March 30, 2021
The Ministry of Economy, Trade and Industry (METI), jointly with the Ministry of Health, Labour and Welfare (MHLW) and the Fire and Disaster Management Agency (FDMA), decided to revise the Guidelines on Assessment of AI Reliability in the Field of Plant Safety. The revised guidelines provide improved user-friendliness, including seven use case examples of leading plant operators that have practically applied the guidelines to their AI development projects. This revision is expected to further encourage plant operators to efficiently utilize the guidelines and to help them enhance the safety and efficiency of plants taking advantage of AI.
1. BackgroundDemand for AI for operators of petroleum and chemical plants is strong as a measure for overcoming labor shortages and for reducing accidents. However, utilization of AI in such plants requires operators to ensure high-level safety and fulfill accountability. These challenges make operators hesitant in the advancement of the utilization of AI.
To overcome this, METI, in conjunction with the MHLW and the FDMA, formulated Guidelines on Assessment of AI Reliability in the Field of Plant Safety in November 2020, and presented approaches to appropriate management of the reliability of AI, i.e., securing the quality of AI expected to improve the safety and productivity of plants.
2. Details of the revised guidelines
The revised guidelines present contents reflecting the results of seven use case examples of leading operators that have practically applied the guidelines to AI development projects. Highlights of the improved points are as follows:
Clarified relationships between the conventional methods for risk assessment, e.g., HAZOP and FMEA, to ascertain plant safety and AI-based safety assessment.
An “Explanations” section is added to the record template for noting down the details of safety assessment in order to improve the user-friendliness of the template.
Provision of explanations in a Q&A format focusing on questions and points that operators may have and face difficulties in making decisions when applying the guidelines to their practices.
Provision of “use case examples” of leading companies that have carried out AI reliability assessment under the guidelines and described the results in accordance with the record template, which comprehensively describe five types of use cases about seven projects in total.
3. Related documents
- Summary of Revision (PDF:398KB)
- Guidelines on Assessment of AI Reliability in the Field of Plant Safety (Second Edition) (PDF:4,140KB)
- Appendix: Template for Reliability Records (PDF:882KB)
- Appendix: Template for Reliability Records (Excel:242KB)
- Summary of Reliability Assessment Practical Examples (7 Examples) (PDF:1,452KB)
- Practical Examples of Reliability Assessment Records (7 Examples) (PDF:2,122KB)
- Practical Examples of Reliability Assessment Records (7 Examples) (Excel:376KB)
Links to related document
- Guidelines and Collection of Case Examples Formulated to Make Petroleum and Chemical Plants Smarter by Taking Advantage of AI (November 17, 2020)
Division in Charge
High Pressure Gas Safety Office, Industrial and Product Safety Policy Group