IoT-Enabled Condition Monitoring: A Starting Point for Digital Manufacturing

Flexible evolution possibilities and expandability make IoT-enabled condition monitoring a perfect starting point for digital manufacturing. Its key role is being a source of real-time data about the health of industrial machinery and the environment in which it operates.

Leveraging this data, IoT-driven condition monitoring allows monitoring the current state of industrial machinery and identifying the combinations of equipment’s health, operational, and ambient parameters that can lead to potential equipment failures or cause product quality deterioration.

However, the solution’s capabilities are even wider than that. With the right implementation approach, IoT-enabled condition monitoring acquires advanced capabilities, including deeper analytics and control, and becomes a springboard for future IoT initiatives, driving improvements at each implementation stage.

The Golden Rule: To drive value, iterate!

To avoid excessive investments and get a fast payoff, a good idea is to start an IoT initiative with a basic architecture. Once it is solidly established, it can be further expanded to amplify the solution’s functionality and tap into more use cases.

Step 1: Build the Basics

Although the peculiarities of implementation differ depending on the manufacturer and the type of equipment they use (as every type of asset has its unique pattern of normal behavior), four basic components of an IoT solution remain the same:

Sensors

IoT-driven condition monitoring utilizes the data about various equipment (e.g., current, vibration, pressure, etc.) and ambient (e.g., temperature, humidity, etc.) parameters. This data can be fetched from SCADA, a machine’s PLC or from sensors attached directly to the equipment components or placed nearby.

Gateways

To reliably connect industrial machinery to the cloud software and ensure an uninterrupted flow of data, field and cloud gateways are used.

Field gateways preprocess sensor data in the field, before moving it to the cloud. Such pre-processing includes message filtering and aggregation. For instance, a welding machine’s sensor transmits relatively static vibration measurements, however, with every reading, the measurements increase and decrease by 0.1%. Field gateways discard the intermediate data points for more efficient data transmission.

The cloud gateway ensures secure data transmission between the field gateways and cloud data storage.

Data lake

A data lake allows storing all of the raw sensor data from multiple pieces of equipment and their components without having to cleanse and process it beforehand.

Big data warehouse

When the data is needed for insights about equipment condition, it is extracted from the data lake, filtered, processed, organized, modeled, and loaded to a big data warehouse.

Along with the sensor data, a big data warehouse stores relevant contextual information (equipment maintenance history, recommended operating parameters, fetched from ERP and other enterprise systems).

Step 2: Ramp Up

Once the basic architecture components are in place, the solution can be ramped up with analytics and user applications.

Basic Analytics

Data analytics applications turn time-series sensor data into insights about equipment condition. For that, sensor readings from the big data warehouse are run through analytics algorithms. The findings are visualized and communicated to the solution users via user applications.

For instance, a discrete manufacturer wants to detect industrial batteries with poor performance and promptly reveal batteries that are soon to discharge. The analysis of the data from temperature and voltage sensors shows that say, batteries X and Y have a descending capacity trend and will need to be replaced soon.

User Applications

User applications, web or mobile, enable bidirectional communication between the condition monitoring solution and its users. User apps display the insights gained with data analytics in the form of charts, diagrams, real-time equipment health dashboards, etc.

Moreover, user applications allow entering additional data about the context in which a machine operates into a big data warehouse. For instance, a maintenance technician can use a mobile app to log a reason for a detected deviation in an equipment parameter.

Step 3: Evolve

Depending on the current business needs, an established condition monitoring solution can be boosted with advanced architecture components. Integrating these components will expand the range of the solution’s capabilities to predictive maintenance and product quality control.

Predictive Maintenance

Predictive maintenance uses the data fetched with continuous equipment condition monitoring to forecast whether the equipment is likely to fail within a certain timeframe performing with a certain load, predict potential failure points, and proactively provide recommendations to maintenance personnel. To enable that, the condition monitoring solution is expanded with:

• Deep analytics

Deep analytics unit uncovers data patterns that signal about potential equipment deterioration. For that, the sensor data (together with the historical and context data) is run through machine learning algorithms. The identified patterns are reflected in predictive models. The models are built, trained and tested for accuracy. Once the accuracy of a model is proven, it is applied for forecasting potential equipment failures. As more equipment data becomes available, the models are revised, updated and retested, so that they become more accurate.

• Control applications

With control applications, an IoT-driven solution can function with a certain degree of automation. For example, if a machine learning model identifies that a particular combination of a machine’s engine temperature and vibration parameters (normal when taken separately) can cause a potential machine failure, a model triggers control applications to send a command to a machine’s actuator and put it in a low-speed mode to prevent deterioration, as well as triggers an alert to inform employees of a potential failure.

Product Quality Control

The architecture components that enable predictive maintenance can also be used to ramp up a condition monitoring solution with the product quality control functionality.

• Deep analytics

Deep analytics unit takes in the equipment condition data, combines it with the yield quality data and the context (e.g., recommended equipment settings, maintenance data, etc.) and runs the combined data set through machine learning algorithms to detect those machine condition patterns that influence the quality of the output products.

• Control applications

Once a machine learning model identifies a machine condition that can influence the quality of the products, it informs the solution users of a potential quality defect and triggers an output. For instance, a machine’s actuators can receive a command to set a machine off in case it enters a pre-failure state to avoid producing faulty products and prevent equipment failure.

Gains and Payoffs

IoT-enabled condition monitoring can drive both process and business improvements, some of which include:

• Reduced equipment maintenance costs

IoT-driven condition monitoring solutions leverage advanced machine learning algorithms to detect those patterns in the equipment data that lead to potential failures. This enables maintenance specialists to address issues before they have a significant impact on asset maintenance costs.

• Extended equipment lifetime

Identifying the combinations of parameters that can potentially lead to equipment failures helps to choose optimal equipment operation mode and adjust ambient conditions to prolong the lifetime of machinery.

• Accurate data for production quality improvements

With the real-time data about how machines are running, manufacturers can detect those patterns in machine health parameters that lead to quality deteriorations. They can use this information to adjust certain operating parameters and thus improve the quality of an output product.

On a Final Note

IoT-driven condition monitoring drives value from three perspectives: it provides a real-time view into the health and condition of the industrial machines, paves the way towards predictive maintenance, and allows manufacturers to make data-driven decisions to improve product quality.

To get these improvements faster and with reasonable investments, it is good to approach the implementation iteratively. Rolled out in iterations, an IoT-driven condition monitoring solution will bring both process and business improvements, as well as establish a solid technological infrastructure for your future digital manufacturing initiatives.