INDUSTRIAL REFINING PROCESS VERSUS THEORY
Georges Joris - MATECH-EUROPE

1. INTRODUCTION

It is very common to hear that the industrial refining process is very mysterious, full of inconsistencies and contradictions. However, most inconsistencies actually stem from a lack of rigor and attachment to the task. This has unfortunately become a growing problem for laboratories, R&D departments and mills in the paper industry. Further, pulp is a non-Newtonian fluid that is still characterized and quantified by ancestral parameters that usually do not fit with the involved process.

1- Pulp sampling is an operation that usually brings errors and misunderstandings.
This operation calls for cautious handling steps that take into consideration the many parameters that could contradict the expected characterization of the relevant pulp sample.

2- The measurement of the no-load power required to calculate the levels of specific energy for each pulp sample is far from being an easy operation to cope with.
Consequently, the effective power could be very different from the expected value.

3- The determination of the slowness (°SR) or the freeness (CSF) is not usually achieved in keeping with the different steps described by the standards. Freeness and slowness have no direct physical meaning and consequently cannot be used to develop models to predict the refining development of a pulp versus the specific energy. Many papers have pointed out some criticism regarding these classical measurements. The specific filtration resistance, SFR [4], should be used but it is really very difficult to fight against papermakers' conservatism.

4- The pulp dynamic viscosity and its gradient play a fundamental role in the refining process, but it is completely ignored at the industrial level. And yet in the case of pulp flocculation or film mat breaking, the efficiency of the refiners can sharply drop.

This paper strives to give an accurate description of some very simple actions from which paper characteristics, energy consumption and, on another level, longer plate lifetime and maintenance could be optimized. Inconsistencies that are increasingly found in the available literature will also be pointed out.

It is easy to demonstrate that from lack of rigor, one can easily develop wrong theories that bring even greater confusion for papermakers. Some false ideas have become wide spread, making it very difficult to rid the papermaking world of them.

The refining process, like any physical process, is not that difficult to understand once the inconsistencies and wrong ideas are swept away. If the mathematical key points of the refining process are complex and beyond the frame of this paper, the ins and outs are at least within understanding. Make it simple, reject sophisticated plate patterns, be very rigorous and proceed step by step. Then, take into consideration that any pulp grade is a non-Newtonian fluid, of which physical properties are gradually altered from the refining process. The pulp from the outlet of the first refiner is no longer the pulp from the supplier. These are very simple facts that are important to bear in mind.

2. PILOT REFINING UNIT
(how easy it is to elaborate wrong theories)

The pilot refining unit mostly consists of a small single disc refiner (9 to 12") or of a small conical refiner working in hydra cycle mode (batch process). In addition, a pump and a tank is present to work as a pulper. In the bottom of the tank there is an impeller, whose velocity should be controlled depending on its function (pulping or mixing) and on the pulp viscosity.

While pulping, the impeller produces shearing actions and must then be slowed down to create an optimal vortex during the refining process. Sometimes, the pilot refining units work in duo-cycle mode with 2 tanks. The right choice depends on the compromise between the homogeneity, in terms of freeness within the refining unit, and the dead bottom volumes in the tanks.

It is very important to get a true representation of industrial refiners in series from a pilot refining unit. Consequently for each batch the pulp must run through the refiner once and once only. This prerequisite is quite impossible to meet but can be approximated. This means that the velocity of the properly shaped propeller should be controlled in relation with the pulp viscosity throughout the refining process to get a light pulp vortex to optimize the freeness distribution from the top to the bottom of the tank. To meet this challenging requirement, pulp recirculation must also be implemented and the recirculation flow must also be controlled. To reach this target, a sophisticated attached program control is necessary, from which the speed of the pump and the opening of each valve must be regulated all together through electronic drivers.

The effective volume of the tank should hold a minimum of 300 litres. For efficient control throughout the trial, many sensors have been set according to a strict layout. Fine gap and axial force sensors are also implemented in the refiner. This is at least the fundamental point that must be considered to carry out reliable trials. Most pilot refining units are, at present, far from meeting these requirements.

The picture aside shows a classical pilot refining unit (supplied in South Africa).
The effective capacity of the tank is 450 litres, and the refiner is a 12" single disc with a specific shaped housing to obtain a more accurate no load power measurement. The inlet pressure is kept constant throughout the trials under a constant flow ranging from 6 to 15 m³/h. The pulp consistency is set up from 30 to 50 g/l. The length of the pipes must be as short as possible.

