Last updated on February 5th, 2024

USE CASE

Automating root cause analysis in EV motors

Objectives

  • Find the root cause of a torque-to-rotate problem in electric vehicle motor manufacturing
  • Monitor the cause of the torque-to-rotate issue to prevent it from occurring again

Challenge

If a high breakaway torque is found at an end-of-line test on an EV motor, it can be challenging to find the root cause of the problem, since the part has been fully assembled and there can be many causes.

Key Results

Root cause of high breakaway torque was found with automated root cause analysis

Future issues were prevented by monitoring the factors that contributed to the problem

Background

One of the most appealing features of electric vehicles is their ability to deliver maximum torque from zero RPM, providing immediate and powerful acceleration. This enhances the driving experience and responsiveness of EVs, making them well-suited for urban driving and stop-and-go traffic. 

Consumers also want their EVs to run efficiently, maximizing the distance they can drive on a single battery charge. It can be a challenge for manufacturers to deliver enough breakaway torque to produce this instant powerful acceleration, while keeping the demand on the battery within reasonable limits. 

Manufacturing and assembling EV motor components so that they rotate optimally fulfills the promise of instant torque and superior acceleration, while maximizing battery life. 

Finding the root cause of torque to rotate (TTR) problems

After an EV motor has been assembled, it is mounted on a test rig to check key performance indicators, including high breakaway torque (the amount necessary to start rotation) and high running torque (the amount of torque required to keep the assembly rotating) of the motor assembly.
Torque

High values in these tests can not only lead to early-life failures in the motor assembly, but they also impact the efficiency of the motor at rated speed, thereby greatly reducing EV battery range. 

During the motor test, gradually increasing torque is applied to the motor using a dynamometer until the motor begins to rotate. The torque curve, current, and speed data are measured.  

If the breakaway torque is found to be too high, this indicates the presence of a defect in the motor components or an assembly problem. However, the motor has already been fully assembled, meaning that rework will be necessary. There are myriad issues that could be causing this problem, such as: 

  • bearing misalignment to either the bearing seat or rotor shaft; 
  • the rotor housing was machined at the extreme low end of/out of tolerance for the bearing pocket; 
  • the rotor shaft was too large, causing the inner race of the bearing to become deformed and increasing the preload on the bearings; 
  • the bearings themselves are defective; or 
  • the rotor was not balanced properly. 

These are just a few of the many possible defects that could occur. 

Instead of disassembling the motor and initiating a time-consuming brainstorming root-cause analysis session, engineers can use LinePulse to analyze the motor manufacturing data to detect possible contributors to the problem in a matter of minutes.

The root-cause analysis function in LinePulse is automated. Machine learning models compare end-of-line test data with data from the upstream signals for each part, linking the data together using the serial numbers on each part. 

By comparing the signal data between parts that failed the test and parts that passed the test, LinePulse identifies which signals differ between the two groups. The signals are displayed in the platform in a list with the estimated percentage of their likelihood to be contributing to the high breakaway torque. This allows engineers to prioritize their investigation of upstream stations to find the issue. This automated feature can greatly reduce the time it takes to find the root cause.  It also provides context for the engineers regarding the size and scale of the problem—was this a one-off occurrence? Has it happened before? And, if so, how often?  Are there parts that exhibit defective characteristics that were not caught at end-of-line testing? Analyzing manufacturing and warranty data together would allow the engineering team to set limits and alerts for those signals that are optimized to catch defects as soon as possible. 

Results: Discovering the root cause

The root-cause analysis module in LinePulse indicated that there was one signal with a high correlation to the failed parts with high breakaway torque: rotor housing bearing seats for the main shaft that were outside of their allowable tolerances.

The root cause of the issue was that the bearing seat gradually became too small over the course of the day. When the bearings were pressed into the seat, the press force was abnormally high. When the rotor shaft was then installed, the preload on the bearing was too high which caused the high torque-to-rotate.

The issue was not detected during production because the machining operations are verified through a CMM process at the start of the day, which verifies that all signals are within their tolerances. After this, it is assumed that the parts are good, as there is no way to inspect every part.

Once this issue was discovered, it was quickly remedied and the problem disappeared with minimal defective parts having been produced.

Results: Preventing future torque-to-rotate problems

Solving a problem is one thing. But how can this problem be anticipated and prevented from happening again? Several steps can be taken within LinePulse to predict when this issue may occur and alert the team in advance.

The engineering team used LinePulse to see that the alignment force, midstroke force, and final force of the failed parts were all outside of their process control limits, but still within their specification limits.

Next, the team ran long-duration capability metrics queries in LinePulse of the alignment force, midstroke force, and final force signals in the pressing station to see how the signals normally behave. LinePulse automatically calculates the study statistics including the mean, sigma, UCL, LCL, Cp, Cpk, etc. Using this information, the engineering team can determine and create “Fixed Control Limits” for each signal within LinePulse.

Then, capability metrics alerts in LinePulse were turned on for alignment force, midstroke force, and final force. Now, each time these limits are breached, LinePulse displays an in-platform alert. It also sends the alert to a Microsoft Teams channel containing the engineers responsible for the pressing station and for the torque-to-rotate test. These alerts are generated in manufacturing real time, so that the team can take action immediately, rather than waiting to find out they have a problem at the end-of-line test, or once their repair units start piling up.

The team also set up alerts in LinePulse for the torque-to-rotate test operation, both for breakaway torque and running torque, so that any other issues could be identified quickly. They also used LinePulse’s anomaly detection module to monitor for any gradual change (“drift”) in the breakaway and running torque signals. This way, the team can act proactively to ensure the quality of the parts shipped, while keeping their throughput as high as possible.

LinePulse can accelerate root cause analysis on your line

How can we Help?

Contact Info

Automate root cause analysis and predict defects in real time

How is that possible?