PLCs improve predictive maintenance
Pulling vital information from field devices and transmitting it to the enterprise boosts maintenance practices.
Maintenance practices on the plant floor in the past were purely reactionary, with many responses to issues or trouble along the lines of, “Something just happened; let’s get out there and fix it.” Even with scheduled preventive maintenance, a common refrain could be, “The schedule says to go out and work on it, so let’s go.”
Reactionary and unnecessary maintenance behaviors are quickly going the way of the fax machine because with today’s advanced technology and increased Industrial Internet of Things (IIoT) connectivity, the understanding and planning of maintenance activities is easier and more cost effective. A strong predictive maintenance program can come to the rescue.
One way to start is by employing IIoT-capable programmable logic controllers (PLCs). PLCs can connect with field input/output (I/O) and operational technology (OT) protocols that tie into myriad deployed sensors, and then communicate seamlessly to enterprise information technology (IT) services via the Cloud to provide information on the condition of devices, machines, and the overall manufacturing environment (see Figure 1).
There is no doubt PLC technology is already strongly established on the plant floor. However, by embedding IT protocols, Cloud connectivity, and security features into today’s PLCs, it is possible to gather data that may have existed idly and use it to provide a much stronger idea as to what condition devices and machines are in to prevent unplanned downtime.
Avoid unplanned downtime
Unplanned downtime is the bane of any manufacturer’s existence. By drawing raw data from the plant floor and promoting it to Cloud resources, a PLC can make it available for plant personnel to review and act upon to avoid unscheduled production stoppage.
A modern PLC can pull data from OT sources and push it—in proper context—to IT-based enterprise and Cloud services, effectively bridging the manufacturing floor to the business enterprise (see Figure 2). Some of the most-used protocols and methods are listed in Table 1.
Any effective predictive maintenance model is based on collecting and analyzing machine data over time. With that data readily available, it will be possible to detect patterns and interpret when a machine could go down. This way, it is possible to schedule a maintenance session before an unplanned event can occur.
The goal is to know when the equipment needs repair or replacement. This way, operators or plant personnel are not gambling with their process, and they have relevant data that puts any issue in context, so it is possible to decide on a solution.
Getting the data quickly and securely is key to the success. To garner plant floor intelligence for predictive maintenance in the past always meant hardwiring sensors to PLCs and then to a PC and then to a network, then to a server and then to the Cloud.
Today, it is much simpler because some PLCs can communicate directly to the Cloud in addition to their traditional roles of controlling equipment, operating complex process loops, and supporting a human-machine interface (HMI). Or a PLC may need not perform any control function at all, and it can act solely as a direct data bridge over intermediate network layers. Some small or medium PLCs are even certified to link to Microsoft’s Azure Cloud Platform, assuring designers the device can work with this Cloud infrastructure.
To perform useful analytics, and eventually machine learning (ML) or artificial intelligence (AI) objectives, a comprehensive and consistent data set is needed. Most organizations looking to implement a predictive maintenance system quickly discover they need more data to feed the ML model to get an accurate prediction, and they need flexible ways to distribute the information to users (see Figure 3).
Traditional PLCs could communicate data, but the content was largely unstructured and required significant programmer effort to propagate communications from the plant floor up through many layers of systems. This can be very costly to configure, deploy, and maintain when using intermediate gateways and classic architectures.
Modern PLCs, on the other hand, can gather OT raw data, preprocess and aggregate it to an extent, and send structured data with context directly to IT systems and the Cloud. This is a more direct and less complex connection than traditional solutions, improving responsiveness and reducing the upstream costs for storage, visualization, and computing. It also makes for a smooth data transition between OT and IT and modularizes and simplifies repeatability of a solution. IT can collaborate with OT to add data points as needed right at the source. This allows OT to maintain ownership of the application and to gain insight on IT technologies.
PLCs working with Cloud platforms like Azure make it even easier to connect data processing and storage closely with the data source, enabling fast, consistent responses with reduced dependency on intermediate resources. Azure tools allow an organization to build its own custom predictive application or use an off-the-shelf solution. Azure offers multiple IIoT capabilities to help users visualize and optimize operations, including:
- Machine learning and analytics to build advanced predictive models that can aid in a maintenance program
- Cosmos DB for data storage
- Power Apps for easily building low code solutions
- Web and mobile visualization.
Two significant benefits of using Azure are scalability and security. Azure provides the infrastructure to scale from a single device to millions of devices without rearchitecting the solution. Azure also enables an organization to aggregate data from multiple locations geographically into a single data store to get a complete picture of how operations and maintenance are performing across the whole organization. Best practices for IT security are applied for applications and communications, and the data store can be redundant.
Data is important, but information provides operational wisdom. For building ML models to predict machine/component failure, Azure offers multiple tools (Jupyter Notebooks), frameworks (PyTorch, TensorFlow, and scikit-learn), and languages (R, Python). These models can be applied to one piece of equipment and scaled to address similar equipment throughout an organization.
Increased connectivity throughout the manufacturing enterprise and up to the Cloud delivers benefits that manufacturers have talked about for years to help them boost productivity and profitability. The catch, though, with that increased connectivity comes a heightened security risk. That is why there should be an enhanced level of security with any new connections through devices. Users need to understand the days of air-gapped systems and networks are gone and increased connectivity is here to stay.
Modern PLCs with IT/OT capabilities should employ built-in security features, which include:
- Not allowing requests from the outside the network
- Storage of username and password credentials should be managed by OT personnel; no default passwords
- IP sharelisting to control which applications are approved for use
- Secure communication over TLS when possible.
In addition, a good idea for any organizations is to allow for greater visibility, which can show what is going on over the network. When it is possible to observe the current level of traffic, understand what the baseline network activity should look like, what is normal, what is abnormal, users may be able to find any anomalous behavior. In short, it is important to discover and be notified about traffic that doesn’t conform to what a normal situation looks like. Many other measures and technical controls are important such as authentication, endpoint protection, backup, and network segmentation, in addition to the right processes and the right governance.
Troubleshooting a problem
Increased visibility not only works to enhance security, it also aids when, despite all predictive maintenance planning, an incident occurs, and plant personnel must troubleshoot the problem.
When a problem occurs, a PLC can quickly address it when configured with built-in troubleshooting tools that indicate issues based on real-time trending data. This way, it is possible to pre-process diagnostic data and make it remotely accessible at any time.
Logging data and being able to understand what happened and when—and why—remains a vital and often misunderstood element to troubleshooting an incident. By enabling greater accessibility and understanding of plant data, users are empowered to undertake predictive maintenance efforts that let them take action to avoid failures and minimize incidents.
Modern IIoT-capable PLCs can be a more active conduit for a user to implement a strong predictive maintenance program that can lower costs and unplanned shutdowns, and improve any plant’s performance.