PiControl’s new MPC controller COLUMBO maximizes plant profits while overcoming common MPC limitations

COLUMBO – New MPC inside PLC

PiControl Solutions LLC is pleased to offer a brand new release of multivariable model predictive controller, called COLUMBO, a full-blown competitor for Aspen DMC, Emerson PredictPro, Foxboro Aveva Connoisseur, Yokogawa PACE and SMOC and other commercial MPC Technologies. COLUMBO does not need new hardware and software, it can run wholly inside a PLC or DCS.

COLUMBO is simpler, practical and powerful

One of the important differentiating facts about Colombo is that it is designed to be practical, powerful and easy to implement compared to any of the current MPC technologies.

COLUMBO relies less on model accuracy

Current MPC controllers rely heavily on model accuracy and when the real dynamic response is different from the model response inside the MPC, then the MPC predictions are poor. COLUMBO is less susceptible and more immune from model errors. COLUMBO provides much more aggressive response to unmeasured disturbances and model inaccuracies compared to Aspen DMC, Honeywell Profit RMPCT, Yokogawa PACE and SMOC, Emerson PredictPro and others.  The response from other current conventional MPCs is weak because of their heavy reliance on the shape of the model.

But, in many chemical processes models are often non-linear, the shape of the dynamic models changes as a function of various factors (e.g., production rates, new catalyst or different operating conditions) or they are just not accurate because of the inability to determine accurate model because of extremely long dead times, non-linearity, complexities or process sensitivity that makes the identification of the model difficult.

COLUMBO reacts more strongly to unmeasured disturbances

PiControl’s new COLUMBO software technology is different from any of the current MPC Technologies in that it relies and focuses more on disturbance rejection and fast compensation of unknowns to return the controlled variable (CV) back to the set point.

MPC Engineers often sit in front of the MPC screen and relish seeing a beautiful future CV prediction how the CV like (say) temperature or online analysis is predicted to beautifully and impressively go back to the setpoint but as time goes by, the CV (controlled variable) is moving in the wrong direction and the prediction is giving a false indication that everything is going to be okay. This is a common occurrence in many MPCs when the models are not correct or the disturbance that suddenly came is large and causing model prediction error.

Current MPC controllers rely too heavily on the accuracy of the shape of the dynamic model but their feedback correction in response to unmeasured disturbances is poor. PiControl’s COLUMBO provides stronger, aggressive feedback corrective action in response to unmeasured disturbances and nonlinear or incorrect models. It does not rely as heavily on the shape of the dynamic model as conventional MPC controllers but it focuses more on how to reject the disturbance and how to restore it back to the set point.

COLUMBO – only MPC to run inside PLC or DCS – no need for new L3 server

COLUMBO runs completely inside the PLC or DCS so there is no need for struggling with OPC problems and Level 3 Level 2.5 connectivity issues and headaches. Many MPC projects start off with a lot of systems work to install the software, get the OPC to work and need to deal with multiple vendors like the DCS vendor, Microsoft and OPC vendor.

Whereas Colombo runs directly inside the PLC and eliminates the need for a new server and eliminates the need for new software. There is no additional new user interface. The time necessary for the process control support engineers to learn, implement and support the system is much shorter.

COLUMBO eliminates chanced of plant upsets

Another advantage of COLUMBO is that each of the manipulated variables (MVs) can be individually turned on or turned off without the fear of a wrong MV suddenly doing something bad/unintended for another CV. There is no single global on/off or prediction/standby switch like in conventional MPC's.

If you flip the mode from auto to cascade for an MV then that MV goes under MPC control. If the operator needs to turn it off he just puts the slave PID MV in auto or manual and that MV gets disabled from the MPC.

Operators do not have to worry about the consequences of disabling any MV or CV for maintenance purposes. In conventional MPCs, disabling one or more MVs can cause operating problems because the wrong MV may be used for controlling a different/unintended CV because now one or more other MVs are no longer available. Many plant upsets have been caused by operators forgetting or not understanding how a complex multivariable system works.

In stark contrast, Colombo eliminates such fears and worries, as disabling one MV will not have any impact on other MVs and will not cause any problems.

