Method of optimizing carbon capture of a fluent body

Abstract

There is provided a method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water, the method comprising: mixing a fluent body with an initial volume of water in a vessel by an actuator for engaging the fluent body, to form a mixture; sensing a physical property of the mixture for a progressive increase in value of the sensed physical property; based on sensing the physical property, predicting a parameter value for carbon capture of the fluent body; and controlling delivery of water to the vessel in accordance with the parameter value. There is also provided a method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water to the fluid body by a machine learning model (MLM).

Description

[0001] METHOD OF OPTIMIZING CARBON CAPTURE OF A FLUENT BODY Carbonatation can be used as a method for sequestering carbon contained in waste products from industrial processes. Waste feedstock for the carbonatation process may include materials such as fly ash, air pollution control residue (APCR) and cement kiln dust (CKD) or cement bypass dust (CBD).

[0002] APCR is a waste produced during a flue gas treatment process that removes contaminants (e.g. heavy metals, chlorides, or other pollutants) from the flue gas. The flue gas may be generated by incineration of a feedstock, commonly by incineration of municipal solid waste in waste-to-energy plants. Typically the APCR will include a mixture of dusts of lime and activated carbon that are added to the flue gas as reagents, together with fly ash generated by incineration of the feedstock and entrained in the flue gas.

[0003] The chemical composition of APCR varies widely with the nature of the feedstock and also the parameters of the incineration process. The makeup of such feedstock often depends on its site of origin. Feedstock for concrete is provided in vast quantities, often in the order of hundreds of kilos, and delivered to or stored at sites in large silos. Batches of feedstock are taken from the silo for transforming into aggregates by processing in a mixer. Quick handling of each batch is thus required for efficient processing of the feedstock into aggregate. At the same time, production of aggregates is required to satisfy quality and process efficiency standards.

[0004] It is known to process a mixture of waste alkaline dusts with a larger particulate additive such as sand or powdered limestone together with water in a mixer. The water reacts with the dust in an exothermic hydration reaction, and the remaining water then forms a film around the particles which, in contact with the carbonic gas, facilitates the subsequent carbonatation reactions with the gas which flows through the mixing vessel. The carbonatation reaction is allowed to proceed, e.g. for about half an hour, before the damp mixture is discharged to a pelletiser in which the carbonatated dust particles agglomerate around the larger particulates to form pellets.Typically the larger particulate additive forms about half or more of the solid content of the mixture, which however can add cost since often it will not be produced as a waste stream.

[0005] The wide variation in surface area and chemical composition, hence reactivity of the dust can also be problematic since it strongly affects the exotherm of the hydration reaction and, in consequence, the viscosity and water content of the subsequent mixture. Typically it is difficult or impossible to predict accurately the reactivity of any given batch of dust, even from a single source, and so the process conditions are often set by rule-of-thumb, frequently resulting in a mixture that is too dry (and so not completely reacted) or too wet (and so not suitable for pelletisation).

[0006] Process time of known methods must also be long enough to obtain complete reaction of the mixture, which unnecessarily reduces throughput (and so efficiency) for more reactive batches.

[0007] Financial viability of a batch-based mixing process, for example for the production of lightweight concrete using accelerated carbon technology, impinges on the speed at which material can be processed and in a manner that renders the output suitable for concrete. There is therefore a need for a manner of processing material in a mixer for concrete formation that is time efficient and compliant with stringent industry standards.

[0008] Summary

[0009] According to a first aspect there is provided a method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water, the method comprising:

[0010] mixing a fluent body with an initial volume of water in a vessel by an actuator for engaging the fluent body, to form a mixture;

[0011] sensing a physical property of the mixture for a progressive increase in value of the sensed physical property;

[0012] based on sensing the physical property, predicting a parameter value for carbon capture of the fluent body; andcontrolling delivery of water to the vessel in accordance with the parameter value.

[0013] In an example sensing the physical property is for a period of time taken for a peak value of the physical property to be sensed.

[0014] In an example the parameter value is a further addition of water. Additionally or alternatively the parameter value is a time duration between predicting the parameter value and controlling delivery of water.

[0015] In an example the method comprises determining a parameter profile based on sensing the physical property, said determining comprising:

[0016] identifying a peak value of the physical property in a duration of time; determining a slope of change of the sensed physical property of the duration of time; and

[0017] calculating an area under the slope over the duration of time.

[0018] In an example the method comprises predicting composition of the fluent body based on the sensed physical property.

[0019] In an example predicting the parameter value comprises predicting at least one of, grain attributes, density attribute, chloride content and carbon dioxide, CO2, and reactivity of the fluent body.

[0020] In an example predicting the parameter value comprises predicting a percentage presence of at least one reactive compound of the input material.

[0021] In an example at least one reactive compound comprises at least one of the following reactive compounds within the fluent body: lime, larnite, calcite, sylvite and portlandite.

[0022] In an example sensing the physical property of the mixture comprises sensing an indication of torque resistance of the mixture against the actuator.

[0023] In an example sensing the torque resistance of the mixture comprises sensing a motor torque reaction of the actuator.

[0024] In an example the motor torque reaction is a motor current reading of the actuator. In an example the method comprises determining a viscosity profile based on an indication of motor torque reaction, said determining comprising:

[0025] identifying a peak value of the indication of motor torque reaction; calculating a slope of change of the indication of motor torque reaction; andcalculating an area under the slope.

[0026] In an example the method comprises:

[0027] mixing the mixture at a predetermined rate, and

[0028] wherein sensing the parameter value comprise sensing power required by the actuator for maintaining the predetermined rate of the mixture.

[0029] In an example the sensed physical property comprises a temperature value of the mixture.

[0030] In an example predicting the parameter value is by a machine learning model, MLM. In an example the method comprises:

[0031] forming the mixture in a first hydration phase,

[0032] sensing a physical property of the mixture for a period progressive increase in value of the sensed physical property during said mixing, at time intervals, to derive a set of time-series data;

[0033] providing the time series data to a machine learning model, MLM; and predicting the parameter value by the MLM, based on the time series data.

[0034] In an example the method comprises:

[0035] simulating, by the MLM, a behavior of the sensed physical property in at least one and the at least one further phase, subsequent to the first hydration phase; and predicting a parameter value based on the simulated behaviour.

[0036] According to another aspect there is provided method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water, by a control system of a mixer apparatus, the method comprising:

[0037] mixing a fluent body with an initial volume of water in a vessel by an actuator for engaging the fluent body, to form a mixture;

[0038] sensing a physical property of the mixture for a progressive increase in value of the sensed physical property; and

[0039] based on sensing the physical property, predicting a parameter value for carbon capture of the fluent body.

