The development of new sensors and mathematical algorithms results in intelligent devices that can model, analyze, and predict the health of rotating equipment
Ideally, a motor should operate to its maximum designed life. Many applications, such as food or chemical processing, can’t afford to shut down a line, even for maintenance, because it is too costly or the process cannot be interrupted. Most motors are designed to operate for about 20 years, with proper maintenance, so longevity should not be a problem.
However, a large number of motors are not reaching their expected life. A 2 hp motor, for example, running a conveyor section at a major food producer’s facility suddenly failed, shutting down the line. Water had seeped into the conduit box during washdowns and eventually ate away at the cabling insulation. It cost the company hundreds of thousands of dollars in lost product and production time.
That company spends $1 million annually to monitor motor vibration on critical systems. Yet none of the complex equipment and maintenance procedures were sufficient to detect problems in that 2 hp motor. Part of the reason is due to the expense of monitoring equipment. Another part is insufficient knowledge about application effects on motor operation.
A solution is at hand. Through recently developed sensors and mathematical algorithms, several motor manufacturers are developing monitoring devices that will model, analyze, and predict the health of motors and other rotating machinery.
These devices will:
• Operate on all size motors, from 2 to over 500 hp.
• Enable users to monitor any and all motors in a facility.
• Monitor winding, bearing and lubrication, rotor, and shaft misalignment parameters simultaneously.
• Provide continuous, real-time, online monitoring of the attached motor.
• Measure and analyze parameters on environment, duty cycles, and installation — monitoring the process and not just the motor.
Sensors: A step beyond control
In addition, these monitoring devices will soon be able to predict the residual life of rotating equipment, a feature long desired by engineers. “It’s no longer sufficient to just diagnose a motor, for example, and say that’s good or that’s bad,” says consultant Dr. Carl Talbott. “Tell me how good, how bad. How long is it going to run? When do I need to relubricate the bearing? How much more time do I have?”
Sensors have been one of the hurdles to monitoring devices with such capabilities. Another hurdle has been the lack of sufficiently powerful diagnostic algorithms.
Part of the problem with sensors has been their cost. A good singleaxis accelerometer, for example, often exceeds the cost of a low-horsepower motor. Also, sensors were difficult to integrate into hardware and condition monitoring systems, until the development of device-level networks and microprocessors. In other cases, the size of a sensor or its sensitivity to noise caused problems.
Reliance Electric Div., Rockwell Automation, has been working with universities and the National Institute of Standards and Technology (NIST) on sensor and software algorithm development to create an on-line motor monitoring device. “The intent is to provide an early warning to incipient motor failure,” says Richard Schaefer, product manager, Reliance Electric. “But it could go beyond that. Eventually, the motor will become an intelligent sensor. It will inform people about application problems and help them manage the process.”
Engineers and scientists are researching several new sensors, one of which is a noise-immune optical sensor. This optical sensor offers potential benefits of low cost, high sensitivity, and low susceptibility to stray magnetic fields. These features are particularly desirable for current sensing. In development are versions to measure current, flux, and torque.
These non-contact sensors are based on the Faraday principle, which says that an external magnetic field can influence the plane of polarization of light passing through a medium. Possible mediums include fiber-optic coil, bulk optical material, or optical thin-film.
Fiber-optic coil has been an uncommon medium for a current sensor because it made such a sensor difficult to manufacture. Stress from bending the fiber into a coil resulted in bi-refringence, an effect that distorts sensor response.
NIST recently developed an annealing procedure that reduces this effect. The procedure makes it possible to manufacture a sensor with more turns. Because the output is proportional to the number of turns, the more turns a sensor has, the higher is its output. In operation, the sensor works with a polarized light source, typically a linear diode. The magnetic field produced by the motor current rotates the plane of polarization of the light source. The amount of the rotation is proportional to the current.
In other research areas, development of bulk and thin-film ferromagnetic iron garnets results in optical current sensors with even more sensitivity. These sensors also have bandwidths of hundreds of megahertz, which enables monitoring devices to sense higher frequencies.
Continue on page 2
Engineers at Reliance have also been working with sensors not normally used for monitoring. “We’ve looked into vibration sensors that are similar to those used in electric guitars,” says Carl Dister, senior electrical engineer, Motor Group R&D. “The frequency bandwidth is well above the audio range we wanted. We’re looking at high frequencies to pick up bearing defects. We’re also looking into thermocouple technology for temperature sensing.”
For torque sensing, engineers are investigating a photo-elastic material that can be painted onto a device, for example, a coupling. When a light source shines on it, it will display a fringe pattern that deforms as the coupling is stressed. Neural nets can be taught to recognize the patterns and, thus, the amount of transmitted torque. All that’s needed is an inexpensive polarized light source. This is still in the lab environment.
Other technologies engineers are investigating for development of low cost sensors include MEMS (micro electromechanical systems technology). A MEMS device is a three-dimensional mechanical structure created through a chemical etching process known as micromachining. Most micro-machined sensors are made of silicon. Several companies offer accelerometers made by this process.
Predicting motor life
Sensors delivering more data are only part of the solution. The next part requires software algorithms that will accurately indicate and predict the condition of the motor.
Engineers, working with several universities over the last three or four years, have developed mathematical equations that accurately define motor behavior under various conditions. These algorithms work with an operating system, all of which are stored on a microprocessor in the monitoring device.
Sensors gather data 24 hours per day. As the data flow in, the algorithms analyze them. Over weeks and months, the analysis becomes more precise because of the volume of evaluated data.
