Applying the APACHE Scoring System to Notebook Health Assessment in Industrial Big Data

Background and Company Overview – Lenovo’s wholly‑owned subsidiary, Lianbao Technology, is the world’s largest PC R&D and manufacturing base, shipping 1.7 billion units by 2020 and operating a national intelligent‑manufacturing demonstration base.

Intelligent Manufacturing Architecture – The architecture consists of four layers: smart‑technology (IoT, edge, digital twin, 5G, blockchain, big data, AI), five capabilities (device, edge, cloud, network, intelligence), four‑fold goals (personalized products, supply‑chain collaboration, proactive service, intelligent decision‑making), and a full‑value‑chain covering design, production, sales, and service.

Industrial Data Landscape – Lianbao collects massive high‑dimensional heterogeneous data across supply‑chain, design, production, testing, repair, and after‑sales, generating over 180 billion product‑test records and 35 trillion time‑series points annually.

PSI Metric Management – Using the self‑built Olympia platform and AI techniques, customer feedback from e‑commerce channels is mined to improve product design, manufacturing, and quality control, thereby enhancing user experience.

APACHE System Introduction – Originally a medical acute physiology and chronic health evaluation system, APACHE provides a multi‑dimensional health‑score model that combines physiological indicators, age, and chronic conditions; the article adapts this concept to notebook health assessment.

Notebook Health‑Assessment Framework – By mapping notebook status indicators (power, temperature, etc.) to APACHE‑style scores, a health‑degree evaluation system is built, including data similarity comparison, a big‑data analysis architecture (data, decision, and application layers), and a health‑analysis workflow.

Decision‑Layer Modeling – Two model categories are presented: (1) statistical distribution models (e.g., normal distribution of CPU temperatures with 6σ thresholds) for anomaly detection and supplier quality monitoring; (2) machine‑learning models (isolation forest, clustering, ensemble learning) for fault detection and predictive maintenance.

Application Layer and Use Cases – Anomaly detection covers component (board, disk) and functional issues (freezes, overheating). Case studies include (a) disk anomaly detection using SMART metrics combined with production data, and (b) motherboard component health scoring using time‑series voltage and frequency data.

Outlook – While AI is mature in internet, finance, and e‑commerce, its application in industrial manufacturing remains a blue‑sea opportunity; Lianbao plans to continue exploring big‑data‑driven AI solutions to advance the industrial internet ecosystem.