Key points
An engineer working at a corn milling production plant receives his daily report on specific opportunities to optimise the facility’s regulatory control systems. The facility is one of dozens owned and operated by this world-class manufacturer, and it relies on 100s of PID controllers to maintain safe, profitable control.
Thousands of data sets generated daily by the plant are automatically captured before being modelled and having their results aggregated. After reviewing the analysis and recommendations the same engineer keys in new tuning parameters for a few underperforming PID controllers and moves on to his next task. As simple as that – notification and assessment, implementation and improvement. Free to move on.
From Food & Beverage and Pulp & Paper to Chemicals and Oil & Gas enhancements to controller performance monitoring technologies are increasingly coming to the aid of manufacturers across the process industries. They’re leveraging innovative modelling capabilities that easily cope with the highly variable dynamics that are typical of manufacturing environments.
They’re automating advanced analysis of existing process data and everyday output changes. They’re empowering engineers to optimise production on the fly and to realise meaningful gains in performance. And they’re helping to resolve mechanical issues that otherwise hamper a manufacturer’s supply chain. This is how Big Data is supposed to work.
Realising new value from everyday events
Loop monitoring technologies have been around since the start of the new Millennium. They actively monitor PID controller performance using a facility’s existing process data. Details seemingly hidden within the data can reveal when a PID controller’s tuning parameters no longer suit a process’ changing dynamics or when a valve starts to operate near its constraints.
These characteristics are the precursors to inefficiency, quality defects, and equipment failure. While the shift from good control to poor control is often gradual the impact on production can be immediate. With so many loops and so many competing priorities it’s often difficult for production staff to know which valve requires attention and which control loop needs adjustment.
The cause-and- effect relationship between the CO and PV serves as the basis for modelling process dynamics and calculating controller tuning parameters.
The ability to actively capture and model everyday output changes was first introduced over a decade ago. Early monitoring technologies sought to capitalise on these events as they routinely occur at/production facilities.
From manual output changes to automated adjustments of a controller’s Set Point, these everyday modifications to a loop’s Controller Output (CO) caused a response by the associated Process Variable (PV).
The cause-and- effect relationship between the CO and PV serves as the basis for modelling process dynamics and calculating controller tuning parameters. Unfortunately common complexities in the process data such as noise and oscillations limited the success of early monitoring technologies.
As a result a relatively small portion of the available data sets could be modelled with accuracy. These early offerings relied on the same modelling methods utilised by traditional tuning software. Output changes needed to both start and end at a steady state in order for an accurate model to be calculated.
Accurately modelling process dynamics is at the core of PID controller tuning and process optimisation. The challenge of noisy, oscillatory conditions was finally overcome in 2009. That was the year that a proprietary method was first introduced for addressing the steady-state requirement.
Called the Non- Steady State (NSS) Modelling Innovation it eliminated the need to both begin and end the bump tests performed during tuning sessions in a steady-state condition. More recently the NSS Modelling Innovation has been adapted for use in conjunction with plant-wide monitoring technologies.
It capitalises on the enormous amounts of process data routinely captured and stored within a plant’s data historian. This new capability was introduced in 2013 and its value as an advanced diagnostic and optimisation utility have been impressive. Indeed, the results confirm that Big Data can drive big paybacks.
Mining models for black gold
An oil extraction company is among numerous manufacturers that have realised significant value from control loop performance monitoring (CLPM) technologies and the dynamic process modelling feature available in select CLPM solutions. Each of the company’s facilities successfully produces over 30,000 barrels of crude oil daily by applying a specialised technique called steam-assisted gravity drainage (SAGD).
The process addresses complexities that are unique to the bitumen found in the oil sands – a composite of crude oil, partially consolidated sandstone, and loose sand. Nearly 1,000 control loops are used at each location to regulate the complexities of the SAGD oil extraction processes.
During one 4- month period a total of 323,612 models were accurately calculated using the facility’s volatile process data and the CLPM technology. That’s an average of 3 models per loop each day.
For each facility that equates to almost 1 million models per year with individual models providing valuable insight into the associated processes’ ever changing dynamics. In aggregate the models reflect a broad range of operational conditions. What’s more, they provide all that’s needed to tune PIDs for optimal performance.
While it’s common for practitioners to limit the number of models when tuning due to the time and impact of traditional step tests, dynamic process modelling provides an abundance of with zero negative impact.
