)
)
)
)
)
)
)
How can we help you?
)
Marija Mijic
HR & Office Allrounder
T +43 660 383 95 46
office@craftworks.at
craftworks GmbH
Schottenfeldgasse 20/6a
1070 Vienna
Improved Acoustic Emissions Testing
with Neural Networks
Through the use of Neural Networks, the reliability and speed for AET was increased.
:quality(80))
Success Story
reading time: 3 min
Manufacturing
Material Industries
Predictive Quality
Industrial Software
58%
Faster process of product certification
Supported
Testers in accurately interpreting data
Seamless
Monitoring of machine learning models
Context
Our client, a reputable certification body, functions as an autonomous entity specializing in testing and certification services. A certification from our client is crucial for many industrial applications and is seen as evidence of high quality and reliability. As such, tests must be conducted with a high degree of accuracy and reliability. Our customer used acoustic emissions testing to certify high pressure tank safety. They wanted to automate the analysis and safety process using Machine Learning.
Challenge
The analysis was conducted through expert evaluation, a process that was both time-consuming and susceptible to errors. Furthermore, the customer faced uncertainty regarding the implementation of a Machine Learning solution to address this issue. Specifically within the realm of Machine Learning, the focus was on identifying the key factors essential for assessing tank integrity and suitability.
Assignment
We proposed to utilize machine learning to autonomously evaluate the results of acoustic testing, allowing for efficient and accurate analysis.
Solution
A collection of Neural Networks was employed to analyze the data, developed using Python and Keras. Additionally, a web application was designed to facilitate the upload of test data and metadata, incorporating authentication features and providing a synopsis of the machine learning model outcomes.
The model interacted through an inference pipeline and a REST Endpoint, enabling seamless engagement.
:quality(80))