COLUMBO design provides highest level of safety and reliability

Colombo also provides very safe and reliable protection against things that could go wrong. If an operator types in an extra 0, e.g., types 30 instead of 300, COLUMBO automatically protects and will not cause a bump in the process. Also, if any of the CVs spike or go to 0 or have any problems, then COLUMBO will protect against harmful large movements in the manipulated variables (MVs). There will not be a plant upset.

COLUMBO does not need conventional Step Tests

COLUMBO can identify open loop dynamic models with complete closed loop data. COLUMBO can even use oscillatory data which is considered bad data and unusable data by other MPC vendors, but it can be used by Colombo. All other conventional MPC's need step tests for one or two weeks, but with COLUMBO, historic data can be used with minimal need for any step tests for model identification. COLUMBO can even use data with slave PID controllers in auto or even cascade mode and then amazingly identify open loop dynamic models with closed loop data.

Isolate unmeasured disturbances while identifying correct model

COLUMBO is the only MPC that has the capability of isolating unmeasured disturbances while identifying accurate models. There are always unmeasured disturbances in a plant, some of these disturbances appear and disappear as steps or pulses and some disturbances appear as ramps that can continue in one direction for extended periods of time.

Current MPC system identification technologies produce bad models due to data contaminated by unmeasured disturbances. Colombo has the amazing capability of isolating unmeasured disturbances while identifying accurate models. COLUMBO will even display to you the residual trend which is the sum of all unmeasured disturbances encountered and rejected from the data.

Embedded Smart AI- Artificial Intelligence coupled with MPC

Some new vendors have applied full-blown AI technology for providing closed loop control. While AI is very powerful and has a promising future, one of the challenges in the chemical manufacturing industry is the presence of complex unmeasured disturbances which are not repeatable. Lack of repeatability poses serious insurmountable challenges for AI technologies as the future path may not be meaningful based on past behavior/happenings. Various new fancy AI controllers have failed in the chemical industry where the AI training problem is far more complex and difficult compared to a simple case of AI application like in the case of Netflix and YouTube. Whereas the behavior of a human being in terms of his selection of movies or GPS travel path can be easy for training for an AI algorithm, the unmeasured disturbances in a chemical plant are often hard to comprehend for even experienced process engineers who know the plant well and extremely difficult for training an AI algorithm. Also, many MPC vendors have tried the use of neural networks for closed-loop control in their MPC but such products also have experienced challenges especially when getting extrapolated outside the range of training of the neural networks. In complete contrast, COLUMBO uses AI in a limited manner, relying more heavily on disturbance rejection and robustness of the controller trajectory and making an optimal usage between identified dynamic models and AI resulting in increase in robustness and safe control action.

Use COLUMBO to improve your existing competitor MPC

COLUMBO has also been used to analyze closed loop data with existing old conventional MPCs active and manipulating the MVs. Colombo was able to improve the MPC models in Aspen DMC and other commercial MPC controllers. Colombo was used to improve MPC models in other MPC's resulting in operational and performance improvements. The amazing functionality in Colombo is the ability to take excel data on the MVPs and CVs with the conventional MPC running and active and then be able to improve the models with this closed-loop data with the MPC active and with all the slave MVs in cascade modes. This functionality is unique to Colombo and is unmatched by any other MPC technology.

Accurate identification of Pseudo-Ramps

Many chemical plant processes have long lead times and extremely long time constants and are often modeled more effectively and more accurately as pseudo-ramps. The accurate determination of the pseudo-ramp model is often challenging but with Colombo, it is possible to identify the pseudo-ramp models accurately and determine the open loop transfer function parameters.

COLUMBO application opportunities

COLUMBO can be applied to chemical, petrochemical, oil and gas, LNG, pharmaceutical, electric power, mining, shale oil, paints and dyes, food manufacturing, fertilizers, plastics and polymers, elastomers and many other related processes.

If there are PID controllers in your plant and if you want to maximize your monetary profits, increase automation, run the plant smoother, run the plant safely at constraints, reduce utilities, reduce alarms, reduce flaring, reduce plant upsets, reduce the stress on the operators, a COLUMBO application can help you.

COLUMBO MPC – faster, more practical, robust and easier solution

The overall time and effort on a typical COLUMBO project are about half or one-third compared to a normal conventional MPC project. Contact PiControl solutions today, visit their website www.picontrolsolutions.com and send an e-mail to- info@psicontrolsolutions.com.