Sorry for such a question. I am new to machine learning. I have gone through some machine learning tutorials. I currently have a dataset from manufacturing sector which include sensor data, process data etc.

Actually I have been given a task to find parameters that can help company produce the least amount of scrap.

The data is coming every minute and include parameters like timestamp, number of parts produced, correct_parts_number, incorrect_parts_number,product_type,machine_status

Can someone help me to know which model should I go for and what should be the approach. I think number of attributes are less. Or if someone can share a simple dataset that is helpful in manufacturing industry to reduce scrap percent. So that I can get an idea.

Your company is probably looking for something similar to what OEE (Overall Effective Efficiency) provides.

There are a couple different factors to this metric:

  • Utilization: How well scheduled is the equipment/work center.
  • Performance: How well the equipment operates upon the schedule.
  • Quality: How well the parts are produced.

To calculate Quality (or Scrap Rate) is simply:

(# of Good Parts) / (Total # of Parts)


In most cases, machine data does not contain relevant data for Scheduled part counts (what's on your Work Order). So, you may need to close the loop on your dataset and validate the Total Part Count (number of parts produced OR correct_parts_number+incorrect_parts_number) against what is in your ERP/MRP system.


Ultimately, this is all just to identify what your scrap percent is, you will need to apply some engineering effort in order to actually reduce your scrap rate. The key with scrap rate is to purely identify where to allocate your engineering efforts.

Each part process is unique and quality issues cannot be magically provided without large and complex data structures to support it. In order to automatically provide options for reducing scrap rate, you would need a multitude of datasets talking to each other from ERP/MRP, Machine Data, Inspection Reports, and CAD/CAM just to name a few.

You may be already working with it, but the machine data standard MTConnect will reach a point that it can accommodate such datasets. It is a semantics-based format, so it just takes time to develop, but hopefully sometime soon it will support ERP-type data. However, don't hold your breath on manufacturing software companies jumping on board in the near future.

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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