[0040] In an example the method comprises controlling delivery of water to the vessel in accordance with the parameter value, by the control system.According to another aspect of the invention there is provided a method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water to the fluid body by a machine learning model, MLM, the method comprising:

[0041] a) providing a batch of a historic fluent body, the historic fluent body having known composition;

[0042] b) processing the batch with an initial volume of water in a vessel, by an actuator, to form a mixture;

[0043] c) sensing a physical property of the mixture for a period progressive increase in value of the sensed physical property during said mixing, wherein said sensing comprises measuring the physical property at time intervals, to derive historic time-series data;

[0044] d) deriving a historic physical property profile for the batch based on the derive historic time-series data;

[0045] e) associating the historic physical property profile with the known composition;

[0046] f) repeating steps b) to e) for at least one additional batch of the fluent body;

[0047] and

[0048] g) providing plural historic physical property profiles to the MLM.

[0049] In an example the method comprises analyzing, by crystallography, portion of the batch of fluent body to determine a composition of the fluent body.

[0050] In an example processing the batch comprises:

[0051] mixing, in a first hydration phase, the fluent body with water, to form the mixture;

[0052] processing, in at least one subsequent phase, the mixture; and

[0053] sensing the physical property during the first hydration phase and during the at least one subsequent phase;

[0054] wherein the historic time-series data is data from the first hydration phase and during the at least one subsequent phase.

[0055] In an example the method comprises training an MLM according to any one of the aforementioned the steps.

[0056] According to yet another aspect there is provided a method optimizing carbon capture of a fluent body of solid particles by controlling delivery of water to the fluid body by a machine learning model, MLM, the method comprising:

[0057] training the MLM by:

[0058] a) providing a batch of a historic fluent body, the historic fluent body having known composition;b) processing the batch with an initial volume of water in a vessel, by an actuator, to form a mixture;

[0059] c) sensing a physical property of the mixture for a period progressive increase in value of the sensed physical property during said mixing, wherein said sensing comprises measuring the physical property at time intervals, to derive historic time-series data;

[0060] d) deriving a historic physical property profile for the batch based on the derive historic time-series data;

[0061] e) associating the historic physical property profile with the known composition; and

[0062] f) repeating steps b) to e) for at least one additional batch of the fluent body; mixing a fluent body with an initial volume of water in a vessel by an actuator for engaging the fluent body, to form a mixture;

[0063] sensing a physical property of the mixture for a progressive increase in value of the sensed physical property;

[0064] based on sensing the physical property, predicting a parameter value for carbon capture of the fluent body using the trained MLM; and

[0065] controlling delivery of water to the vessel in accordance with the parameter value.

[0066] Summary of the Drawings

[0067] Fig. 1 shows schematically one example of an apparatus for carbonatation of a fluent body;

[0068] Fig. 2 shows pictorially XRD analysis of three different batches of an input residue;

[0069] Fig. 3 shows graphically behaviour of a motor current of a mixer;

[0070] Fig. 4 shows graphically time series data for a first hydration phase;

[0071] Fig. 5 shows an example subset of training data for the MLM;

[0072] Figs. 6A and 6B shows graphically the same set of data points obtained by X-ray diffraction testing for the proportion of lime, and the corresponding data points inferred by a model;Fig. 7 shows graphically modelled temperature, carbon dioxide capture and mass of the mixture;

[0073] Fig 8 shows schematically an overview of a process for optimising carbon capture; and

[0074] Fig. 9 shows schematically a process for optimising carbon capture of an input material according to an embodiment.

[0075] Detailed Description

[0076] An process of preparing aggregates in a mixer in accordance with the invention for lightweight aggregate from an input residue is discussed below. Input material of varying compositions require different volumes of water for proper hydration as part of carbonatation. The time for which hydrated input material needs to be exposed to carbonic gas for carbonatation also varies based on the makeup of the input material.

[0077] It has been realised that by observing certain physical properties of the mixture as it undergoes the carbon capture process, it is possible to identify each of the phases, and to subsequently calculate how much water to add to the mixture, and when, in order to transform the material into an aggregate.

[0078] According to an aspect of the invention there is provided a method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water. The method comprises: mixing a fluent body with an initial volume of water in a vessel by an actuator for engaging the fluent body, to form a mixture; sensing a physical property of the mixture for a progressive increase in value of the sensed physical property; based on sensing the physical property, predicting a parameter value for carbon capture of the fluent body; and controlling delivery of water to the vessel in accordance with the parameter value.

[0079] Fig. 1 shows schematically one example of an apparatus for carbonatation of a fluent body 101 of solid particles 102 by reaction with a carbonic gas G, wherein at least some of the solid particles 102 are dust particles 102'.The apparatus is operable to process multiple batches 10T sequentially, each in accordance with the process, wherein each batch 10T is formed from a different fluent body 101.

[0080] The control system

[0081] The process may be embodied in a control system 10 of the apparatus, which as illustrated may include a processor 13 and a memory 14. The processor 13 may execute instructions stored in non-volatile memory 14 to control the apparatus in accordance with the process.

[0082] The steps of the process may be performed by the remaining elements of the apparatus which include an actuator 40, a water flow control means 60, a gas flow control means 70, and a sensing means 51.

[0083] The mixer

[0084] The apparatus includes a mixer comprising a vessel 20 for containing the fluent body 101 , and a rotor surface or surfaces 31 arranged within the vessel 20 for rotation about an axis X31. The vessel 20 contains both the mixture 100 and the gas G which is in contact with the surface of the mixture 100 during carbonatation. Together, the vessel 20 and the rotor surface 31 define a mixer.

[0085] The actuator 40 may be a motor (e.g. an electric motor or a hydraulic motor) and is operable by the control system 10 to drive the rotor surface 31 in rotation about the axis X31.

[0086] Preferably the vessel 20 defines an enclosed volume that contains the gas G during the process, as shown.

[0087] The rotor surface 31 may be a surface of a rotor 30 that is movable relative to a static vessel 20, as exemplified by the illustrated embodiment.

[0088] Alternatively (not shown), the rotor surface may be an interior surface of the vessel, wherein the vessel is driven in rotation. The interior surface of the vessel may include fins or blades or other elements that agitate the mixture as the vessel rotates. Alternatively or additionally, a fixed surface or an array of fixed surfaces (e.g.fixed blades) may be arranged within the vessel so that the mixture set in motion by rotation of the interior surface of the vessel will impinge on the fixed surface or surfaces, causing a stirring action which exposes fresh surfaces of the mixture for reaction with the gas. Alternatively or additionally, the vessel may include moving rotor surfaces that move relative to the vessel as the vessel rotates.

[0089] In each case, the rotor surface (e.g. surfaces 31 as illustrated) or the fixed surface or surfaces may be arranged to extend down through the top surface of the fluent body 101 within the vessel, so that a moving cavity is created in the mixture 100 behind the rotor surface 31 or fixed surface as the top surface of the mixture 100 is disrupted by movement of the mixture 100 past the respective surfaces. The cavity exposes fresh surfaces of the mixture 100 for reaction with the gas G.

