TECHNICAL FIELD

The present application relates to a method performed by a control unit for controlling a process of drying free-flowing particulate material in a continuous flow dryer. Further aspects of the application relate to a control unit for performing the controlling of drying processes and a continuous flow dryer comprising such a control unit.

BACKGROUND OF THE INVENTION

Controlling continuous drying processes wherein one wants to obtain certain moisture levels for the dried material is difficult. Most drying processes for seeds and grain within the agricultural area control the output intensity of the material in order to obtain the desired moisture level for the dried commodity. If the input material is moister than usual, the output intensity is lowered and if the input material is drier the output intensity is raised. Other parameters that influence the end result are the temperature of drying air, air flow, ambient temperature and atmospheric humidity.

Today's systems for continuous drying of free flowing particulate material face the problem of dryers containing a large amount of material that require a relatively long time to reach the desired moisture level. If the moisture level of the ingoing material remains the same throughout the drying process, it is easy to obtain a desired value for the moisture level in the material exiting the dryer by using feedback from a sensor located somewhere in the dryer. The signal from the sensor reflects in some ways the obtained moisture level in the dried commodity. The sensor may be a temperature sensor, sensor for atmospheric humidity, or other types of sensors that directly or indirectly reflect the moisture level in the commodity inside the dryer.

When the commodity passes through the dryer it will be exposed to a drying effect proportional to the throughput speed (i.e. residence time) in the dryer. Materials with different ingoing moisture level will require different residence times to reach the desired moisture level.

Present techniques use preferably one sensor, located either at the entrance to the dryer, somewhere inside, or at the exit of the dryer. The sensor will sense the moisture level of the material and the control unit will adjust the residence time of the free-flowing material in the drier according to the obtained measurement. A sensor located at the entrance of the dryer can control the output capacity to suit the material being loaded into the dryer, but if the moisture level of the ingoing material shifts during the drying process, the moisture level of the outgoing material will vary as the residence time is optimal only for one certain ingoing moisture level. If the ingoing moisture level suddenly is increased compared to the moisture level of the previous loaded material, the output intensity will decrease and the material already inside the dryer will become overdried as the residence time for this material will be longer than required for its ingoing moisture level.

A sensor located somewhere inside the dryer has the advantage that it can correct the outgoing rate for the already dried material that reaches the sensor. If the sensor is located before the exit, a change in output intensity may correct the moisture level towards the desired level as there is still some residence time during which the material may be dried. The drawback with this solution is that the controlled output intensity influences the material in the dryer before the sensor and the moisture level of the material reaching this sensor may be a surprise that puts the control process into oscillation if variations in the ingoing moisture level appear. Some systems are designed not to release material that is too moist which means that when the sensor detects a moisture level that is too high, the output intensity will be strongly reduced. This means that the material at the top of the dryer will be overdried, and when this material at a later stage reaches the sensor, the sensor will send a signal to increase the output intensity considerably. An increase in output intensity will again affect the material located above the sensor, which now will become too moist and when it reaches the sensor again the output intensity will be decreased. The controlling system enters an oscillating pattern that is difficult to break. This may lead to overdrying of the commodity by several percent. A sensor located at the exit of the dryer can only direct the output intensity towards a plausible level, as the material when it reaches the sensor already obtained its output moisture level. It will only provide information whether the material exiting the dryer should spent more or less time inside the drier. It cannot detect the water-content of the material located above the sensor.

Overdrying wheat by 1 %, from 20% to 13% instead of from 20% to 14% will increase the energy consumption 19.6% per kg ingoing material, and lower the drying capacity by 16%. Furthermore, the dried commodity will lose 1 .15 % in weight which in turn will decrease proceeds when the material is sold. Ovredrying wheat by 2%, from 20% to 12% instead of from 20% to 14% will increase the energy consumption by 40% per kg ingoing material, and lower the drying capacity by 28.5%. Furthermore, the dried commodity will lose 2.3 % in weight which in turn will decrease proceeds when the material is sold.

An additional dimension regarding the complexity of trying to control the drying process is that the residence time inside the dryer to obtain the desired water-content not only depends on ingoing water-content. Other factors which will influence the process are type of material to be dried, ingoing temperature of the material to be dried, maturity of grains, the presence of undesirable particles such as dust or dirt, outside temperature, outdoor air moisture, drying temperature and air flow. These parameters often vary with time and will influence the optimal residence time for the material spent in the dryer.

The article "Mathematical Simulation of Corn Drying - A New Model by Thompson, T., Peart, R., & Foster, G. (Saint Joseph, Michigan: Transactions of the American Society of Agricultural Engineers, 1968) discloses a simulated drying process based on a

mathematical model for drying thin layers of corn under ideal conditions as found in a laboratory. This model is clearly not optimal for estimating the drying process in a grain dryer containing several tons of grain. However, to the surprise of the inventors it was found that despite the fact that different cereals have different properties the course of drying is similar for different crops, and the original mathematical formula could after some modification be adapted to account for many of the factors affecting the drying process in a continuous flow dryer.

