Predictive maintenance

Equipment downtime is extremely expensive. However, most modern equipment has sensors to detect faults. While these can identify problems early, this reactive maintenance can still be costly. Using the sensors, history of both that piece of equipment and similar equipment, organizations can start a predictive maintenance program to identify at risk parts and machines and replace/repair them at a convenient time before they fail.

Overview

Identify at risk parts or machines before they fail and schedule maintenance at a convenient time before the failure takes the machine out of operation.

TARGET

Will a machine/part fail in the next period of time.

Challenge

Starting with the machine list from the ERP and merging it with the maintenance logs to determine the target, this data then needs to have the machines maintenance and failure history from the maintenance logs merged in. Next, machine sensor data should be added to gather a full picture of the machine's state. If the machine is exposed to the weather, external weather data will be useful to identify machines in hostile environments. All this data needs to be cleaned, joined and transformed into valuable ML features before going into model training. This pre-modeling prep process can be frustrating and time consuming. We are here to help.

Modeling techniques and libraries

Machine learning analysis

Build machine learning models to predict if a machine will fail or need maintenance as a function of the independent variables. Use model interpretability packages to evaluate the impact of the independent variables on the prediction.

Packages:
  • Sklearn
  • ELI5
  • LIME
  • SHAP

Data features

Machine Type
ERP
Data Type
Categorical
Target
No
Yes
Machine Utilization in last day
ERP
Data Type
Continuous
Target
No
Yes
Machine Utilization in last month
ERP
Data Type
Continuous
Target
No
Yes
Machine Utilization in last quarter
ERP
Data Type
Continuous
Target
No
Yes
Machine Utilization in last year
ERP
Data Type
Continuous
Target
No
Yes
Machine failure/repair
Maintenance Log
Data Type
Binary
Target
No
Yes
Maintenance in last day
Maintenance Log
Data Type
Continuous
Target
No
Yes
Maintenance in last month
Maintenance Log
Data Type
Continuous
Target
No
Yes
Maintenance in last quarter
Maintenance Log
Data Type
Continuous
Target
No
Yes
Maintenance in last year
Maintenance Log
Data Type
Continuous
Target
No
Yes
Rain
Weather
Data Type
Binary
Target
No
Yes
Severe Weather
Weather
Data Type
Binary
Target
No
Yes

Related accelerators

No items found.

Get your data science on.