[0090] As illustrated, the rotor surface(s) or fixed surface(s) may include two or more blades, spaced apart radially with respect to the rotation axis X31 of the rotor, wherein each blade extends down through the surface of the mixture 100 towards the base of the vessel 20, so as to cut through the mixture 100 as the rotor surface 31 rotates about the axis X31.

[0091] Although the rotor may move with a planetary motion if desired, it is found perfectly satisfactory in practice to arrange the rotor surface to move in simple rotation about a fixed axis X31 without planetary motion. This considerably simplifies the drive assembly when compared with prior art planetary arrangements.

[0092] The rotor may be an assembly that moves in rotation about a fixed axis in a static vessel. The fixed axis X31 may be vertical, which is found to be very effective in assisting the nucleation and pelletisation of the mixture.

[0093] In alternative arrangements (not shown), the rotor may include a local axis about which the rotor surface rotates, wherein the rotor axis is acollinear (i.e. is not collinear) with the fixed axis about which the rotor rotates. The fixed axis may be vertical. The local axis may be vertical (so the rotor surface defines a planetary motion) or horizontal (so the rotor surface does not define a planetary motion).

[0094] Two or more such rotor axes may be spaced apart around the fixed axis, with one or more rotor surfaces moving in rotation about each respective local axis as the localaxes move in rotation with the rotor assembly about the fixed axis. For example, the rotor could have two or more local rotor assemblies extending from a central hub. For example, two such local assemblies may extend in opposite directions in diametrically opposed relation from the central hub and rotate about a common horizontal axis perpendicular to the fixed vertical axis about which the hub rotates. The mixer may include both a stirrer unit which rotates at a relatively low speed (e.g. about 80 RPM) to mix the water and dust, and a high speed agitator unit which rotates at a higher speed than the stirrer unit (e.g. about 400 RPM) to break down dry lumps of dust that tend to form when the dust is hydrated, as known in the art. Both units may be mounted on a common rotor that rotates at lower speed, e.g. about 20 RPM so that both units sweep the volume of the vessel.

[0095] However, it has been found that a satisfactory result can be obtained from a mixer that does not include a high speed agitator unit, and moreover, that any of a wide variety of conventional mixers can be used satisfactorily, particularly the types used for combining cement, aggregate and water to produce batches of concrete.

[0096] It has been observed that after the first use cycle, the agitator blade assembly will often become coated in hydrated dust to such an extent as to form in effect a solid, rotating lump in which the blades can no longer be discerned.

[0097] When the mixer is then used to produce a further batch of carbonatated material, the condition of the agitator is observed to have little or no effect on the quality of the material produced.

[0098] Surprisingly, in tests, it was found that when the high speed agitator failed, this had little effect on the process, and nucleation and pelletisation was unimpaired.

[0099] It is believed that this is due at least in part to the seed particles formed in the nucleation phase P4 which effectively break down any remaining dry lumps of dust in the mixture, even where no high speed agitator is provided, and even where the fluent body 101 consists entirely of dust particles 102' without any coarse fraction as commonly used in prior art processes.Since the agitator requires a powerful motor, both capital and running costs can thus be reduced by simply dispensing with the agitator.

[0100] It is found that mixers that move the rotor surface 31 in a constant direction of rotation about a fixed axis can provide effective nucleation and pelletisation. The fixed axis X31 may be a fixed vertical axis, as shown.

[0101] Each batch 10T may include around 750kg or more of particulate material, so the vessel may have a capacity of about 1500 litres. A rotating vessel of this size will typically tilt in order to discharge the batch, which adds significant complexity and cost. Therefore it is preferred to employ a moving rotor surface in a static vessel, as illustrated.

[0102] The interior surface 21 of the vessel may be a surface of revolution about the rotation axis X31, as illustrated, which is simple and effective for nucleation and pelletisation.

[0103] In the illustrated example, shown schematically, the rotor 30 includes multiple blades 3T, only some of which are shown, each defining a respective rotor surface 31, mounted on a central hub 33 that rotates about a fixed vertical axis X31 in a static vessel 20 with a cylindrical interior wall 21. The blades 3T are spaced apart angularly around the axis X31. Each blade 3T extends downwardly from the distal end of a respective shaft 32 that extends radially from the hub 33. Each blade 3T may have a scraper portion 34 at its lower end near the base of the vessel 20. Each radial shaft 32 is rotatable about its respective length axis relative to the hub 33, and is biased to a rest position (as illustrated) by a spring within the hub 33. In use, the torque reaction of the mixture 100 deflects the blade 3T and so rotates the radial shaft 32 against the restoring force of its spring (rotation R32), lifting the scraper portion 34 as the blade moves through the mixture 100. The blades 3T are arranged at different radial distances from the rotor axis X31' so that they sweep the whole volume of the vessel 20 as the rotor 30 rotates.

[0104] Mixers of the general type illustrated are available from Sicoma S.r.l. of Ponte Valleceppi, Perugia, Italy.The fluent body 101 may be admitted through an inlet 25 at the top or side of the vessel, e.g. from a supply conveyor 26. The inlet 25 may be sealed by an inlet door (not shown) to contain the gas G within the vessel.

[0105] The vessel 20 may have a discharge door 22, e.g. in the base of the vessel as shown or in the sidewall, that is selectively opened and closed to discharge the mixture 100, e.g. onto a discharge conveyor 24. The door may be operated by a door actuator 23 controlled by the control system 10.

[0106] The gas may be introduced via a gas inlet 74 and exhausted via a gas outlet 74. The vessel 20 may be substantially fluidly sealed from the ambient environment during at least the carbonatation phases of the process, except for the gas and water inlets, and openable before or during the initial period P1a to introduce the fluent body 101 via the inlet 25, and after the process to discharge the nucleated or pelletised mixture 100 via the discharge door 22. The discharged mixture may be directed to an end use (e.g. to make building blocks) or may be discharged for further pelletisation to an external pelletiser of known type, e.g .a drum or pan type pelletiser.

[0107] The sensing means

[0108] The sensing means 51 is arranged to sense a viscosity of the mixture 100 within the vessel 20, and to generate viscosity data 81 as a time series representing the sensed viscosity of the mixture 100 over time, and the control system 10 is arranged to receive the viscosity data from the sensing means 51.

[0109] The sensing means can be any sensor or sensing arrangement that generates a signal indicative of the viscosity of the mixture. The sensing means may include a contact part in contact with the mixture that moves relative to the mixture 100, and an sensor that measures (either directly or indirectly) a force between the contact part and the mixture, e.g. as a torque reaction or as applied power or speed or deflection or any other measurable parameter.

[0110] The sensing means 51 may be arranged to sense variations in power supplied to the actuator 40, wherein the power is arranged to vary responsive to a varying torque reaction of the mixture 100 at the rotor surface 31 , and so represents the viscosity ofthe mixture 100. In the illustrated embodiment, the sensing means 51 senses the current I flowing to the actuator 40, represented on the Y axis in Fig. 1 over time t on a scale from 0 (zero) to 1 (maximum).