The paper Automatic Control of Crossflow Grain Dryers: Design of a Model-Predictive Controller by Qiang Liu; F. W. Bakker-Arkema (J. agric Engng Res. (2001 ) 80 (2), 173- 181 ) describes a concept of grain drying processes using model predictive control (MPC) including the measurement of ingoing and outgoing moisture. The difference between the measured moisture output and the moisture predicted by the model is used to

continuously modify the mathematical model. The authors of the paper have tested their method in a drying simulator where it has been proven to be accurate and sufficient. Also a field testing on a commercial cross-flow grain dryer has been performed.

However, although the MPC-approach seems to be effective and accurate in theory, it is not that efficient in reality. Even if the model is continuously modified by comparing the measured values with the values predicted by the model, the model predictive control may still be insufficient. It turns out that the mathematical model alone is not able to cover the sometimes chaotic conditions in a grain dryer. For the model predictive control to work as expected, at least the following most important preconditions must be met. For instance:

1 . The moisture content over a single layer of material inside the dryer must be

roughly the same.

2. The dryer is expected to perform fairly equally in all drying sections, i.e. the drying process must be steady and smooth, in particular over the width of the dryer.

3. The material flow through the dryer must not be disturbed.

4. The material inside the dryer must be fairly homogeneous, i.e. in case the material to be dried is a mixture of different materials, or if the material is very dirty.

If one or more of these preconditions are not fulfilled the dryer doesn't perform in a way that is predicable by the mathematical model. Even if the mathematical model would include all variables that theoretically can affect the drying result, it is very difficult to adjust one or more of these variables by measuring the difference between predicted moisture and real moisture, because the difference in moisture doesn't contain any information about why the model's predictions are wrong.

This drawback could be overcome by placing a lot of different sensors all over the dryer. The sensors could provide detailed information about the prevailing conditions in the dryer. This way, it would be easier to adjust the mathematical model according to the information provided. In theory, that might work, but it is not very practical for both technical and economic reasons. The model as described herein has been complemented by a regulation process that will take care of errors not covered by the control process and the mathematical model.

SUMMARY OF THE INVENTION

A first aspect of the present invention relates to a method performed by a control unit for controlling a process of drying free-flowing particulate material in a continuous dryer. In continuous flow dryers the free-flowing particulate material is normally continuously flowing through the dryer without stopping. However, the free-flowing particulate material may also move through the dryer in a semi-continuous flow wherein the loading and unloading of material takes place intermittently thereby forming layers of free-flowing particulate material inside the dryer.

Continuous flow dryers are often categorized according to the direction in which air flows in relation to the free-flowing particulate material. Suitable continuous flow driers according to the invention are e.g. Cross-Flow Dryers wherein the air flow is generally perpendicular to the free-flowing particulate material flow. The free-flowing particulate material flows by gravity down a column as the heated air blows across the column. In Concurrent-Flow Dryers both the free-flowing particulate material and air are moving in the same direction. In Counter-Flow Dryers the free-flowing particulate material and air flow are in opposite directions and in Mixed-Flow Dryers the air flows in both counter and concurrent directions. However the invention is not limited to the listed continuous flow dryers.

The method of the invention is particularly useful for drying any kind of free-flowing particulate material such as grains and seeds (e.g. wheat, maize, barley, rye, oat, rape seed, rybs, peas, sunflower, soya beans, paddy rice, cocoa beans and coffee), but also pellets, wood chips etc. or any other free-flowing particulate material in need of drying.

A continuous flow dryer comprises an inlet into which the free-flowing particulate material enters the drier, and an outlet through which the dried free-flowing particulate material exits the drier. The inlet is normally located at the top of the dryer and the outlet is normally located somewhere below the upper inlet. A first humidity sensor is located at, or in near proximity of the inlet, and according to the method of the invention the control unit receives at least a first moisture value (M _in ) from the first humidity sensor, which is indicative of the moisture content of at least one layer of free-flowing particulate material entering into the dryer via the inlet.

The free-flowing material moves from the inlet towards the outlet in a continuous or intermittent flow. A second humidity sensor is located at or in near proximity of the outlet from the dryer, and according to the method of the invention the control unit receives at least a second humidity value {M _out ) from the second humidity sensor. Said humidity value {M _out ) is indicative of the moisture content of the at least one layer of free-flowing particulate material when exiting out of the dryer via the outlet.

The humidity sensors as used in the present invention may be any kind of sensor that directly or indirectly will detect the moisture content of the ingoing free-flowing particulate material. Examples of suitable sensors are capacitive sensors or temperature sensors. The person skilled in the art is knowledgeable of suitable sensors that can be used in the present invention.

