Condition monitoring - an intuitive approach with a little theory

Machines and structures fail. Its a fact of life! The reasons are many and most are avoidable. It is important to understand the modes of possible failure so that their early symptoms may be detected, analyzed and warnings sounded. Modes of failure include:

• wear - loss of material by an abrasion process such as friction
• fatigue - loss of strength due to crack formation under cyclic loading
• corrosion - loss or change of material by an adverse chemical process
• overload - exceeding materials capabilities causing plastic deformation
• ageing - material changing properties with time
• environment - exposure to conditions such as temperature that modify a material's characteristics
• operator abuse - rough or inappropriate use of equipment
• combination of the above

Ideally none of these should occur, however, engineering is the art of compromise and sometimes the compromises are too great, their impact not fully understood or the machine is mistreated.

Direct or indirect methods

The progress of a pending failure can be monitored by direct and indirect means. For example, in measuring fatigue one can measure the acoustic emissions generated by crack growth or one can measure the stress cycles and relate them to a theoretical fatigue model to predict the fatigue. The first method is more direct as it relies on signals generated by the actual fatigue process while the second is indirect as it relies on measuring a parameter likely to cause fatigue. However, both methods require laboratory calibration so the value of one method over the other is not always obvious.

Sometimes the failure mode cannot be tracked or is too expensive to monitor effectively. For example, friction is much more likely if the lubrication system is compromised. Monitoring oil pressure is a form of low cost indirect preventative condition monitoring. Monitoring suspended particles in oil is a more advanced indirect lubrication system monitoring method. An alternative friction monitoring technique is to measure the temperature rise due to friction. This is more direct than oil monitoring but has the disadvantage of requiring potentially damaging friction before a a problem is detected.

These pages cover all aspects of machine monitoring but focus in more detail on areas where real-time monitoring is appropriate. As technology advances, more and more tests that where conducted in the laboratory are being automated and conducted continuously in real-time.

Cause and effect

Machines and structures are monitored in order to predict failures that would have a negative impact on productivity, usefulness or safety. By having early warning of the potential failure, preventative or corrective maintenance can be scheduled for the least disruptive time. Sometimes there are also warranty issues that need to be decided. So, the principal aim in Condition Monitoring is to prevent unexpected failure by the detection of early warning symptoms of the likely failure modes. Detection of early symptoms is the key, and that is the emphasis of these pages.

Often, components of a system are known to fail in their normal use. For example, the brake pads on a vehicle wear and will fail. That is a certainty. We avoid this mode of failure either by regular inspection or by break lining thickness detection, and then replacing the pads before failure occurs. Due to the critical roll of brakes, there are well developed procedures for ensuring actual break failure due to wear does not occur.

Take another automotive example - a vehicle's wheel bearings. Bearings will wear, but the rate at which they wear varies considerably. Bearing load, shock load, lubrication, dust penetration, manufacturing quality all impact the wear rate. Also, a wheel bearing is not easily inspected without disassembly and re-assembly expense. In a ball bearing, the wear is rarely evenly distributed, rather a microscopic area will begin to wear and progressively grow in area and depth. As the balls roll over this area, vibration is created with a fundamental frequency equal to the rate at which the balls drop into the wear pits. Harmonics will also be generated due to the abrupt nature of these wear pits. As a driver of the vehicle, you become aware of the whining noise that the worn bearing generates. It is this type of "noise" that is used to detect the bearing wear by vibration sensors.

Failures always have a reason for happening. As we have noted, lubrication failure will cause a bearing failure that will fail because heat and abrasion from friction will cause accelerated regenerative wear. There is a cause and effect:

Despite the number of causes and effects, well designed machines will have well understood failure modes with well understood causes. Condition Monitoring develops on this knowledge to become an effective preventative methodology.

Looking for symptoms

Condition Monitoring is, more than anything, the art of Identifying early symptoms of probable failure. In may ways condition monitoring is more an art than a science. Even relatively simple mechanical systems can be complex to formally analyize for likely failure modes, so one is often left to apply experience, 'feel" or intuition.

Symptomes can include the following:

• vibration - traditionally the most common symptom monitored
• stress cycle or rainflow counting - particularly useful in machines and structures with an uncertain range of loadings such as aircraft, ships and wind powered generators
• acoustic emissions - ultrasonic sounds emitted by crack propagation events.
• temperature rise - most useful for bearings and seals where friction is likely to be involved
• lubricant - oil analysis for machines with lubrication systems
• operating conditions - operator abuse, environment

Most of these symptoms have characteristic that can be measured, or can be deduced by indirect measurement and calculation. An open mind combined with intuition can often identify symptoms that are not in the text-books!

Condition monitoring tools

Condition monitoring covers such a wide field that as you may expect, there is an equally wide range of tools:

Vibration analysis

The analysis of the unique patterns of vibration created by specific components of an electro mechanical system. These unique patterns can provide early indications of problems such as machinery imbalance, misalignment, bearing wear, worn gears, etc.

Temperature analysis

Identifying an abnormal rise in the temperature of machinery that could result from problems such as bearing wear, lack of lubrication, poor electrical connections, etc.

Oil analysis

Analyzing the physical and chemical properties of lubrication to identify contamination, wear particles, as well as changes in viscosity and other chemical properties. Controlling these critical parameters is key to preventing premature failure of mechanical systems such as bearings, gears, etc.

Motor current signature analysis

Analyzing the unique signature patterns in the motor supply current created by fluctuations in load, high resistance joints in rotor bars and end rings or uneven air gap between the rotor and stator.

Operating state dynamic analysis

The correlation of machinery operating parameters such as load, pressure, speed, etc to the dynamic characteristics of the machinery such as vibration and motor current signature.

Sonic leak detection

The use of sonic and ultrasonic measurements to identify changes in sound patterns caused by gas or fluid leaks, bearing wear or other deterioration.

Acoustic emissions

The use of ultrasonic measurements to monitor the location and growth of fatigue cracks in a structure.

Balancing

The use of vibration measurements to reduce the mechanical imbalance of a rotating piece of equipment. Mechanical imbalance is one of the primary causes of excessive vibration and premature mechanical component failure.

Qualifying symptom measurements

In determining the accuracy of a detected symptom, it is usually necessary to qualify the conditions under which it is being measured. For example, a fan bearing temperature may increase due to thermal conduction after the fan is shut down, giving a spurious result. Qualification rules must be defined, so the rules for the fan may be:

• after the fan has been on for five minutes, measure the temperature difference between the bearing and the fan mount
• if the gradient exceeds the gradient measured yesterday by 1.5°C for more than five minutes, flag a warning
• if the gradient exceeds 10°C sound the alarm

Different symptoms will require different qualifying rules. Sometimes a failure mode will produce multiple symptoms that can be compared in some way as part of the qualifying procedure. Qualifying will be discussed in more detail with each symptom measurement method.

Capturing, analyzing and effectively using machine condition information provides a strategic and competitive advantage, allowing companies to maintain consistent quality and maximize their return on investment.