.
The obvious characteristic of this era is that expert systems are driven from an objective level, while also shifting to the micro level, characterized by providing substantial assistance in promoting the application of high-tech. Methodologically, the collaborative design approach of computers is employed, promoting parallel design while integrating new methods, technologies, and models of modern computing to achieve integrated application efficiency, optimizing product design processes and quality through specialized product structure modeling.
This shared object, collaborative activities, and dynamic alliances provide a good platform for the miniaturization of expert systems, such as in graphene, nanotechnology, superconducting technology, laser technology, virtual reality technology, quantum communication, quantum photonic computing, 5G promotion, microbial applications, artificial applications of carbon dioxide and methane, and advanced carbon fiber, all playing a beneficial role.
Endogenous Stage: This represents a future or currently developing ecological expert system model, also known as the fourth generation expert system. It mainly indicates a microcomputer simulation and integration control-based approach, promoting collaborative development between expert systems and intelligent inner systems such as neural networks, dissipative structures, chaos theory, and fuzzy control, which may help artificial intelligence absorb more nutrients, enhance its comprehensiveness and applicability, and through collaborative development, highlight self-organizing and self-learning functions, where many specialties exhibit brain-like behaviors and often surpass human intelligence. Some aspects display the relatively independent thinking of expert systems, which is a profound manifestation of various intelligent thoughts merging, thus resulting in the effectiveness of specialized expert systems, meeting opportunities for their higher-level development. This new type of endogenous expert system can also be called a generalized expert system, and its promotion will play a positive role in various aspects of our modern high-quality development society.
●Here I will first discuss some expert system-related examples I have experienced or heard about. Automotive air conditioning is a crucial matter for driving in both winter and summer. Currently, some vehicles have realized temperature arrangements in different zones, but in recent years, the international automotive community is designing and testing a personalized distributed air conditioning supply system. When driving in summer with air conditioning, the driver may be accompanied by elderly parents, a middle-aged person, and children aged five or six. The vehicle is equipped with advanced sensors and electronic execution systems that measure each person’s skin to determine the appropriate temperature and airflow needed to ensure that every passenger feels comfortable. This distributed device is called an automotive air conditioning expert system, which I heard about at the University of Michigan.
Currently, some cars are equipped with air spring suspension and come with a smart electronic control and air compressor system. When driving on mountainous or uneven roads, the air suspension system raises the vehicle chassis to adapt to the driving requirements of the mountain roads. When driving on the highway, the air suspension lowers as much as possible to reduce air resistance and enhance safety during high-speed travel.
To promote lightweight applications in automobiles, the body and related components will ultimately use all-carbon fiber materials, which weigh only ¼ that of steel, yet have a tensile strength of over 3500Mpa, seven times that of steel or more. This means that in the event of a collision, it can exhibit certain impact resistance and achieve rebound elasticity, benefiting human protection. It is understood that currently, the main carbon fiber plate connection applications use highly adhesive agents instead of welding, which is insufficient, and additional rivets are added. I have seen prototype cars at General Motors in the USA.
In the U.S. and Europe, drivers primarily drive themselves. I saw at Ford a smart driver’s seat made of full cowhide, equipped with more than ten miniature motors and air compressors. When purchasing a car, the driver first undergoes ergonomic measurement, and the computer adjusts to your most suitable state for driving. If your spouse is also measured, then when switching drivers, a press of a button will adjust the seat to fit your spouse’s ergonomic condition. Furthermore, with the continuous enhancement of automotive intelligence, cars also implement OTA (Over-The-Air) updates like smartphones and computers, especially in the cockpit improvements, which can be very attractive. Different versions of OTA may incur costs or be free, significantly enhancing the personalization of automotive applications and increasing profits for manufacturers and after-sales services.
Automatic steering and braking systems in vehicles. Achieving fully automated driving (L5) in the world’s vehicles may take about twenty years, around 2040, but we should first implement the two intelligent tasks mentioned earlier. Recently, there have been reports of multiple incidents where drivers mistakenly pressed the accelerator instead of the brake pedal, leading to accidents where cars fell into rivers or crashed into buildings. I witnessed a situation on the German autobahn where more than a hundred cars collided, and similar incidents occurred in the USA with two or three hundred cars crashing together. The automatic steering application of expert systems and neural network intelligent control systems can identify erroneous directional inputs from drivers, achieving intelligent replacement behaviors to guide vehicles in the correct direction.
The automatic braking system can recognize incorrect directions and, even if the driver presses the accelerator, the expert system and neural network adaptive controller will convert the accelerator function to that of a brake pedal, utilizing dynamic braking characteristics to achieve the desired deceleration or braking performance. This demonstrates the expert system’s error correction capabilities.
We have conducted extensive research on preventing head-on collisions, but there has been insufficient research on preventing side collisions. In the early years, I observed a few professors at the University of Michigan who formed a software company dedicated to researching the prevention of side collisions and mitigating their hazards. After repeated validations, General Motors acquired their patents and collaborated on reducing the risks of side collisions, applying digital KBE technology in product research, which combines knowledge systems with engineering systems. KBE differs from computer-aided design (CAD) systems, which are primarily secondary development tools for custom program editing; KBE utilizes the integration, openness, and scalability between knowledge programming and engineering design programming.
