Recently, CASCO's Intelligent Operation and Maintenance Platform has been deeply integrated with the DeepSeek large models, becoming the first in the industry to transform from a "Q&A assistant" to an "assisted diagnosis tool", with a fault diagnosis accuracy exceeding 99.7%. Currently, this feature has completed on-site verification in Jinan Metro, promising to establish a new paradigm for intelligent operation and maintenance of rail transit in the AI era.

Three Technological Innovations Leading the Industry

The quality of equipment operation and maintenance is an important guarantee for ensuring the safety and efficiency of metro operation. CASCO's self-developed IT-PHM Intelligent Operation and Maintenance Platform aims to solve many industry pain points faced by traditional operation and maintenance, providing users with out-of-the-box and flexibly configurable applications. It serves as a "family doctor" in rail transit equipment maintenance.

Through the private deployment of the DeepSeek models, CASCO has successfully connected AI with the data models, mechanism models, and business modules in the intelligent operation and maintenance platform. It has taken the lead in achieving three new technological breakthroughs in the field of rail transit intelligent operation and maintenance, injecting AI superpowers into the "family doctor" of intelligent operation and maintenance.

Accurate Enhancement of Fault Data Sets: Utilize DeepSeek to generate high-quality fault data sets, effectively solving the problem of few equipment fault data with poor quality, and enhancing the iterative training of the platform.

Accurate Diagnosis with Small Models: Through refined prompt design and model fine-tuning, it achieves accurate fault diagnosis with small parameter models, with an accuracy exceeding 99.7%, significantly reducing operation and maintenance computing costs.

Deep Integration of Agent Technology: Comprehensively utilize knowledge bases, retrieval-augmented generation (RAG), and agent technology to enhance the platform's reasoning and analysis capabilities, providing more professional and accurate operation and maintenance decision-making guidance, promoting the "Q&A assistant" to a "diagnosis expert".

Two Typical Applications with Significant Effects

After on-site verification, CASCO's two intelligent operation and maintenance scenarios based on DeepSeek have successfully demonstrated their effectiveness, showcasing high-performance intelligent applications.

Autonomous Equipment Health Analysis: Breaking through the traditional paradigm of mechanism analysis and constructing a closed-loop system of "autonomous learning - case accumulation - capability evolution". Through the intelligent analysis engine driven by large models, it effectively solves core pain points such as the long cycle of expert experience accumulation and low system iteration efficiency, enabling continuous self-evolution of equipment anomaly recognition and health analysis, and providing accurate decision-making support for complex operation and maintenance scenarios.

Fault Emergency Organization and Guidance: Innovating the traditional knowledge base search model and establishing a dynamic response mechanism of "knowledge update - scenario matching - intelligent generation" to achieve intelligent coordination of personnel division, maintenance guidance, tool reminders, and passenger announcements during fault emergencies. Relying on the multimodal understanding capabilities of large models, it not only ensures comprehensive coverage of maintenance solution libraries, but also achieves dynamic adaptation to personalized scenarios, effectively resolving the contradiction between fragmented maintenance experience and precise guidance needs, significantly improving emergency response quality.

In the future, CASCO will continue to deepen the application of AI technology with large models as the core, to directly address pain points of concern to clients in the equipment operation and maintenance field, and expand the depth and breadth of "AI + rail transit" integration, thus further empowering the intelligent upgrade of the rail transit industry.