[0111] To ensure that the current I is a faithful analogue of the viscosity of the mixture, the control system 10 may be arranged to supply power (or in accordance with the method, power may be supplied) to the actuator 40 to operate the actuator 40 to drive the rotor surface 31 in rotation at a substantially constant speed A2 during at least a part of the process.

[0112] This arrangement is illustrated in Fig. 1 , wherein the actuator 40 is an electric motor and the actuator speed A(s) is plotted over time t. Voltage is constant, and so current I varies with viscosity to maintain constant speed. Of course, other parameters could be monitored depending on the motor control arrangement.

[0113] It should be noted that in the example of Fig. 1 , actuator speed A(s) was reduced at the beginning of the second time period P5b of the fifth, pelletisation phase. The sudden drop in actuator current I between the first and second time periods P5a and P5b of the pelletisation phase reflects the reduction in actuator speed A(s) rather than an reduction in viscosity. Moving average viscosity traces V5a' and V5b' thus indicate approximately constant viscosity through the pelletisation phase P5, further discussed below.

[0114] In alternative arrangements, the sensing means could be separate from the actuator or motor that drives the rotor. For example, the sensing means could be arranged to sense angular deflection of the blade shaft 32 of one or more of the rotor blades 3T.

[0115] Water flow W

[0116] The water flow control means 60 is operable by the control system 10 to admit water W into the vessel to form a mixture 100 with the fluent body 101.

[0117] The water flow control means may be any arrangement for controllably starting and stopping or increasing and reducing a flow of water, such as a water pump that can be turned on or off, or (as illustrated) a valve operable by the control system 10 to admit water W from a pressurised water supply, e.g. a water main or a tank.In the illustrated example, the water is admitted via the valve 60 to an array of jets or sprinklers 61 inside the vessel. In practice, the jets or sprinklers 61 can be used also to clean the vessel after operation, e.g. on plant shutdown.

[0118] Gas flow G

[0119] The gas flow control means 70 is operable by the control system 10 to control a flow of the gas G into the vessel 20 at a gas flow rate G2, G3.

[0120] The gas flow control means may be any arrangement for controllably starting and stopping or increasing and reducing a flow of gas, such as a pump or a valve or moveable baffle or deflector arrangement.

[0121] The gas G may be stored, for example in a pressurised tank, and released selectively into the vessel via a valve controlled by the control system 10. The vessel may be configured to contain the gas at atmospheric pressure or slightly above or below atmospheric pressure, and to replenish the gas from the tank to maintain a constant pressure in the vessel as the gas G reacts with the mixture 100.

[0122] In many applications however, the gas flow control means 70 is configured to control the flow of gas G so that it flows through the vessel 20, where some or all of the carbon compounds in the gas are captured by reaction with the fluent body, with the remaining fraction of the gas flowing out of the vessel again.

[0123] Alternatively, as illustrated, the gas flow control means 70 may include a pump or impeller, e.g. a fan 71 driven by a fan motor 72 as shown, to urge the gas flow G through the vessel. The speed of the fan can be controlled by the control system 10 to control the flow of gas G through the vessel.

[0124] The fan may be arranged to exhaust the gas G out of the vessel via outlet 74, causing gas G to be drawn in through inlet 73 and thus creating a slight underpressure in the vessel 20 relative to ambient atmospheric pressure, so that the gas G cannot escape from the vessel 20 in case a leak should develop.

[0125] Once an input residue is loaded from a silo to a vessel for mixing, a carbon capture process occurs in four general phases:Phase 1 Hydration - Adding the input material (feedstock / carbonatable substance / fluid body) and a volume of water to the mixer in a first hydration phase.

[0126] Phase 2 Hydration and Carbonatation - adding at least one of water and carbonic gas to the mixer. Where CO2 is added to the hydrated mixture, the material is mixed whilst the exothermic carbonation reaction takes place. Additional water is added during this phase as per the required ‘recipe’ for that input residue.

[0127] Phase 3 Carbonatation - passing carbonic gas into the mixer in a third carbonatation phase; and

[0128] Phase 4 Pelletisation - by mixing in the mixer, forming seed particles in a fourth nucleation phase. An agglomeration process converts the carbonated input residue into pellets. These pellets must match a given specified size and strength characteristics based on their intended use.

[0129] It has been found that the variables involved in transforming feedstock into aggregate depend significantly on the composition of the feedstock. Such variables include the quantity of water required for, and duration of, any or all of the stages of the above process. Therefore basing the quantity of water and / or the duration of phases of the process on composition of the input material can dramatically speed up processing of the feedstock to aggregates.

[0130] Composition of the input material varies from site to site. In the case of accelerated carbon technology this is particularly based on the specific waste products that are available for use for carbonatation in processes outlined in this description.

[0131] Composition of the input material can vary within site, and even between batches. In order to accommodate such a vastly variable composition of the input material, processes can often take long periods of time, thereby limiting the number of batches of input material that can be processed each day.

[0132] Volume of water to be added to the feedstock, and duration of and between addition(s) of water to a mixer are difficult to know ahead of time. In particular, the ability of a compound to capture carbon is often only measurable retrospectively, after carbonatation itself is complete.In an example arrangement for use in carbonatation, a vessel starts with a input residue, the reactivity of which dictates the process. The physical and chemical properties of these input residues have been analysed by the applicant so as to be able to characterise the input residue in such a way that can be used for optimizing carbon capture from future input residues.

[0133] Analysis of the input residue can include study of the following attributes:

[0134] Mineralogy composition - Reactive examples: Lime, Larnite, Calcite, Sylvite, Portlandite, Mumme, Gehlenite; Non-reactive examples: Anhydrite, Brownmillerite, Corundum, Periclase, Quartz, Amorphous content

[0135] Grain attributes - size distribution, surface area and shape

[0136] Density attributes - loose, tapped, particle

[0137] Chloride content

[0138] Carbon Dioxide, CO2, reactivity

[0139] The input residue may be delivered in vast quantities and are often housed in silos on-site. Attributes for the input residues can be examined in ‘batches’ and delivered to a lab as a reflection of silo delivery on-site - i.e. the lab has a sample of material held in the input-silo at any one time. The residue may undergo, for example, x-ray diffraction, XRD, at the lab in order to determine attributes of the sample.

[0140] A complex interdependency relationship exists between the aforementioned attributes. Further, factors such as the way in which component of the input residue were added to the silo, movement within the silo and settling can all cause variations between and even within batches of input residue from the same site, or within the same silo.

[0141] Relationships between reactive components in particular is non-linear within and each batch, between batches and over time.