Furthermore, a sampling device is highly recommended to get pulp samples that can be linked afterwards to the calculated specific energy. A volume of 2 to 4 litres should be taken within a very short period of time (2 to 4 seconds) at moments well determined and calculated by the computer. All these requirements could seem too sophisticated and consequently not very necessary. However even while operating with such sophisticated pilot refining units, reliable results cannot be obtained as yet. Further factors must be considered.

In addition to the implementation of sensors, valves, drivers and their relevant controls, it is quite impossible to take accurate pulp samples alongside the specific energy simply because the homogeneity of the pilot refining unit is not perfect. Even with an optimal recirculation flow, and an optimal vortex of pulp in the tank, the freshly refined pulp cannot be properly mixed with the pulp inside the tank to get a freeness that could increase linearly from the top to the bottom of the tank.

Even under these strict conditions, the error in determining the relevant specific energy of the pulp samples can easily reach 10% and much more with classical pilot refining units. In this paper we are going to demonstrate that an error of only 5% could already lead to light inconsistencies. An error of 10% could lead to contradictions. Thus, even with a sophisticated pilot refining unit, reliable results cannot be verified. Furthermore, the state of freeness pulp distribution within the refining pilot unit depends on the level of freeness itself.

At first, the virgin pulp cannot be properly mixed with the "one pass refined pulp" according to a constant gradient of freeness within the tank. Then, the dynamic viscosity of the mix drops very quickly, which eases the mixing according to a more homogeneous distribution. The grade of homogeneity is then in close relation with the freeness development. And consequently it is not only a problem of inaccuracy to cope with, but also a drifting problem which is much more serious.

The number of pulp samples taken during one trial should be high enough (For example1 pulp sample every 4 minutes for a 20 minute trial). This is a compromise between the accuracy and the working time inverted in achieving the measurements. However, to truly solve this problem, physics and mathematical models to predict pulp characteristics versus specific energy must be elaborated.

These models, called first order models, are only valid for a well determined pulp grade refined in a well determined refiner under well determined conditions of plate patterns, flow, pressure, temperature and so on. For example, the weighted fibre length development can be accurately characterized by the equation:
(1)
In which E stands for the net refining specific energy, Lfo stands for the weighted fibre length of the virgin pulp and where s and p are the parameters. These parameters must be determined for each trial. From 6 pulp samples taken throughout a refining trial we have a system of 6 linear equations in the 2 unknowns s and p. Through appropriate mathematical methods in the least-square sense, these parameters can be determined. From this relation, we can then anticipate the development of the weighted fibre length for a well determined pulp grade, refined in a given refiner with determined plate patterns and working under determined hydraulic conditions and applied power. The relative error should be less than 2% if the trial is carried out properly and if the measurements of the fibre lengths are achieved carefully.

In the same way, the determination of the freeness versus the specific energy must be
anticipated through a model. A physical parameter that is the specific filtration resistance is required to start the development. From the Kozeny Carman law:
(2)
where σ (m²/kg) stands for the specific surface of the fibres, c (kg/m³) for the hand sheet consistency, α (m³/kg) for the specific volume of the fibres and Kk for the Kozeny factor it has been established from the Ingmanson hypothesis [4]:
(3)
From the relations (1) and (3), there is a link between the specific filtration resistance and the weighted fibre length under any level of specific energy (Ej) for a given pulp grade, refined in a given refiner under well determined conditions (plate patterns, flow, applied power etc). The specific filtration resistance will then be turned into freeness or slowness to respect papermakers who cannot do without them.

Refining is a process involving mechanical behaviour of matter (fluid and solid) that can be fully described. The effects of any change in their settings can be accurately assessed by the continuum physics models (1) and (3) with the help of mathematical methods.
These models are required for a better understanding of the refining process to which idealization of the link between SFR and weighted fibre length leads to maximize their coherence versus the net specific energy E.

Through mathematical iterative procedures, one can bring the required corrections on the net specific energy E under the best coherence conditions. In doing so, every drift or fluctuation of the relevant anticipated levels of net specific energy E for each pulp sample are corrected accordingly.

Nevertheless, the conversion of the specific filtration resistance SFR into °SR or CSF to meet the papermakers' requirements is not perfect, but is quite acceptable. Now, the pilot refiner must be controlled by the effective power, which means that the no-load power must be accurately determined throughout each trial.

