Overcoming IT/OT convergence challenges for predictive maintenance in process manufacturing
Bridging the information technology/operational technology (IT/OT) system divide is crucial for process industries for successful predictive maintenance (PdM) utilization.
- vUnderstand predictive maintenance’s (PdM) role in process manufacturing and how it can reduce unplanned downtime for companies.
- Understand three challenges manufacturers face bringing information technology (IT) and operational technology (OT) systems together.
- Learn about gathering the right data to make PdM successful and how companies can develop a system to get the right data they need.
Predictive maintenance insights
- Predictive maintenance (PdM) helps prioritize and schedule maintenance more by providing real-time monitoring and informing companies when they will fail.
- PdM is critical for process manufacturing applications because it can help reduce unplanned downtime, which can cost companies a lot of money and endanger workers.
Predictive maintenance (PdM) helps prioritize and schedule maintenance more effectively by monitoring assets in real time and using the data to predict when they will fail, allowing maintenance teams to plan ahead of time. The consequence of this up-to-the-minute intelligence is significant: It can significantly increase overall equipment effectiveness (OEE), helping manufacturers put a dent in the estimated $50B in annual losses they face from unplanned downtime.
Traditional maintenance programs are costly and inefficient. Some studies have shown up to 40% of maintenance work is considered unnecessary, tying up surplus capital for extensive spare parts inventories for no valid reason. A small chemical plant, for example, may have more than $4 million in parts on its shelves at any one time, which is a substantial sum of money that could be better allocated elsewhere.
Accessing suitable data – and subsequently analyzing it to make predictions and taking action – is one of the significant struggles with PdM implementation. Manufacturing plants are challenging because they’re comprised of many physical assets (e.g., pipes, valves, compressors, pumps, motors, bearings), which are all being monitored and controlled by operational technology (OT) systems. Getting this data extracted from OT systems and into a usable format for PdM is a complicated and sometimes futile task.
While businesses have seen the rewards from digital transformation efforts in information technology (IT) infrastructure, many struggle with bridging the chasm between IT and OT systems in manufacturing. The flow of information between the factory floor and the back office is critical in PdM and relies on close coordination between IT and OT systems.
Integrating existing OT infrastructure into modern IT systems is one of the most significant barriers standing between plant operators and the implementation of a successful PdM program. It’s expensive to integrate systems and challenging to manage, which is why leading manufacturers often use third-party vendors. However, many don’t understand the data because it lacks the proper context needed to drive insights and then action for OT systems.
Three IT/OT system integration challenges in manufacturing
Several challenges in IT/OT system integration in manufacturing exist, particularly in process industries. First, the technology in most OT systems, including industrial control systems (ICS), supervisory control and data acquisition (SCADA), and plant data historians, introduces substantial complexity, including time synchronization issues, data compression, and calibration problems. Second, most process manufacturing plants are only instrumented for process control – they don’t have auxiliary instrumentation that might be used by PdM applications. Finally, the output of a PdM program relies on a more holistic view of every asset in the system; this requires a combination of asset data and maintenance management systems, which not every factory has.
1. Connecting control systems
OT systems are the critical backbone to the entire manufacturing process, ensuring visibility and control over operations. However, every plant has a patchwork quilt of disparate OT systems that help manage, monitor and control physical procedures across a vast range of assets. These disparate OT systems may include:
Programmable logic controllers (PLC)
Distributed control systems (DCS)
Manufacturing execution systems (MES)
Computerized maintenance manufacturing systems (CMMS)
Other control systems.
Gaining access to these systems in an IT environment is not as simple as adding an ethernet cable to the back of each asset. Each control system is like an island on its own, driven by proprietary, incompatible, and often antiquated protocols. Furthermore, most plants are “air-gapped” from the internet due to security considerations. All these factors render intersystem communications within a plant difficult and communications between facilities nearly impossible.
2. Collecting sensor data
Collecting data from sensors, devices, and all the equipment assets needed for processing for PdM presents another challenge. For example, sensors built for control systems are different from sensors used for PdM applications. In one case, sample rates for a temperature or pressure control might be collected at a rate of one per second. In contrast, the vibration data needed to assess pump health requires sampling rates of several thousand times per second, far too much for control systems but much better suited for the IT network.
