Peter W. Hills, Mechanalysis (India) Limited, India
Proactive condition management of rotating machinery is not new to the power sector and is applied widely but with varying degrees of success. The financial benefits have long been recognized and widely reported, but the cost of implementation, required expertise and continuity of the systems remain as constraints to its broader use. To date, the focus of condition monitoring of rotating equipment has been on detecting the mechanical aspects of a machine, such as imbalance, alignment, etc, with little attention being paid to the on-line detection of its electrical system.
However, with the development of the Artesis Motor Condition Monitor (MCM), which applies advanced modelling techniques to the machine’s three-phase electrical supply, on-line detection of both electrical and mechanical faults can be detected.
Condition Management Strategy
When a maintenance director is looking at introducing a new condition management strategy into an organization there are a number of questions that first need to be answered. They include:
- What do you want to achieve?
- What is your budget for equipment and training?
- How many and what type of machines are to be monitored?
- Do you require integration with plant processes and maintenance management systems?
It is also important to identify whether the process plant is strategic, so that it demands continuous protection, or is of lower priority requiring only periodic evaluation when determining maintenance priorities. Taking such decisions should be part of an overall company strategy to improve plant availability. The commitment of the executive office, as well as the operations and maintenance managers is essential in returning the benefits and meeting the expectations of all stakeholders.
Condition Monitoring Technologies
Acoustic-based condition monitoring goes back to the birth of the railways when wheel tappers listened for differences in sound to identify cracks in the wheels. In the 1960s, delicate laboratory scientific instrument were transformed into robust field tool for use by mechanical engineers and technicians and in the early 1980s, the advent of the PC and microprocessor produced a powerful signal processing capability. This made portable vibration data collectors and on-line monitoring systems excellent machinery diagnostic tools. With this development came increased complexity and therefore the need for greater expertise. In addition, an increased understanding of machinery performance and limitations of fault detection was recognized.
While most vibration analysis systems are sensitive to the development of mechanical defects, electrical faults are largely ignored or at best left to an occasional check. Although the cost of electronics has come down, the installation costs of on-line systems remain high, particularly with on-line sensor cabling, which is often more expensive than the detection instruments themselves.
However, a proven affordable system that addresses both mechanical and electrical monitoring on-line is now available. The MCM applies advanced modelling techniques to the driver and driven machine’s electrical supply to detect electrical and mechanical faults. Electrical motors, generators and transformers, with their associated equipment, can now be monitored on-line without specific sensors hard-wired to the machine. In essence the electrical machine becomes the sensing device, thereby eliminating system components and items prone to damage during emergency machine replacement.
It is natural when introduced to a new technology to make comparisons with what is presently known. A solution that appears to overcome many of the constraints of existing systems will always be viewed with a degree of scepticism. Even though there is no single technology that can detect all rotating machinery faults, progressively more and more integrated fault detection systems will be developed to meet this objective. The MCM takes machinery condition management a significant step forward by combining electrical and mechanical fault reporting.
On-Line & Offline Vibration Protection Systems
Off-line and on-line vibration systems both have a number of strengths and weaknesses. Off-line vibration monitoring provides unlimited readings, is portable, non-invasive and on the spot visual faults can be noted, but it can be prone to reading errors, limited diagnostics with critial machines and is labour intensive, especially data analysis. The strengths of on-line vibration monitoring include continuous, consistent data acquisition and the ability to communicate with the Distributed Control System, but it has a high initial investment and is limited to utilizing installed sensors.
Both vibration-based condition monitoring systems also tend to be designed to only detect mechanical faults, which are generally the result of contamination, lubrication loss, wear and thermodynamic changes in the machines. The electrical condition of rotating plant is seldom considered. Although faults can be detected by vibration measurement when at an advanced stage this is far from ideal for a proactive maintenance strategy.
A typical traditional vibration on-line protection monitoring installation will consist of numerous customized elements. Bringing these elements together often involves quite a complex and time-consuming process. For a relatively simple 18-channel system, it can take from start to finish at least six months before it is installed and operational.
In contrast, the MCM is simply one meter per machine. Each unit is simply installed in the existing motor switch panel and connected to the existing current transformers, where it measure both mechanical and electrical stresses based on ‘learned’ performance.
As we have discussed, traditional vibration based condition monitoring and protection involves mounting sensors on the motor and driven machine and measuring the overall vibration energy. Analysis of the spectrum using an external device with expert skills is very effective and can identify many faults developing.
Meaningful analysis often requires the observation of measurements made over a long period of time, most often manually. Portable motor current signature analysis has sometimes been used to complement mechanical vibration-based programmes. This approach is based on analysis of the line current supplied to a motor.
The variances in the stator-rotor air gap are reflected back in the motor’s current through the air gap flux affecting the counter electromotive force, so that the current carries information related to both mechanical and electrical faults. Faults will therefore exhibit a change in the frequency spectrum of the current at specific frequencies.
Although data acquisition is simple in current signature analysis because only electrical signals are measured, it is invasive and in many instances the plant has to be switched off to connect the sensing devices. Interpretation of the data also requires expert personnel and is as time consuming as vibration analysis. Just like vibration analysis, current signature analysis is an output assessment only. It can be difficult to determine whether an abnormal signature is due to a problem in the motor or to unexpected harmonics in the supply voltage or some process change.
MCM was developed to eliminate the shortcomings of both on-line and off-line vibration and the current signature analysis systems, by taking a radically different approach using a model-based fault detection and diagnostics technique where the expected dynamic behaviour (model) of the three-phase system (Figure 1) under varying conditions, such as load, is determined and compared with the measured dynamic behaviour to monitor abnormalities.