Here, we have to face a very complex problem because one cannot measure the no-load power simply with a wattmeter connected on the motor when the rotor is discarded from the stator. The no-load power is not necessarily given by a wattmeter under these conditions (the gap between the rotor and stator is high enough).

The measurement of the no-load power must not be confused with the true value of the no-load power. This is a really well established matter of confusion in the industry. For example, when a double disc refiner is unloaded no one can prove that the floating rotor is set on its central position in the housing of the refiner. In the case of single disc refiners, the clearance between the rotor and the motor side housing must be taken into consideration. In any refiner, the rate of pulp turbulences at the entrance of the plate refining area depends on the refiner gap and bar arrangement. When the gap decreases, the rate of turbulence reduces and consequently the no-load power also decreases. Under a small gap, the refiner starts consuming effective power while the no-load power continues to decrease. Consequently, it is quite impossible to accurately measure the no-load power. To partly solve this problem, the housing of a pilot refiner should be shaped to get a laminar flow at the refiner entrance. In doing so, the power measured under no-load conditions is slightly lower but is more sensitive to the rate of flocculation of the pulp. The measurement must be repeated several times, with properly deflocculated pulp that is not refined. In case of partial plugging of the plate grooves , the measurement of the no-load power is lower, but this is not the true no-load power. Actually, total or partial plugging actions are more frequent with virgin pulp and in hydra cycle mode the plugging effect could disappear after a short period of time or, on the contrary, after being reinforced.

During the measurement of the no-load power, the pumping action of the refiner must be under control or diagnosed to discard all measurements when plugging effects appear. The no-load power also depends on the flow, thus making it imperative that it is constantly maintained throughout the trial. The no-load power further depends on the dynamic (3) pulp viscosity which means that the no-load power lowers throughout the trial. This means that we will encounter an error and a drifting problem which must be totally controlled to avoid inconsistencies in the results. This also means that the applied power must decrease in function of the pulp temperature, to maintain a constant effective power.

Actually, the applied power should decrease versus the dynamic pulp viscosity that cannot be measured on line. Nevertheless, the temperature and pulp grade are the main parameters to be taken into consideration to estimate the development of the dynamic pulp viscosity. The figure below shows this development in relation with the temperature and pulp grade in hydra cycle mode. The pulp is prepared with clear water at 20°c. Under other conditions the development of the dynamic pulp viscosity can be widely altered.



Finally while loading the refiner the frequency of power measurements must be high enough (an oscilloscope is required) to accurately determine the no-load power as described by the figure below.


After only 2 minutes of refining (14 to 21°SR or 730 to 580 CSF) the measurement of the no-load power went from 12.6 to 12.3 KW, which may seem very small. This difference is already big enough to create some false conclusions. Actually, this is the selected result from several measurements of the virgin pulp and after 2 minutes of refining with the same pulp grade.

The incertitude is +/- 0.5 KW on the virgin pulp. The incertitude gap is lower with hardwood or with refined softwood. The no-load power is usually measured at the loading time that is the same as the virgin pulp.

For a well determined pulp composition, refined through a given refiner with determined plate patterns under the same conditions of flow and pressure, the no-load power only depends on the dynamic pulp viscosity. In the case of industrial refiners in series, the no-load power is increasingly lower from the first to the last refiner under the same state of plate wearing provided that the above conditions are confirmed. In case of a pilot refiner working in hydra cycle mode, the no-load power decreases in relation to the refining time.
When the effective power (and not the applied power) is perfectly controlled and maintained constant throughout each trial, the net specific energy can be accurately calculated versus the refining time.

Now the question is: under all these prerequisites, could we rely on the results given from each trial? Yes, as long as we do not compare trials to each other. To compare refining trials under different effective powers or with different plate patterns, the formulation of physics or mathematical models is again required. These models can only be built up from correct parameters that have a true physical meaning. It has been proven [6][7] [8] that the classical concept of specific edge load SEL must be proscribed. The true physical concepts such as the reference specific edge load SEL₀ and g* must be carried.

From the fundamental expression:
(4)
we can establish:
(5)
with:
(6)
 
in which the factors λ , δ and γ >1 are the parameters to be determined from the refining trials for a well determined pulp composition with λ (0) = 0.

