In some industries, the efficacy of maintenance can have a significant bearing on the company’s competitiveness. The more the maintenance costs the lower the margins, and the higher the downtime, the lower the revenues. This is especially true of sectors like manufacturing, oil and gas (O&G), power, construction, and cement. 

Suboptimal maintenance strategies can lower machine productivity by up to 20%. As a result, businesses have been adopting more and more effective maintenance strategies over time. With the Industry 4.0 paradigm, predictive maintenance has emerged as the unabated winner when it comes to maintenance efficacy. As a result, its adoption is growing rapidly. 

However, predictive maintenance is not a single technology. It refers to a category of use cases that are activated by multiple technologies – internet of things (IoT) being one of the most important of them. In this article, understand the role of IoT technology within the larger ambit of predictive maintenance, and how it contributes to the most efficient maintenance strategy today.

Key components of IoT-based predictive maintenance

Predictive maintenance systems make use of data from various machines to forecast their remaining useful life (RUL). This makes it possible to schedule maintenance activity before the machine is likely to fail – and keep it running optimally for the maximum time. The source of the data used for forecasting can be numerous – for instance, data in a SCADA system. IoT-based predictive maintenance makes use of IoT sensors to generate this data. 

Understanding the digital architecture of IoT-based predictive maintenance

IoT-based predictive maintenance systems are essentially built with five key components. These include:

  • IoT sensors, which generate and supply data for the PdM system.
  • Data pipelines, which stream data from the sensors to servers in the cloud.
  • Storage technologies (typically a data warehouse or a data lake) that house the machine data.
  • AI and ML models, use data to predict the time to failure and/or the likely failure mode.
  • A computerized maintenance management system (CMMS), to which these insights are delivered.

PdM systems may employ varying architectures, depending on the use case. For example, some may continuously compute the likelihood of failure, whereas others may update it periodically – depending on the criticality of the machine in the given scenario.

The function of IoT in predictive maintenance

IoT refers to the network of sensors which collect the data to be supplied to the PdM application. It constitutes the sensor technology, which measures the required parameters, and the network technology which transmits this data to the application. 

In older machines, the ability to collect critical data is not available natively. For example, a decade-old conveyor belt is unlikely to be equipped with temperature or vibration sensors, which would help infer if the belt is mistracking, misaligned, or slipping. IoT sensors enable the collection of such data from legacy machines. In some industries, the age of equipment is over 2-3 decades. IoT technology helps organizations collect valuable data from such machines, and maximize the ROI on them. 

A closer look at sensor technology

In IoT-based predictive maintenance, the type of sensor used is determined by numerous factors. Multiple techniques may be available to predict the failure of the same asset, each requiring differing instrumentation approaches. Moreover, multiple types of sensors may be available for the same instrumentation approach. As an example, vibration sensors make use of accelerometers, laser-displacement techniques, or micro-electromechanical systems to capture vibration data from machines. 

The selection of sensor technology depends on the precision required for the use case, operating conditions, costs, and networking limitations. Finally, a variety of network protocols are leveraged to build the IoT network. These include Advanced Message Queuing Protocol, Bluetooth Low Energy, 4G/5G, Message Queuing Telemetry Transport, and Z-Wave.

Building insights with AI and ML

The data streams from the sensors are typically ingested into a data lake through ETL pipelines. The data is usually cleaned, normalized, or transformed before it is used for making inferences. In some systems, an edge server may be used to filter the data to be transmitted to the cloud for the PdM application.

In PdM, the data streams from IoT sensors are ingested into trained models, which can recognize the optimal running state of a machine based on the input variables. The number of input variables may be considerable. For instance, an AI model may make use of hundreds of vibration data streams, and pressure and temperature signals to predict the failure point of a machine. Using these signals, the algorithm may also be able to pinpoint the likely cause of the failure – thereby enabling precise, targeted repair interventions. Advanced predictive maintenance solutions like UptimeAI can make use of unlimited parameters to generate highly precise forecasts.

Bringing it together

The generation of precise insights into the time to failure of a machine is not the end goal of a PdM application. Instead, it is to optimize maintenance schedules, and make maintenance workflows as efficient as possible. To this end, the insights generated by the AI models are delivered to CMMS or MRO platforms, to automate maintenance schedules, speed repair scheduling, and assist technicians in carrying out precise repairs.

What next?

IoT-based predictive maintenance solutions currently represent the cutting-edge in machine maintenance. They offer numerous advantages over SCADA-based PdM systems. For one, they enable the capture of granular machine data (like vibration signatures) that may not be available in SCADA systems, but is of immense use in an AI-based forecasting solution. Secondly, they eliminate the need for building data connectors that integrate legacy OT applications to new-age architectures.

However, some situations may demand the use of hybrid designs that exploit both SCADA data and IoT networks to power PdM applications. Moreover, SCADA systems may still serve as valuable sources of historical data for some use cases, and eliminate duplication of effort.

Elevate the impact of your maintenance organization on your bottom line with UptimeAI, a leading AI-based predictive maintenance solution. Contact us now to get started.