Solving Non-Expert Domain Issues with Data Within Siloed Organizational Structures

gpt-4-turbo has translated this article into English.

Background

At the beginning of my employment, I was in charge of App and GUI development when the product frequently stopped working in the field, prompting the company to assign additional field support tasks.

The product was based on a control system and embedded environment, and at that time, I had no experience in embedded or control domains.

Additionally, the embedded and control system departments operated in a silo structure, which hindered information sharing and collaboration.

Problem Definition

• The product stopped more than 3 times a day on average
• Recovery was only possible through manual reset
• The root cause had not been clearly identified

Problem Approach and Execution Strategy

1. Data-driven approach
   • Directly collect status data from the field
   • Log sensor values, control statuses, and location information at the time of issue occurrence
   • Extract common patterns
2. Collaboration with domain experts
   • Continuous communication with the system control development team
   • Learn about control state flow and the operating principles of safety switches
3. Anomaly detection and analysis
   • Identify conditions where location/sensor values exceed thresholds leading to stops
   • Attempt to adjust thresholds → root problem not solved
   • Plotting sensor data reveals a symptom where magnetometer readings are periodically incorrect
4. Maintaining problem definition amidst organizational resistance
   • Embedded team repeatedly responds with: “Sensor issue, no improvement possible”
   • Document, quantify, and analyze the phenomenon based on reports grounded in numerical analysis

⚙️ Execution Results and System Improvements

• Developed an event definition and alarm system for symptom detection
• Set up a data transmission pipeline for collecting field sensor data
• Developed a subsystem capable of remote reset
• Improved service stability by automating and remote controlling repetitive field reset tasks

Final Outcome

• Decrease in incident occurrence, reduced field response burden
• Enhanced operational efficiency with remote recovery capabilities

• Problem fully resolved months after resignation through firmware modification,

  with confirmation from internal employees that the analyzed cause was correct

Key Competencies

• Proactive problem definition and execution in a non-expert domain
• Facilitating collaboration and demonstrating execution skills within organizational barriers
• Complete experience from data-based analysis to operational automation design