If the Canadian Standard Freeness is preferred to the Shopper-Riegler, the conversion formula will be then applied:

where u and v stand for the specific parameters of the pulp grade and the white water.
To reach an acceptable accuracy, at least three trials should be carried out under different effective powers that must be properly selected to cover a wide range of reference specific edge loads. From the relation (5) the fastest freeness development (CSF or °SR) can be determined for a well determined pulp composition being refined through a well determined refiner with well determined plate patterns.

Actually, from the relation (5) under a well determined level of net specific energy Ej we can write:

but:

that gives :

and consequently, the reference specific edge load to be applied to each industrial refiner in series to get minimal energy consumption is given by the relation:
(7)
in which Ej is the net cumulated specific energy (KWH/T)

From the relation (6), it follows that the function d (Ej) is continuously increasing which means that the reference specific edge load must imperatively decrease with the freeness development. In hydra cycle mode, the refiner must be progressively unloaded versus the refining time. With industrial refiners in series the reference specific edge load must imperatively decrease from the first to the last refiner according to the law 1/ d (Ej). Unless the function d (Ej) is well know, any attempt to decrease the reference specific edge load from the first to the last refiner could bring some disappointments. Furthermore, this law only works if the refiners are proven to work perfectly and are properly sized versus the flow, which is unfortunately not very common at industrial level.

It is easy to understand that during the refining process the fibres must first be delaminated before any fibrillation and hydration treatment. The required shearing actions to delaminate the fibres greatly depend on the wood species and the morphological parameters of the fibres. Once delaminated, the fibres become more fragile and consequently should be treated increasingly gentler, to avoid a harsh cutting effect and extra energy consumption. This way of operating will not only save energy but will also develop better strength characteristics.

Nowadays, environmental protection and energy saving are new challenges to deal with. The above described method is quite simple to put into practice, provided that the study is carried out with great care and rigor such as described in this chapter. With a good pilot refiner, the function ( ) j d E can be accurately calculated and transposed to any industrial refining unit, after the refiners have been properly diagnosed and the characteristics of the white water have been taken into consideration. In so doing, the financial aspects are also very important.

Lower energy consumption (50% and sometimes much more), less expensive chemical additives, better stability, higher runnability and less maintenance are usually the results from this technique, called "bijective diagram technique".

Let us consider the case of bleach kraft softwood from southern European pines, prepared in clear water at 20°C and refined under optimal hydraulic conditions, through correct disc refiners with pseudo-sectorial plate configurations characterized by a grinding code 3-3-4 and g* =24^o.



At the beginning of the refining process (the first refiner in the line), the pulp must be refined under 1.63 Ws/m to reach an optimal development of the delamination process without damaging the fibres. When the freeness reaches 25°S R (about 500 CSF) the fibres are more fragile and the reference specific edge load must be reduced to 1.4 Ws/m (second refiner in the line for example). Under 40 °SR or 310 CSF the fibres are still slightly more fragile, and as a result the reference specific edge load must be decreased again to now reach 1.3 Ws/m (third refiner in the line, for example). The delamination process should be completely achieved (possibly in the second refiner in the line), so that the effect of the pulp dynamic viscosity should predominate from which gap reduction arises under the same effective power. To maintain the same gap or, better still, to optimize the gap versus the fiber fragility to prevent a harsh cutting effect while maintaining a fast freeness development, the reference specific edge load must be controlled in keeping with the function 1/ d (Ej).

It is highly recommended to avoid using hit or miss techniques to approach the above function , unless the refining unit has been properly diagnosed and clear key points about the pulp composition have been investigated. The above characteristic greatly depends on the pulp temperature, pulp consistency and the state of white water [9].

3. INCONSISTENCIES AND ABERRATIONS

The purpose of this chapter is to prove how easy it is to obtain inconsistencies or to produce contradictions, simply from lack of rigor. Consider a pilot single disc refiner 12" that is supposed to work perfectly with correct plate patterns under optimal hydraulic conditions. Assume that the laboratory staff are wonderfully efficient and reliable, and work under strict and accurate procedures. As a consequence, the freeness, weighted fibre length and hand sheet paper characteristics from each pulp sample are correct. Finally, assume that the sensors deliver values without any relative error. Under these dreamlike conditions, who could suspect that something wrong could happen in the line? And yet, as mentioned above, the measurement of the no-load power cannot be perfect.