3. Access to historical data
By design, control systems often do not provide long-term data storage, so manufacturers turn to data historians to record and retrieve production and process data. Data historians are software programs designed to capture and record operational data in a time-series database for fast retrieval. Connecting the data historians to the IT network provides historical data on plant assets. Maintenance records are another critical area needed for effective PdM. Beyond data historians, computerized maintenance management systems (CMMS) need to be connected to align the maintenance information with the historical performance of each asset.
Orchestrating IT/OT convergence by finding the right partner
There is no shortage of businesses out there promising an easy way to integrate and converge IT/OT systems and implement PdM. The reality is this is a time-intensive, expensive and complicated process many plants are not prepared to handle. That means finding a technology partner willing to work in the trenches and helping the company achieve long-term success.
The process begins with a complete audit of the plant, control systems, historical data, maintenance records and PdM goals. All this information is critical to reducing unplanned downtime with high prediction accuracy and with as much lead time before failure as possible. The audit also will uncover specific challenges associated with connecting control systems to the IT environment. From there, a technology plan can be developed to help overcome these significant barriers, all in the name of extracting the highest quality data possible, including:
Connecting control systems: Identifying and connecting each system using the right mix of connectors and software to connect and communicate while extracting real-time data.
Collecting relevant data: Augmenting control system sensor data by integrating sensors to capture the right data attributes and frequency needed for PdM.
Connecting historical data and maintenance histories: Connecting data historians and CMMS systems to get a comprehensive overview of asset health and history.
Establishing communications: Building a communications network that pulls sensor data into edge gateways to feed into industrial computer systems, either wired or wirelessly.
Why the right data matters in PdM
PdM improves the availability of equipment by anticipating when a component may fail. The upside to this foresight is the ability to schedule maintenance at a more opportune time. The quality of the data feeding into the system directly correlates to PdM predictions’ effectiveness, especially in process industries where downtime can be disruptive and expensive. That’s why having the right data at the right time is so essential.
For example, if new sensors were installed, it may take several months or longer before there is an adequate baseline to understand the assets. The data from the OT systems give the added context from the control, historian and maintenance systems to expedite the bootstrap period. Real-time data and historical data are crucial in providing insights into the aging of a particular asset. The more rich, variable and nuanced information available, the more significant the insights and the higher the predictability for a PdM application.
Control valves case study
Control valves are critical equipment in process manufacturing systems. These valves may be hydraulically- or electrically-operated and are used to direct the flow of various process fluids between various systems. They are subject to many different failure modes such as internal or external leaks, clogging and erosion. In sensor-rich environments such as oil exploration, it is common to see many OT system sensors around control valves including flow, pressure, temperature and even vibration in order to detect anomalies such as cavitation that might lead to valve failures.
However, many process manufacturing plants do not have the luxury of comprehensive instrumentation because there are often hundreds or thousands of valves all around each plant, and maintenance or instrumentation budgets are typically strained. As such, a PdM system that is developed for data-rich environments would be useless in most process manufacturing settings.
In one particular case, the company needed to monitor multiple control valves, but there was almost no OT system data related to these valves except for open-close commands and open-close feedback (i.e., sensors that indicate when the valve is fully closed (or opened) once it is given a close (or open) command. Instrumenting these valves with additional sensors was out of the question due to budget concerns.
Anomaly detection and prognostic algorithms are designed to work with a wide range of available data. By deploying algorithms that work with the existing instrumentation, we were able to detect anomalies such as clogging (leading to a gradual increase in open-close times) and predict when the valves would fail. These prediction capabilities allowed the customer to prioritize valves for inspection and repair and therefore eliminating unscheduled shutdowns due to valve clogging issues.
Advancing predictive maintenance for process manufacturing systems
The PdM solutions of today have made great strides in helping plant operators plan for equipment maintenance. However, they’re still not where they need to be. At best, existing solutions only give a few days’ lead time to act on their predictions. Yet, PdM failure predictions of one or two weeks in advance are not adequate for most chemical plants.
It will be critical to look at approaches that extend the time horizon on failure predictions, giving plant operators the ability to deploy their maintenance resources more efficiently. Doing that requires understanding the data requirements from the ground up, and taking appropriate measures to make sure that the data is available to the PdM system, whether it comes from IT or OT-connected devices.
Do you have experience and expertise with the topics mentioned in this content? You should consider contributing to our CFE Media editorial team and getting the recognition you and your company deserve. Click here to start this process.