Quick solutions for intelligent condition monitoring
The Industrial Internet of Things (IIoT) has been talked about for so long that it’s easy to forget that implementation is in its infancy.
In the future, we will look back and marvel at the progress achieved. Right now, many companies are still learning about the opportunities and contemplating how to make a start. Collecting a small amount of data from a limited range of sources can be an effective way to begin. As the proverb says, a journey of 1,000 miles begins with a single step.
In principle, capturing a body of data should enable companies to generate insights into how to do things better. It becomes possible to analyze trends and spot anomalies that can help optimize manufacturing processes, manage inventory, drive waste out of operations and minimize equipment maintenance costs and downtime.
But there are some caveats. Collecting appropriate data and analyzing that data in the right way are essential to yield actionable information. This can be seen as a barrier to companies that may feel they lack suitable data-science skills and analytics tools.
Time and cost also need to be considered. In many enterprises, IIoT still needs to prove its worth and there may be skepticism within the organization. Any budget allocated to IIoT projects may be small, while convincing results are likely to be expected quickly. The IIoT champions in the enterprise may need a fast solution to demonstrate the advantages for the business and thus secure the high-level commitment and further investment needed to scale the company’s IIoT solutions.
The first step
Intelligent machine-condition monitoring can provide a starting point to demonstrate the potential for increased productivity and lower equipment ownership costs. By continuously assessing the health of the system and individual components, condition monitoring can transform traditional maintenance policies — which either correct faults after they arise or try to prevent faults by replacing specific parts at pre-determined intervals — to a predictive model.
Predictive maintenance enables parts to be replaced only when needed thereby reducing equipment stoppages and also helping to avoid excessive spare-parts inventory as well as expenditure. When replacement is needed, the downtime necessary to perform the replacement can be conveniently scheduled. Condition monitoring based on information from sensors attached to the equipment also eliminates time spent diagnosing faults and makes it possible to detect misuse of equipment, either deliberate or inadvertent. Changes in operator training or supervision can then be implemented to prevent losses and increase process yield.
To implement condition monitoring with a rotating machine, for example, relevant measurement parameters include the motor speed, load torque, voltage and current, which can usually be obtained from the drive. Combining this data with information such as acoustic, vibration and temperature data from additional sensors enables abnormal conditions such as bearing wear or imbalances to be detected.
Some industrial drives make provision for connecting external sensors and may even embed condition-monitoring algorithms within the software. To begin condition monitoring, a typical approach is first to characterize the equipment under normal operating conditions, sweeping the full operating speed range, to establish baseline values. Thresholds can then be determined to establish a margin above and below those values. As these thresholds are approached or exceeded while the machine is running, the monitoring system can generate alerts to indicate that action is needed. There are some problems with this approach. If thresholds are set too low, large numbers of false alarms and unnecessary actions will be triggered. If the thresholds are too high, failures will likely occur before an appropriate alert is generated.
Infusing artificial intelligence into the solution can improve the results. Using machine learning techniques, the behavior of a healthy system under normal conditions can be characterized in terms of multiple parameters simultaneously. Unusual performance, such as vibration that is excessive in relation to the values of other parameters such as torque and speed can then be detected quickly and with a high degree of certainty. Without AI, manually calculating the relationships between all parameters, to set dynamic threshold values across the full range of operations, would be impractical.
Modular solution
Avnet brought together temperature, vibration and acoustic sensors to create a multi-sensing module with an integrated machine-learning core (MLC). The MLC simplifies the use of machine-learning techniques to enable predictive maintenance with support for algorithms such as activity detection and audio-scene classification. With built-in wireless connectivity, this module simplifies prototyping and testing condition monitoring systems capable of intelligent predictive maintenance.
The module is based on the STMicroelectronics wireless industrial node (STWIN) core system board. With industrial-grade inertial sensors, acoustic sensing up to 80 kHz allowing sensing in the ultrasound range, environmental sensors for absolute pressure and relative humidity, it is suited to various applications in addition to condition monitoring, including crack detection in ball bearings and detecting leakage in pressurized systems such as gas pipelines.
When creating any connected sensing solution, power management and security also need to be taken into account. The STWIN module takes care of those, too, including an embedded secure element as well as a lithium-ion battery charger. It can also do acoustic scene classification. The module supports Bluetooth Low Energy out of the box and allows alternative connectivity such as Wi-Fi, cellular, or low-power wide-area network using expansion modules. Software is available to connect the node to the cloud and non-data scientists can get a condition monitoring solution within a couple of hours using the STWIN module.
Avnet also provides cloud-based condition monitoring and tracking tools that let users view their equipment status in real time and manage assets remotely. There are dashboard features that let users view equipment operating status and assess performance, including deviations in normal patterns that can indicate machine problems. These tools simplify quantifying machine performance to detect anomalies and provide help with creating a predictive maintenance plan to minimize downtime and boost productivity. Other features include a ticketing system that helps track the handling of issues such as equipment repairs, enhancing traceability and measurability of maintenance activities.
Wireless industrial node for condition monitoring and predictive maintenance
The industrial-grade sensing module is based on
the STMicroelectronics STWIN.
Takeaways
Thanks to the availability of modular IoT sensing nodes and ready-to-use software packs for essentials such as machine learning and cloud connectivity, intelligent condition monitoring is one of the “low apples” among the wide variety of potential IIoT solutions. Quick to implement without requiring significant up-skilling in data science and AI, a modular condition-monitoring solution can deliver powerful results that demonstrate the value of IIoT applications and can strengthen the case to commit to additional and more ambitious projects.