Consider 35 g/l softwood pulp grade consistency being refined through the refiner. With pseudo-sectorial configuration plates (grinding code 3-3-4 °SR 9 and g* = 24°), the no-load power from this 12" refiner (1500 RPM) is 13.2 KW. This value has been determined accurately under the conditions depicted in the above chapter. With a perfect oscilloscope, without considering the problems of turbulence, flocculation and refiner gap control, the 20 no-load power measurements (repeated from 20 refining trials with the same pulp grade, T°c, flow and consistency ) range from 13.5 to 14.5 KW as depicted in the figure below.


At industrial level with a 22" double-disc refiner, the no-load power measured on one refiner under the same above mentioned conditions through an oscilloscopic method integrated in the DIATRONIC (control and auto-diagnostic system produced by MATECH-EUROPE) for one week is given by the diagram below.

We can notice that accurate measurements on a refining pilot unit or on an industrial refiner show the same dispersion, but not entirely for the same reasons. Actually, the flow at the inlet of a suitable pilot refiner is laminar, so the power measurement is very sensitive to the rate of flocculation of the pulp. At industrial level, the dynamic pulp viscosity is not as stable from the variation of the white water characteristics. Much higher fluctuations can even arise when the floating rotor does not stick to its central position.

Coming back to the pilot refining unit, the four refining trials under 32 KW, 24 KW, 20 KW and finally 18 KW (applied or total power) are carried out. The relevant effective powers are given in the table here below.

Applied power (KW) Effective power min (KW) Effective power max (KW) 32 17,5 18,5 26 11,5 12,5 20 5,5 6,5 18 3,5 4,5

Also take into consideration the no-load power drifting versus the dynamic
pulp viscosity, which means that the above mentioned effective powers are decreasing unless under control that is never implemented with classical pilot refining units.

The following table depicts what would happen in terms of effective powers and their relevant relative errors if, for example, 12 kg of dry pulp is refined for 20 minutes and a pulp sample is taken every 4 minutes.

T refining (min) 0 4 8 12 16 20 True no-load Po 13,2 13 12,8 12,5 12,2 12 True Peff(1st trial) 18,8 18,8 18,8 18,8 18,8 18,8 Peff max under 32 KW 18,5 1,6% 18,7 0 ,5% 18,9 0,5% 19,2 2,1% 19,5 3,7% 19,7 4,8% Peff min under 32 KW 17,5 6,9% 17,7 6,1% 17,9 4,8% 18,2 3,2% 18,5 1,6% 18,7 0,5% True Peff (2nd trial) 12,8 12,8 12,8 12,8 12,8 12,8 Peff max under 26 KW 12 ,5 2,3% 12,7 0,8% 12,9 0,8% 13,2 3,1% 13,5 5,5% 13,7 7% Peff min under 26 KW 11,5 10 ,2% 11,7 8,6% 11,9 7% 12,2 4,7% 12,5 2,3% 12,7 0,8% True Peff (3rd trial) 6,8 6,8 6,8 6,8 6,8 6,8 Peff max under 20 KW 6,5 4,4% 6,7 1,5% 6,9 1,5% 7,2 5,9% 7,5 10,3% 7,7 13,2% Peff min under 20 KW 5,5 19,1% 5,7 16,2% 5,9 13,2% 6,2 8,8% 6,5 4,4% 6,7 1,5% True Peff (4th trial) 4,8 4,8 4,8 4,8 4,8 4,8 Peff max under 18 KW 4,5 6 ,3% 4,7 2,1% 4,9 2,1% 5,2 8,3% 5,5 14,6% 5,7 18,8% Peff min under 18 KW 3,5 27% 3,7 23% 3,9 18,8% 4,2 12,5% 4,5 6,3% 4,7 2,1%

Note that the decreasing law of the no-load power is not the same for each trial but for simplification, the values have been averaged. In practice, the refining times for each trial are calculated to get similar distributions of specific energy that gives birth to conditioned matrixes. In so doing, we reach maximum accuracy.

Coming back to the example, this simplification can be dealt with due to the fact that the no-load power is classically always measured at the beginning of each refining trial. At this step, it is possible to have an error of 27% under an applied power of 18 KW. The lower the applied power, the higher the relative error. Some laboratories did carry out similar trials under 16 KW to study low SEL refining process, from which a relative error of 47% is quite realistic. Other laboratories are working with smaller single disc refiners ( 9 ") that can easily deliver erroneous results (more than 50% of relative error only from the no-load power considerations).

