PIDAnalytix.COM
Plant Wide Optimization Made Easier
The most profound findings of Multi-Loop AI-PID tuning is that individually every loop in the 512 cases of AI-PID Tuning demonstrated herein produces a linearized response in accordance with eqn 2 .
This kind of performance across multiple control loops makes it easier and more effective to do optimization across these loops as constituting plant wide operations.
This can be done more directly and efficiently than what is being done with the various MPC controllers such as DMC, RMPCT etc.
Most MPC currently implementation require that all of the PID loops related its manipulated variables, namely (SP) are tuned responsively and tightly. In the case of 2 unit control depicted in the diagram below, there are in total 19 PID Loops whose SP will be manipulated by the MPC.
Loop Tuning is the weakest point of failures for maintaining MPC applications
Ask any MPC engineers about how critical is PID Loop Tuning and as yet how well it is done in practice. Most of these expert MPC engineers will throw their hands in the air and want to run away and hid. Most MPC engineers know from the bitter first hand experience that PID-Tuning is the weakest link in their carefully designed MPC. The basic fact is that the MPC models all have embedded in them the effects of certain PID Tuning at the time of plant testing. Normally it takes a good amount time in between the plant testing and commissioning of the MPC (2-3 months), during this time, it is most likely that the embedded PID loops would have been re-tuned for one reason or other, re-tuning of PID loops is an on going efforts. Therefore, at the first instance of commissioning the MPC, the reality of plant situation will jeopardize their success. There is no way the old pid tuning be restored to help. Even if this situation did not happen, the MPC with its own action to optimize plant operation would cause the PID loops to respond differently to SP change. This is the hard fact of the so called "Advanced Process Control" and yet from its inception in 1990 to this date, they have no workable solution to "re-tune" PIDs. The reason for this is simple, PID-Tuning is considered to be outside the scope of MPC. Consequently, the bitter experience of this state of affairs is that most of MPC soon stop working, awaiting new testing and re-doing.
AI-PID MPC for Plant Wide Optimization - A better way to do APC
A new framework of Advanced Process Control can be built using AI-PID Tuned regulatory control of a plant with multiple PID loops. This new framework is called "AI-PID MPC". At the outset, all of the PID loops will be tuned as per AI-PID Tuning. This can be done very quickly in a matter of a few days. Followed by this plant testing is done to collect data of other variables for optimization such as product Impurity in overheads and bottom products etc. The plant test data is then used to build MPC. As part of this APC project, the plant will have in-line version of AI-PID tuning to maintain and adapt their tuning. These new tuning will not affect other variables model as the new tuning response would be the same as at the time of plant testing was done. This will solved the inherent problem of model mismatch in the MPC methods as done by others.
Any model mismatch as it relates to other variables will be corrected as normally done in a MPC.
This will make APC implementation in record time and most critically without the presence of weakest point of failures as in the other MPCS.
Reinventing Advanced Process Control (APC)
In the last 30 years, APC has evolved for control and optimization of process plants involving multiple PID control loops. In particular, this was spur headed by Dynamic Matrix Control in around 1990 and now this is generally termed as being Model Predictive Control (MPC) for plant wide optimal control. There are a few of MPC system available for a number of DSC vendors such as Honeywell, Emerson, ABB, Yokogowa and others.
All of these APCs are built on the same shaky and unreliable PID controllers. Average life expectancy of APC is less than 3 months, and most of them operate on life-support system.
Both re-tuning of the PID loops and re-modelling of the models used in the APC require timely update as the operating region of the plant is changed by the MPC. Both of these two requirements of system maintenance is simply lacking.
The root cause of the loss of controllability of MPCs is that the PID-loops are not easy to re-tuned just in time.
New Paradigm In APC Design...
AI-PID Tuning can provide a robust and adaptive platform for multi-loop wide optimal operation. AI-PID can on its own can perform self-tuning based on its own internal performance loop, monitoring loop performance and performing re-tuning just in time. In doing so, it will greatly eliminate the root cause of MPC mal-performance. Thus, AI-PID tuning can aid in supporting and extending life cycle of any MPC system.
More importantly and profoundly, with AI-PID tuning, a new design of APC can be devised that will yield a more robust and non-linear APC system implementation.