The control unit will control the speed of the free flowing particulate material exiting out of the dryer via the outlet based on the received at least first and second humidity values (M _in , M _out ) and an internal set point for an average outgoing moisture content ( ) regulated around an ideal average moisture level ( ) set by a user of the control unit (100)..

The inventive concept of the drying process of the present invention is based on measured values for ingoing and outgoing humidity values (M _in , M _out ) for a free-flowing particulate material. Said measured input and output humidity values {M _in , M _out ) are used in a mathematical model, i.e. a transfer function which is able to estimate the conditions inside the dryer. As mentioned above the free-flowing material normally moves from the inlet towards the outlet in a continuous flow. In continuous flow dryers the free-flowing particulate material is continuously flowing through the dryer without stopping, i.e. free- flowing particulate material is continuously being loaded into the dryer, while dried free- flowing particulate material is simultaneously unloaded from the dryer at the same rate.

However, the free-flowing particulate material may also move through the dryer in a semi- continuous flow such that a certain volume of material is loaded into the dryer, while simultaneously a certain volume is unloaded through the outlet. In this way a number of layers of free-flowing particulate material are formed inside the dryer, wherein each layer has an ingoing moisture content (M _in ) when entering the dryer as determined by the first humidity sensor, and an outgoing moisture content (M _out ) when exiting the dryer as determined by the second humidity sensor.

Also, for the sake of describing the inventive concept of the present invention, it is proposed that the free-flowing particulate material is moving through the dryer from the inlet to the outlet as an infinite number of "layers". At any time (t) during the process an estimated moisture level M(t) can be calculated for said particular layer of free-flowing particulate material by means of the transfer function. The same transfer function may be used to estimate the moisture level Λ Z for said layer of free-flowing particulate material at the end of the drying process, provided that the output intensity remains constant. This way it is also possible to estimate an optimal residence time (t _r ) for the free-flowing particulate material inside the drier. The measurement of the outgoing moisture content {M _out ) is only performed as a verification of the estimated moisture level M.

Before the start of the drying process the user will set an ideal average moisture level ( ) for the dried free-flowing particulate material exiting the dryer at the end of the drying process. The ideal average moisture level ( ) may vary according to the type of free- flowing particulate material being dried, i.e. whether it is wheat or maize or some other type of products or material. Preferably the user also determines maximum moisture content {M _max ) that is acceptable for the obtained free-flowing particulate material. A suitable value for M _max may be M + 1 .

The ideal average moisture level ( ) is entered into the control unit, and the regulation by the control unit is performed based on the difference between the average of the measured outgoing moisture contents (M _out ) of a fixed number of sections of free-flowing particulate material when exiting the dryer and the ideal average moisture level ( ) as set by a user of the control unit.

There may be several reasons why the average of outgoing moisture content (M _out ) of a fixed number of sections of free-flowing particulate material when exiting the dryer and the ideal average moisture level ( ) as set by a user of the control unit do not coincide. All mathematical models are based on idealized conditions, representing some kind of simplification of the real world. Even if the model is continuously modified by comparing the measured values with the values predicted by the model, the model may still be insufficient. Deficiencies in the humidity sensors may be one reason, but also large variations in moisture content (M _in ) of the free-flowing particulate material going into the dryer may influence the average outgoing moisture content (M _out ) of the material leaving the dryer (i.e. there are large differences in moisture content between different layers of free-flowing particulate material being loaded into the continuous dryer during the run of a drying process).

However, the largest factor contributing to a difference is deficiencies in the mathematical model, i.e. the transfer function may provide an estimated outgoing moisture content M that differs significantly from the measured moisture content (M _out ) under certain circumstances. This calculation error may be temporary, caused by unpredictable fluctuations in the dryer. Long term errors though can be caused by influencing factors that change over time, e.g. variations in ambient humidity or temperature.

It is therefore of uttermost importance to adapt the mathematical model continuously to the prevailing conditions, covering both small temporary fluctuations and larger long-term errors.

The control algorithm is designed to provide optimum flow speed aiming for an internal set point for the average outgoing moisture content ( ). However, even though the mathematical model is adjusted continuously, there are still cases when the average for the measured outgoing moisture contents (M _out ) differs from the desired value. It is therefore necessary that the average outgoing moisture content (M _out ) for all sections of free-flowing particulate material exiting the dryer is continuously regulated around the ideal average moisture level ( ) as set by the user. This regulation is realized by a simple P-controller, manipulating the internal set point for the average outgoing moisture content ( ) used by the control algorithm. Although a P-controller has drawbacks in comparison to a PI- or PID-controller in terms of a remaining steady-state error and settling time, it has been proven to be efficient enough to meet the requirements. The great advantage of a simple P-controller is the relative immunity for changes in the dynamics of the system to be controlled, considering the very complex dynamics of a continuous flow dryer. It is emphasized that this invention is not limited to a P-controller though. Any kind of effective regulation process would be suitable. It should be noted that a P-controller is commonly also referred to as a proportional controller, and a PID-controller is commonly referred to as a proportional-integral-derivative (PID) controller.

