July 2019

The Industrial Internet of Things

Drive IIoT success with advanced analytics

Refineries, petrochemical and other process plants face similar challenges in daily operations, as each facility must mitigate risk, anticipate maintenance, optimize operations and minimize operational expenses.

Refineries, petrochemical and other process plants face similar challenges in daily operations, as each facility must mitigate risk, anticipate maintenance, optimize operations and minimize operational expenses. To achieve these goals, plant personnel rely heavily on data to drive decisions.

The excitement around the Industrial Internet of Things (IIoT) is its ability to deliver this data at lower cost and in greater volumes than ever before. This is accomplished when IIoT implementations take advantage of advancements in the areas of wired and wireless sensors, networking and data storage technologies.

The driving forces for these innovations are declining costs and increased connectivity options, enabling plants to measure far more parameters of interest, many of which were too expensive to monitor with older solutions.

Most of these sensors will be connected to a plant’s existing regulatory control or asset management system. Wired sensors with 4-20mA outputs require spare inputs on these systems, but many newer sensors and instruments will use fieldbus digital communications, simplifying wiring and lowering costs.

Wireless sensors can be connected to a plant’s control or asset management system via a gateway. The sensors connect to a gateway that is hardwired to the host, usually via an Ethernet connection, eliminating the need for additional input points.

Access to this data by engineers and plant management may be enabled worldwide to anyone with intranet or internet connectivity and proper security credentials. The proliferation of high-speed Wi-Fi and cellular networks has made this type of connectivity ubiquitous and reliable.

However, collecting, storing and disseminating data is just the starting point for IIoT implementations. Process and reliability engineers, often referred to as subject matter experts (SMEs), must analyze this data and derive results to provide operational guidance.

This is best accomplished by using advanced analytics software to connect to the myriad of data sources. The right software is a very effective tool in the hands of SMEs, enabling them to create and share insights using a workflow process, as shown in FIG. 1.

The following examples show how advanced analytics software can be used in refineries and petrochemical plants to improve operations.

Preventing vessel brittle fracture

Vessels in refineries and petrochemical plants require scheduled inspection and maintenance because many different factors can cause degradation of vessel integrity. A major risk requiring attention during transient operations is brittle fracture.

Brittle fracture occurs when a vessel experiences high stress, such as elevated pressure, without being sufficiently preheated. The minimum temperature at which a vessel can withstand pressure is often represented by a minimum allowable temperature (MAT) curve plotted on an x-y diagram, with pressure on the x-axis and temperature on the y-axis. Awareness and careful monitoring of pressure and temperature during vessel heat up and cool down can mitigate risks of brittle fracture.

Plant personnel often use the MAT diagram appropriate for a piece of equipment given its metallurgy, and transpose these as limits, particularly for startup and shutdown procedures.

A better approach is to use advanced analytics software to convert the relationship between pressure and temperature, as depicted on the MAT diagram, to a new signal based on the vessel pressure. This allows a desirable operating range to be established and used to closely monitor operations based on live pressure data.

MAT diagrams differ based on a vessel’s metallurgy and other factors, and the new signals created for comparison must also be unique to each vessel to align with operating pressures and temperatures.

The left side of FIG. 2 shows a MAT vs. pressure diagram applicable for a vessel.1 This diagram represents operating conditions as one focused point on the trend. During startup or shutdown conditions, that point may vary significantly. To maintain integrity, it is important to ensure that this point will always remain above the MAT line.

As shown on the right side of FIG. 2, a SME can use advanced analytics software to create a piecewise function for this line, which now becomes an inferred variable calculated as a function of the vessel’s pressure. This allows plant personnel to track operation in real time by observing—and alarming, as required—this inferred signal, which varies based on the measured pressure variable.

Protecting coke drums

Like vessels, coke drums experience temperature fluctuations on a regular basis. Metallurgists or vessel vendors can identify acceptable heat-up rates that are potentially separate from brittle fracture pressure requirements. This data can be used along with advanced analytics software to identify drum cycles, calculate the rate of temperature change for each cycle and trend those aggregated rates to determine when and if they change over time.

This same approach can be used to immediately identify when a heat-up or cooldown was faster than recommended. Identifying these deviations can dictate what operational adjustments must be made to meet desired vessel life, and to drive inspection schedules in response to deviations.

Analyzing coke drum cycles is an area where reliability and process engineers must often collaborate. Reliability engineers may be interested in the heat cycles, while process engineers may be interested in target temperatures or the duration of coke drum cycles.

With traditional analytics tools such as spreadsheets, reliability engineers usually rely on process engineers to identify the times/durations of each cycle. As new cycles occur, someone must be responsible for documenting each cycle and its associated parameters, and then disseminating that information.

Rather than passing off spreadsheets or updating spreadsheets manually with new information, advanced analytics software can be used to facilitate ongoing documentation and sharing of cycle information.

Predicting heat exchanger performance

Process engineers are often responsible for justifying maintenance on fixed equipment. In advance of a planned outage, these engineers are tasked with identifying the heat exchangers to be serviced. The goal is to minimize maintenance costs and outage time while ensuring that successful operations can be carried out until the next planned downtime.