[0142] Fig. 2 shows pictorially XRD analysis of three different batches of an input residue. Considering lime only for the purpose of discussion, there is 11.7% limeconcentration in batch #64, 15.8% in batch #73, and 16.1% in batch #64. Other reactive components in the input residue also vary across the three sample batches.

[0143] Once the mineralogy of the input residue has been determined, the input residue can be processed.

[0144] Water effectively acts as a transmission medium surrounding the particles that make up the input material. Similarly, reactivity of the input material is at least in part based on physical properties of components of the input material. Such physical properties include particle size and surface area of the particle that is available for reactions. In this way, determining composition of the input material can be a useful indication of reactivity of the input material.

[0145] A carbonatation phase (discussed in further detail below) can last a substantial amount of time to provide sufficient carbonic gas, to an excess, to the hydrated aggregate for carbonatation of all of the input material in the mixture. For this reason carbonic gases are passed into a mixer for an extended period time, in some instances to an excess, for a mixture to fully undergo carbonatation. Thus the hydrated feedstock / input material often undergoes carbonatation for a longer period of time than required for that specific batch of feedstock / input material.

[0146] Fig. 3 shows graphically behaviour of a motor current of a mixer during a process for carbonatating an input material in a mixer. Herein input material is caused to undergo the aforementioned phases, each phase having characteristic patterns as it is processed within a mixer vessel over a period of time:

[0147] In this embodiment mixer motor current is an indication of a physical property, namely viscosity, of the input residue-water mixture. Whilst various processparameters can be measured during the aforementioned four phases, it has been found experimentally that mixer motor current gives a good indication as to what stage the process has reached. In other embodiments the physical property may be a temperature reading of the mixture.

[0148] At the first hydration phase, a fixed initial amount of water and fixed weight of an input material is added to a mixer. The input material and water may be added simultaneously to the mixer to minimise the time taken for this first initial wettingperiod. As the mixture in the mixer gains weight over this period and is mixed, a motor of the mixer draws on an increasing amount of current to maintain sufficient power to mix the mixture.

[0149] In this embodiment the motor is a variable motor operating at a known constant rotation per minute, rpm. Depending on the weight and viscosity of the mixture in the mixer at any one time, the amount of current required to maintain an rpm value of the mixer varies. In this way, based on the contents of the mixer, the power required by the motor to maintain mixing at a known rpm directly relates to current drawn on by the motor. This motor current serves as the indication of torque reaction of the mixture in the mixer.

[0150] Generally speaking, observing torque reaction of the actuator can be difficult in practice. The torque reaction is nuanced and can often be difficult to detect with conventional equipment. Therefore in some examples a machine learning model, MLM, may be used for observing the torque reaction. Such an MLM may be trained to accommodate for environment factors, such as equipment behaviour, that may impact torque reaction of the mixer. In an example the MLM is arranged take measurements corresponding to the observed torque reaction for use in deriving a predicted viscosity profile of the input material.

[0151] This first hydration phase ends once the progressive rise in motor current reaches a peak value.

[0152] The initial volume of water added at this stage is a fixed, predetermined amount. This stage typically takes around 4 minutes.

[0153] At a second hydration and carbonatation phase, a variable volume of water is added to the mixture (if any at all, depending on the input residue) for complete hydration of the mixture. Optionally carbonic gas may be passed over the mixture in an additional carbonatation step within the second phase. This second phase continues until the motor current reading eventually reaches a substantially constant value.

[0154] At a third stage carbonic gas is provided to the mixture. A stream of carbonic gas may be caused to pass over, and thereby carbonatate, the hydrated input material. In some embodiments a fan may be caused to operate for causing gas to be pulledinto, through and out of the mixer. The carbonatation phase can last a substantial amount of time to provide sufficient carbonic gas, to an excess, to the mixture for thorough carbonatation.

[0155] Following this, a fourth nucleation phase occurs. The carbonated mixture is formed into seed particles, typically of about 1mm - 2mm diameter. These seed particles produce a fluent mixture that can be used for a variety of useful purposes, typically as a bulk building material. Optionally the seed particles may be further pelletised to form larger pellets having a diameter of several millimetres up to one or two centimetres or more, in subsequent phases that are not discussed herein for brevity.

[0156] Fig. 3 can be considered to show fingerprint data of a batch for a known composition of input residue. A relationship can thus be established between the composition of the input material, as determined by, for example XRD, and corresponding fingerprint data for each batch of input residue. The fingerprint data may be measured for all or part of the process of transforming input material into an aggregate.

[0157] It has been recognised that a mechanism for predicting the hydration and carbonatation ‘needs’ of the input residue would greatly improve the efficient of the entire carbon capture process. Presently, there is heavy reliance on knowing the composition or mineralogy of the input residue for knowing how much water to add and when.

[0158] Lab analysis required for determining mineralogy of the input material can take in the region of months to be completed. In contrast, carbon capture processes can take, generally speaking, approximately 30 minutes. In order words processing of feedstock at a site is inherently limited by the time taken for mineralogy information to be returned.

[0159] Water is added to the mixture at various phases throughout the process, and in controlled amounts, to ensure that the mixture is not over or under-wetted. If the mixture is too dry, carbon capture may be restricted and the aggregate will not properly form. Too wet, and the mixture will turn into an unusable sludge with dramatically reduced carbon capture efficiency.It has been realised that batched addition of water and mixing facilitates production of lightweight aggregate from feedstock that is almost entirely made up of waste products. This efficiency can be further improved by being able to predict the quantity of water to be added during the process, and / or deriving a schedule for said addition of water.

[0160] There is required a method of conducting carbon capture process that does not rely on time-consuming lab analysis and which is able to predict and optimal volume of water and / or schedule for water delivery. To this end, there is provided a method of optimising carbon capture which uses a combination of measured properties and predicted values, using the smallest possible amount of initial data for each batch of input reside, whilst providing sufficient information to indicate how much water to add and when for optimal carbon capture from the input residue.

[0161] Returning to Fig. 3 and taking the first, hydration phase, time series data can be derived by taking motor current readings at regular intervals until a peak value is reached (see boundary of phase 1 / phase 2). Based on the time series data, it is possible to identify a peak value of the indication of motor torque reaction (or, more generally speaking, the sensed physical property), calculating a slope of change of an indication of motor torque reaction (sensed physical property); and calculating an area under the slope. Based on the peak value, slope, and area under the slope, a parameter value for carbon capture of the fluent body can be determined.

[0162] Said parameter value may comprise a volume of water to be added at a subsequent phase of the process, and / or a schedule for adding a water.

[0163] In this embodiment sensing the physical parameter of the mixture comprises sensing an indication of torque resistance of the mixture against the actuator. In some embodiments sensing the torque resistance of the mixture comprises sensing a motor torque reaction of the actuator. Herein the motor torque reaction is a motor current reading of the actuator. Sensing motor current may comprise recording current values over a predetermined period of time, and / or until a number of currenttime samples have been recorded.

