Which element does AI need to predict electric equipment fault? Harmonics wave patterns.
It is difficult to predict the fault of electricity-powered equipment, such as motors, pumps and air-conditioning systems. When a piece of equipment which has been working just fine until yesterday suddenly breaks down, we are forced to bear unexpected costs for repairs.
There is insurance against breakdowns. Many suppliers suggest an option of replacing parts periodically after certain years of usage. However, insurance requires recurring expenses, and the periodic parts replacement indiscriminately replaces all the parts at once including sill usable ones. This is a source of concern for the managers in charge of building and factory facility maintenance.
Inverters are used to power motors for energy-saving operation. They are commonly used in air conditioners and pumps. Inverters have the advantage of easily changing the operating speed of equipment, but they also have the disadvantage of releasing ‘electrical rubbish’ called harmonics, which have a negative impact on the power supply system. Harmonics can be troublesome, but there is an endeavour to use the nature of harmonics for fault prediction.
Sudden equipment failures or breakdowns may cause unplanned downtime and loss.
What are harmonics in electrical circuits?
The type of electricity we receive from power companies is alternating current (AC) electricity. The AC electricity is generally characterised by the periodic switching of positive and negative electricity. The supply frequency, such as 50 Hz and 60 Hz, is determined according to the number of switching times (by the way, in Japan, there are two frequencies: 50 Hertz in Eastern Japan, including Tokyo and the regions northwards, and 60 Hertz in Western Japan, including Nagoya and westwards). The supply voltage and frequency of the electricity from power companies are regulated at the national or regional level worldwide.
There is a type of electricity called harmonics. Harmonics are a phenomenon in which a number of electrical frequencies that are integer multiples of the supply frequency are superimposed on the same electrical wiring. For example, harmonics of 50 Hz (fundamental) are, being multiplied by an integer, 100 Hz (second harmonic), and 150 Hz (third harmonic), and all these frequencies flow overlapping on the same wiring.
When inverters are used as a power source, harmonics disrupt the pure sine wave AC of the commercial frequencies. In the electrical engineering, harmonics are therefore considered a problematic element that needs to be controlled. Currently, in many countries around the world, including Japan, there are tolerance standards in the power supply system for harmonics that come from semiconductor equipment such as inverters. Engineers who install inverters must take measures to suppress harmonics to a level that does not exceed these standards.
Do harmonics indicate possible equipment fault?
When electrical equipment such as drainage pumps and air-conditioning systems are driven by inverters, the power current always contains harmonics. In recent years, it has been understood that harmonics describe different wave patterns before an equipment fault occurs.
Let us look at the current which is supplied to air-conditioning equipment, for example. Suppose that the third harmonics generated from the fundamental wave of the current (in the case of 50 Hz, the current at 150 Hz) have risen in an unusual manner, compared to the way immediately after the new installation. If the air-conditioning equipment then suddenly suffers a bearing burnout, this increase in the third harmonic can be inferred as a sign of fault.
This is the idea to use harmonics, which are usually treated as a nuisance, to detect the signs of fault. Faults of electrical equipment occur suddenly. It is difficult to detect clear indications by hearing the sound of machine or looking at the inside. Therefore, if you are able to predict a fault, you can take preventive actions such as replacing the relevant components before an actual incident happens and eliminating machine downtime.
The AI-powered fault prediction monitor by harmonics of Panasonic. Image source : Panasonic
Panasonic’s AI-powered fault prediction by harmonics
The proportion of harmonics in the electricity current supplied to electrical equipment can be easily measured with specialised instruments. By now, such measuring instruments have been used for verification in order to suppress the harmonics.
There is an endeavour now by the Japanese electronics manufacturer Panasonic to detect signs of fault by monitoring the measured harmonics at a fixed point and observing significant changes in the wave pattern.
Reference: https://www3.panasonic.biz/ac/j/service_solution/ai-diagnosis/index.jsp?ad=press20210208
This is a method of estimating the probability of fault by observing changes in the harmonics and using statistical methods. No theoretical law has been discovered as to which parts are deteriorating if any of the harmonics change. Judgments are therefore largely made based on empirical rules.
The service offered by Panasonic uses the AI to compare the acquired harmonics data with those that has been compiled in the past. The AI has been pre-trained by the vast amount of harmonics data on the cloud server, and it refers to the harmonics distribution that is classified as the sign of high probability of fault and determines what type of fault is likely to occur. Customers can receive a fault prediction notice simply by sending their harmonics data to the server.
It is a rare attempt to apply the harmonics wave patterns for such a purpose, and Panasonic’s solution is also a novel initiative; there is no need to attach a sensor device to the machine since the monitoring is conducted on the electric cables in the equipment’s power control panel. The data is obtained without being influenced by the vibration and disturbances of the machine.
Statistical methods and rules of thumb are often used in the field of facilities management. As this fault prediction method has only recently been introduced, the quality of training data is required to be constantly updated. Further accumulation of a large number of past cases will contribute to more accurate decisions.
The application of AI-powered fault prediction monitor : conveyors , fans , pumps, etc.
What is quantifiable will be replacing the human eyes?
Until now, the human eyes have carried out routine inspections of electrical installations; inspectors check for possible problems through visual inspections, temperature measurements, and the detection of unusual noises. Capturing changes in the wave patterns of harmonics during normal operation as a sign of fault means that the inspection criteria of the past have been replaced by algorithm.
The new technique relies on an electrical parameter: the harmonics. Yet both the previous and new methods are similar in the sense that the previous workflow and the new AI-powered one both identify signs of fault from the past examples.
What people can judge can be judged by AI as long as the parameters can be quantified, and this notion would suggest that many workflows will be automated by AI. There is a possibility in the future that another technique such as measuring the sound of frequencies that human is not capable to hear to detect a sign of equipment fault.