October 2021

Trends and Resources

Business Trends: How to achieve industrial autonomy in the refining and petrochemical industries

The process industries are moving from automated to autonomous operations.

Millette, P., Honeywell Process Solutions

Driven by advances in digitalization, the convergence of operations technology and information technology and the digital transformation of infrastructure and industry, process industries are gradually moving from automated to autonomous operations. 

Within refineries and petrochemical complexes, the behavior of the material inside pipes and equipment is governed by five well-understood principles: conservation of mass, conservation of energy, conservation of momentum (fluid mechanics), thermodynamics (equilibrium operations) and chemical reaction (kinetics and catalysis). Each of these principles has been used by engineers for generations to design and operate plants safely and efficiently. 

The tools used by engineers to undertake this work have evolved considerably over the years, from manual slide rules and calculators to mainframe computers and, finally, to purpose-built chemical engineering design and simulation applications. 

Likewise, in operations, we can observe a historical and maturing technology set—from manual processes (Level 0) to fully autonomous operations (Level 5). These levels are used industry-wide to track the progression of operations (FIG. 1). 

FIG. 1. The progression of operations. 

However, there are multiple domains of governance in large, complex process plants, ranging from the safety of people and the environment to quality, logistics and regulations. Each of these domains has its own set of decisions and challenges. A few of these will be examined in the following section. 

Cross-domain linkages

The key to progressing performance is to look at cross-domain linkages and ultimately render these as intelligent, resilient and autonomous. 

For example, walking around a plant, a noisy bearing in a rotating machine can attract attention, particularly if the machine is critical. Linking noise to equipment performance can help anticipate problems. Another example is an unexpected smell that hints that something needs attention. 

Making such links across the control system can help drive more autonomous operations. Additional indicators—such as an interface level or water boot level—can be observed by fixed or mobile cameras, enabling operational decisions to be made based on the observations. 

To this end, there are plenty of sensing methods to mimic sight, hearing and smelling via cameras, gas detection and acoustic sensors. Some of these sensors can be combined with sensors that feed directly to higher-level optimization. 

Another example shows the linkages across functions. Switching crude oil feeds to a crude distillation unit remains an ongoing challenge at many sites (e.g., reducing disturbances to product qualities and minimizing operational action). A more autonomous approach would identify key inputs such as accurate or predicted crude tank contents—including any mixing or layering effects—and then link this information to unit controls. More broadly, a direct link from plant scheduling to the autonomous start/stop of a crude switch—to meet scheduling goals with little or no operator intervention—would be beneficial. 

Defined procedures that cut across domains are key to achieving intelligent, resilient and autonomous operations. The primary goal is not eliminating human operators, rather it is deploying technology that allows those humans to make better decisions. 

Control to optimization

In the refining and petrochemical sectors, progress has been made in recent years to improve plant outcomes in terms of safety, reliability and efficiency. The application of technology has played a key role in this effort. 

Process control loops run in automatic mode, continuously correcting deviations in single process values. Alarms provide warning signs of where the process is at the current time, prompting operators to try to understand the nature and cause of the incident and to mitigate unwanted effects. This has been the mainstay of process control since the days of pneumatic control and wall-filled panels of controllers and alarm lights. 

The progression of technology means that the scope of a single operator today is easily three times what it was before digital control systems. Combining multiple event occurrences by the control system into concluded conditions can help operators achieve more intelligent operations. Automated or semi-automated operational procedures also help guide operators, along with context-sensitive documentation. 

Advanced process control (APC) is used to manage multiple interactions and control complex processes to reduce energy consumption while maximizing product yields. Industry is now benefitting from APC models that update themselves, even in closed-loop ones, with no interruption. This helps achieve the resilience of advanced controls. 

The oil and gas industry also understands the importance of optimizing multiple units—particularly in refineries and petrochemical complexes where there are multiple processing options for different streams—to meet production goals or maximize the output of higher-value products. 

Refineries and petrochemical complexes typically have run lengths of 3 yr–4 yr between major shutdowns. During that time, plant equipment (e.g., reactors or distillation towers) gradually fouls and/or degrades over time. Catalyst performance degrades, as well. This means that reaction severity, as expressed by temperature, must rise closer to metallurgy or refractory limits. Presently, this long-term cycle is managed manually, with multiple decision points. 

However, much of this can be supported using simulation (e.g., using a digital twin) to achieve better performance over the run-length cycle of a plant. When linked to the overall economic optimization of a site, this enables the enforcement of run-length driven objectives, underscoring intelligent and resilient operations. 

There is also optimization across multiple units in which the interaction is more economic than process-driven. Technologies to express this relationship of macro-level economics and micro-level stream flow targets are part of the latest developments, helping to make overall optimization more autonomous. A site-wide or refinery linear program model can be linked to individual controls through clever transformation of the decision space from one level to another. Therefore, the governance of plant economics can be made more autonomous. 

Similarly, models that run dynamically can help instruct and train operators to understand the state of their unit and what actions may be required, enabling them to react in the best possible way. Again, this underscores resilient operations. 

