Multivariable control is always explained using very complex terminology that involves concepts such as detailed process models, real-time optimization, and matrix mathematics. This means that few people other than advanced process control (APC) engineers understand it. A better understanding of multivariable control in industrial process operations has brought more people into the process automation business, while bringing operational benefits including timeliness, consistency and fewer alarms.
Popularizing ARC Knowledge
Many people involved in projects can be in a state of confusion, often accepting APC projects without fully understanding the goals, benefits, impacts and success rates. Moreover, this situation leaves the industry relying once again on APC engineers to explain the many unexpected shortcomings of APC, such as high costs, short life cycles, and higher maintenance costs, which in most cases have not been satisfactorily explained.
With nearly 40 years of experience with multivariable control, a more intuitive and qualitative understanding of multivariable control and its role in industrial operations has emerged. This may have several beneficial effects on APC and process automation, including simpler and more powerful software tools, better defined applications and greater involvement of all stakeholders.
What is multivariate control?
Multivariable control can be defined as the automation of single-loop controller setpoint and output adjustments that would otherwise be left to the operations team to perform manually. If the setpoint and output adjustments are made by the operations staff during the duty cycle, this is manual multivariable control. Automatic multivariable control techniques, such as model predictive control (MPC) or model-free multivariable control, automate this task.
Automatic multivariable control (or closed-loop multivariable) offers the same benefits as individual closed-loop controls, including more timeliness and consistency, fewer alarms and limit exceedances, and better optimization. It often results in significant operational improvements and economic benefits.
Role in Process Operations
The role of multivariable control in industrial operations can be understood as the difference between automated multivariable control and manual multivariable control. Manual multivariable control has always existed in industry because almost every process operation is a multivariable control proposition. Just ask any operations person and you'll see.
Automatic multivariable control automates or takes over the task of making setpoint and output adjustments for the associated group of controllers. This typically results in more consistent and timely adjustments, fewer alarms and limits exceeded, and more optimization. These benefits can also be understood as inherent to closed-loop versus open-loop control. In the realm of single-loop control, these benefits have always been easy to understand, and in fact, this applies to multivariable control as well.
The traditional constraints diagram illustrates their differences. With manual multivariable control, the operator can keep an appropriate amount of buffer or margin of error between the ongoing operation and the limits in case of unexpected changes or disturbances in the process. The buffer usually translates into economic loss relative to a fully optimized operation.
In manuals, if an unexpected process change or disturbance occurs, the operator maintains a buffer or margin of error between the ongoing operation and the constraint limits. The buffer usually translates into economic loss relative to a fully optimized operation.
Utilizing automatic or closed-loop multivariable control, operations can be kept closer to actual limits and buffer zones can be used as an advanced control advantage. This is possible because multivariable control means that automatic responses can be relied upon to take action when process conditions change. Similarly, multivariable control can automatically fall back within the constraint limits, allowing greater benefits to be captured because it can operate in both directions.
Multivariable control applications remain "under the radar" of the conventional large matrix MPC paradigm because the high cost of MPC has not been successfully justified and is too large for the limitations of Advanced Regulatory Control (ARC).
The "Low Altitude Radar" example shows a multivariate control application that has been "under the radar" of MPC. It shows the number of interventions made by the operator at the console in response to a setpoint, output or mode change in a given time period. It shows the 25 worst "roles", i.e. those controllers that require the most operator intervention. This is an easy chart to draw for any modern control system console.
Effective metrics provide meaningful measurements that visualize progress over time. Has the industry been neglecting metrics that can justify automated multivariable controls? The 25 worst controllers (those requiring the most operational staff intervention) can be easily charted for any modern control system console.
Micromanaging APC
To be sure, most of these interventions represent manual multivariable control scenarios, where the operations team is overwhelmed with micromanaging the group of controllers in question.
The goal of multivariate control is to automate the manual multivariate control scheme and to close the multivariate loop while reducing the number of variables.
Missing APC metrics
Recently, the industry has adopted similar best practices at least twice - manually managing loops (in addition to the manual multivariate loops we're talking about now) and managing alarms for "bad roles" (in addition to the bad roles we're talking about now that require frequent intervention by operations staff). roles).
Manual multivariate loops and frequent operations staff interventions have a number of undesirable consequences, including more alarms and limit exceedances, and less focus and optimization by operations staff on higher-level tasks. Manual interventions are often inconsistent, untimely and sub-optimal.
Effective metrics provide meaningful metrics that visualize progress over time. Figure 2 meets these criteria and reflects the fundamental aspects of successful process automation and the quality of console operations. Is industry overlooking this natural and potentially important metric?