[0164] In some embodiments, processing the input residue and measuring motor torque reacting comprises mixing the mixture at a predetermined rate sensing powerrequired for maintaining the predetermined rate of the mixture. In some embodiments the predetermined rate is a number of rotations per minute, and the measure motor current is a measure of power that is required by the actuator for maintaining mixing of the mixture at the predetermined rate.

[0165] Relationships between reactive components are often non-linear, and use of Artificial Intelligence, Al, can greatly assist detection and understanding and interpretation of observable, physical properties of the mixture. Specifically, a machine learning model, MLM may be introduced for predicting an optimum volume and / or schedule for addition of water to the mixture, based only on data relating to the first phase of an input residue.

[0166] A focus of the MLM is to derive an indication of an estimated mineralogy of an input residue as fast as possible prior to the second hydration and carbonation phase. For this reason, data of from the first hydration phase is of particular interest as the input feature to the model. This has been chosen to give the fastest-response with the least amount of data possible.

[0167] Various fixed conditions can be applied at the first hydration stage. In an example 750kg of input material and 70 litres of water are poured into the mixer.

[0168] Fig. 4 shows graphically time series data for a first hydration phase. Specifically, this diagram shows time against mixer motor current (A) as an indication of viscosity of the mixture. In this example, motor current was measured at 10 second intervals, t. In some embodiments the MLM is an Artificial Neural Network, ANN regression model. Of various suitable Al architectures, an ANN regression model was found to be particularly suited for handling the complex relationships, high-dimensionality, and non linear transformations involved in using MLM for determining water supply during a carbonatation process.

[0169] The ANN is made up of multiple layers of interconnected artificial neurons, including the input layer, with a number of hidden layers, and an output layer. Each neuron is connected to every other neuron in the subsequent layer.

[0170] The model in the first instance has an input array of plural values and an output of a single number. In the present embodiment the input array is time series data.

[0171] Effectively each of these data points in Fig. 4 may be provided to the model. Theoutput of the model is a single number, for example representing lime concentration within the input residue.

[0172] The relationships between the multimodal input array and output data value are complex, adding to the aforementioned complexities when considering reactivity of input residues. Further the nature of the input is noisy data, with outliers, and the model preferably adapts and generalises to account for irregularities.

[0173] It has been found experimentally that an input array having 160 samples is the maximum input-array of the first hydration phase data. The ANN may comprise four successive hidden layers prior to the single output layer. Each hidden layer has half as many neurons as the previous layer. In a preferred embodiment the model comprises 116 inputs interconnected to 56 neurons, which in turn are interconnected to 29 neurons, with these interconnected to 15 neurons, then interconnected to 7, and finally these are connected to a single output neuron. This neural network tree structure is exemplary. Other structures and combination thereof may be used in different embodiments of the invention.

[0174] Once the first hydration phase data is provided, say, as an input array, to the ANN, an output value from the ANN in this embodiment is an indication of the concentration of material from which carbon capture is possible in the input residue as a whole. This output value may, for example, be lime concentration of the input residue.

[0175] The MLM is able to provide such valuable information based on first phase data by training the machine learning model, MLM, trained on historic carbon capture data associated with a historic sensing of the physical property of the historic input residue.

[0176] Vast quantities of such historic data has been used to train the MLM. In an embodiment historic carbon capture data comprises:

[0177] crystallography data of the historic input residue;

[0178] data from a first hydration phase of a historic input residue; and

[0179] data from subsequent processing of the hydrated historic input residue.

[0180] Data from subsequent processing of the hydrated historic input residue may comprise data from a second, hydration and carbonatation phase of the historic input residue; data from a third, carbonatation phase of the historic input residue; and data from a fourth nucleation phase of the historic input residue.

[0181] Data for training the MLM may resemble the data provided at Fig. 2. Deriving such data may first comprise determining a composition of a historic input residue. There may be a step of processing the historic fluent body of solid particles, of mixing the historic input residue with an initial volume of water in a vessel, by an actuator, to form a mixture, during the first hydration phase. During at least one subsequentphase, a volume of water and / or a volume of carbonic gas may be added to the mixture. Following this, a physical property of the mixture is sensed at time intervals during the first phase and the at least one subsequent phase, thereby deriving time series data for the historic input residue. In the present embodiment the sensed property is motor current. From hereon a physical property profile (viscosity profile in this embodiment) is formed for the historic input residue based on the time series data for the input residue. The physical property profile is associated with the composition of the historic input residue. These steps are repeated for plural batches of the historic input residue.

[0182] Subsequently the MLM can be trained using values of time series data to the MLM; and the composition of the historic input residue. Training the MLM in accordance with the above may comprise deriving at least one coefficient of the neural network, based on the historic carbon capture data.

[0183] From hereon, during a use phase of the MLM, time series data from a first hydration phase off an input residue is provided to the trained MLM.

[0184] Fig. 5 shows an example subset of training data for the MLM, wherein for four batches motor mixer currents from each run of the batch are plotted over timestep. Plural runs within that batch are shown superimposed onto a batch graph, with an average of the runs shown in darker plotted points.

[0185] Each batch corresponds to its adjacent XRD result indicating mineralogy of that batch. It is these unique reactive-compound ratios that give rise to nuances / features within the graph data, and it is these features the MLM has been trained to identify by feeding data relating to all runs of the batches of input residue.

[0186] Fig. 6A shows graphically the same set of data points obtained by X-ray diffraction testing for the proportion of lime (i) as a percentage by weight of the respective sample number S#. Fig. 6B shows graphically (in a slightly lighter shade) the corresponding data points inferred by the model, based on the sensed viscosity data captured during processing each corresponding batch of the tested material in accordance with the process, and superimposed onto the actual test Hence, Fig. 6B shows a pair of values (lighter, inferred value; superimposed on darker, test data value) for each sample. Each of the data points relates to Lime concentration (%) is plotted for each delivery over time (as batches).

[0187] After extensive training and refining of parameters of the model, plotting the estimated lime concentration against the ground truth shows an accuracy of approximately 85% for the model. This level of accuracy is sufficient for commercial employment of the model, for the manufacture of aggregates. The accuracy may be further improved by modifying the model to reflect other, optionally additional, input parameters such as: initial water added during hydration (litres), volume of residue (kg) and mixer temperature over the hydration-phase (degrees Celsius) measured over the same time-steps as the mixer motor current.In some embodiments and output of the MLM is a predicted composition of the fluent body based on the sensed physical property (motor current in the aforementioned example). Predicting the parameter value may comprise predicting at least one physical and / or chemical attribute of the input material. For example the attribute may be least one of, grain attributes, density attribute, chloride content and carbon dioxide, CO2, and reactivity of the fluent body. Predicting the parameter value may comprise predicting a percentage presence of at least one reactive compound of the input material. In an example the following reactive compound is at least one of lime, larnite, calcite, sylvite and portlandite.

