The chemical engineering professional organization, AIChE held their 2021 AIChE Spring Meeting & 17th Global Congress on Process Safety as a virtual conference in April.
Emerson’s James Beall presented Novel Techniques to Improve Polymer Reactor Tuning based on a project he worked on. Reactor control is complex with many interactions between control loops. The way to tackle this challenge is to start by identifying the process dynamics and create a realistic process simulation to identify improvements in the control strategy to address the complexities.
James opened his presentation showing the slurry loop reactor temperature control project on which he consulted.
He highlighted the challenges in this process:
- Temperature controller tuning led to temperature variability causing reactor downtime due to reactor kill by the safety system
- Long Reactor startups (waiting on temperature to stabilize)
- Tight temperature tolerance for shutdown
- Multiple product formulas require different tuning
- Operations adverse to step testing and tuning changes
Improving the control strategy could drive performance improvements, such as:
- Minimized downtime associated with coolant system tuning (swapping coolers, catalyst mud pot empty, etc.)
- Faster startup since temperature stabilizes quicker
- Reduced frequency of trips from temperature variability
For this project, James and the plant engineers used a model-based tuning methodology, Lambda tuning to address the complex dynamics. The process began by using the DeltaV Entech Toolkit to measure the dynamics with a manual output step test. From the process response to this output change, tuning could be calculated with a desired closed loop response time and a sufficient stability margin. The project also required a medium fidelity simulation of the interacting loops to be able to test the loops together to better predict the interactions.
The coolant process had 2nd order self-regulating, lead with overshoot dynamics and the step test measured deadtime (Td), process gain (Kp), time constants Tau1 & Tau2, and feedback factor beta of the process variable (PV) response to the step change. The reactor temperature had a second order self-regulating, overdamped response with a measured deadtime, process gain (Kp), time constants Tau1 & Tau2.
He explained the path to a solution by taking the following steps:
- Measure temperature responses from setpoint (SP) changes of the temperature controller (TC) in automatic mode
- Develop medium fidelity dynamic simulation in control system
- Estimate improvement & plan step tests based on simulation
- Measure temperature response dynamics from manual step tests for each polymer group
- Calculate recommended tuning and revise simulation
- Compare existing and recommended tuning using simulation
- Implement recommendations and revise as needed
- Confirm performance with all polymer formulas
They used the simulation for the reactor and coolant loop to test the original versus new tuning parameters. For example, when increasing the ethylene feed to the reactor, the original tuning recovered the process to steady conditions in 50 minutes. With the new tuning using the Lambda tuning method, the recovery time was reduced to 15 minutes.
In the live process, recovery when taking a cooler offline required an hour. With the tuning changes made to the coolant temperature control loop, recovery took less than 15 minutes. With the new tuning in the coolant and reactor temperature controllers, the reactor startup time was reduced by 50% and spurious reactor trips were virtually eliminated.
Visit the Production Performance Consulting section on Emerson.com for more on the services to help you assess, improve, and sustain efficient and reliable operations. You can also connect and interact with control and variability reduction experts in the Control & Safety Systems group in the Emerson Exchange 365 community.