Yes, the title of this post is correct. In 2017, ARC Advisory Group estimated the global downtime in manufacturing industry is in the range of one trillion dollars annually. That is a lot of money, and to put it into a perspective, the global GDP in 2019 according to World Bank was 87.8 trillion dollars. It is not surprising that reducing the downtime is one of the most attractive outcomes industrial IoT can provide.
What options exist in reducing downtime? Predictive maintenance has proven a cost-efficient application to address downtime challenges and provide ROI to justify projects. IoT Analytics forecasts that the predictive maintenance market is growing at 39% CAGR to $23.5 billion dollars in 2024. What makes predictive maintenance so attractive is that it addresses two key issues at the same time. If the machinery or components like motors, pumps and bearings are run until they fail, there can be more costly damages done to the equipment due to the failure. In addition, there is the time spent by the staff trying to get replacement parts on site and then working overtime to fix the issue. All of this adds to the final cost of an unplanned downtime event and contributes to lost production. On the other hand, if the equipment is over-serviced by changing wearing parts too often or too early, the downtime also increases because of the too-frequent scheduled service breaks. In predictive maintenance, the algorithms use sensor data collected from the machinery and components to warn the operator of a future failure condition ahead of time, allowing ample time to schedule and plan for the maintenance before the failure occurs.
Predictive maintenance solutions commonly rely on detecting anomalies in vibration fingerprints of motors, pumps, bearings, and other devices that run the industrial and commercial processes. Because cabling costs for adding vibration sensors are immensely high, these sensors are typically leveraging wireless communications and powered from a battery. We have some unique advantages for predictive maintenance solution developers. Our products include industry-leading low-power consumption wireless SoCs and modules. Using the built-in low-power modes, the sensors can benefit from fast wakeup times and balancing time between sleep and active modes. This power optimization translates into longer battery life, which means lower total cost of ownership (TCO) for the end customer because the sensors require less maintenance during their lifetime.
The environments in which predictive maintenance solutions are deployed vary to a large degree. This is why the solution developer should partner with a communications expert like we that can support a wide range of wireless technologies in multiple frequency bands. For longer-range needs, technologies such as Wi-SUN, Mioty, or other sub-GHz options are more suitable. Local networks within a factory or a plant could benefit from Bluetooth and mesh technologies, or leverage existing dual-band Wi-Fi infrastructure to connect the sensors.
Artificial intelligence and machine learning (AI/ML) has extended its reach from being a cloud-level application requiring massive computing resources to something that can be efficiently executed at Cortex-M level microcontrollers. Silicon Labs' AI/ML partners have built tools that allow predictive maintenance algorithms to run on just a few kilobytes of RAM memory. The edge pre-processing means that the local radio can be turned off until there is an anomaly that needs to be reported to the back office system and the operator. This can further conserve the precious battery capacity and enhance the TCO.
If you want to take part in solving this trillion-dollar question, a good place to start is by exploring our Thunderboard Sense 2 Evaluation Kit. This kit integrates wireless communications with an array of sensors, including accelerometer and temperature, which are the most common in predictive maintenance applications. You can also browse our Design Network for partners, who can help you to design solutions that run on our wireless SoCs and modules. Finally, take a look at our recent case study on Sensemore, which chose our pre-certified Bluetooth modules for its predictive maintenance sensor. This decision allowed them to fast-forward their development efforts and get to the market quicker.