To do this, process engineers must transform historical data into a forecast of reasonable expectations. This is usually done through the application of first-principles equations to calculate heat transfer coefficients based on historic temperatures and flowrates. Fluid properties from equipment data sheets or historical lab data may also be included in these calculations.

The result of these calculations is a model representing how the heat transfer coefficient is expected to change over time. This usually requires the data and raw calculations to be filtered first to create an accurate model, often referred to as data cleansing.

SMEs can perform data cleansing with traditional spreadsheet tools, but this is a cumbersome and time-consuming process. For example, a SME may need to remove data if an instrument was out of service for some time, or if an earlier process outage occurred where the calculations do not apply. Using advanced analytics software to perform this cleansing can simplify and streamline the task of refreshing a heat exchanger end-of-run forecast.

Once the appropriate data is identified and cleansed, the decay of the heat transfer coefficient can be accurately modeled based on time in service and other applicable variables, such as target temperature or anticipated flowrate (FIG. 3).

This model can be used to project the heat transfer coefficient decay into the future to determine when the exchanger will fail to meet its required heat transfer rate. This information can then be used to predict heat exchanger performance and schedule maintenance only as needed.

Modeling success

Process engineers are actively engaged in monitoring product quality and yields, so they must closely observe operating conditions. Many process units can produce different grades of products when parameters are adjusted. For example, increasing reforming severity can result in lower liquid volume yield but high-octane product, and these types of operations will sometimes be necessary to meet product specifications.

To anticipate how much an operational target must be adjusted, a SME is often asked to create a model for the process. Carrying this reformer example further, it is possible—but not likely—that one model for reformate yield can encompass the wide array of inputs.

In most cases, the most accurate models will instead be based on narrower subsets of data. For example, the process engineer will likely create low- and a high-severity models, and use each model as needed to meet operational goals.

Advanced analytics software empowers engineers to create multiple models and then combine them into one model representative of operations at any point in time. While process engineers will generate these models and use them to understand specific situations, such as how the catalyst may be degrading over time, others in the organization can also benefit from their detailed efforts.

Since this new model will be of interest to many different experts within an organization, collaboration and knowledge sharing help increase efficiency. For instance, operations planning personnel must understand these models to provide appropriate guidance to meet product requirements, or to evaluate the plant’s capability to make new products (FIG. 4).

Using advanced analytics software with built-in sharing functionality allows others within the organization to access the models developed by SMEs, supporting data-driven decisions across the enterprise.

Cutting operating expenses

Advanced analytics software can also be used to reduce operating expenses. Most process plants have a planning engineer or other individual responsible for communicating intermediate or utility requirements to service providers on a routine basis. These are often referred to as nominations and can be applied to electricity, natural gas, hydrogen or other utilities or intermediates purchased by the plant from a third-party provider, such as an electric utility.

These nominations are necessary for service providers to plan and adjust their operations appropriately. For example, a natural gas utility must plan pipeline operations and supply to allow for required amounts of natural gas deliveries. Commercial contracts typically require a process plant to consume some percentage of the nomination, with a penalty charged for either over or under consumption of the nominated volume.

To minimize operating costs in environments where production, and therefore utilities consumption varies widely, planning engineers must combine their knowledge of planned operations in the short term with current and historic actual consumption.

This can be done by using advanced analytics software to create a model for intermediate consumption, using process historian data from past operations. Data variables may include flowrates, temperatures and other process parameters from many units. Much like the reformer model mentioned here, multiple models must often be combined to more closely represent operations and produce the most accurate results.

While the model is based on historical data, it must be applied to current and expected future data to provide an appropriate look ahead to anticipated operations. Traditional tools would require manually copying and pasting both recent historical and planned data into a spreadsheet as part of this analysis. Using advanced analytics software that can connect to various data sources, an SME can skip these data access steps to quickly and simply view historical and planned data.

Another advantage of creating the model with advanced analytics software is the highly visual nature of reviewing the results and determining when to retrain or adjust the model (FIG. 5). While this kind of activity can be carried out with traditional tools, model refinement is more complicated with those approaches and can lead to lagging when adjusting for error.

Depending on the size of these lagging corrections and the accuracy of the traditional models, these lagging nomination errors can cost operators millions of dollars annually. Advanced analytics software improves the quality and timeliness of models, allowing plant personnel to more accurately predict utility requirements.

Takeaway

For each of the examples discussed here, traditional tools such as spreadsheets can be used, but these general-purpose tools require much more effort than using advanced analytics software specifically designed to create insights from time-series process data.

Many of these examples involve advanced modeling. An important aspect of effective predictive models is the incorporation of subject matter knowledge, and this expertise must also be shared to produce maximum benefit throughout a company. Advanced analytics software simplifies sharing and collaboration, allowing engineers to invest their time and knowledge in activities driving their organizations to higher profitability.

As data provided by IIoT implementations continues to increase, corresponding demands on SMEs will only increase, driving the need for advanced analytics to find more insights faster. HP

LITERATURE CITED

  1. Benac, D. J., N. Cherolis and D. Wood, “Managing cold temperature and brittle fracture hazards in pressure vessels,” Journal of Failure Analysis and Prevention, Vol. 16, Iss. 1, February 2016.

The Author

Related Articles

From the Archive

Comments

Comments

{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}