top of page
landing_image_placholder.png

Accoustics based Anomaly Detection for industrial Assets

Technology

created

11/15/22

Job description

The capability of robustly detecting machine failures in a premature state are a critical component to data-driven prognostics and health management (PHM) systems in manufacturing. They allow to:
• optimize the cost and resource efficiency of maintenance operations,
• prevent unplanned machine downtime and secondary damages due to critical failures,
• detect quality issues at an early stage and take corrective action, and thus reduce waste.

Unsupervised anomaly detection algorithms are particularly promising in most real-world use cases, due to the extreme unbalance between healthy and unhealthy samples and their ability to produce models, that can detect yet unseen classes of machine faults. The use of acoustics data promises a retrofittable solution, accessible to a broad range of existing machinery. Training models for such a broad range of scenarios currently still requires hand-tuning by skilled data scientists.

A solution based on ultrasonic microphones has been developed and tested at CSEM in a laboratory setting. For the solution to be effectively industrialized, it is necessary to demonstrate, that it can be applied to industrial use cases of varying nature in terms of machinery and fault types.

remote or Alpnach, Switzerland

study project

Study Project, Master Thesis, Bachelor Thesis

June 30, 2023

To apply send us an e-mail to:

Responsibilities

Your task will be to apply unsupervised anomaly detection to multiple datasets acquired with the same acoustics sensor retrofitted to different machinery of multiple Swiss manufacturers. Based on theory from time-series analysis and heuristics from the field of PHM, you will develop a system to characterize such use cases and derive a procedure to train and validate robust anomaly detection models with minimal need for manual optimization. Your aim will be to demonstrate this procedure within the presented scenarios.

Essential requirements

• good communication skills and a willingness to closely collaborate with our team,
• an understanding of machine and deep learning, particularly applied to time-series data,
• proficiency in Python and Pytorch
• a basic understanding of version control (Git) and databases (SQL)
• a desire to produce results with a lasting impact

Bonus
(plus not must)

What we offer you

Our CTO, a skilled practitioner in the industrialization of deep learning solutions, will be your direct supervisor. We hope to grant you a great learning experience as a machine learning engineer with an interesting real-world use case at your fingertips.

If you believe our expectations to match, contact us. We will be happy to hear from you!

All qualified applicants will receive consideration for employment without regard to color, religion, sex, sexual orientation, gender identity, national origin or disability.
EU Citizenship is required for most positions.

bottom of page