Predictive Maintenance · Risk Management
Shengtong Technology needed a way to make industrial inspection models more accurate and easier to adapt across power, mining, and chemical sites without buying large amounts of local GPU infrastructure.
Reported impact
Shengtong improved model accuracy and generalization, sped up data processing and model iteration, and reduced the need for expensive on-premises compute in industrial inspection projects.
Shengtong Technology used Huawei Cloud and the Pangu mining foundation model to build a layered industrial inspection platform that turns generic AI into site-specific inspection applications. At the base, the company trained a 3-billion-parameter 'Shengtong industrial inspection large model' on top of the Pangu mining model, then produced 700-million-parameter power, coal, and chemical industry models and finally fine-tuned L2 application models for specific customer workflows. Huawei Cloud supplies the integrated train-and-infer development environment, elastic compute, and storage needed to process raw inspection data, run pretraining and fine-tuning, deploy models, and iterate them continuously. This multi-level architecture lets the same stack handle personnel safety, equipment condition, production process, and site-environment inspection while reducing dependence on local GPU hardware and making anomaly recognition more generalizable across industrial sites.
Inspection data and images are collected into Huawei Cloud, where the train-and-infer platform processes them for pretraining, industry adaptation, and scenario fine-tuning. The resulting application models are then deployed back into field inspection workflows to identify abnormal people, equipment, process, and environmental conditions.
"Huawei Cloud provides the train-and-infer integrated platform, elastic compute, and storage that speed data processing and model iteration while lowering local hardware cost. (original: 基于华为云部署训推一体化开发平台，提供了模型预训练、微调、部署和迭代的功能，形成了AI开发的完整闭环。... 在云环境下，软件平台及算法模型的部署、更新和维护变得更加简单和高效。)"
"Shengtong built a 3-billion-parameter industrial inspection large model on top of the Pangu mining model and then produced 700-million-parameter power, mining, and chemical industry models. (original: 基于盘古矿山大模型底座及华为云算力支持，圣瞳研发了30亿级参数的预训练模型"圣瞳工业巡检大模型"，并基于此推出了电力、煤矿、化工等7亿级参数的行业大模型。)"
"Shengtong uses a multi-level model hierarchy: the Pangu model as L0, industry models as L1, and application models as L2 for customer deployment. (original: 基于盘古大模型（L0）提供通用基础能力，利用特定行业数据预训练（电力、煤矿、化工等）行业大模型（L1），经过细分业务微调生成应用于业务场景部署应用模型（L2）。)"
"The company provides end-to-end industrial AI services for data management, model training, model application, and device access, covering personnel safety, equipment condition, production process, and site environment inspection. (original: 圣瞳科技为工业客户提供全链路AI解决方案，通过建设基于工业AI的数据管理、模型训练、模型应用、设备接入等服务，一方面满足工业场景下对于人员安全、设备状态、生产流程、现场环境的巡检需求。)"
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Huawei Cloud Sample Archived · Jan 15, 2026 Conch Group Conch needed to scale AI across cement production, but traditional model development was slow, hard to generalize, and too weak to support real-time quality, energy, equipment, and safety optimization at industrial… 28 equipment-monitoring scenarios AI training center · Huawei Cloud Stack · NLP large model Huawei Cloud Sample Archived · Aug 26, 2025 Fengtu Technology Fengtu Technology needed to run GPU-heavy inspection and safety models in the cloud without wasting compute, while also improving fatigue detection and commercial-vehicle risk response for logistics customers. around 10% reduction in escort labor input DMS fatigue detection · FunctionGraph · ResNet image classifier 
Snowflake Sample Modified · Aug 7, 2026 Uniper Volatile energy markets demanded faster access to trading, risk, and operational data than Uniper's aging reporting platform and fragmented systems could provide. 18.5 gigawatts of power-generating capacity Snowflake · Snowflake CoCo · Snowflake CoWork 