It is the right time to reduce costs and increase efficiency. How can intelligent diagnosis be popular?

Artificial intelligence is leaping from the academic field to a new stage of practical application, and swarm intelligence technology driven by large models is becoming the core driving force for innovation.

For the automotive industry, the use of large models in vehicles is still in its infancy. As of June, more than 20 domestic automakers have implemented large models in vehicles, mainly covering smart cockpits and smart driving. In addition, a small number of large models are used in design, production, and sales.

What the entire industry is facing is to provide users with more personalized and cost-controlled service solutions more efficiently through the integration of man and machine throughout the entire user's full touchpoint driving experience.

Lotus CEO Feng Qingfeng revealed that Lotus' intelligent driving team has less than 400 people, compared with other car companies with 2,000 or even 3,000 people. The win-win situation of "minimizing personnel and maximizing effects" is thanks to the intelligent system.

It can be seen from this that comprehensive cost leadership has undoubtedly become the basis of competition among automobile companies, and intelligence enabled by artificial intelligence will become the deciding factor in the decisive battle.

Therefore, He Sicong, vice president of Airlabi, stated at the Automotive Blue Book Forum that the simplest application of artificial intelligence in automotive software is to supplement the human service knowledge base for automotive diagnosis. As there will be more and more problems with the electronics of our entire vehicle, the biggest application of artificial intelligence for after-sales service engineering will be to reduce costs and increase efficiency.

At present, it seems that how traditional diagnosis is integrated with new and more intelligent technologies is on the eve of an explosion.

New technologies bring about changes in service models

With the increasing maturity and popularity of OTA+ technology, diagnostic and analysis technologies are playing an increasingly important role in the automotive industry. From remote diagnosis, vehicle diagnosis to flexible data collection, big data visualization, and the collection and analysis of vehicle intelligent driving logs, these technologies have not only promoted the development of automotive intelligence, but also prompted OEMs to make profound transformations in traditional after-sales and service areas.

In the context of rapid technological iteration, OEMs are facing unprecedented opportunities and challenges. Traditional after-sales and service models are gradually being replaced by new service models such as data early warning, proactive service, and interruption of fault propagation chains. The new model relies on in-depth understanding of data and signals, as well as accurate association between failure modes and data information, to achieve real-time monitoring of vehicle status and predictive maintenance, which means a qualitative leap in the proactiveness and automation of diagnostic technology.

However, it is difficult for traditional dealers and maintenance personnel to have a deep understanding of complex technologies such as data, signals, logic, and strategies. How to quickly integrate intelligent maintenance and diagnostic technology solutions into current vehicle maintenance and diagnostic practices and use intelligent means to reduce costs and increase efficiency is the focus of the entire industry.

Huawei Released VHR Cloud Service 3.0 Qiankun Yunque Big Model, and intelligent vehicle diagnosis has been brought into the public eye. It is reported that Qiankun Yunque Big Model supports question-and-answer interaction. After entering the fault description, Yunque Big Model can automatically understand the problem through semantic analysis, conduct intelligent triage, formulate diagnostic plans, and generate diagnostic conclusions and repair suggestions. The whole process is automated, further reducing the original hour-level diagnostic time to minutes, which is a significant improvement compared to the existing remote diagnostic capabilities.

Is AI a strong assist?

Under the new competition background of annual model change, cars are no longer subject to minor or major changes in the traditional sense. Different brands and models of new energy vehicles are frequently iterating, bringing with them numerous software versions and complex data structures, which has become one of the challenges faced by intelligent diagnosis.

The keyword of Huawei's solution is undoubtedly AI. On the other hand, OEMs use artificial intelligence to improve their system efficiency when building and managing their own intelligent platforms. It is not easy to make full use of the value of AI in the field of diagnosis.

"In the past six months, our team has read a large number of AI papers and tried different large models at home and abroad to analyze quality and after-sales issues. The results are not satisfactory and the after-sales department is not satisfied." A technician from an OEM discussed this with Airlabi, and such lament is not an isolated case.