The residence time (t _r ) for each layer of free-flowing particulate material in the drier is estimated by means of the transfer function such that the estimated average moisture level for all layers is as close to the internal set point for the average outgoing moisture content ( ) as possible. None of the layers may have an estimated moisture level M above M _max .

The P-controller calculates the present error err and adjusts the internal set point for the average outgoing moisture content ( ) such that the ideal average moisture level ( ) set by the user is achieved for the dried free-flowing particulate material exiting the outlet:

M = M + K _p * err wherein

M is the internal set point for the average outgoing moisture content provided by the P- controller ; and

M is the ideal average moisture level (%) as set by the user; and

err = M— M _out , and

M _out is the average moisture content (%) of the last n sections that have been discharged from the dryer (n is a fixed number) as measured by the second humidity sensor;

K _p represents the gain factor of the P-controller. Thus, the regulation of the internal set point for the average outgoing moisture content ( ) may be based on the difference (err) between the ideal average moisture level ( ) set by a user of the control unit (100) and a measured average moisture output (M _out ). Furthermore, the regulation of the internal set point for the average outgoing moisture content ( ) may be performed using a proportional controller.

A mathematical model is used to calculate the optimal residence time (t _r ) of each layer of free flowing particulate material in the dryer. Said calculation of residence time (t _r ) is based on the received first humidity value (M _in ) resulting in an estimated moisture value ( ) representing the moisture of each layer in the dryer at any time. At discharge the received second moisture value (M _out ) is compared to the estimated value ( ).

Depending on the difference M _err = M— M _out ) the transfer function of the mathematical model is modified. The transfer function indicating the residence time (t _r ) is determined by:

t _r = x - [A - ln( ff) + B ^■ [ln( ff)] ² ]

wherein

M - Me

M _in = first humidity value (%) of the at least one layer of free-flowing particulate material; and

M = estimated moisture level of the at least one layer of free-flowing particulate material; and

Me = equilibrium moisture, the lowest possible value obtainable for the moisture level of the at least one layer of free-flowing particulate material as determined by:

wherein

RH: reference value for air moisture,

ϋ : drying temperature in °C

such that

MR = e wherein

A = -1.70584 + 0.008784 ^■ ϋ;

B = 427 A■ _e -°-°594 - 0-1.056

x: variable

The parameters x and RH are determined numerically and adjusted continuously, depending on the error ( _err ). Thanks to the regulator, there is no need for the mathematical model to cover all conditions in the dryer. Even if one or more of the preconditions is not fulfilled, the regulator is able to compensate for that error. At the same time, the grain dryer can be equipped with only two moisture sensors which are cost- effective and easy to calibrate.

The verification of the transfer function is performed by comparing the measured outgoing moisture content (M _out ) with the moisture level ( ) as estimated by the transfer function. The difference (M _err = M— M _out ) represents the current error in the mathematical model. As a quick response to the error, the control algorithm adjusts the estimated moisture values M for each layer in the dryer by applying the calculated error proportionally:

Madj = M _in - (M _in - M) * err_adj for each layer in the dryer, where

Min-M

for the latest discharged layer.

As a direct result even the estimated optimal residence time of each layer in the dryer will be adjusted. This simple linear adjustment is working well for smaller deficiencies in the mathematical model. But even larger but temporary errors can be caught this way. Larger long-term errors though cannot be treated efficiently with this approach.

If the error ( _err ) is too large the transfer function will have to be adjusted such that by finding new values for x and RH the difference between estimated moisture level ( ) and measured outgoing moisture (M _out ) is minimized. Thus, the controlling is performed such that the difference (M _err ) between an estimated moisture value (M) and the at least a second humidity value (M _ou t) is minimized.

A second aspect of the present invention relates to a control unit for controlling a process for drying free-flowing particulate material in a continuous dryer. Said dryer comprises an inlet and an outlet and the control unit comprises a processing circuitry configured to receive at least a first humidity value (M _in ) from a first humidity sensor indicative of the moisture content of at least one layer of free-flowing particulate material entered into the dryer via the inlet. The control unit will receive at least a second humidity value {M _out ) from a second humidity sensor indicative of the moisture content of the at least one layer of free-flowing particulate material when exiting out of the dryer via the lower outlet. The control unit will control the speed of the free-flowing particulate material exiting out of the dryer via the outlet based on the received at least first and second humidity values (M _iri , M _out ) and an internal set point for an average outgoing moisture content ( ) regulated around an ideal average moisture level ( ) set by a user of the control unit (100)..