The determination of the relevant specific edge loads is also affected by the error from the no -load powers. The table below gives the reference specific edge loads for each trial.

1st trial 2nd trial 3rd trial 4th trial True SELo (Ws/m) 2 .25 1.53 0.81 0.58 Calculated SELo (Ws/m) 2.09 to 2.21 1.38 to 1.50 0.66 to 0.78 0.42 to 0.54

The lack of homogeneity of any pilot refining unit even under sophisticated controls must be also taken into consideration. This is a second source of error usually ignored by most paper laboratories. How can we control the gradient of freeness in the tank? How can we manage to get the freeness to drop linearly from the top to the bottom of the tank at any moment during the trial?

To approach this situation, the velocity of the impeller and the recirculation flow should be controlled versus the pulp dynamic viscosity. In practice, a perfect homogeneity cannot be obtained. Without the above mentioned controls, the pulp sample could be very far from its expected representation. Under optimal control, a drifting error of +/-5% is already a reliable situation. Furthermore, the virgin pulp will not easily mix with the "one pass refined pulp" and the mixing process cannot start before the "one pass refined pulp" comes back to the tank. Actually, this phenomenon cannot be observed because the pulp is already running through the refining unit during the loading operation.

To get a minimal drifting process, the volume of pulp running through the main pipe should be very low compared to the effective volume of the pulp in the tank. Some paper laboratories use very small tanks, which shortens the refining trial time but it also worsens the problem of lack of homogeneity. The freeness development appears to be slower than it should be and the drift can reach 10% (under optimal conditions) during the first run, after which homogeneity is achieved. To solve this problem, implementation of sophisticated controls and corrections through a reliable attached software (first order models) is a prerequisite. The diagram here after depicts a classical situation of the CSF development (blue characteristic) compared to the true development (red characteristic) after correction from the attached software.



Now, we can go on with the determination of relative errors relevant to the calculation of the levels of net specific energy for each pulp sample from our four trials. One must also presume the freeness and the slowness measured by our wonderful laboratory staff as being completely reliable.

E1 E2 E3 E4 E5 1st trial (KWH/T) 104 208 312 416 520 1st trial range [102,118] [186,217] [282,334] [389,452] [491,572] 2nd trial (KWH/T) 71 142 213 284 355 2nd trial range [67,81] [123,136] [188,225] [263,315] [335,399] 3rd trial (KWH/T) 38 76 114 152 190 3rd trial range [32,42] [61,81] [94,127] [138,176] [179,226] 4th trial (KWH/T) 27 54 81 108 135 4th trial range [21,29] [37,56] [63,87] [90,130] [126,168]   CSF1 CSF2 CSF3 CSF4 CSF5 1st trial 518 313 157 97 60 2nd trial 622 428 298 207 143 3rd trial 720 632 552 478 408 4th trial 734 693 634 601 550   °SR1 °SR2 °SR3 °SR4 °SR5 1st trial 24 40 59 68.5 75 2nd trial 18.5 30 41.5 52 61 3rd trial 14.5 18 22 26.5 31.5 4th trial 14 15.5 18 19.5 22

The range has been calculated from the above information taking into consideration the initial drift of 10% under the first level of energy E1. At first glance, the errors do not seem to be high enough to bring about confusion. Let us examine the classical fundamental characteristics °SR(SEL O, E) and CSF(SELO, E) plotted here after from the reliable values and analyzed by the attached software called FIBROLOGIC partly described here above. Let us consider both characteristics.




From the diagrams, we can see that the reference specific edge load must decrease from 2.1 to 1.7 Ws/m within the range [13 ; 40] °SR or [750 ; 320] CSF. Now, if we enter into the attached FIBROLOGIC program, the lowest level of specific energy obtained from inaccurate no-load power measurement and lack of corrections from the heterogeneity of the pilot refining unit, we get the diagrams here below:





Although the freeness and slowness characteristics have been corrected by the relation (5), the first contradiction appears. The refiners must be loaded from 1.2 to 1.4 Ws/m during the refining process, which is completely absurd. However, note that only the errors from the no-load power and the lack of homogeneity in the refining pilot were taken into consideration. In practice, there are plenty of further sources of error. If we consider the characteristics before being corrected through the relation (5), the results are in complete contradiction with the physical reality of the refining process as illustrated by the next figure.

