Omron Field Engineering Systematizes Highly Personalized Maintenance Operations with AI Judgment Model Co-developed with Matsuo Research Institute

Omron Field Engineering Co., Ltd. (Headquarters: Chuo-ku, Tokyo, President: Taisuke Tateishi, hereinafter OFE) has jointly developed a new system with Matsuo Research Institute Co., Ltd. (Headquarters: Bunkyo-ku, Tokyo, Representative Director: Tobuku Kawakami) that automatically judges inspection photos taken at work sites using AI.

This system is an AI judgment technology jointly developed with the aim of realizing field DX, merging Matsuo Research Institute's generative AI and image recognition technology with the site knowledge of maintenance operations cultivated by OFE over many years. As high effects have been confirmed through practical operation at internal sites, full-scale introduction began in October 2025.

In recent years, against the backdrop of a declining working population, it has become an important challenge in the field of maintenance and inspection operations to balance the maintenance of work quality with the efficient utilization of human resources. OFE has been responsible for maintenance and operation services in social infrastructure areas such as railways and finance, and has provided high-quality services for many years. On the other hand, the work of checking inspection photos, being an important double-check process, has also been an operation that has heavily relied on manual labor until now.

In this project, we focused on "verifying equipment setting values (large amount of text information)," which is particularly prone to personalization and has a high work burden. By merging OFE's accumulated site know-how with Matsuo Research Institute's advanced AI technology, we constructed an automatic judgment model using generative AI. We have established a foundation for realizing high-quality maintenance operations with reduced resources by systematizing the judgment process that traditionally relied on manual labor.

Effect Verification in Practical Operation

This system conducted an effect verification for approximately 4 months at internal sites, confirming the following results:

• AI judgment count: 8332 cases

• AI accuracy (*1): 89%

• System error rate (*2): 0.2% (achieving high stable operation)

From these results, we confirmed that this system possesses accuracy and stability that can withstand practical operations. From October 2025, we are prioritizing the introduction, focusing on inspection tasks where the AI application effect is particularly high. For on-site use, operations are conducted after preliminary verification using actual inspection images used in the field. In addition to improving work efficiency through AI utilization, it also contributes to quality improvement by reducing the risk of oversight that is unavoidable with manual work.

（注釈）

*1 AI精度：

AIが出力した結果が、あらかじめ定義した正解データとどの程度一致しているかを示す指標。本数値は実際の業務データを用いた検証結果に基づいて算出しており、現場業務における実用性の高さを示すもの。

*2 システムエラー率：

本システムの運用・検証過程において、想定された処理フローが正常に完了しなかった割合を示す指標。通信障害や処理停止、結果が出力されないケースなど、システムとしての動作不具合を対象としており、安定的な稼働性を評価するために用いられる。

Technical Features: Reproducing Human Judgment Processes with AI

The biggest feature of this system is that it goes beyond image pattern recognition, as seen in conventional visual inspection, by reading text information contained in photos and verifying if its content matches the correct setting value...