With the advent of Industry 4.0, as the Germans call it, manufacturers are using various types of sensors to collect information on asset health which are used for performing predictive analytics such as generation of preventive maintenance work orders and predicting potential asset downtime.
When working with integrating sensor data from manufacturers on our Factory.360 platform, some of the most prominent sensors that we came across include temperature, voltage, vibration, electricity, and humidity sensors. This article delves into the usage of acoustic sensors in factories to diagnose problems and develop risk mitigation plans.
We often diagnose problems in machines dependent on whether we can hear a noise or not. However, sound that is out of human hearing range can also say a lot about machines. Humans can only hear sound in the range of 20 to 22,000 Hz. This leaves a lot unheard. Using machine learning and deep learning, it is possible to analyze sounds that cannot be even heard.
Problems with usage of light or ultrasound sensors
Most of the faults can be detected using sounds since machines consist of moving parts that grind against each other leading to friction and noise. Other means like visible light cannot be used for such machines since light cannot pass through asset components and hence would fail in determining any critical issues.
Some interest has also been shown on the usage of Ultrasound as a sensor. However, Ultrasound, which can also detect minor sounds, is expensive and would require movement of a receiver and transmitter around the machinery similar to an ultrasound machine in hospital. It is for this reason that this method is not preferred.
Additionally, industries do not prefer invasive solutions and hence, usage of acoustic sensors allows a non invasive setup with minimal intrusion into the workspace.
Acoustic Sensors in the market:
In combination with machine learning algorithms and predictive algorithms, non invasive acoustic sensors can help in the detection of such inaudible noise long before the machine begins to fail. Frequency analysis can be done to analyze the frequency of a slight aberrant noise which might not be heard by us.
Another way of detecting sounds real time is through the usage of acoustic cameras that pick up noise and visualize it as heat images. This information can then be analyzed using complex computational algorithms to determine the root cause of failure. For example, in power transmission systems, such cameras could determine specific points of unusual sound and use it to predict early stages of part failure. In pressurized piping systems carrying air or liquids, such cameras could detect exact points of leakages.
This acoustic information can be collected by placing multiple sensors at target points in machines and connecting them to a wireless edge device which directly transmits and uploads the data to a cloud server where it can be analyzed. Combined with an asset management system and predictive analytics, this can give us detailed insights into key efficiency asset parameters.
Acoustic Diagnosis helps in sustainability
According to research, upto 40% lof energy costs in a factory could be caused by air leaks. When a motor starts getting damaged, the efficiency of the machine reduces. To make up for the reduced efficiency contributed by the fault, the motor consumes more energy. This leads to additional electricity consumption and higher electricity bills.
Detection of defects based on inaudible sound allows manufacturers to fix the machines, reducing downtime and electricity costs and extending the lifetime of machine. This can total to a significant savings in operations. For instance, we have seen some companies reduce their electricity consumption by upto 10% which is significant considering that almost 500 million motors are being operated globally. Combined with avenues for predictive analysis, these benefits could get compounded.
Considering all the above advantages, acoustic sensors and cameras can be a powerful tool for predictive analytics.