Discipline: Technology and Engineering
Subcategory: Civil/Mechanical/Manufacturing Engineering
Matlock Mennu - University of South Florida
Computer numerical control (CNC) machine tools are essential tools in the manufacturing of various components in the automotive and aircraft manufacturing industry. Thus, degradation of machine tool linear axes has a huge impact on the quality of parts manufactured through this process. Billions of US dollars are lost every year due to degradation of machine tools during production. Currently, there are direct methods of measuring geometric errors using laser-based and other standard methods. However, these methods are time-consuming and complicated for many users. In addition, such methods halt production, which usually equates to lost revenue, so manufacturers prefer not to shut down their machines for tests. Therefore, this project focuses on using an inertial measurement unit (IMU) with relatively inexpensive sensors for measuring changes in geometric errors of linear axes efficiently and with sufficient accuracy. The IMU-based method has been tested with much success on a linear axis testbed. Now, a smaller IMU is being set up for placement and practical usage on machine tools. The linear axis testbed relies on acquiring axis position data from the motor encoder in the system. The main challenge was to derive nominal position data from an accelerometer rather than relying on other sources, so that the IMU can be placed in a ‘plug and play’ fashion on any machine tool without needing to acquire controller position data. This process involved writing MATLAB functions that can derive position data from acceleration data from the accelerometers in the sensor box. Once a robust method was developed, the estimated positions were compared with the measured positions (from the testbed encoder) to determine the method accuracy. The mean errors for the slow, moderate and fast speeds of the test bed were 3.16, 15.65 and 94 microns respectively. After the accuracy was determined to be sufficient for analysis purposes, the function was integrated into the main analysis code that estimates linear axis errors as a function of nominal axis position. Finally, metrics were tested for their ability to distinguish various level of degradation from the linear axis error motions. Future research involves creating a separate MATLAB-based GUI that uses the analysis subroutines for automatic detection of degradation, resulting in plots, metrics, and other aspects of a user-friendly GUI. In addition, the MATLAB-based GUI will use the estimated errors to visualize the error motions along a linear axis.
Funder Acknowledgement(s): National Institute of Standards and Technology, United States Department of Commerce
Faculty Advisor: Bernard Batson, bbatson@usf.edu
Role: I worked on developing the Matlab functions to derive the position data from the accelerometer data of the IMU. I was involved in creating these functions from scratch and identifying the right methods in order to determine nominal position data. I worked on finding the right filters in order to reduce the noise in the data while maintaining the integrity of the data. In addition, I worked on determining the accuracy of the method by applying the Matlab function to multiple data points and doing statistical analysis. Once the method was proved for test bed data, the sensor was applied to an actual machine tool where additional data was collected using the Matlab functions I created. I helped identify any errors that occurred during the data acquisition process on the machine tool.