The working points move leftwards (lower SELO) and downwards (lower E). Further, the lower the reference specific edge load, the greater the drift of the working points. As a consequence, the °SR appears to be overestimated (or CSF underestimated). This also means that some asymptotic limit conditions could not be met. For example:

If one takes into consideration the drift of the no-load power, the above wrong shaped characteristic °SR (SEL O) could even turn its maximums into minimums, as shown in the figure below, which is nonsense.

A dreamlike pilot refining unit under the care of a wonderful laboratory staff is not the only one prerequisite to avoid contradictions and inconsistencies. Further important parameters govern the correct achievement of a refining study. Namely, it is imperative that the refining pilot unit be controlled by the effective power throughout each trial and that the results are diagnosed through ad hoc attached software. This will link the required corrections dictated by the sampling operations to the lack of perfect homogeneity of any pilot refining unit. To compare plate patterns, the operation is still more complex because some specific parameters far beyond the frame of this paper are also involved in the process.

At an industrial level, any attempt to compare refining units or to analyze results can also easily lead to misunderstandings and contradictions. Actually, we estimate that 70% of industrial refiners are not working properly and these extra parameters (floating rotor, film mat breaking, partial plugging, bad hydraulic conditions, compacting, soaping, parallelism, cavitations, balancing and so on) must also be considered.

It is easy to understand why most papermakers consider the refining process as a mysterious topic. In general, most theories or optimization techniques are not trusted. This is probably the main reason why most industrial refining units are far from optimal. Stock preparation is really the weak point in most paper mills, and unfortunately paper is produced under high levels of energy consumption with plenty of expensive chemical additives to compensate the poor fibre development from the refining process. It is high time that this philosophy is changed to not only preserve our environmental conditions of life but to also obtain a fruitful financial feedback. As we all know, protection of the environment costs a lot of money. How extraordinary would it be to propose saving the planet and money at the same time, by simply spending some time and thought on the stock preparation.

4. CONCLUSIONS
When a pilot refining unit is accurately controlled and the results are analyzed and corrected through adequate attached software and when an industrial refining unit is also controlled under the same conditions, reliable and consistent results are obtained. From this, it is possible to optimize the stock preparation unit and to anticipate most problems. In so doing, it appears that the pulp composition is constantly under physical developments of its dynamic viscosity and fibre morphology that are modifying the pumping characteristic of the refiner, its gap clearance, its shearing actions on the fibres and the film mat stability.

This means that without adequate control against the state of the pulp composition, a huge potential of characteristic development is simply lost and energy consumption can consequently rises very sharply. From 102 industrial diagnostics (that represents 482 refiners) carried out in papermills all over the world, the loss is on average $12 to $25 US per ton of paper produced. Under some circumstances, with the help of a DIATRONIC, the savings could exceed $50 US per ton of paper produced.

5. ACKNOWLEDGEMENT
Many thanks to Doctor Denis Curtil from the French Paper School (University of Grenoble), for his fruitful collaboration in the conception of the DIATRONIC and the control of the pilot refining units.

6. REFERENCES
[1] Clark J.D. Freeness fallacies and facts Tappi J 53, 1, 1970
[2] Pires E.C.A.M. Computational model for water drainage in Fourdrier Paper Machine
Tappi J 71, 4 1988.
[3] Gless J-M A new drainage analysis system Tappi J 67 , 3 1984
[4] W.L. Ingmanson Filtration resistance of compressible pulps Chemical Engineering
Progress Vol. 49 , 11 1953.
[5] Eléments d'analyse pour l'optimisation du raffinage de la pâte à papier, Thèse de
doctorat présentée par Mayade Thierry en 1995 à l'E.F.P.G.
[6] JC Roux , G. Joris Angular parameters beyond specific edge load , Tappsa journal
July 2005
[7] G. Joris, JC Roux, Specific edge load Japanese journal of paper technology, 10 ,
2003
[8] J-C Roux, G. Joris, G Caucal, Quelques écueils de la charge spécifique d'arêtes dans
le raffinage à basse concentration. ATIP Vol 53 n°1 Janvier- mars 1999
[9] G; Joris, JC Roux Optimizing interactivity of chemicals and processes in the wet end. 4th International Wet End conference PIRA Nice 2004