Figure 1: MCM connects to current transformers of 3-phase supply
The MCM first learns about the motor-driven system to which it is connected to for a period of time to acquire and process real-time data. The data are analyzed using a set of advanced system identification algorithms in order to allow the calculation of expected dynamic behaviour and model parameters. Changes in the parameters of the system can be used to indicate any abnormalities developing in the system. Further processing of these parameters is used for fault diagnosis.
In contrast to traditional vibration and current signature analyses, this approach is based on a cause-effect (input-output) relationship that is immune to ambient or input noise. Additionally, the difference between expected and actual behaviour filters out and enhances only abnormalities generated by the system allowing the presentation of both earlier and more accurate alerts. The expert system approach eliminates the need for database or record keeping, expert personnel, and time-consuming data gathering and analysis. It provides comprehensive fault coverage (mechanical and electrical as well as the driven system), even though it measures only voltages and currents.
How MCM works
MCM uses model-based fault detection and diagnosis techniques (Figure 2). The system first learns the characteristics of the motor-based system for a period by acquiring and processing the motor data. The results are stored in its internal database and a reference model established. This is represented by the values of a number of model parameters, in terms of both mean values and standard deviations. While monitoring, MCM processes the acquired motor data and compares the results to the data stored in its internal database. If the results obtained from the acquired data are significantly different from the reference model, MCM indicates a specific fault level taking into account the magnitude and the time duration of the difference.
MCM monitors and compares 22 different model parameters that are classified into three groups.
Group 1: The eight ‘electrical parameters’ are the network equivalent parameters and are correlated to the physical parameters of the motor, such as inductance and resistance. These parameters are sensitive to electrical faults developing in the motor. MCM evaluates and analyzes the differences between the model parameters at any moment in time and the average value of the same parameters that are obtained during the learning stage. These differences are normalized with respect to their standard deviations obtained during the learn stage. The processed values indicate the number of standard deviations they are away from the average values obtained during the learning stage. If they exceed threshold values an alarm is given. Changes in their values are associated with faults that are developing in the system, for example an isolation problem in winding affecting the parameters associated with resistance. This would allow MCM to detect the isolation problem at an early stage.
Although MCM is primarily used to detect electrical problems, it also can indicate incipient mechanical faults. For example, an imbalance or gear problem would cause dynamic eccentricity in the air gap, resulting in a change in the induction parameters and therefore in the model parameters. By monitoring the changes in these model parameters imbalances can be detected very early, avoiding damage to other machine components, such as bearings.
Group 2: In addition to the above parameters MCM also monitors the supply voltage, as well as the load conditions. If the supply voltage changes abnormally, has an imbalance or very high harmonic content then it issues a ‘Watch Line’ alarm. Similarly, if the load conditions do not match with the conditions observed during the learning stage then again a Watch Load alarm is issued. The Watch Load alarm means that either the load conditions have changed or there is a fault developing in the system. If the user determines that there is a change in the process, they can add this new load condition into the conditions observed during the learning period.
Group 3: Using the measured three phase voltage and current signals, MCM also calculates a set of physical parameters such as ‘rms-values’ of the three phase voltage and current, power factor, etc. This group also includes parameters such as total harmonic distortion, harmonic content of the incoming signal and voltage imbalance that give an indication of the quality of supply power. Active and reactive power parameters in this group might be used for energy consumption estimations. It combines many measurements that are of interest to both production and maintenance operators in a single device. For all its sophistication, MCM is a small meter that is ideal for installation on motor control panels. Selected measurements can be displayed on the LCD screen of the device and alarms are given using LED traffic lights for ease of use.
Ease of Integration
MCM can be integrated into plant-wide operations information system with its own desktop application MCM SCADA used for trending and diagnostics. The MCM SCADA also provides the user with reports outlining fault status, diagnostics, as well as relevant parameters about the operation of the equipment during a selected period. In addition to trending, it obtains the frequency intervals of mechanical parameters and determines the corresponding faults, such as bearing problems, imbalance and looseness that are then presented to the user.
Average values obtained for energy consumption (voltage, current, active power, reactive power and power factor), as well as the quality of the power supply (THD, harmonics, voltage imbalance and current imbalance) are also provided (Figure 3).
Figure 3: A MCM diagnostic report
MCM can automatically send this report when an alarm occurs and at selected periods using e-mail. The MCM SCADA also allows diagnostic data to be monitored from several different computers at remote locations.
Alternatively, MCM can be integrated into automation and maintenance management systems using the standard Modbus protocol.
Successful Applications in the Field
This condition monitoring technology has been successfully used in a wide range of industries, some of which like the power industry have applied condition monitoring for many years. They have found MCM to be complementary to existing systems, particularly for remote or inaccessible machinery prone to electrical faults. In addition, where equipment is difficult to justify the higher cost of traditional approaches, the MCM is an ideal solution. An example of this are hydro turbines, especially older units that do not justify expensive on-line diagnostic systems yet experience many electrical winding failures.
MCM has also been successfully used to monitor compressors, and in one example correctly identified a developing bearing fault and provided a 3three month warning to allow the successful planning of a maintenance intervention. The measurement trend also showed that MCM was aware of the incipient defect for a full month before considering it sufficiently serious to alert maintenance staff.
The automated diagnostic capabilities of MCM have allowed users to extend their condition monitoring programmes at minimum cost and without increasing the load on already stretched analysts.
New applications for MCM include wind turbines, where existing condition monitoring systems tend to consist of an array of sensors, wiring, etc, that all add to the cost of the installation. MCM as a single instrument offers dual role monitoring at minimal cost and no intrusion to the rotating machinery.
The article is based on the paper ‘Intelligent Condition Management On-line’ presented at POWER-GEN India & Central Asia 2008, 3-5 April, New Delhi, India.