Boolean logic is overrated.
Whether in the field of philosophy, science or politics, statements about True and False, good or bad, black and white, friend or foe, are posed as eternal truths, but lacking the sense of time and place. Einstein introduced the space-time and relativity by changing the perspective on the universe. I would like to change the perspective on implementing a control. A standard control is often characterized by a conditional statement like: If some condition is True Then act accordingly. Enumerate a lot of these expressions and the control is implemented. But why is the process asking for that action? Why is it almost impossible to grasp the reasoning behind these statements? By putting the process in the focal point of our attention, the perspective changes and the implementation implodes to understandable, readable and maintainable expressions.
This website introduces the concept of the Process Timing Diagram, PTD for short, to implement all aspects of a control including operator interface. The control is assumed to be performed by a process computer executing all instructions repeatedly, like a PLC or DCS.
The word "control" in this text refers mostly to logic control in contrast to analogue control, but also implies the safeguarding logic and operator interface. Normally these four disciplines are separated for historical reasons, but I strongly recommend integrating them by using the PTD philosophy. The separation of these disciplines has led to a standardization of control based upon the equipment energized by the process computer. The operator is given a popup window for every PID controller, valve, pump or motor. These windows are often referred to as "faceplates", offered by the DCS manufacturer in an attempt to shape its standard approach. This standardization however is ignoring the process to be controlled. To integrate these normal standard components, a lot of instructions have to be added to do justice to the process. It is like using ten times more cement than bricks when building a house. It doesn't look elegant and you will have regrets. The PTD is elegant.
All problems with commissioning, especially its not predictable duration as a result of poor quality, are a product of this standardization with the wrong focus. The PTD method could be described as "back to the basics" and is designed to create smooth and short commissioning. The structure of the instructions forces the quality to be high.
By redesigning the standard implementation structure, making other building blocks and integrating analogue and logic control plus the safeguarding with focus on what is intended to control: the process, the implementation effort and number of instructions are reduced ten times whilst increasing the reliability and quality with at least a factor five.
Of course, the operator interface has to change as well, but he will benefit also. Offering him the exact process state and showing him what "the control is thinking", most human errors will be problems of the past. Large systems tend to offer too much unstructured information to the operator. This induces that human error. Therefore, we are seeking for methods to present the process as clear as possible by using an uniform structure. If an operator is familiar with 5% of an installation, he should be able operating the whole system.
The basic idea of the PTD is the State Transition Diagram (STD for short), but now extended with special use of timing and strict implementation rules. The definition of a state using a PTD is well defined towards implementation, but with the process representation in mind, the definition depends on its elapsed time.
The latter may seem strange, but it is logical due to the fact that the measurements of the physical process state are assumed faulty. So, any combinatorial logic with these inputs should be avoided. Without filtering the inputs against the current state, the errors would propagate like an avalanche.
If the strict implementation rules are obeyed, you can enjoy these benifits:
- The response time of changes in the field towards the equipment energized is guaranteed to be a single cycle time of the processor. Faster is only possible by using different hardware.
- The control is made "proven correct". In all cases, you can predict the result of the control. This is not only true for the implementer, but also for the commissioning engineer and the operator.
- The reliability of the previous statement will hold for decades.
Normally after some years the process will show a different response due to aging, like corrosion. Combinatorial solutions will suffer from this and showing different responses than during commissioning. The PTD remains reliable and predictable.
Personally I made a control that lasted for 17 years while the software remained unmodified. Another example even lasted for 32 years, but the hardware was twice renewed. The PTD is proven design.
- During commissioning alterations can be made within the context of the PTD states without influencing other states.
This means that after a modification only the altered states needs testing.
- The integration of the safeguarding filters all superfluous alarms and trips.
In case of emergency the operator is properly informed and can make his decision in a split second without studying elaborate alarm lists.
The "Cause and Effect Diagram" can be generated automatically and can be compared with a design document with the same name, but lacking the proper process context.