[0188] Discussion up to now has focussed primarily on the first hydration phase. Using the information thus far it is possible to model chemical kinetics for ultimately controlling water delivery for an input residue provided to a mixer.

[0189] In an example the MLM is used to estimate a percentage of compounds in an input material. From hereon molar masses can be estimated.

[0190] It may be useful to consider a representation of a typical reaction occurring during the process. For simplicity and convenience only lime, the most reactive compound found in the input material, is considered.

[0191] During a hydration phase:

[0192]

[0193] Wherein:

[0194] CaO = Calcium Oxide (Lime);

[0195] H2O = Water;

[0196] Ca(0H)2= Calcium Hydroxide (intermediate product)

[0197] Taking an example wherein a 750 kilogram, kg, batch of an input material is found, use the MLM, to contain 16% lime is estimated in a sample, classical calculation can be used to determine that approximately 2140 moles of CaO are present in the batch.

[0198] From hereon, with reference to the equations above, considering a 1:1 ratio lime of to water, it can be determined that 38.56kg of water is required to fully hydrate the 120kg of lime present in the 750kg batch.

[0199] Reactants are generally expected to make up approximately 50% in one form or another (by reaction or by simply dissolving in the water), of a 750kg batch and an initial batch of water measuring 40 kg. In systems not using the steps outlined in the present application, approximately 300 litres, I, of water is typically used as part of the hydration stage.Subsequently, during carbonatation:

[0200]

[0201] Wherein

[0202] CO2 = Carbon Dioxide

[0203] CaCO3 = Calcium Carbonate

[0204] Considering the same 1:1 ratio of moles of carbon dioxide and calcium carbonate, and the previous calculation of 2140moles of calcium carbonate, it is possible for a maximum of 94kg of carbon dioxide to be absorbed by the hydrated mixture.

[0205] These estimations form the basis of a chemical reaction simulation that is used to estimate how much water to add, and when, to the mixture in order produce an aggregate from the input material. The above values and chemical equation represent an ideal scenario with assumption made in order to best simulate the chemical kinetics of carbonation.

[0206] The chemical reaction simulation comprises a series of differential equations. Herein rates are provided by the Arrhenius equations, and parameters are selected based on calcium oxide and calcium hydroxide.

[0207] Based on the estimated mineralogy of the input material, and by simulating a chemical reaction based on the estimated mineralogy of the input material it is possible to predict carbon capture capability of the fluent body. The simulation may also consider the exothermic nature of the reactions taking place within the mixture. From hereon, it is possible to model predicted behaviour of the mixture.

[0208] A programmable logic controller, pic, may carry out at least part of the chemical reaction simulation. In some embodiments, the MLM conducts at least part of the chemical reaction simulation. The simulation may based on observed time series data during a first hydration phase in which an input material is mixed with an initial volume of water.

[0209] Fig. 7 shows graphically modelled temperature, carbon dioxide capture and mass of the mixture, achieved by solving the aforementioned differential equations. Following from the above example 120kg of lime is estimated to be present in the 750kg batch of input material. The maximum weight approaches 253 Kg, which made up of 120 kg of calcium oxide, plus the 38 kg of water, and 93 kg of carbon dioxide. As outlined previously, a 120 Kg batch of fully hydrated calcium oxide should capture 93 Kg of carbon dioxide. This is seen asymptotically over time, showing the reaction converges to the maximum within a matter of minutes. This may also be used as an indicator for the end of the carbonatation phase.

[0210] As somewhat expected, lime shrinks exponentially whilst the calcium carbonate product grows at a given rate. Calcium hydroxide grows with the depletion of calciumoxide and shrinks with the formation of calcium carbonate. Reactions herein are temperature dependant and also exothermic in nature, which further feeds into the reaction. This may be seen by the exponential increases in temperature and then slow decline as the calcium oxide depletes. The temperature then begins to return to the ambient surrounding temperature. Any of these considerations may be incorporated into an MLM for increasing accuracy of at least one of predicted mineralogy of an input material and the predicted parameter.

[0211] In some embodiments a method may comprise comparing predicted mineralogy with a pre-determined probability density reference, PDF, plot for the composition.

[0212] Fig 8 shows schematically an overview of a process for optimising carbon capture of an input material by controlling a mixer. There is shown a process 800, wherein at a first step 802 indication of torque resistance is received, for example as a set of current readings from the mixer at time intervals. Subsequently in a second step 804 a relationship of indication vs time is determined. In an example the second step comprises plotting indication value overtime. Based on this relationship it is possible to determine a slope rate of the indication increase overtime as the input material is mixed with water, a peak value for the indication, and the area under the slope.

[0213] In an example wherein the indication of torque resistance is a mixer motor current, a slope rate of current increase over time as the input material is mixed with water, a peak current value during hydration, and the area under the slope are determined. In a third step 806 a viscosity profile is determined, based on the relationship between the received indication over time 804. Finally, based on the viscosity profile, there is a final step of predicting a parameter value 808, wherein such a parameter value includes volume of water to be added and / or duration of carbonatation. In another embodiment, a temperature profile may be determined based on observing temperature behaviour of the mixer.

[0214] Further processing of the input material is then possible based on the predicted parameter value. The at least one step of the process may be carried out at or by a control system of a mixer or mixing arrangement, for example as described with respect to Fig. 1. A programmable logic controller, pic, may be controlled by the control system to adjust supply of water and / or carbonic gas to the mixer.

[0215] In some embodiments as an intermediary step, a predicted composition of the input material may be derived by the MLM.

[0216] Fig. 9 shows schematically a process for optimising carbon capture of an input material according to an embodiment.

[0217] The process 900 comprises receiving a sensed indication of torque resistance 902. Based on the indication input variables are determined 904. In an example wherein the indication of torque resistance is a mixer motor current three variables can bedetermined: a slope rate of current increase over time as the input material is mixed with water, a peak current value during hydration, and the area under the slope. In this embodiment the input variables are provided to an MLM for processing.

[0218] Specifically, in a next step the MLM determines a viscosity profile for the input material 906, based on which the MLM predicts parameter value 908. Said parameter value may be an estimation of the volume of water and / or duration of a carbonatation period. In this embodiment an output of the MLM is subsequently used for determining corresponding controls for the mixer / mixing arrangement 910, so as to ultimately control the mixer / mixing arrangement 912.

[0219] As indicated by the dashed boxes, all or part of the process for carbon capture can be controlled by an MLM. In the specific example where all of the steps 902 to 912 are automated by an MLM, little to no human intervention is required from the point of providing a mixture in a first hydration phase to controlling water and CO2 delivery of the mixer to provide seed particles. In other embodiments the MLM is arranged to additionally automate the first hydration phase.

[0220] The following definitions are provided for clarity.

