In the locally deployed DeepSeek dialog box at LiuGong, a line was typed: "How can AI large models empower the construction machinery industry?" Twelve seconds later, it gave a deeply reasoned answer covering construction machinery design, production and manufacturing, equipment operation and maintenance, predictive maintenance, construction efficiency, safety and low carbon, and service model innovation.

And these contents have gradually been implemented and applied at LiuGong. At the Loader New Technology Research Institute, engineers are designing new product solutions through the LiuGong large model; at excavator service and repair sites, the LiuGong AI repair expert is guiding service personnel to accurately locate faulty parts; at the simulation laboratory, engineers are collaborating with the large model to write simulation experiment code; in the R&D department, colleagues currently engaged in new product planning and design are using the large model to extract product market demands from tens of thousands of bid documents... This intelligent manufacturing revolution triggered by AI large models is reconstructing a new paradigm for construction machinery, and the large model has long quietly become a digital employee, learning about and participating in everything at LiuGong.

Among them, LiuGong's project "Service Intelligent Question-Answering System Based on Unstructured Data Assisting Rapid Fault Diagnosis" won the Ministry of Industry and Information Technology's "2024 Typical Practice Case of Data Element Application Scenarios in the Industrial Field," and the project "Application of Generative Artificial Intelligence (AIGC) in Knowledge Question-Answering in the Construction Machinery Industry" was selected as a regional "2024 Guangxi Enterprise Typical Innovation Case," becoming a model application in the industry.

Transforming manual experience into digital assets

With the advancement of LiuGong's "comprehensive internationalization" strategy, how to overcome language and time zone barriers and provide global customers with timely and effective after-sales service for the full product series is a severe challenge facing the enterprise.

The AIGC team of the LiuGong Digital Research Institute injected LiuGong's accumulated product fault repair records, maintenance manuals, and operation manuals into the large model, and combined them with connected vehicle data to create the industry's first AI fault diagnosis and repair expert, "Liu Xiaoyuan." After a question is entered, this "AI expert" will even ask follow-up questions to further clarify the fault point, and then quickly and accurately provide fault diagnosis and repair solutions. After application in the loader, excavator, and industrial vehicle business units, its product fault diagnosis accuracy exceeded 87%, helping to improve repair efficiency by 14%.

This not only optimized the fault handling process and resolved the prominent contradiction between the rapid growth of LiuGong's global equipment sales and the shortage and insufficient capability of service personnel, improving the service experience of global customers, but also effectively preserved, organized, and utilized LiuGong's valuable experience accumulated over many years, encoding decades of practical experience into reusable digital assets.

Leaping the value of knowledge

Facing LiuGong's vast R&D knowledge reserves, the Digital Research Institute used the large model as a foundation and, by injecting LiuGong's R&D knowledge assets, formed a closed-loop knowledge chain and tools covering the entire R&D process, achieving improved design efficiency and maximized knowledge value, raising the knowledge utilization rate from less than 30% to 92%.

From "parts" to "whole," and from "whole" to "parts." The large model organizes scattered R&D knowledge assets and then uses the vast knowledge base to output "smart moves," helping improve the efficiency of product design and development from components to complete machines. It is worth mentioning that in the parts standardization work, LiuGong's R&D design team used the large model to automatically classify parts and extract key parameters, increasing overall standardization efficiency by more than 5 times, thereby effectively utilizing existing resources and reducing the company's overall design and development costs.

In the process of LiuGong's globalization, the expansion of enterprise scale has brought multiple challenges in talent pipeline development, management system upgrading, and the integration of massive amounts of operational data. With the help of the large model, the efficiency of massive data processing and result analysis output has improved significantly, not only breaking down data silos between departments but also precipitating scattered information into strategic assets of the enterprise.

Subsequently, the Digital Research Institute will gradually expand the coverage of the local large model at LiuGong, build a complete LiuGong knowledge graph, and, on the premise of ensuring data security, create an omniscient intelligent agent for LiuGong knowledge.

Standing at the intersection of the Fourth Industrial Revolution, when large model technology knocks on LiuGong's door, what LiuGong does is not only a single-point breakthrough empowered by AI, but also an intelligent ecosystem covering the entire value chain of "research, production, supply, sales, and service." LiuGong will overcome the "last mile" of implementing large models at LiuGong, bring the unique "talent" of large models into every complex scenario, and comprehensively build a digital and intelligent LiuGong.