AUSTIN, Texas -- Emerson Process Management’s advanced process control specialists have developed an adaptive model-predictive controller (AMPC) that is optimizing production and improving profitability at Sabic Innovative Plastics’ Bergen op Zoom facility in The Netherlands. With a certain part of the process having long dead times varying between 6 and 24 hours, standard advanced process control techniques were not effective. Emerson created a new process model that has ended many years of manual control and produced an immediate improvement in production efficiency.
Sabic Innovative Plastics, one of the world’s largest manufacturers of plastics, produces a range of products including Lexan (a popular polycarbonate alternative) and Noryl (polyphenylene oxide) at its Bergen op Zoom plant. Historically, when setting up a specific set of batch processes, the runtime was calculated based on downstream demand, the available (interim) storage capacity and the efficiency of the various processing facilities. Overcompensation either way meant that production progress tended to be either too fast or too slow, with less than optimal yields. A model-predictive control (MPC) is often applied to challenging applications such as this, but is unsuitable for processes with large dead-time variations.
“We are constantly looking for improvement opportunities,” says Rein de Jonge, global advanced manufacturing systems leader at Sabic. “We have known for some time that one of our processes presented a control challenge which was not easy to solve. We approached a number of automation suppliers, and Emerson came up with what we believed to be the simplest and most elegant proposal.”
To address this problem, Emerson developed a physical model of the complete facility and implemented control methods based on the Smith predictor -- a model-based control tool that can handle the dead-time phenomena of simple processes. The dynamic model was based on mass balance, integrating the variable dead times of the Sabic process. Given the variable dead time, traditional model correction control tools also proved insufficient and were replaced with adaptive Lambda tuning, thus obtaining the adaptive model-predictive controller.
Emerson looked at all the production data from the past few years and used this to verify the behavior of the AMPC, using the subsequent results for further fine tuning. The result is a control system that renders the process much more efficient, with less left-over product and waste, and more economical use of expensive base materials.
“Production is more profitable now, plus the system provides us with information which was unavailable before, helping us to optimize processes even further. The AMPC is not just a technical success; it is also an economic success!” said de Jonge.
The AMPC is a dynamic online mass balance. It is a physical model which Sabic uses to configure the process, with the predictive part providing them with reliable information about how much product will be ready at any given time. A predicted trend line has been integrated into the AMPC that is able to anticipate events and change specific settings at exactly the right time. If actual production ends up being lower than expected, this may be compensated for by higher production at a later time. The AMPC is aware of this and does not intervene. Operators would previously have been tempted to intervene, which would require a reversal later on, resulting in lower efficiency and increased product loss.
At Sabic Innovative Plastics, the AMPC runs as a component of Emerson’s Provox control system, but it also runs on other platforms.