[0221] The carbonic gas

[0222] The carbonic gas may consist of carbon dioxide or a mixture of carbon dioxide with one or more other gases, typically nitrogen, nitrogen oxides, carbon monoxide and / or other combustion products. The carbonic gas may contain between 10% and 100% CO2. In some instance the carbonic gas may contain more than 50% but less than 100% CO2, e.g. up to about 75% CO2 by volume.

[0223] The dust

[0224] In this specification, a dust means a powder having a particle size small enough to be entrained in a waste gas stream from which the dust is extracted. Typically the dust particles have a maximum diameter substantially less than 1mm, such as about 300 microns or less, often about 100 microns or less, down to as small as a few tens of microns or even 1 micron or less. The surface area of the dust particles will vary with particle size and shape, which in turn depends on how they are produced.The reactivity of the dust particles will vary with their surface area and chemical composition. The chemical composition is often highly variable, even between different batches of dust from the same source, depending for example on changing feedstock and process conditions.

[0225] The dust may be extracted from a flue gas generated during an industrial process such as incineration or calcination of a feedstock. The dust may be a fly ash, which is to say, an ash that is entrained in flue gas.

[0226] Dusts such as APCR, CKD and CBD may be highly alkaline and typically have a small particle size and high surface area, resulting in a relatively low bulk density, often less than about 1000kg / mA3, for example, around 500 - 750 kg / mA3.

[0227] Advantageously, the dust particles may form at least 40%, 50%, 60%, 70%, 80%, 90% or 95%, or even up to 100% by weight of solid particles in the input material. In this way the novel process may be used to convert waste dusts into a useful aggregate without, or with relatively little, addition of larger particulates such as sand or powdered minerals, which have been necessary in some prior art processes.

[0228] Optionally however, the solid particles may include a proportion of non-dust particles, such as sand or powdered minerals, for example, where it is desired to produce a final aggregate product that includes those other materials. The non-dust particles may form a minor proportion of the fluent body by weight, for example, less than 50%, 40%, 30%, 20%, 10%, or 5% by weight.

[0229] A method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water. A fluent body is mixed with an initial volume of water in to form a mixture. Following this is sensing a physical property of the mixture for a progressive increase in value of the sensed physical property. Based on sensing the physical property, a parameter value for carbon capture of the fluent body is predicted, and delivery of water to the vessel is controlled in accordance with the predicted parameter value.For the avoidance of any doubt, the terms “a”, “an” and “the” are intended, unless specifically indicated otherwise or the context requires otherwise, to include plural alternatives, e.g., at least one.

[0230] "Optional" or "optionally" means that the subsequently described event or circumstance can or cannot occur, and that the description includes instances where the event or circumstance occurs and instances where it does not.

Claims

CLAIMS1. A method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water, the method comprising:mixing a fluent body with an initial volume of water in a vessel by an actuator for engaging the fluent body, to form a mixture;sensing a physical property of the mixture for a progressive increase in value of the sensed physical property;based on sensing the physical property, predicting a parameter value for carbon capture of the fluent body; andcontrolling delivery of water to the vessel in accordance with the parameter value.

2. A method according to claim 1 wherein sensing the physical property is for a period of time taken for a peak value of the physical property to be sensed.

3. A method according to claim 1 or claim 2 wherein the parameter value is a further addition of water.

4. A method according to any preceding claim wherein the parameter value is a time duration between predicting the parameter value and controlling delivery of water.

5. A method according to any preceding claim comprising determining a parameter profile based on sensing the physical property, said determining comprising:identifying a peak value of the physical property in a duration of time; determining a slope of change of the sensed physical property of the duration of time; andcalculating an area under the slope over the duration of time.

6. A method according to any preceding claim comprising predicting composition of the fluent body based on the sensed physical property.

7. A method according to any preceding claim wherein predicting the parameter value comprises predicting at least one of, grain attributes, density attribute, chloride content and carbon dioxide, CO2, and reactivity of the fluent body.

8. A method according to any preceding where predicting the parameter value comprises predicting a percentage presence of at least one reactive compound of the input material.

9. A method according to claim 8 wherein the at least one reactive compound comprises at least one of the following reactive compounds within the fluent body: lime, larnite, calcite, sylvite and portlandite.

10. A method according to any preceding claim wherein sensing the physical property of the mixture comprises sensing an indication of torque resistance of the mixture against the actuator.11.A method according to claim 10 wherein sensing the torque resistance of the mixture comprises sensing a motor torque reaction of the actuator.

12. A method according to claim 11 wherein the motor torque reaction is a motor current reading of the actuator.

13. A method according to any one of claims 10 to 12 comprising determining a viscosity profile based on an indication of motor torque reaction, said determining comprising:identifying a peak value of the indication of motor torque reaction; calculating a slope of change of the indication of motor torque reaction; and calculating an area under the slope.

14. A method according to any preceding claims comprising:mixing the mixture at a predetermined rate, andwherein sensing the parameter value comprise sensing power required by the actuator for maintaining the predetermined rate of the mixture.

15. A method according to any preceding claim wherein the sensed physical property comprises a temperature value of the mixture.

16. A method according to any preceding claim wherein predicting the parameter value is by a machine learning model, MLM.

17. A method according to claim 16, comprising:forming the mixture in a first hydration phase,sensing a physical property of the mixture for a period progressive increase in value of the sensed physical property during said mixing, at time intervals, to derive a set of time-series data;providing the time series data to a machine learning model, MLM; and predicting the parameter value by the MLM, based on the time series data.

18. A method according to claim 17, comprising:simulating, by the MLM, a behavior of the sensed physical property in at least one and the at least one further phase, subsequent to the first hydration phase; andpredicting a parameter value based on the simulated behaviour.

19. A method of optimizing carbon capture of a fluent body of solid particles by controlling delivery of water to the fluid body by a machine learning model, MLM, the method comprising:g) providing a batch of a historic fluent body, the historic fluent body having known composition;h) processing the batch with an initial volume of water in a vessel, by an actuator, to form a mixture;i) sensing a physical property of the mixture for a period progressive increase in value of the sensed physical property during said mixing, wherein said sensing comprises measuring the physical property at time intervals, to derive historic time-series data;j) deriving a historic physical property profile for the batch based on the derive historic time-series data;k) associating the historic physical property profile with the known composition;l) repeating steps b) to e) for at least one additional batch of the fluent body;andm) providing plural historic physical property profiles to the MLM.

20. A method according to claim 19 comprising analyzing, by crystallography, portion of the batch of fluent body to determine a composition of the fluent body.21.A method according to claim 19 or claim 20 wherein processing the batch comprises:mixing, in a first hydration phase, the fluent body with water, to form the mixture;processing, in at least one subsequent phase, the mixture; andsensing the physical property during the first hydration phase and during the at least one subsequent phase;wherein the historic time-series data is data from the first hydration phase and during the at least one subsequent phase.

22. A method of optimizing carbon capture of a fluent body according to claim 16 or claim 17, comprising training an MLM according to the steps of any one of claims 19 to 21.

Images