AI Calibration Optimisation Agent

Overview

Industrial equipment often follows fixed calibration schedules that are not optimised for the true reliability profile of the system. This project explores how statistical reliability models can be used to dynamically estimate optimal calibration intervals, improving operational efficiency while maintaining high reliability standards.

The Problem

Many calibration schedules are based on conservative heuristics rather than data-driven reliability analysis. This can lead to:

The goal of this project is to determine calibration intervals that balance cost efficiency with reliability targets.

Methodology

The system models equipment reliability using tools such as Weibull reliability analysis. Key parameters such as the shape parameter (β) and scale parameter (η) are estimated from failure data to generate reliability curves over time.

Calibration intervals are often suboptimal - leading to useless spending and suboptimal rates

Results

The model produces a reliability curve that estimates the probability of successful operation over time and identifies the optimal calibration interval required to maintain a 95% reliability target.

Reliability analysis chart showing probability of success over time

Key Metrics

The tool analyses user input and determines the appropriate tools to use to accomplish the user's goal. Then, it writes the appropriate python code based on the user input, analyses the output of the code, and delivers appropriate recommendations.

Potential Applications

This approach could be integrated into predictive maintenance systems to dynamically optimise maintenance schedules in industries such as healthcare technology, manufacturing, and industrial instrumentation.

This project was inspired by my internship at GE Healthcare, where I observed data and processes first-hand, confirming the potential of the solution. I have since then completed IBM's Build an AI Agent online course and plan to further refine the project.