Lastly, for offsites operations, where there is remarkable linkage between feed logistics, process units, product blending in refineries and finished product logistics, more can be achieved by taking new approaches to tank farm valve management or the capture of measurements from the field. The use of wireless communications over a tank farm or marine loading dock can help achieve more intelligent logistics operations, enabling field devices to provide additional information. An operator may then be guided to the exact location for a needed manual operation. 

Alarms and process safety

Alarm management and the analysis of alarm logs can help operators and engineers understand the nature of over-frequent events. There are also events captured automatically by safety instrumented systems. Safety event logs have been the subject of manual reviews, while alarm management has progressed in scope and breadth of supporting analytics. Enforcement of mode-based alarms has also expanded the scope of alarm management, but these remain same-unit mode indications. Cross-unit or transitional alarm settings are not generally implemented. 

Process and functional safety have—for several years—had well-defined rules and governance principles, from hazard and operability study (HAZOP) and layers of protection analysis (LOPA) to defined safety integrity requirements. However, verification of function remains largely described in concept as part of standards such as IEC 61511. Achieving more closed-loop verification and flagging the need for remediation of safety elements, such as sensors, valves or logic, are also part of achieving more resilient operations. Whether the scheduling of valve and sensor verification or the notification of results, capturing such information and linking with the process control domain are part of an autonomous approach. Lastly, autonomously managing temporarily degraded safety loops can benefit a plant by removing the requirement for human intervention and ensuring consistent, safe operation. 

Process and equipment asset health and performance

In recent times, the health and performance of equipment assets (e.g., pumps, heat exchangers, compressors and filters) have been undertaken using representation models of structure and behavior. For example, design and real-time performance data of a pump, including its pump curve, along with information on the motor driver and its characteristics, are used to model pump behavior. This has helped give rise to a templated model to benefit all instances of the same type of equipment. While the health and performance of individual equipment are better observed and flagged automatically to engineers for observation and maintenance (including the link to a maintenance system), this is still undertaken on single assets. 

The industry has reached a point in engineering oversight where the next step is to benefit from cross-asset behavior flags. Does the temperature of incoming fluid from an upstream exchanger flag an event to the pump? Does a deficiency in the pump achieving enough discharge head get flagged to a downstream reactor? Autonomous operations in engineering surveillance mean cross-asset conversations in which signals of health and performance are shared. 

As described earlier, the performance of process and equipment degrades over the run length of a plant. The adaptation of performance criteria over that time can help achieve resilient operations within the scope of what is possible at any point in the run. Likewise, asset health and performance can benefit from better information on the activity surrounding them, including ambient temperature, pressure, gases detected in the area and any known process safety degradations. 

Remote operations

In the scenarios described, operator actions in the field have reduced extensively. Within refineries and petrochemical complexes, operators should be focused on performance improvement rather than data collection and manual analysis. The achievement of control center locations away from the plant site is now more possible than ever before. 

However, remoteness is also for project activities within a facility, as well as for new units, expansions or revamped units. The changes in infrastructure of control, safety, and security (including cybersecurity) can be achieved remotely, with minimal plant impact, up to point of cut-over. This is achieved by separating the design of control, safety and security infrastructure from physical equipment and using universal I/O technology that renders every control cabinet the same and every channel able to be configured for any input/output purpose. 

With remote capabilities also comes the prospect of remote support for control and safety systems infrastructure. This can be achieved in a cyber-secure way, with the control and safety system assets doing the talking on the state of their health and performance. 

Enabling systems infrastructure

In industrial process control, risk assessment methods identify where critical faults can occur with a high consequence, and personnel design and implement ways to verify that a single failure does not cause disruption. 

Being resilient means failures can occur, but the system or operation continues to operate normally, and recovery is automated. If something fails, a redundant controller takes over the computing load. 

For example, the author’s company allows multiple process controllers to operate as a distributed mesh—like a data center of controllers. This approach enables simpler project engineering because control strategies no longer need assignment to a specific physical controller. Instead, they can be deposited in a community of controllers that will automatically distribute control to wherever there is available compute across process controllers. 

This contrasts with traditional redundancy in which a single fault triggers a backup failure, and a second fault causes an outage. With resilience, operation continues even with multiple faults, running until the computing is exhausted. 

Simultaneously, resilience must be considered across the entire operation, ensuring that operations can continue in the face of equipment failures, power outages and weather events. An important lesson from current uncertain times is that the industry was not fully prepared for a common-cause issue that affected humans. Having backup control centers helps increase automation resilience. 


This article has examined how intelligence, resilience and autonomy can be achieved across functional domains. As much as possible, the decision space or governance domain between functions should be linked to benefit operations. There are always new technology developments targeting more autonomous operations, and these also play a role in interactions with operational activities. 

The gradual maturing of operations across multiple domains towards more intelligent, resilient and ultimately autonomous operations must remain an industry imperative. HP

The Author

Related Articles

From the Archive



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