An ideal average moisture level ( ) which is optimal for the specific type of free-flowing particulate material to be dried is decided by the user and entered into the control unit. The ideal average moisture level ( ) may vary according to the type of free-flowing particulate material being dried. Preferably the user also determines maximum moisture content {M _max ) that is acceptable for the obtained free-flowing particulate material. A suitable value for M _max may be M + 1 . The processing circuitry is configured to control the speed of the free-flowing particulate material exiting out of the dryer such that the difference {M _err ) between an estimated moisture value (M) and the at least a second humidity value {M _out ) is minimized.. The processing circuitry is configured to calculate the optimal speed of the free-flowing particulate material exiting out of the dryer by using an algorithm that calculates the optimal residence time (t _r ) of each of the at least one layer of free-flowing particulate material in the dryer using a mathematical model. On top of that a simple P-regulator is used to keep the average of all measured outgoing moisture contents (M _out ) for the fixed number of sections as close as possible to the ideal average moisture value ( ) set by the user.

The mathematical model is used to estimate the residence time (t _r ) of each of the at least one layer of free-flowing particulate material in the dryer indicated by the received at least one first humidity value {M _iri ).

One or more parameters in the mathematical model is updated based on the received at least one second humidity value {M _out ).

The mathematical model used to indicate the residence time (t _r ) of a layer of free-flowing particulate material in the drier is determined by

t _r = x - [A - ln( ff) + B ^■ [ln( ff)] ² ] wherein

M - Me

MR =

M _in - Me

M _in = first humidity value (%) of the at least one layer of free-flowing particulate material; and

M = estimated moisture level of the at least one layer of free-flowing particulate material; and

Me = equilibrium moisture, the lowest possible value obtainable for the moisture level of the at least one layer of free-flowing particulate material as determined by: ln(l - RH)

Me =

0.0000382 ^■ (1.8 ^■ ϋ + 82.0) wherein

RH: reference value for air moisture

ϋ : drying temperature in °C

such that

wherein

A = -1.70584 + 0.008784 ^■ ϋ;

B = 427 A■ _e -°-°594 - 0-1.056

x: variable

The parameters x and RH are determined numerically and adjusted continuously, depending on the error ( _err ).

The processing circuitry (101 ) of the control unit (100) may be configured to regulate the internal set point for the average outgoing moisture content ( ) based on the difference (err) between the ideal average moisture level ( ) set by a user of the control unit (100) and a measured average moisture output (M _out ). Furthermore, the processing circuitry (101 ) may be configured to regulate the internal set point for the average outgoing moisture content ( ) using a proportional controller. A third aspect of the invention relates to a continuous flow dryer for drying free-flowing particulate material, said dryer comprising an inlet, an outlet, and a first humidity sensor at the inlet. Said first humidity sensor is configured to provide at least a first humidity value (M _in ) indicative of the moisture content of at least one layer of free-flowing particulate material entered into the dryer. Said second humidity sensor at the outlet is configured to provide at least a second humidity value {M _out ) indicative of the moisture content of the at least one layer of free-flowing particulate material when exiting out of the dryer. The continuous flow dryer further comprises a control unit comprising a processing circuitry configured to receive the at least a first humidity value (M _in ) from the first humidity sensor, receive the at least a second humidity value {M _out ) from the second humidity sensor, and control the speed of the free-flowing particulate material exiting out of the dryer via the outlet based on the received at least first and second humidity values (M _in , M _out ) and an internal set point for an average outgoing moisture content ( ) regulated around an ideal average moisture level ( ) set by a user of the control unit (100).

BRI EF DESCRIPTION OF THE DRAWINGS

Figure 1 illustrates a continuous dryer in which embodiments of a method and control unit for drying free-flowing particulate material as presented herein is implemented. Figure 2 illustrates a schematic process diagram for controlling a process of drying free- flowing particulate material in a continuous dryer according to embodiments of the method and control unit described herein.

