0
$\begingroup$

I am a student who will finish my studies next year and I want to analyze the job market in advance. I have found an interesting job where it says:

The department of assemblies and systems of the X deals with the simulation, testing and evaluation of components, assemblies and complete systems mainly from the automotive sector. In the future, machine learning methods will be used to detect changes in components or test benches at an early stage

I find the last sentence very interesting, but I don't quite understand it. If you expect a change to a component (e.g. the component starts to break), you can hard code it to detect this change. Could somebody explain to me with a simple example how machine learning can be helpful in this area?

$\endgroup$
0
$\begingroup$

Defect Detection and Failure Detection/Predictive Maintenance are two areas of modern manufacturing and engineering where machine learning can provide substantial and even radical solutions.

For example, intelligent systems may detect defects at an early stage with enough accuracy and generality so as to save both costs and increase reliability.

On the other hand, intelligent systems may monitor operation logs of systems (eg an aircraft) and predict component failure exactly when needed with accuracy and generality so as to reduce downtime, save costs and increase safety.

Both these approaches require sophisticated learning methods (ie machine, deep, learning) so as to handle complexity and at the same time be general enough and accurate enough for practical purposes.

Sample references:

  1. AI-Based Visual Inspection For Defect Detection
  2. Study on Machine Learning Based Intelligent Defect Detection System
  3. How to Find the Right Machine Learning Techniques for Predictive Maintenance?
  4. Machine Learning Techniques for Predictive Maintenance
$\endgroup$
0
$\begingroup$

This is an interesting question about how machine learning can be used to changes in components. In this case, I am going to infer that is maybe talking about detecting defect components, which of course should not be put into a new vehicle.

In this case, you might use images (maybe taken from video of the assembly floor) to the detect potentially defect components. So, hard-coding this would be extremely hard, because:

  1. Firstly, you would need to the code to detect the component and separate this from the background (this task is known as semantic segmentation).
  2. You then have to compare the component with images of a "good" component and a "bad" component (preferably multiple examples of each). This also isn't an easy task to hard-code, because you need might observe the component in different angles (bird's eye view or front on, or example), which are not discrete. Therefore, some level of generalisation is required to do this (hence machine learning)

I look forward to other's interpretations and answers to this question.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.