Equalis Predictive Motor Maintenance (PMM) - Beta Program Summary

Abstract

Electric motors are used in a wide variety of industrial and transportation applications. Reliability of their operation in most cases is of critical importance and any failures that occur can yield significant costs and downtime, or even lead to disastrous consequences. Traditionally, the problem of ensuring motor reliability is dealt with using a regular time interval of maintenance, whereby trained service personal at a regular time intervals check the machines and their state of operation. As technology emerges, and control hardware size and power requirements are reduced, we now have the ability to remotely monitor critical motor conditions and report their overall condition back to a system manager thus optimizing the machines up time.

Of the various monitoring techniques commonly employed, vibration measurement remains an effective approach. This is due to the fact that structural failures in a machine system cause changes to the dynamic characteristics of the machine. These changes are reflected in its vibration signals and signatures. By implementing appropriate signal decomposition and representation, features hidden in the vibration signals can be extracted, and an assessment of the machine health status can be made well in advance of a critical machine event horizon.

An attractive alternative is now in emergence from Equalis. We have begun to leverage high performance embedded controllers with a low profile accelerometer for vibration measurement and a GSM cellular module to upload data to a server. This cost effective technology can monitor on a regular, ongoing, basis the mechanical condition of critical motors and thus report on the overall health of a system. The health is available in simple terms by evaluating the machine spectra signature in terms of a “Green” indicator revealing normal running conditions, a “Yellow” indicator as outside normal, and a “Red” indicator as requiring maintenance within a known and limited time.

This approach allows continuous up-to-date information about the state of the critically important equipment, proactive planning for maintenance accordingly, reduce overall costs, minimize undesirable downtime, and prevent unexpected and possibly disastrous equipment failures.

Methods & Approach

A practical condition-based monitor system designed on this approach involves the following basic objectives;

  • Strategic high fidelity motor vibration measurements

  • Collection of the multiple measurements from motors critical in the system

  • Uploading of that data to a remote server via cellular

  • Download and analysis of the collected data

  • Detection and prognosis of motor faults, and generation of motor maintenance plans

Equalis has completed a beta (prototype) system design and testing, showing an application that can collect real -time data, and now offers a higher performance package to be used in the field for ongoing deployments.

The beta (prototype) system (Figure 1a & 1b) consists of a commercially available low power microcontroller combined with an accelerometer package that provides an overall update rate of 3.2 kHz., a maximum of 16 G capacity, and draws .1 micro amps in standby mode. This system communicates with the PC via USB cable and serial data. It is intended for overall characterization and initial data collection on a NEMA 250 frame motor in lab conditions.

Figure 1a - Motors are Equipped with Equalis Motor Monitor Beta System (Prototype) for Characterization and Initial Laboratory Data Collection Tests

Figure 1b

The embedded controller system was characterized using an ELWE Model U8556001 Vibration Generator combined with a Digital Function / Arbitrary Waveform Generator. The sensor assembly is securely fastened to the vibration generator shaft. The waveform generator is set to deliver sine wave signal of a known frequency to the sensor assembly (Figure 1c and d). The PC is loaded with an Equalis serial communication module to collect streamed data from the controller. The module allows update rate and file size to be adjusted as required. The data rate and file size is adjusted as required for evaluation. The unfiltered data is then saved to file and an FFT module is run to produce a spectral signature of the sample. The FFT signature is then graphed for visualization and analysis.

To evaluate overall performance data sets are collected and evaluated beginning at 60 Hz (Figure 1e). Each data set is increased by 20 Hz and evaluated until the magnitude response is decays -3db down.

Figure 1c

Figure 1d

Figure 1e

Following the system characterization tests and the calibration data below:

  • 1081 Data Samples

  • 2 mils amplitude = 1200 lines, 4 mils amplitude = 2400 lines

  • 60 hertz = 142 bins

We began tests on the NEMA 250 frame machine. We collected unfiltered data on one axis of the tri-axial accelerometer and in the radial direction of the motor shaft. The accelerometer is mounted by center boring and tapping a bearing retainer bolt just adjacent to the single row ball bearing of the AC induction machine (Figure 2a & b).

Figure 2a

Figure 2b (Typical Bearing Retainer Cross-section showing hardware)

Past field and customer experience with cage failures in this machine frame size made data available from the bearing manufacturer to contrast and compare our tests and our spectra model for this type of defect (Figure 3 & Photo 3). We call attention in Figure 3 to the 2 major peaks in the FFT image. The larger of the peaks representing the fundamental frequency or machine speed and the smaller peak representing the cage defect as it begins to emerge.

Photo 3

Figure 3 (SKF Cage Fracture FFT)

Bearing manufacturers widely publish data of cage and other rolling element defects. This information reveals that accurate models can be developed and then quantified (Figure 4). A model of this defect with our drive end bearing yields a cage defect frequency between 35 and 40 % of the fundamental frequency depending on bearing details and speed according to published information.

Figure 4 Rolling Element Bearing (1, 2, 3, 4)

Review of the spectra signature over an entire bearing breakdown time frame reveals multiple stages that occur during the failure process. These stages may occur over a very substantial period of time (Figure 5). Our method of inducing one cage defect and then a second served to shorten the process between zones or call out specific zones. Our data collected was characterized against this type of model.

