Can AI Predict When Industrial Sensors Need Maintenance?
What if sensors could tell us when they are starting to fail, before their measurements become unreliable?
Description
Industrial sensors are often expected to operate for years in demanding environments. Over time, however, aging, contamination, temperature changes, and environmental exposure can affect their accuracy.
The challenge is determining whether a changing measurement reflects a real change in the environment or a sensor that is gradually degrading.
At Knowit Connectivity, we are offering a thesis opportunity to investigate how Edge AI can detect sensor degradation and predict when maintenance, recalibration, or replacement is needed.
About the Project
The objective is to develop and evaluate an AI-based approach for monitoring sensor health over time.
You will investigate how sensor signals change due to aging, environmental conditions, and other degradation mechanisms.
By combining controlled experiments, sensor data analysis, and machine learning, the project aims to identify early warning signs of sensor degradation.
An important aspect is developing models that can operate directly on resource-constrained embedded devices rather than depending entirely on cloud processing.
What You Will Explore
Your work may include:
Investigating how different sensors degrade over time.
Designing controlled experiments to simulate environmental stress, sensor drift, and calibration loss.
Collecting and structuring sensor measurements into a labeled dataset.
Developing machine learning models for anomaly detection and sensor-health classification.
Exploring techniques for predicting maintenance or recalibration needs.
Comparing traditional statistical approaches with Edge AI models.
Optimizing models for embedded hardware using TinyML techniques.
Evaluating detection accuracy, false alarms, inference latency, and energy consumption.
Investigating how models can adapt to changing operating conditions without overlooking genuine degradation.
Who We're Looking For
This project could suit students studying Machine Learning, Computer Science, Embedded Systems, Electrical Engineering, Engineering Physics, or related fields.
Relevant interests include:
Artificial intelligence and machine learning.
Signal processing and time-series analysis.
Sensor technology and measurement systems.
Embedded software and TinyML.
Experimental design and data analysis.
Predictive maintenance and industrial applications.
You should enjoy combining practical experiments with data-driven problem-solving.
Expected Outcomes
The thesis aims to produce a structured sensor-health dataset, a validated AI model for identifying degradation, and an embedded proof of concept demonstrating sensor-health monitoring.
You will investigate how accurately degradation can be detected and whether maintenance requirements can be predicted before sensor performance becomes unacceptable.
The results should also contribute to a reusable methodology for evaluating additional sensor technologies.
The project will use controlled and potentially accelerated experiments. These can provide valuable insights but will not replace long-term validation in real industrial environments.
The final research scope will be defined together with your supervisors.
Interested in This Thesis?
Want to use machine learning to solve practical challenges involving industrial reliability and intelligent sensor systems?
Apply through our thesis application page and indicate your interest in "Edge AI for Sensor Degradation".