Improving Heat Exchanger Operation with Predictive Analytics
Predictive analytics supplement a shortage of skilled workers and provide actionable information.
Industrial processing facilities face an interesting balancing act. On one hand, they face the challenge of ensuring safety, quality, profit, environmental compliance and product reliability. On the other hand, they face the challenge of applying the right knowledge across organizational and geographic boundaries. And of course, these diverse objectives must be managed within an environment that includes mitigating costs and reductions in available experienced staff.
The operation and monitoring of fixed and rotating assets play key roles in facilities that are trying to find the right balance. One asset — heat exchangers — tends to get less attention because there are no moving parts and there exists a traditionally slow rate of fouling. Yet, exchangers play a significant role in energy balancing for the plant and overall energy efficiency. Many facilities struggle with optimizing energy consumption, and a major cause of poor energy efficiency and production loss is heat exchanger fouling.
Older facilities tend to have just enough instrumentation to safely — though not necessarily reliably or optimally — operate the plant. Without these additional measurements, it can be difficult to fully analyze heat exchanger performance related to:
- Trending the rate of fouling.
- Detecting when the rate of fouling has accelerated.
- Deciding when to clean a fouled heat exchanger bundle.
Determining how to manage fouling can be time consuming because data collection often is done manually through potentially inaccurate processes, sometimes in hazardous environments. After data is gathered, homegrown Excel-based tools make analysis cumbersome.
Facilities across many industries are searching for methods to gather data safely, automatically and accurately. For example, wireless measurements can enable sensors to be added easily and cost effectively where and when they are needed. The updated gathering and analysis processes can be highly efficient, allowing skilled employees to work on higher-value tasks.
Setting up the infrastructure of sensors and enabling automated predictive analytics allow for timely awareness and planned preventive maintenance and repair. Although analyzed data is good, the organization cannot forget to train people to use the new information in a timely manner for corrective action. This means that when presented with analyzed information, personnel know what corrective action to take and are not confused.
Heat Exchanger Challenges
One of the first challenges in uncovering fouled heat exchangers is gathering accurate and meaningful data. Without frequent data collection, analysis is difficult. Without accurate data, meaningful analysis and successful solutions are less likely. Especially in older processing facilities, which might have been built with the minimal amount of instrumentation required to operate the plant safely, measurement devices might not be in place where standard designs would recommend them.
For example, many heat exchangers are composed of several exchanger bundles together (figure 1). If heat transfer decreases in one heat exchanger bundle, facilities have a difficult time knowing which bundle. Because fouling across heat exchanger bundles is not linear, one bundle can have a majority of the fouling and require the most immediate attention and cleaning. So the question becomes, which exchanger?
The bundled exchangers create a series that can be difficult to troubleshoot without measuring flow, pressures and temperatures in and out of both sides of each heat exchanger bundle. To determine which exchanger needs attention, a maintenance department needs a measurement device between each of the bundled exchangers. Often in these situations, however, the only available measurements are recorded at the initial input of material and media, and at the final output of material and media (figure 2). For facilities with older instrumentation schemes, the additional measurements often are not installed between each heat exchanger bundle.
To obtain the necessary intermediate measurements, some facilities use empty thermowells on the tube or shell sides between bundled exchangers. In this situation, personnel manually insert a measurement device and hold it there to get temperature and pressure readings.
Depending upon the environment, the task can put personnel at risk. Regardless of the environment, however, the expense in time to hook and unhook a device for multiple measurements can be prohibitive. In addition, manually inserting a device to read measurements can lead to inaccurate data. The measurement might be taken in the wrong place and result in a poor reading. For example, the measurement might be of a bare pipe where insulation is not showing.
When a facility lacks measurements, they limit their ability to quickly identify abnormal operation.
Evaluate Fouling and Other Equipment Conditions through Automated Predictive Analytics
Automatically collecting process measurements makes business sense. You cannot analyze and optimize what you do not measure. In today’s competitive market, instrumentation costs are relatively inexpensive, but wiring costs can be prohibitive.
One solution is adding wireless sensors for flow, temperature and pressure measurements in and out of each side of the heat exchanger bundles. These wireless measurements are only the beginning, allowing the opportunity to turn these additional process data points into timely actionable information with the use of predictive analytic software. Finally, the staff needs to be trained on what actions are required from this new information.
For those facilities that have the required process and asset health measurements, automated predictive analytics, standard processes and procedures to act on this new information, and trained personnel, the result is a predictive work culture that operates with key benefits:
- Reliable equipment.
- Nearly failure-free operation.
- Low maintenance costs.
- Consistently high quality product.
- Low product unit cost.
When selecting a wireless solution that would include a wireless transmitter (figure 3), consider products that use an open standard such as WirelessHART (IEC 62591). Look also for an automation provider with experience in determining the right instrument for the application.
Adding measurements is only the beginning. If you simply add measurement points, the historian will be loaded with data to be used after an incident rather than for improvements or for incident avoidance. Technology can provide facilities with a way to be proactive with data, changing data into information. Predictive analytics turn data into information that can be used to make more informed decisions that, in turn, become successful actions. In a sense, the data can be used to look forward rather than only looking at it historically.
Predictive analytics can capture multiple data points simultaneously, analyze automatically and alert personnel to abnormal operation or imminent failure. As part of predictive analytics, key performance indicators (KPIs) show whether the device is working properly at all times. An automated monitoring strategy provides online indication of an asset’s health, which in turn provides advanced warning and allows time for changes, thus eliminating process upsets, off-specification products and safety incidents.
Training Personnel to Use the New Actionable Information
The goal is to turn that collected data into useful information, whether displayed in KPI dashboards, or alerting of abnormal operation or imminent failure. In order for the information — or analyzed data — to be effective, it must be presented to the right person in a timely manner, allowing sufficient time to take corrective action.
When installing additional measurements, it is a good practice to review standard operating procedures (SOPs) and standard maintenance procedures that describe how staff should react to this new information. Without proper training or updates to procedures, an abnormal operation alert may actually cause confusion and inaction, resulting in no benefit from the addition of sensors and predictive analytics.
So, when starting a program that includes predictive analytics information, train the staff to know what to do with the information that tells them the process is going into an abnormal operation. Supplement the staff with analyzed data, or information, that they can act on.
Predictive analytic software can be generated in-house or acquired from reputable solution providers. Although many facilities have the capabilities to develop their own predictive analytic tools, most just do not have the time to develop analysis software.