Therefore, regardless of the digital approach, we should establish more specialized micro software engineering companies to enrich the multi-level professional advantages of our automotive industry. In regions like Silicon Valley and Detroit in the USA, there are numerous micro software engineering enterprises worth learning from and promoting. We need to regard automotive software design engineering as an independent profession, focusing on a microcomputer specialty product, where a small team can delve deeply and produce software engineering programs that others do not know or cannot achieve. The more such micro-enterprises there are, the stronger the foundation for innovative prototypes in automotive technology, as the market and wealth are created through the artistic creations of these micro-enterprises. We cannot view these micro software engineering companies merely as traditional small businesses; they are essential for the development of our automotive industry, which is continuously evolving.
In the current automotive industry, the entire process has applied digitization, programming, logic, and flexibility platforms, significantly enhancing the design and manufacturing processes. For instance, CAD (Computer-Aided Design), CAM (Computer-Aided Manufacturing), CAE (Computer-Aided Engineering), and CAPP (Computer-Aided Process Planning) have all been implemented. At the end of the last century, CAT (Computer-Aided Testing Engineering) was further developed, greatly improving the digital verification process of product design and saving labor costs.
For example, in the early 1980s, General Motors had to conduct three to four crash tests on new vehicle designs almost daily. Later, they switched to CAT crash application engineering, allowing them to conduct physical crash tests once every week or two. At the General Motors research center, I saw engineers conducting vehicle test field longevity tests on computers, but to ensure accuracy, they often had to stay at the computer station, transferring data at night to Bangalore, India, for continued testing. Since it was night in the USA, it was daytime in India, allowing for continuous operation.
Currently, CAT applications are quite complex, including many new energy vehicles, requiring the establishment of a new set of testing methods.
Concept of Expert Systems in the Automotive Industry as a “Forest” and “Tree Root”
The so-called “forest” concept refers to including all fields of the automotive industry, where expert systems, particularly generalized expert systems, can be developed and applied based on their unique characteristics and effectiveness, striving for better and more effective outcomes. It is especially important to focus on the fact that the modern automotive industry is undergoing major transformations and intense competition, leading to the continuous emergence of new automotive-related fields. We should establish comprehensive and specialized expert system think tanks to professionally address the automotive expert system industry.
Currently, it is widely believed that the development of the automotive industry “must be software-centered,” which is essential, but ultimately, it must be implemented in the real economy of the automotive industry, promoting the automotive industry towards high-quality, faster development. This is the necessary path for us to transition from being a major automotive country to a strong automotive nation. The establishment, processing, and application of the “forest” expert systems must be promoted in various fields such as automotive design, research and development, testing, planning, supervision, forecasting, diagnosis, repair, digitization, control, analysis, maintenance, standards, troubleshooting, and emergency response.
The meaning of “tree root” refers to a towering tree with dense branches and a high stature, which is well supported by unseen roots underground, which are essential for the tree’s healthy growth; the deeper the roots, the more flourishing the tree, and if the roots decay, the tree withers.
If the large tree represents a luxury car, then the roots consist of thousands of components that make up the car, especially the key core technology components. For example, the four core technologies of intelligent automatic driving include environmental perception, precise positioning, path planning, and linear execution. For electric vehicles, the core technologies are the “three electrics”: battery, motor, and electronic control. In connected vehicles, software design and application are core technologies, with software becoming the evolving DNA of automobiles; while hardware is essential, the invisible software is needed for enhancement. Data is the core power driving automotive evolution, and a major concern is the automotive chips, mainly referring to automotive-grade functional chips, power semiconductors, and sensors, most of which are imported, becoming a critical bottleneck for the automotive industry.
We must be very clear that key technologies cannot be obtained, bought, or begged for; they must be earned through our hard work. We should strive to achieve results similar to the glorious achievements of “two bombs and one satellite” during those relatively backward years. This reflects our long-standing approach in developing the automotive industry, focusing on quantity rather than quality. The emphasis on main engines over components is particularly evident in many local governments that only build car factories, showcasing long assembly lines of one thousand seven or eight hundred meters for leaders and foreign friends to visit, while neglecting small, focused, remote component enterprises. This traditional mindset has persisted for seventy years without fundamental improvement, reflecting formalism and bureaucratic thinking in the automotive industry.
We currently have one hundred seventy or eighty automotive manufacturers, and the number is still increasing, while the USA originally had three major automotive manufacturers. However, Chrysler has undergone several reorganizations due to insufficient scale, and now only two and a half traditional automotive manufacturers remain in the USA. In contrast, the automotive industries in Japan and Germany are thriving, fundamentally supported by strong component enterprises.
I have heard foreign friends say that the future development of personalized mass production business models will eliminate the long assembly lines of over a thousand meters, opting instead for multiple shorter assembly lines of forty to fifty meters. The assembly lines for large components will also be converted to U-shaped single-process production lines, which are being implemented in many of our component enterprises as well. The contemporary automotive industry is undergoing a significant transformation, but with increasing resistance from de-globalization, we must enhance our independent capabilities. We must dare to overcome local protectionism and bureaucratic thinking and methods, under the guidance of building a new structure for the unified national market, focus on domestic demand, actively engage in “double demand” activities, promote automotive technology innovation and industry upgrading, and enable the automotive industry to enter a new modern development pattern, ensuring that our goal of becoming a strong automotive nation is achieved.
In summary, expert systems should be widely promoted in the automotive industry, with an emphasis on deepening applications at specific points, and the combination of points and areas will yield more experience and results.