Motor manufacturers have or are developing algorithms that can use motor specific information such as:
• Ball passing frequencies of the bearings.
• Bar passing frequencies of the rotor.
• Stator slot configuration.
• Friction and windage losses.
• Natural or resonance frequency.
• Winding or rotor thermal capability.
• Load data for analysis of potential motor installation problems.
• Environmental factors and their effects on motor operation.
Once the motor-monitor package is installed in the application, engineers will still be able to add newly developed algorithms to the monitor device. Users can download the new algorithms through the Internet.
“With all these data,” says Schaefer, “you can improve the functionality of the motor over time, something we’ve never had the ability to do in the motor industry before. We believe that in five years, about 20% of motors will ship with some form of microprocessor on-board intelligence like this.”
How long do motors really operate?
According to a recent survey, motors don’t live as long as they are designed to. Reliance Electric, Rockwell Automation questioned over 100 engineers about their motor problems. One question asked how long the engineers expected conventional ac motors to last. A majority, 86%, said their motors rarely lasted longer than 5 to 10 years. No one indicated that they had a motor survive to its 15th year. In fact, most were experiencing significant motor failures in the second or third year of a motor’s life, getting about 25% or less of the life designed into a standard motor.
Topping the list of causes were bearing failure and motor abuse. Grease contamination was third, followed by improper motor installation. Poor motor design and weak electrical design ranked last.
An intelligent motor
An example of these new developments in motor monitoring is the IQ PreAlert from Reliance Electric. The device encloses four printed-circuit boards, the sensors, and cabling. Of the boards, one supports the microprocessor and its operating system, algorithms, and memory; another provides the interface to the sensors; a third controls communications through RS232 or a device-level bus; and the fourth board powers the monitoring device.
The sensors look at current and voltages of all three phases; monitor winding and bearing temperatures; and keep tabs on vibration and speed in adjustable-speed applications. The data are analyzed by the algorithms stored in the microprocessor, and if necessary, alert operators or maintenance personnel to potential problems.
In the future, the company plans to offer algorithms that will measure current in adjustablespeed applications and possibly sense torque. Also in the offing are algorithms for lubrication, shaft misalignment (based on current signature), residual life prediction, start-stop information, as well as methods to avoid problems from phase loss, ground fault, and overload without standard protection devices. Engineers will be able to download these future algorithms from the Internet.
The monitor package is integrated with the motor, usually mounted to the side. It is not possible to retrofit the monitor to existing motors. This combined motor-monitor package is available for ac motors, in frame sizes from 180 to 449, and horse-powers from 2 to 500. Users install the IQ motor like any other motor.
Continue on page 3
Sensors monitor bearing health
The developments in sensor technology are not just for motor monitoring. Other rotating equipment, such as free-standing bearings, benefit from sensors increasing diagnostic abilities. One example is the EZlink monitoring system developed by Dodge. In this system, sensors embedded in the free-standing bearing housings measure bearing temperature and vibration. An optional speed pickup is also available.
The system operates as a node on a DeviceNet network. The network provides power to the sensors, which transmit data that the EZlink system analyzes. This analyzed information regarding bearing health is transmitted serially to an industrial control, such as a PLC, or a PC monitor. With the on board memory, the system can analyze just the changes of state and communicate those to a controller or monitor only as necessary.
Use of DeviceNet eliminates the need for discrete power supplies and controls, dedicated input cards, transmitters, power conditioners and substantial wiring.
Temperature measurements are taken by an embedded thermocouple wired to the system. To measure vibration, the system uses a piezoelectric accelerometer. The sensor continuously monitors amplitude and frequency. All components mechanically attached to the bearing will be part of the signature. This allows monitoring of complete pulley assemblies.
Any mechanically attached component that experiences a failure will be felt and reported by the system. With the speed pickup option, a dc proximity probe measures rotations of the locking device. Because the speed is counted, conveyor speeds can be monitored continuously for belt slip, sequence, or breakage.
This system is available for a variety of mounted bearings with bores over 2 7⁄16 in. and on worm, worm and helical, and helical geared speed reducers.
A new spin on torque measurement
New manufacturing technologies, such as MEMS or annealing processes, are not the only tools developers use to create new sensors. The innovative use of materials, such as magnetoelasticbased materials, is another technique.
One example is the TorqStar series of noncontact torque sensors, from Lebow Products Div., Eaton Corp. The sensor consists of a stainless steel shaft that is diametrically sized for a specific torque range. A metal ring fits tightly on the shaft, and an enclosure blocks outside noise interference. The shaft has a journal machined with a precise taper. The ring, which is assembled onto the journal, is a titanium nickel material that is magnetoelastic. Thus, the ring is a pseudo-magnet that does not produce a force when at rest. When the shaft rotates, the distortion in either direction transmits to the ring, expanding it and creating a magnetic flux.
The size of the magnetic force is proportional to the stress on the shaft and subsequent flexing of the ring. A flux-gate magnetometer, strategically placed around the ring, measures the magnetic force. The magnetometer converts the force to an electrical signal, which is then amplified and transmitted to a control. These sensors respond in the frequency range of 0.1 msec.
Engineers can adapt these inline rotary and reaction type sensors to fit within the holder of a rotating cutting tool or attach them to the output shaft of electric motors. The rotary sensors measure torque from 10 to 1,000 lb-in., the reaction sensors from 50 to 1,000 lb-in. Accuracy is to ±0.5% of full sensing scale.