AI Controlled Decanter

Abstract

The present invention relates in particular to a computer-implemented method for optimizing the output of a decanter in operation using an enhanced artificial intelligence (AI) engine, comprising the steps of: a. operating the decanter according to a plurality of operating parameters; b. processing a physical input comprising sludge and polymer through said decanter to produce a physical output comprising centrate and cake; c. determining a plurality of material parameters based on the physical output; d. passing the plurality of material parameters and the plurality of operational parameters to the enhanced AI engine; e. determining, by said enhanced AI engine, a quality value for each of said plurality of material parameters; f. predicting, by said enhanced AI engine, a plurality of adjusted driving parameters; g. further operating the decanter based on the plurality of adjusted operating parameters.

Description

[Technical Field]

[0001] The present invention relates to a computer-implemented method for optimizing the output of a decanter centrifuge, an associated computer-implemented method for training a reinforcement / supervised learning artificial intelligence (AI) engine, an associated trained reinforcement AI engine, an apparatus, and a computer program. [Background technology]

[0002] Decanter centrifuges (also referred to herein as decanters) are widely used in the field of liquid processing, particularly water processing, such as wastewater retreatment. Decanters are used to separate different phases and / or precipitates contained in a fluid (e.g., a liquid substance such as sludge) from one another. For this purpose, the rotation volume of the decanter is filled with the fluid, and the different phases / sediments can be separated from one another, for example, by centrifugal forces that act differently on the different phases / sediments in the fluid depending on their respective masses. Thus, phases / sediments with a larger mass are relegated to the radially outer regions of the rotation volume of the decanter compared to phases / sediments with a smaller mass, while the latter remain in a radial region closer to the central axis of the decanter around which the decanter rotates.

[0003] When operating a decanter in the field of wastewater treatment, it is assumed that a purified output fluid (e.g., clean water, also called centrate) is obtained by separating suspended sediment from the water contained in the sludge and removing the separated sediment from the rotating volume of the decanter. The separated phase / sediment, also called cake, is disposed of as waste. Disposal is generally expensive, and this cost is often proportional to the weight of the cake. Therefore, to minimize disposal costs, it is preferable to keep the weight of the discarded cake low. Furthermore, the heavier the cake, the more complex its transportation becomes, since more powerful conveying equipment is required. Therefore, the smaller the cake weight (depending on its water content), the easier it is to transport. Therefore, reducing the cake mass and / or weight is desirable. This can be achieved at least in part by minimizing the amount of liquid in the cake, i.e., its moisture content, which can be achieved, for example, by appropriately setting the respective operating parameters of the decanter (e.g., the speed of the decanter bowl).

[0004] This adjustment is typically performed manually by the decanter operator (and / or by a (manually set) PID controller). To do so, the operator generally needs to repeatedly sample the cake to determine its moisture content, and based on this moisture content, decide which operating parameters should be used next, for example, to further reduce the cake moisture content. However, such manual cake moisture content optimization procedures are costly due to the effort required and may not produce satisfactory results. The latter may result in the cake remaining with higher moisture than expected, resulting in an excessive cake weight and / or mass. Therefore, a time-consuming iterative optimization performed by an operator procedure is required.

[0005] Furthermore, the initial start-up of the decanter can take some time (e.g., about 30 minutes or more) to find and reach the desired range of optimal operating parameters. This means that during this initial start-up, the decanter is not operating under ideal operating conditions, potentially resulting in cake with undesirable moisture content and unreasonably increased associated disposal costs.

[0006] The situation becomes even more complicated when optimizing multiple parameters (e.g., cake dryness and water purity). More specifically, depending on the situation, more than just minimizing cake moisture is desired. In some cases, it may be even more advantageous to also optimize, for example, the quality (purity of the water produced in the decanter, also known as "centrate quality"). These two aspects lead to an interaction between cake moisture and centrate quality, which requires adjusting multiple operating parameters of the decanter to achieve the desired optimized output of the decanter. This further complicates the manual optimization procedure in terms of time, effort, and effectiveness.

[0007] The prior art has only partially addressed these issues by proposing the use of artificial intelligence based tools to determine the optimum operating parameters of the decanter and optimize the decanter output.

[0008] As an example, WO 2019 / 150002 relates to a method for providing at least one input parameter of a sludge dewatering process of a wastewater treatment plant, the method comprising the steps of acquiring data representing process data and / or plant configuration data of a wastewater treatment plant, providing at least a portion of the acquired data to at least one model formed by combining at least historical process data and plant configuration data collected from a plurality of wastewater treatment plants and properties of chemicals applied to the wastewater treatment plant, and predicting at least one input parameter and / or at least one output parameter of the sludge dewatering process by the at least one model for adjusting the sludge dewatering process of the wastewater treatment plant.

