The control chemistry for most ethylene and propylene polymerisation processes as well as for many other types of polymers follows two general systems: –
The control of free radical polymerisation is achieved by regulating the introduction of a Free Radical Agent (FRA) such as organic peroxide, to a reactor and so polymer chain formation.
The control of catalytic polymerisation is achieved by regulating the introduction of a Chain Transfer Agent (CTA) such as propane or hydrogen, to a reactor and so polymer chain formation.
At first glance it would appear that control would be made from in-reactor measurement but it turns out that ex-reactor measurement is more than adequate provided the chain properties are accurately determined. Polymerisation of the alpha olefins started out as a manually operated batch process. Over the last 70 years’ huge progress has undoubtedly been achieved. There have been numerous developments that include continuous process for a wide range of polymer types but a significant advancement of process automation has been held back. This can change if the industry adapts the control system with a better measurement of chain properties.
Control of a continuous process
Polymerisation of ethylene and propylene is always a random and competitive process between the main initiator/catalyst and subsidiary chain modifiers. Also, any given polymerisation can produce more than one type of chain structure. Despite these complications the polymer manufacturing industry has found it possible to constrain a continuous manufacturing process to produce polymer that has very consistent properties.
This has been done by using a control strategy based on the relationships shown above backed up by a selective process technique. In steady state production, consistency has usually been achieved by harnessing some form of feedback action that can home the production on to the desired grade. This control action has been greatly enhanced by augmenting the wanted chain groups and suppressing the unwanted groups. Under optimal conditions this strategy makes it possible to consistently produce polymer that has physical properties that do not vary more than 2-3% from the standard specification.
Intense and diverse process development has seen the evolution of manyproprietary solutions that produce the wanted and suppress the unwanted chains. The main thrust of development has been in catalysts but separation techniques still prevail in some of the older processes. Each solution is special to the particular plant or polymer. In the class of materials, broadly known as mono-modal polymer, the wanted chains outnumber the unwanted chains by a factor that can be greater than 50:1. These materials behave as if the distribution of chain lengths were strictly Gaussian. The polymerisation processes can therefore be characterised and controlled by a measurement of Average Molecular Weight (AMW) or a measurement that faithfully represents AWM over the entire range of polymer. Other polymers classed broadly as bi-modal have two groups of chain lengths, which can overlap. These polymers can be produced by chain growth or by blending two polymers. In the case of controlling chain growth, the feedback measurement needs to characterise the populations of one or both of the groups of chains. Control of blends is easier if the two constituent parts are characterised before the blend takes place. Blending of a single stream output is often used to produce a homogenous grade. This technique can be of negative value to the downstream user and represents additional costs to the producer. The batch size will be limited by the capacity of the blending silo(s) used and the molecular weight distribution will be broadened in a random fashion, according to the reaction history.
For more than 50 years, Melt Flow Index (MFI) has remained the measurement of choice for process control: MFI has also remained the measurement of choice for Quality Control and Quality Assurance. The early polymer plants were controlled from ½ or 1 hourly MFI tests made by a large laboratory staff. The majority of polymer plants retain the manual MFI test for control purposes but in a changed context. In the late 70’s the emphasis was on expansion of production and diversification of product. From that time to the present computer models that represent AMW, specific to each plant process have been developed for control. AMW is calculated from the reaction conditions and fed back into the process to control the inner loops. Manual MFI spot test and other off-line tests, as often as necessary, are employed to reset the model. For many polymerisations this has worked remarkably well but elsewhere this style of control now needs to be overhauled in order to meet current values in energy/materials usage and product consistency.
In fact, all of the elements of the ‘overhaul’ were explored in the mid-1970s (LDPE) and the mid-1990s (Gas Phase PP). The plants were equipped with early examples of the MFI analyser. These early analysers were well advanced, they were available full time and maintained in calibration with ASTM D1238 standard. The plants were also equipped with gas analysers and aids to control gas composition. Each part of the control system was built up from the existing manually operated sequences. This involved imposing process excursions in order to establish more exact control coefficients and process delays. Every product run was set into closed loop control using the MFI analyser result as the feedback measurement. By far the most arduous task was to make every product transition unconditionally stable and to achieve the minimum of changeover time to the new target grade. With the limitations of the delay of the fastest MFI analyser this involved making and tuning each transition control model.
