Acerta AI deploys machine learning approach to reduce fuel cell testing time by 76%
Production implementation with a hydrogen fuel cell manufacturer will demonstrate practical application in a real manufacturing environment while maintaining stringent quality requirements.
TORONTO, April 30, 2026 /PRNewswire-PRWeb/ — Acerta AI, a provider of operational AI solutions for discrete manufacturing, announced the results of a production deployment of a machine learning and AI solution to reduce final test time for hydrogen fuel cell stacks. When fully integrated at scale, this solution is expected to reduce test times by up to 76% and increase production throughput while maintaining stringent quality requirements.
End-of-line testing is one of the most costly and throughput-limiting steps when scaling up fuel cell production. Reducing test time while maintaining quality directly increases available capacity and reduces cost per stack. The initiative is based on a two-year collaboration with a leading hydrogen fuel cell manufacturer, where the system was developed, tested, and deployed in a real-world production environment. By identifying early signs of failure, this approach reduces test times from 2+ hours to 15-30 minutes while maintaining quality assurance.
Greta Kuturenko CEO of Acerta AI said:
In a production environment, model performance alone is not enough.
The differentiator is not just model accuracy, but the ability to operationalize AI output into production decisions that increase throughput, reduce costs, and protect quality.
Greta Kuturenko CEO of Acerta AI said:
In a production environment, model performance alone is not enough.
“The challenge is to turn predictions into reliable decisions that optimize throughput and cost without compromising quality.”
Acerta’s unique approach clearly separates prediction from decision-making and transforms model output into decision-making policies that define trade-offs between expected cost, test coverage, and resource usage. These policies can be adjusted to reflect different modes of operation: conservative, balanced, or aggressive, depending on your operational requirements.
Sergei StrelnikovVP of Engineering at Acerta AI.
There is no “perfect” model in manufacturing.
“Therefore, it becomes important to go beyond prediction and explicitly link model outputs to production metrics such as throughput, cost, and resource usage. Our approach focuses on policy-based decision-making, where the trade-offs between cost and risk are clearly defined.”
The system is intensively trained on large datasets, deployed at the edge of production, and operates under constraints on latency, reliability, and integration with physical systems. This cloud-to-edge deployment model ensures consistency between training pipelines and production operations.
The results were shared with the global powertrain and propulsion community at the 2026 International Vienna Automotive Symposium, highlighting the industry’s relevance and operational readiness. The paper “Accelerating end-of-line testing of fuel cell stacks with machine learning: Early failure detection and cost reduction in production” details the complete system architecture, including data ingestion, model training, edge deployment, and monitoring.
The symposium, organized by the Austrian Association of Automotive Engineers (ÖVK) in cooperation with the Vienna University of Technology, is widely recognized as the leading forum for powertrain and propulsion technology, bringing together OEMs, suppliers and researchers.
read Latest news shaping the hydrogen market at Hydrogen Central
Acerta AI deploys machine learning approach to reduce fuel cell testing time by 76% (source)