DETAI LED DESCRI PTION OF EMBODIMENTS

Referring to the figures, wherein like reference numerals designate identical or corresponding parts throughout several views. Figurel illustrates for the purpose of describing the method of the invention one type of continuous flow dryer 10 in its simplest form suitable for drying free-flowing particulate material such as grains or seeds. Said continuous flow drier comprises a main body 1 1 for receiving and circulating free-flowing particulate material 20 to be dried, a chamber 12 for entering hot air 30 into the main body 1 1 and a leading exhausted air chamber 13. At least one inlet 14 is located at the upper portion of the main body 1 1 , and at the inlet, or in close proximity to the inlet a first humidity sensor 15 is located. The free-flowing particulate material 20 containing moisture enters through the inlet 14, moves downwardly through the main body 1 1 . Air from the outside is entered into the chamber 12 and is heated by a plurality of burners that may be operated by any suitable heat source and located in several locations. The heated air is blown by a blower or a fan into the main body 1 1 containing the moist free-flowing particulate material, blows through the mass of free-flowing particulate material 20, exits the main body 1 l and enters the leading exhausted air chamber 13. The free-flowing particulate material 20 exits through an outlet 16 at the bottom of the main body 1 1 . A second humidity sensor 17 is located at, or in near proximity of the outlet 16. One or more partitions may be provided inside the chamber 12 for directing the hot air 30 as desired. Also, depending on the type of flow dryer, the main body 1 1 may contain partitions for directing free-flowing particulate material 20 through the main body 1 1 . It is emphasized that the dryer illustrated in Figurel demonstrates the principle of continuous flow drying and should not be limiting to the invention.

During continuous flow operation, free-flowing particulate material continuously enters through the inlet at the upper part of the main body, and continuously exits through the outlet, for example by using augers or the like. At the same time the blower or fan blows hot air heated by burners through the mass of free-flowing particulate material moving downwardly from the inlet to the outlet. For semi-continuous flow processes, known volumes of free-flowing particulate material 20 are loaded into the dryer 10 at intervals, and known volumes of dried free-flowing particulate material 20 are unloaded from the dryer 10. A certain time as set by the user passes between each loading/unloading such that a number of layers or sections of free-flowing particulate material 20 form throughout the dryer 10 between the inlet 14 and the outlet 16.

At the inlet 14, the first humidity sensor 15 measures the moisture content of the free- flowing particulate material entering the main body 1 1 . According to the method of the invention the first humidity sensor 15 sends at least a first humidity value (M _in ) indicative of the moisture content of the free-flowing particulate material 20 entering the continuous flow dryer, to a control unit 100 connected to the continuous flow dryer 10. The free- flowing particulate material 20 moves through the main body 1 1 towards the outlet in a continuous or semi-continuous flow, and at the outlet 16, the second humidity sensor 17 measures the moisture content of the free-flowing particulate material exiting the continuous flow dryer 10. According to the method of the invention the second humidity sensor 17 sends at least a second humidity value {M _out ) indicative of the moisture content of the dried free-flowing particulate material exiting the continuous flow dryer 10, to the control unit 100 connected to the continuous flow dryer 10.

The control unit 100 controls the process of drying free-flowing particulate material in a continuous dryer 10. Said control unit 100 comprises a processing circuitry 101 configured to receive the at least first humidity value (M _in ) from the first humidity sensor 15. The first humidity value (M _in ) is indicative of the moisture content of at least one layer of free- flowing particulate material 20 entering into the dryer 10 via the upper inlet 14. The free- flowing particulate material 20 moves from the inlet 14 to the outlet 16 in a continuous or semi-continuous flow, and when it reaches the outlet 16, the processing circuitry 101 will receive at least a second humidity value {M _out ) from a second humidity sensor 17 indicative of the moisture content of the at least one layer of free-flowing particulate material when exiting out of the dryer 10 via the lower outlet 16.

The free-flowing particulate material entering the dryer 10 has to be dried to a certain ideal average moisture level ( ) which is optimal for each type of free-flowing particulate material 20 and is set by the user. Preferably, the user also decides a maximal moisture level {M _max ) that is acceptable. No layer of free-flowing particulate material leaving the dryer may have a moisture level superseding the maximal moisture level {M _max ) as set by the user.

The processing circuitry 101 of the control unit 100 is configured to control the speed of the free-flowing particulate material exiting out of the dryer 10, i.e. the residence time (t _r ) each layer spends inside the dryer, such that the difference between the average of the measured outgoing moisture content for a fixed number of layers (M _out ) exiting out of the dryer 10 and an ideal average moisture level ( ) as set by a user of the control unit 100 is minimized.

The mathematical model used to estimate the residence time (t _r ) of each of the at least one layer of free-flowing particulate material in the dryer 10 is determined by:

t _r = x - [A - ln( ff) + B ^■ [ln( ff)] ² ] , wherein

M _in is the first humidity value (%) indicating the moisture content of the at least one layer of free-flowing particulate material entering the inlet (14) of the dryer (10); M is an estimated moisture level (%) of the at least one layer of free-flowing particulate material at the time when exiting the dryer (10); and

Me, equilibrium moisture, is a lowest possible value obtainable for the moisture level of the at least one layer of free-flowing particulate material as determined by:

-ln(l - RH)

Me =

0.0000382 ^■ (1.8 ^■ ϋ + 82.0) wherein

RH is a reference value for air moisture;

ϋ is a drying temperature in °C;

such that

A+ —r^+A ²

MR = e wherein

A = -1.70584 + 0.008784 ^■ ϋ;

β = 427.4 . _e -o.o594 - tf-i.o56. _and

x is a variable scaling factor for time.

