Motion System Design
Predicting machine failure

Predicting machine failure

Electrical signatures combined with ingenuity help engineers create new analysis tools for motor-operated devices. These tools can predict component failures and which parts of the the components are about to go

Amotor-operated mechanism fails and you must shut down critical operations. Unfortunately, to analyze why it failed, and to fix it, you have to go to its location, sometimes miles away. You were just there a week ago. If you had known the device was about to fail, you could have repaired it then and avoided this problem.

The engineers of Duke Power Co. have many devices to maintain, from motoroperated valves to large pumps. “An unplanned shut down is critical as well as costly, particularly when a component failure requires purchasing electrical power from other sources,” says Bobby W. Roberts, senior technical specialist in the Predictive Maintenance & Analytical Dept. at the plant. “Periodic testing of these devices catches some potential problems, but not enough of them. Thus, we sought other methods to improve predictive analyses.”

Analyzing problems

Initially, the engineers looked for an off-the-shelf device to handle all their testing. Failing to find one, they created a solution that combines off-the-shelf components. These include software from HEM Data Corp., an ordinary personal computer, and data acquisition boards. Concurrent with their search for a readymade solution, they discovered that certain electrical signatures of motor-operated devices yield useful diagnostic information.

The engineers recognized that:

• They can correlate instantaneous values of power in a motor to load changes.
• A motor-driven device consumes energy repeatably and predictably during normal and expected mechanical actions and load events.
• They could use these power consumption characteristics to spot problems.

“With motors, power is eight to ten times more sensitive an indicator than just current to changes in a motor’s load,” says Mr. Roberts. “With this information, we at the Nuclear Performance Group defined the ideal solution’s features. It must be fast enough to acquire, display, and record sine waves of three-phase voltage, current, power, and power factor, all in real time.”


In their PC-based solution, the data acquisition boards (DA) execute analog to digital (A/D) conversions. The Snap- Master software from HEM Data Corp., executes the math functions for data reduction and analysis, including time and frequency domains, phase angle, and Fast Fourier Transform (FFT) calculations. (The engineers found FFT waveforms useful in detecting rotor problems in motors). The software also supports the DA board’s sampling rates to 1 MHz.

The software samples each channel at exactly the same instant, and simultaneously displays multiple windows on the PC screen. Recorded and calculated data are displayed in the same way.

Device testing

Prior to the engineers’ new solution, motor-operated devices underwent regular testing, particularly the motor-operated valve (MOV). Water flowing through the valve cools various equipment in the plant. An electrical motor drives a gear train (valve operator) connected to the valve stem to open and close the valve.

Regular mechanical testing ensures that the valve opens and closes properly and efficiently. For these tests, engineers cycle it through its full range of motion. They note the time it takes to move from fully open to fully closed. They also measure stem thrust and the torque needed to seat the valves.

These tests can also indicate future problems. “For example, if valve-stroke time, stem thrust, or torque are excessive, the MOV may soon fail,” says Mr. Roberts. “However, mechanical testing may not detect more subtle problems. For these problems, we have found power consumption characteristics — the MOV’s “signature” — useful.”

The engineers took their off-theshelf components and assembled a prototype test system. Then, they conducted a series of tests on valves implanted with common problems to determine whether the test system detected meaningful load changes. Problems included a bent stem, worn gears, improperly set torque and limit switches, tight packing, loose stemnut, excessive spring-pack gap, motor single phasing, and obstructions in the valve seat.

“In each test, the prototype system correctly found the problem,” Says Mr. Roberts. “But we noticed an interesting phenomenon. During a valve cycle, an instantaneous power waveform creates a detailed picture of the operating conditions.” Figure 1 shows an example. Because an electric motor produces mechanical events, each event reveals itself as a change in power level with respect to time. A specific valve cycle matches each event; such as inrush power, area of lost motion, hammer blow, running load, and torque switch trip. “This signature is repeatable and unique for each valve type and size,” adds Mr. Roberts.

The engineers then set up a Motor Power Signature Analysis (MPSA) system. The voltage and current inputs feed into the DA system, Figure 4, where the software digitizes them for processing. The software calculates instantaneous volt-amperes as the sum of the V x I products from all three phases. The MPSA then plots the data in graphical waveforms.

The software saves the power waveform, voltage, current, and power factor signatures as a data file for later use. Engineers identify data points with a cursor, annotating specific points or large data regions. This information is listed in a table with corresponding X and Y values, Figure 2. These values are automatically sent to a report generation spreadsheet, in this case Microsoft Excel, using the Dynamic Data Exchange (DDE) feature of Windows. DDE is the standard method for exchanging data between Windows programs.

A predictive maintenance specialist uses the spreadsheet to compare baseline data, typically by placing the new power waveform plot over the baseline plot of the same component. As shown in Figure 3, signature analysis reveals power transmission problems. The motor and pump misalignment cause the motor to consume more power compared with the properly aligned pump. Thus, technicians can spot trends that point to impending failure.

The motor control center collects voltage and current data for these analyses. For 600 Vac lines, clamp-on transformers directly measure current. For higher voltage lines, permanently installed current and potential transformers step down the voltage to a level compatible with the MPSA system. For a three-phase motor, six of the A/D converter card’s 16 channels are used for measurements. Power and power factor transducers use another pair of A/D channels. No signal conditioning is required because all circuitry is isolated. Data acquisition rates may vary but are normally set between 1,000 and 3,000 samples per second per channel.

Bearing clearance, packing friction and other problems

“In measuring power consumption on a 3-in. gate valve, we noticed a slight variation in the power curve at certain increments of the valve cycle,” says Mr. Roberts. The valve was dismantled, revealing bearing clearance as the problem. Older measurement methods failed to detect bearing clearance and the sound level of normal plant operations obscured valve noise.

Tests on the effects of packing friction found anomalies as small as packing load changes. The effects of valve obstructions were also detectable. Furthermore, torque-actuated limit-switch analysis reliably detected whether a switch tripped normally or because of misadjustment.

For problems originating in motors, engineers use FFT analysis. The software uses instantaneous current data to create frequency domain plots. If the motor is at least 75% loaded during data collection, specific frequency harmonics of the waveforms reveal broken rotor bars. Power factor provides additional diagnostic information. Power signatures can also detect voltage imbalances, single phasing, and reduced source voltage produced by improper transformer size or tap setting.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.