Factory process DX at a major chemical manufacturer
Example 1: Factory process DX at a major chemical manufacturer
We provide end-to-end support for AI system implementation with zero proof-of-concept (PoC) using Human in the Loop technology, for factory automation.
Background/issues
-
Leaving corroded piping in chemical plants unattended poses a risk of serious accidents such as leaks , and thorough inspections are necessary to prevent any omissions or oversights.
-
Operations staff took a huge number of pictures of the pipes, and maintenance staff reviewed these images to assess the degree of corrosion. Because this placed a heavy burden on the staff , the use of AI in chemical plants is being considered.
-
In a job where mistakes are unacceptable, the challenge arises that insufficient data makes it difficult to achieve 100% accuracy with AI.
ABEJA's Value Proposition
The process incorporates a human (maintenance worker) making the final corrosion assessment, and implements image recognition AI in a zero-proof-of-concept (PoC) scenario. This creates a system that allows for early streamlining of operations even if the AI's accuracy is insufficient .
A system has been built on the ABEJA Platform that allows the AI to relearn based on what maintenance personnel have taught it during operation.
As AI accuracy improves, the burden on "humans" will gradually decrease.
In chemical manufacturers' plants, neglecting corroded piping can lead to serious accidents, making rigorous inspections essential.
At our client companies, operators would take a huge number of photos of the piping, and maintenance personnel would visually inspect these images to determine the degree of corrosion and decide on countermeasures.
Given the insufficient training data making it difficult to achieve 100% accuracy with AI, and in this task where overlooking corrosion is absolutely unacceptable, we developed a concept of "using AI to support human judgment" with the aim of reducing the burden on humans. Together with the client company, ABEJA built a new system on the ABEJA Platform.
approach
Expanding the scope to factory automation
Selection of business processes to be targeted for DX
Tobe formulation
requirements definition
AI system construction
Operation/accuracy improvement
We collaborated with on-site personnel to formulate a business concept where "people and AI work together."
The scope of the AI will be narrowed down to "determining the location of corrosion" and "determining the degree of corrosion."
We will build an AI system using the ABEJA Platform.
AI undergoes retraining and accuracy improvement while being operated in conjunction with "humans".
Based on on-site needs, the project was selected to focus on pipe corrosion inspection services.
results
【short term】
Reduce labor costs for the target operations.
Learn how to use AI at the practical level and move towards further applications.
[Long term]
Factory automation (construction of digital value chains)