Both RH and x are unknown. RH is a value which is related to the atmospheric humidity and affects the lowest moisture level (Me) attainable for the dried final product and x is a scaling factor for time.

Due to deficiencies in the mathematical model and the influence of environmental parameters, the transfer function has to be continuously calibrated. The accuracy of the mathematical formula is therefore constantly re-evaluated by comparing the second humidity value {M _out ) as measured by the second humidity sensor 17, which reflects the actual dried product exiting the dryer 10, with the estimated moisture level (M) as estimated by the transfer function resulting in the current error M _err = M— M _out ).

An error adjustment factor (err_adj) may thus be applied to adjust the estimated moisture levels (M) in the dryer 10 including the estimated optimal residence time (t _r ) for each layer, e.g.

Madj = M _in - (M _in - M) * err_adj

for each layer in the dryer 10 , where

for the latest discharged layer.

This linear adjustment is suitable for small or temporary errors.

An acceptable difference (M _err ) could be around 1 .0 % moisture content. Below that value, linear adjustment alone is effective enough. If the difference is larger, the mathematical model has to be calibrated.

The mathematical model is calibrated using the measured humidity values (M _in , M _0Ut ) obtained from the first and second humidity sensors 15, 17, and the time (f) for the layer of the free-flowing particulate material spent in the dryer 10. The drying temperature (ϋ) is assumed to remain constant throughout the process. In other words, f represents the time it takes for the free-flowing particulate material to be dried from moisture level M _in to moisture level M _out at drying temperature ϋ. The tuple of measured moisture contents M _in and M _out , drying time f, and the drying temperature ϋ is saved in a data base.

The transfer function of the mathematical model is determined based on the last n saved tuples ( _in , M _outi t, ϋ)ι for i = 1. . n, by finding values RH and x such that

is minimized, wherein f(x, RH i = M _{i =} MR _t * (M _ini - Me) + Me;

, - ln(l-RH)

Me = Me _t = '

0.0000382-(1.8 ^■ i9; + 82.0) J

MRi = e for which A = A _t = -1.70584 + 0.008784 ^■ i9 _;

B = B _t = 427.4■ e-0.059 - tfi-i.056

As the drying temperature ϋ is assumed to be constant it is true that ϋ = d _t and therefore A = A _{i t} B = Bi and Me = Me _t for all i = 1. . n. It should be noted that n may be chosen to be between 5 to 120 measurements, preferably between 10 to 60 measurements and most preferably 30 measurements.

By means of the mathematical formula a moisture level (M) may be estimated for every layer at any time throughout the drying process as well as at the time when the same layer exits the dryer 10. From the same mathematical formula it is possible to estimate an optimal residence time (t _r ) for each layer of free-flowing particulate material 20 in the dryer 10, such that the obtained product is neither over-dried nor under-dried.

The output intensity (i.e. the residence time (t _r ) for the free-flowing particulate material in the dryer 10) is estimated by the mathematical formula. The ideal average moisture level ( ) as set by the user and type of free-flowing particulate material to be dried, is a value which the control unit is trying to achieve. Due to deficiencies in the mathematical model, fluctuations in the moisture content of the ingoing material, and unpredictable disturbing factors during the drying process, the average of the outgoing moisture content (M _out ) for a fixed number of sections as measured by the second humidity sensor 17 may differ from the estimated average moisture content ( ) predicted by the transfer function of the mathematical model.

To keep the average for all measured outgoing moisture contents (M _out ), as measured by the second humidity sensor, as close as possible to the ideal average moisture level ( ) as set by the user, a regulation process is necessary. This is achieved by introducing an internal set point for the estimated average outgoing moisture content ( ) for the control algorithm. The estimated average outgoing moisture content M is continuously regulated around the ideal average moisture level ( ). The processing circuitry 101 is configured to control the speed of the free-flowing particulate material exiting out of the dryer 10 by using a P-controller as determined by:

M = M + K _p * err , wherein

M is the internal set point for the estimated average outgoing moisture content provided by the P-controller to the control unit; and

M is the ideal average moisture level (%) as set by the user; err = M— M^;

M _out is the average moisture content (%) of the last n sections that have been discharged from the dryer 10 (n is a fixed number) as measured by the second humidity sensor 17; and

K _p is the gain factor. Figure 2 shows a schematic process diagram for controlling a process of drying free- flowing particulate material in a continuous dryer 10 according to embodiments of the method and control unit 100 described herein. The process depicted in Figure 2 may be implemented in and performed by the processing circuitry 101 of the control unit 100 as described above. It should be noted that the dashed marked area 200 indicates the regulation performed by the P-controller and the dashed marked area 210 indicates the control process.