On the face of it would appear that the modern polymerisation control schemes have been built in very similar ways but it is worth noting some important features that would lift the value of any new improvement. The PP scheme made its advancement in control technique by employing an offline analyser that could sample on demand ex-reactor or finished product and so accurately track AWM in real time. This helped establish reaction dynamics for each part of the process. From the basic reaction model it was possible to refine it with data of gas composition, catalyst characteristics etc., so that the process computer could correct for small changes and avoid problems such as model drive-off. Fine control of the process gave information that led to installation of gas conditioning systems, which are now commonplace.
These control projects gave ground breaking improvements incurred a heavy expenditure of effort and took a long time to reach a satisfactory level of completion. Since the LDPE and PP projects were completed the precision of the MFI analyser has improved by a factor of approximately 2. Also there has been strong progress made in the generic computer modelling facility. As a result of these significant advances, it is now possible to complete the large scale control project ‘overhaul’ in a very productive and cost-effective manner.
The well known LDPE process has been chosen as it gives the clearest example of the workings of free radical, chain transfer and copolymer reactions. This complexity can best be described by first considering the free radical reaction and then adding the other two types of reaction at a later stage.
Fig 1 Schematic for MFI/FRA controlled LDPE Polymerisation
Fig. 2 Schematic for MFI/CTA LDPE polymerisation controller
Fig 1 shows a typical LDPE process in a simplified Free Radical Agent (FRA) form. There are two inner loops that are in cascade control with an outer loop. The primary inner loop concerns the temperature control of the flow rate of the FRA peroxide initiator, commonly known as the catalyst. The other inner loop concerns the maintenance of reactor pressure. In this particular reactor configuration, the pressure is controlled by an extrusion valve at the base of the autoclave. Once set in motion the two inner loops will stay in tight control. Any small changes, such as gas composition, will cause a change in chain length. This can be corrected by the outer loop, which acts as a muti-variable controller on the two parameters of pressure and temperature. An increase in pressure, or a reduction in temperature, will cause an increase in chain length. In LDPE this will be routinely made by a Melt Flow Index (MFI) test on the finished product. Fig 1 also shows an At-Line Melt Indexer providing the prime measured variable in a computer controlled closed loop system. This scheme dates back to c1975.
The basic LDPE reaction is often augmented by the introduction of a Chain Transfer Agent (CTA), such as Propane of Butane, which is commonly termed a Modifier. Fig 2. shows the additional components in the typical LDPE process. It is necessary to obtain a representative sample of the gas entering the reaction chamber. The inner loop is controlled from the modifier gas concentration that is bled from the high pressure feed to the reactor. An increase in modifier concentration will cause a decrease in chain length. The outer loop becomes more complex but it will still be controllable from the same MFI value as in the basic system. In order to maintain continuous control of a certain chain length, the temperature, pressure and CTA set points must be moved together in the correct proportions.
The versatility of an LDPE polymer can be further enhanced by the introduction of a comonomer such as Vinyl Acetate (VA). The polymer known as Ethylene Vinyl Acetate (EVA) is valued for its adhesive properties and film clarity. The monomer in liquid state is normally metered into the compressor chain with a rate-controlled positive displacement pump, operating at high pressure. The outer control loop for EVA incorporates a separate anlyser which needs to measure the combined VA content and reject the free VA. Fortunately such an anlyser, which derives VA content from the change in capacitance, can share the same pellet sample stream as the MFI analyser.
The LDPE reaction produces a wax-like low polymer as well as various gaseous impurities. Without process treament these impurities will heavily impinge on any control system. The low polymer can be extracted immediately prior to the pellet extruder but the rogue gases are brought down to a less intrusive level by ‘purging’ back to the refinery.