In fact, the current AI large-scale model's logical deduction capabilities are obviously insufficient. It is unable to provide accurate information for areas with professional requirements, such as failure analysis and maintenance plans, and is not technically feasible.

This shows that the underlying data and knowledge base capabilities directly affect the intelligent effect of AI in diagnosis.

When discussing the areas of R&D quality and after-sales technology, Chen Zixin from Airlabi Diagnostics BU elaborated on the key to how AI and computers can accurately understand and analyze vehicle logic and signals. He pointed out that the core of success does not lie in AI technology itself, but relies on three key technical prerequisites:

1. Improvement of knowledge graph: After more than ten years of development, knowledge graph has made great progress. Its application in the field of diagnosis can better understand and analyze the associations and dependencies between various automotive components, integrate information such as various components, fault types and solutions, and form a comprehensive knowledge base.

It should be emphasized that building a knowledge graph suitable for the automotive industry requires rich project experience and continuous iteration and maintenance by engineers with experience in vehicle subsystem development and operation. On this basis, combined with the AI ​​big model, the powerful combination of knowledge graph + AI can complement each other to improve the development efficiency of the knowledge graph and achieve a leapfrog improvement in diagnostic time from hours to minutes. However, it cannot be ignored that the difficulty and cost of building a knowledge graph should not be underestimated.

2. Platform integration capabilities: The complex and independent systems of OEMs also pose challenges to diagnosis. The intelligent diagnosis platform needs to have strong integration capabilities to connect and integrate data from dozens of systems and have a deep understanding of the OEM's business scenarios and underlying designs. It must have professional know-how in diagnosis, network messages, MPU logs, OTA, accessories, maintenance information, etc., and know how to combine these data to assist AI in efficient analysis.

3. Refined implementation of data governance: In the process of comprehensive integration and governance of source data, special attention is paid to the detailed sorting of key information sources such as SSTS (service support system), DEFMA (design failure mode and effect analysis), maintenance manuals, engineering schematics, etc. Through advanced atomic splitting technology, the complex data set can be refined to the most basic component units, and a close linkage mechanism is formed with vehicle master data management to ensure that each model data strictly follows the vehicle-specific configuration attributes and detailed breakpoint information.

In addition, in order to maximize the value of data, it is necessary to deeply integrate diagnosis with the vehicle's full life cycle software management platform, and closely link problem analysis with high-frequency OTA activities to enhance data value.

Without a solid foundation, even if the OEM invests a lot of resources in AI development, it will often be ineffective. Chen Zixin emphasized that the intelligent management and application of vehicle data is not achieved overnight, but a long-term and continuous process that requires continuous optimization and updating based on dynamically changing vehicle information, R&D information, and after-sales information. Therefore, a long-term, stable partner with deep involvement in the automotive industry will be an indispensable and powerful assistant for OEMs in the field of AI applications.

Three can

Healthy competition is to move away from the lower-level volume price and move up to the upper-level volume technology. The trend of intelligent diagnosis has arrived, and AI manufacturers are flocking to it. However, from the perspective of industry practice, Internet and AI manufacturers that are not deeply integrated with the automotive industry will inevitably face the problem of acclimatization. This is also the reason why the current OEMs and AI have not quickly joined hands.

Ailabi believes that intelligent diagnosis is a technical shortcut to reduce costs and increase efficiency. However, to do a good job in intelligent diagnosis, we need to ask "three questions": can the platform be integrated, can intelligence be implemented, and can data strengthen the foundation?

With the increasing complexity of automotive systems, traditional decentralized platforms can no longer meet the needs of efficient and accurate data processing. By integrating TCE and BOM, Airlabi's A6 platform can seamlessly connect dozens of system platforms of car manufacturers, break data silos, and achieve data interconnection. At the same time, A6 can collect, organize, and analyze data from different data sources to achieve data standardization, unification, and visualization. It provides a very valuable data governance and integration foundation for the construction of an intelligent diagnostic platform.