EXAMPLE 1

In the following example, the advantageous embodiments of the method of are illustrated, wherein about 150 m ³ of maize was dried during a total process time of 7 hours and 17 min in a continuous mixed flow dryer.

The ideal average moisture level ( ) for the particular type of maize to be dried was set to 14.0% and the maximum moisture level M _max (i.e. the highest acceptable moisture content for any of the dried layers of maize leaving the dryer 10) was set to 16.0%. The drying temperature ϋ was set at 1 10°C and remained constant throughout the run.

A continuous mixed flow dryer having a volume of 47.6 m ³ was loaded with moist maize. This particular dryer was loaded with maize such that 17 separate sections or layers of maize formed inside the dryer 10, each section having a volume of about 2.8 m ³ . The moisture content (M _in ) of the loaded maize was measured by the first humidity sensor to be 27.7%, and the dryer 10 was unloaded from the outlet in volumes of about 0.25 m ³ each time. At the start of a run the dryer 10 is filled completely with moist material and consequently the first 17 sections of free-flowing particulate material were not collected as they have not spent enough time inside the drier to reach the ideal average moisture level as set by the user. These sections of maize are discarded and preferably reloaded into the dryer 10.

In this particular run five humidity values (M _in ) as measured by the first humidity sensor were averaged for each section and sent to the control unit 100. By means of the mathematical formula the process circuitry of the control unit estimated a residence time (t _r ) for each section in order to obtain an ideal average moisture level ( ) of 14% as set by the user. Table 1 illustrates the estimated moisture contents for each section at the given times, as well as estimated moisture level for the same section when it exits the dryer 10, provided the output rate remains the same.

Table 1

The values x and RH were calculated to be 1 .5 and 70% respectively. The actual moisture content M _out for section 17 as measured by the second humidity sensor at the exit of this section from the dryer 10 was 13.6%. When the moisture level Λ Z for section 17 (see table 1 ) as estimated by the mathematical formula (i.e. 14.1 %) is compared to the actual moisture content (M _out ) of 13.6 % it is seen that the maize has been slightly overdried, and therefore the transfer function has to be adjusted. RH och x are therefore adjusted by the processing circuitry such that the difference between M and M _out is minimized. When the next section (i.e. section 16) exits the dryer 10, the estimated moisture level M is compared to the measured moisture content (M _out ) for section 16 and if the difference is too large the parameters x and RH are adjusted once more to minimize the difference.

The ideal average moisture level ( ) is set by the user to be 14.0%. The average of all outgoing moisture contents (M _out ) for all sections as measured by the humidity sensor was 12.9%, i.e. a difference (err) of 1 .1 %. This value is entered into the P-controller which provides an internal set point ( ) such that ( - M _out ) is minimized. The internal set point ( ) provided by the P-controller was set to 15.0%, which will result in an adjustment in the residence time (t _r ) for each of the sections such that they will spend a slightly shorter time in the dryer 10 and consequently will not be overdried. The average of all estimated moisture levels (M) for all seventeen layers from Table 1 is 14.8 % which is very close to the adjusted internal set point ( ).

EXAMPLE 2

The following example illustrates the flexibility and robustness of the self-adjusting method as disclosed herein, wherein a large amount of wheat was dried over a period of several days in a continuous mixed flow dryer.

Here, the dryer 10 was mechanically constructed in a way that the number of drying- and cooling sections was variable. Initially, the dryer 10 was set up to 13 drying sections and 4 cooling sections, and the drying control system delivered a satisfactory homogenous moisture output very close to the set point ( ) set by the user to be 13.3%. After about one day of drying the user decided to change the ratio between drying- and cooling sections in the dryer 10 to 14 - 3 (i.e. to 14 drying sections and 3 cooling sections), which would increase the drying capacity. However, the user modified the dryer 10

mechanically, but he forgot to tell the drying control system, i.e. the control unit 100, about the change. Thus, the drying capacity increased significantly and rapidly. As a result the outgoing grain became over-dried.

As the drying control system realized the growing discrepancy between predicted outgoing moisture ( ) and measured outgoing moisture M _out ) the transfer function of the mathematical model was successfully calibrated and the outgoing moisture stabilized relatively quickly. After several hours the user suddenly remembered his mistake and changed the ratio of drying- and cooling sections also in the drying control system. Now, that the drying control system had already adapted to the changed behavior of the dryer 10, the same situation occurred again, but in reverse. The grain became under-dried for a short period of time. But again, the drying control system adapted relatively quickly and compensated for this error.

This example is obviously a worst case scenario that proved the ability of the control system described herein to quickly adapt to sudden and influential changes in the drying conditions. Similar scenarios would be a change in the drying temperature or a change in the material to be dried without informing the drying control system. Both scenarios could be observed in field studies and in both cases the drying control system could adapt to the changed conditions.