Practitioners familiar with both modelling a process’ dynamics and tuning PID controllers appreciate the benefits of multiple tests and the time that’s typically involved. Indeed, most industrial processes demonstrate some degree of non-linear behaviour – they act differently as they transition across operating ranges.
As such tests performed above and below a control loop’s normal operating range help to reveal the underlying dynamics and to calculate model parameters that account for the full range of behaviour.
These tests can involve extensive amounts of time which is a resource few practitioners have in abundance, and they are in addition to the other output changes that occur daily at a typical production facility. The incorporation of advanced and automated modelling capabilities within CLPM solutions represents a meaningful step forward.
Dynamic process modelling is a Big Data utility that facilitates continuous, plant-wide process optimisation. In addition to modelling everyday output changes select CLPM solutions aggregate the resulting data and analyse the results on an individual loop basis. With aggregated data these technologies empower engineering staff to determine which results correspond with a process’ normal range of operation.
Other relevant controller details such as the existing tuning parameters can be made available thru a plant’s existing data historian. Access to that information can be used to simulate the controller’s performance, contrasting its findings with the performance of existing tuning parameters and providing recommendations that align with requirements for loop responsiveness.
While it’s common for practitioners to limit the number of models when tuning due to the time and impact of traditional step tests, dynamic process modelling provides an abundance of insight with zero negative impact.
Using the treasure trove of process models and analysis staff made regular, incremental adjustments to the extraction plant’s regulatory control systems and steadily improved the process’ ability to unearth ‘black gold’.
Alerts and reports kept production staff aware of opportunities to optimise were uncovered. Details of each model fit and associated analysis appeared with a simple click on the CLPM solution’s user interface along with new tuning recommendations.
From fast responding pressure and flow loops to slower temperature controllers, solutions like Control Station’s PlantESP supplied all that the engineering team needed. Most importantly, no additional bump tests were required to realise the gains in both performance and production. The company’s SAGD operations are setting the pace for efficiency among producers in the region.
The continuous march toward process optimisation
CLPM solutions are among a growing portfolio of technologies that link IT and OT and improve a manufacturer’s operational effectiveness. Whereas many Big Data technologies applied to manufacturing focus narrowly on the reliability of key production assets, CLPM solutions cast a wider and complimentary net by targeting the performance at a facility’s expansive network of regulatory control systems.
From increases in output and quality to decreases in energy consumption and waste, the financial gains associated with improved regulatory control have been shown to impact a plant’s top- and bottom-line performance. A recent case study involving the deployment of CLPM technology at an automotive parts manufacturer documented a 13.5% increase in production capacity.
Those gains were made possible by identifying dynamic process behaviour that unnecessarily extended production cycle time and by correcting the issues with adjustments to the facility’s regulatory controllers. Such improvement potential is sure to be of interest to those process manufacturers who view continuous improvement as the foundation of their competitive advantage.
Whether value comes in the form of increased throughput, reduced downtime, or just a simpler means of maintaining safe operations, Big Data is empowering manufacturers to march steadily forward.
Innovation has a long history of driving down the costs of both production and distribution. Whereas control technologies such as the PLC and DCS helped to automate production processes, today’s Big Data solutions are equipping manufacturers with the insights needed to optimise the broader supply chain.
They’re revealing the cause and effect of harmful variability, and they’re improving a manufacturer’s ability to meet the lead times so critical to efficient distribution. Even the lowly PID controller has captured the attention of Big Data.
Indeed, web-enabled technologies like CLPM are accelerating the digitisation of supply networks – an increasingly complex and global network that’s on- demand and always-on. This empowers manufacturers to squeeze additional value from the production resources within their control.
By leveraging existing data, performing advanced analytics, and connecting management more closely to the supply chain CLPM solutions offer transformational value. That is the promise of Big Data. Whether value comes in the form of increased throughput, reduced downtime, or just a simpler means of maintaining safe operations, Big Data is empowering manufacturers to march steadily forward. Alert and assess, implement and improve. Free to move on.
About the Authors: Damien Munroe is General Manager of Control Station Limited located in Roscrea, Ireland (damien.munroe@controlstation.com).
Peter Thomas, C.Eng. is Managing Director of Control Specialists Limited headquartered in Warrington,United Kingdom (peter.thomas@controlspecialists.co.uk).
