Reducing Operating Costs And Improving Safety

To ensure safe operation of the nuclear plant, the Institute of Nuclear Power Operations sets the performance objectives and defines the criteria (PO&C) to meet those objectives. The plant personnel manually compare each line item in the CAP report against the PO&C to evaluate the seriousness of the event. A spilled cup of coffee in the wrong place might not be too serious, while a critical pump leaking that has the potential to shut down the plant is another matter entirely. This manual process of classifying the seriousness of each item in a CAP report lies at the heart of the continuous effort to improve the safety and operation of the nuclear power plant, but requires significant time and resources. Our client needed a way to automate the classification process to increase its efficiency and to extract information from years of reported events to better understand their causes and anticipate possible adverse events.

As operators create each CAP report, Saffron automatically applies the attributes of all PO&C objectives to each line item within the CAP report and then suggests an appropriate classification. An experienced operator reviews each Saffron recommendation to confirm the recommendation or to change it. The validated report then becomes part of Saffron’s memory base, providing the input for Saffron to continuously improve by learning and remembering the specific variables upon which the classification decision is made. Not only does Saffron continuously learn from new cap reports, Saffron also implements this learning to prior cap report classifications to ensure that the most advanced knowledge is applied retrospectively.

• [Data Management - Data Analysis]
  Saffron enables plant operators to see causal relationships between events and root causes and have immediate access to converging, trending or repeating patterns.
• [Efficiency Improvement - Operation]
  Operators can now easily identify and eliminate causes of adverse events. No models are required, saving time to insight and driving high Return-On-Investment.

• Saffron automatically classified 18,000 CAP items (sample dataset), with 92% accurate and 18 false positives and 15 false negatives against this data set.