- The PTD diagram resides in the process computer, but can easily and automatically be converted into the operator interface image. This is the main reason for high quality and elegant control.
I am using the word "philosophy" in the title for a good reason. If the concepts are understood you will love knowing the what, how and why of the PTD method.
One of the fundamentals is concentrating memory to the PTD itself and avoiding all other memories.
A memory is a source of problems because it is hard to know how valid her representation is. If the memory is recently set, it will probably be valid. But after time elapses the uncertainty will increase.
Therefor I am thanking George Stephanopulos for his chapter about analogue control in a discrete environment.1981: Chemical Process Control: An Introduction to Theory and Practice; Page 634.
He describes an algorithm for a PID controller using the "velocity form". No memory is used for the integral part of the algorithm which is based upon the difference between the actual and desired process values. So, the problem called "integral windup" does not exists!
The lack of memory is turned into an advantage. Only short-term memory is used: the previous difference is needed to calculate the proportional part of the PID output. Using a value of the previous cycle is the kind of memory which represents a reliable value because it is very recent history. The derivative part of the algorithm requires the difference of two cycles ago. It is still a very recent value, so you could successfully argue that these values are not the long-term memories you should avoid.
Like the detection of a flank of a Boolean variable. You need to know the value of the previous cycle, but not a value of an hour ago or worse. After a longer period of time you can no longer define a memory reliable. Always keep in mind that you'll have to deal with a dynamic process. The majority of commissioning problems are caused by combinatorial logic based on ancient history and present. It makes no sense or is very difficult to grasp.
The concept "velocity form" is an integral part of the PTD philosophy. The current state represents what the control is "thinking" about the state of the physical process. It is not important how and when this state became active. You must always ignore the past. The only thing that matters is the decision to take the next step, preferably in the right direction. Each journey starts with a single step. And while traveling the next step is the only step that matters.
Creating logic for a "Condition to Advance" (CTA for short or "transition" in STD terms) is always defining this next step with the actual state as the given context in mind. Using the actual state as implicit context implodes the combinatorial logic. The best way of expressing that is by saying that it is "not polluted by situations out of context" and therefor readable.
The simulation language I use to test the control realistically is based on the same principle. Many simulation languages are collecting equations for process states around the diagonal of a matrix and are solving it as a set of linear equations.
The PTD approach is different. The next cycle in the simulation is only 100 milliseconds later (using my HCADwin), so we are just calculating the derivative of a process state variable and use that as an increment. The next cycle will correct the error we made. As long as the bookkeeping of mass and energy is based on the preservation laws, all process states will tend to find the equilibrium after some cycles regardsless of the initial conditions.
The physical law for the preservation of impulse is replaced by a new law: preservation of numerical stability. The resulting system may deviate in exact timing behavior, but is pleasantly robust and shows a high performance. Very well suited for testing the control within the timing restrains of the process computer. And by adding some delays the simulation becomes surprisingly realistic.
Remember that a model of reality is always wrong. Reality is always right. But the model can serve its purpose.
PTD method in a nutshell.
The PTD method is structuring the automation of processes with the intent of maximising realiability and predictability.
Therefor commissioning effort is minimised and the operator can enjoy to be informed complete and adequate with maximum operability during the full life cycle of the installation.
The PTD implementation method is based upon the two-stage rocket of the perfect process representation. The second fundament is associating equipment with the represented states while ignoring all relations with measurements. Remember that measurements are faulty. The process representation takes form in a simple graphic: the diagram. Now the essence of the control is condensed to a single page or shown in a single window. This diagram is the ideal start for functional description.
The logic control for the process fits on a single page.
Maintaining a perfect administration (=process representation), is only possible using the correct implementation technique. For that the control processor and its language must be known. Regardless what DCS or PLC, even a failsafe PLC is used, it's basic instruction set allows a proper implementation.