Figure 5

The test motor is a DPG machine rated at;

- 14 HP

- 3780 RPM

- 422 Volts AC

- 19.6 Amps AC

- 127 Hertz

The initial tests were with no connected load and powered by a G130 Siemens Inverter. Subsequent tests were also conducted with the motor connected to a 125 HP DC dynamometer and loaded to 100% rated torque. All tests were run at 30 hertz and subsequent speeds just under 1800 RPM.

The tests were run with 4 sets of data collected at 30 minute intervals as listed below;

  • 3 tests with new bearings installed on the drive end and opposite drive end and no connected load

  • 3 tests with drive end bearing cage defect introduced at one location (Figure 6) and no connected load

  • 3 tests with the drive end bearing cage defect at two locations 180 degrees apart with no connected load

  • 3 tests with drive end bearing cage defect at two locations 180 degrees apart under full dynamometer load conditions

Figure 6 Single Cage Defect Visible in Lower Right

Summary & Results of Beta (Prototype) Testing

The data presented below is listed in three segments. Each test is listed below for reference.

  1. The machine with new bearings installed and not connected to any load. Three frequency nodes are visible.The left most node represents 1X or operational speed.No peaks below 1X are visible and thus no evidence of bearing defects in zone 1 or 2. Higher frequency peaks between 3 and 6X are visible.These peaks, changing for each machine, drift in frequencies with operating temperature and are a natural machine characteristic integral to the design. (See Figure 7)

  2. The machine with one cage defect as shown above and no load. 1X visible and 1 major peak with side bands in formation at approximately 30 bins or .38 x 1X. (See Figure 8)

  3. The machine with two cage defects and no load connected. One visible under 1X at higher level. (See Figure 9)

  4. The machine with one cage defect on the DC dynamometer

  5. The machine with 2 cage defects on the DC dynamometer. Substantial activity below 1X (See Figure 10)

Figure 7 No Connected Load, New Bearings, FFT For Test 1

Figure 8 No Connected Load, 1 Drive End Cage Defect, FFT For Test 2

Figure 9, No Load, 2 Defects, FFT For Test 3

Figure 10 100 % Load, 2 Drive End Cage Fractures, FFT For Test 5

PMM System Design

Following the results of the Beta (Prototype), an industrial system was developed (Figure 11a, b, c, & d). This system utilizes a higher performance microcontroller and accelerometer while maintaining the small form factor and current draw to deliver 10 kHz overall update rate and an overall bandwidth nearly 300% that of the Beta System. This system also adds a GSM cellular module and will upload serial data to a remote cloud server and then to the PC for analysis thus allowing it to be deployed to an actual vehicle or field application.

Figure 11a (Vehicle Example)

Figure 11b (Remote Location Pump Example)

Figure 11c (Prototype Embedded Control Build Up)

Figure 11d (Prototype Package for Deployment)

Refinements

- This beta study was conducted using familiar FFT to detect bearing variance. Alternative analytical methods including learning systems will produce results that allow earlier bearing defect detection. The follow up system and its analytical approach will include a STFT model as well as a Hidden Markov machine learning model and their data analysis.

- The initial study was limited to laboratory tests only. Thus results are under very controlled conditions and excluding a substantial component of white noise present in real world applications. Follow up field deployments will include substantial filtration and de-noising of data. Field tests will be application specific as vehicular applications verses stationary are likely to develop very different white noise levels and signatures.

- Field deployment will show further statistical quantification.

- The Beta system allowed for the detection of our steel cage ball bearing defect and produced an overall bandwidth of approximately 6 – 7 times the operational speed.A higher bandwidth system is required to improve fidelity and include higher frequency components.This higher bandwidth system will allow for earlier detection and thus a more comprehensive predictive solution.

Equalis welcomes PMM customer initiatives

Equalis welcomes feedback and is available to work with operators on a live Predictive Motor Maintenance project. Please respond to info@equalis.com for more information.

Citations

1/

Article title: KB Results

Website title: Stiweb.com

URL: http://www.stiweb.com/kb_results.asp?ID=53

2/

Article title: NTN Bearing Frequencies

Website title: www.ntnamericas.com

URL: http://www.ntnamericas.com/en/website/documents/brochures-and-literature/tech-sheets-and-supplements/frequencies.pdf

3/

Article title:Understanding Bearing Vibration Frequency

Website title: electromotores.com

URL: http://electromotores.com/PDF/InfoTécnica/EASA/Understanding%20Bearing%20Vibration%20Frequencies.pdf

4/

Article title: Spectrum Analysis, The Key Feature of Analyzing Spectra

Website title: www.skf.com

URL: http://www.skf.com/binary/tcm:12-113997/CM5118%20EN%20Spectrum%20Analysis.pdf

Featured Posts
Recent Posts
Archive
Search By Tags
No tags yet.
Follow Us
  • Facebook Basic Square
  • Twitter Basic Square
  • Google+ Basic Square

© 2020 Equalis LLC

Columbus, OH, United States

513-266-3862

  • LinkedIn Social Icon
  • Twitter