[0009] Japanese Patent Application Laid-Open Publication No. 2021-102195 relates to a machine learning device for a centrifuge, which includes: a training dataset storage unit that stores multiple training datasets each consisting of input data including the slurry concentration of the treated liquid, the moisture content of the dewatered solids, the concentration of the separated liquid, and the torque value of the screw conveyor, and output data including control parameters of the centrifuge system associated with the input data; a learning unit that inputs multiple training datasets to train a learning model that infers the correlation between the input data and the output data; and a trained model storage unit that stores the learning model trained by the learning unit. The control parameters include at least one of the amount of additive added to the treated liquid, the centrifugal force of the bowl, and the differential speed controlled by a differential speed generator.

[0010] Japanese Patent No. 6994330 relates to a dehydration system including a dehydration device that generates a turbid residue by removing liquid from a suspension, and an analysis device that stores suspension data indicating the state of the suspension, operating data including multiple operating parameters of the dehydration device, and a calculation model that is pre-constructed based on the turbid residue data indicating the state of the turbid residue. The analysis device calculates a predicted state value of the turbid residue by inputting current values ​​of the suspension data and the operating parameters into the calculation model.

[0011] AI-based approaches known in the art are unable to provide the desired effect for operating a decanter, i.e., the decanter's output is not fully optimized in a satisfactory manner. Furthermore, these AI-based approaches lack flexibility, as it is not easy to further train the respective AI models during operation. Therefore, there is a need for an approach that improves optimization results and allows for (further) optimization during operation. Meanwhile, there is also a need to provide a method that achieves adequate optimization results with less input data and reduces the overall process complexity. Summary of the Invention

[0012] These problems are at least partially addressed by one aspect of the present invention, a computer-implemented method for optimizing the output of an operating decanter using an enhanced artificial intelligence (AI) engine. The method may include: a. operating the decanter according to a plurality of operating parameters; b. processing a physical input, including sludge and polymers, with the decanter to produce a physical output, including centrate and cake. The method may further include: c. determining a plurality of material parameters based on the physical output; and d. passing the plurality of material parameters and the plurality of operating parameters to the enhanced AI engine. The method may further include: e. determining a quality value for each of the plurality of material parameters with the enhanced AI engine; and f. predicting a plurality of adjusted operating parameters with the enhanced AI engine. The method may further include g. operating the decanter based on the plurality of adjusted operating parameters.

[0013] The physical output can be understood as the materials originally contained in the physical input that have been separated from each other by the operation of the decanter, and the quality value can be understood as a numerical value indicating whether the respective material parameters are within a predetermined tolerance parameter range.

[0014] The adjusted operating parameters may relate to operating parameters that are determined after it has been determined that a previous operating parameter did not result in optimization of one or more of the plurality of material parameters.

[0015] Operating a decanter based at least in part on the computer-implemented method described above facilitates automated operation of the decanter, since no human operator input is required, or at least minimal input, to operate the decanter. Reinforcement-based AI control of decanters is particularly suited to improving AI-based control of decanters, allowing the decanter to be quickly set to optimal operating conditions (since the operator does not need to "trial and error" to find the decanter's optimal operating parameters). This at least reduces or minimizes disposal costs of the decanter's physical output. Automated operation of decanters also prevents excessive resource consumption due to potential operator inattention (e.g., when an operator fails to quickly recognize that a decanter is not being operated at its optimal settings).

[0016] The plurality of operating parameters and the plurality of adjusted operating parameters may include one or more of a decanter scroll differential speed, a decanter bowl speed, a sludge feed rate, and / or a polymer feed rate.

[0017] The decanter bowl speed can be understood as the rotation speed (rpm) of the internal (processing) volume of the decanter where the actual separation of different phases (e.g., liquid and solid) and / or sediment takes place. A higher rotation speed results in a higher centrifugal force acting on the different phases / sediments contained in the internal volume, whereas a lower rotation speed may result in a lower centrifugal force (less separation between the different phases / sediments). Thus, the heavier phases / sediments accumulate in the radially outer part of the internal volume of the decanter bowl and can even be expelled from the internal volume of the decanter bowl.

[0018] The decanter bowl may be at least partially surrounded by a scroll adapted to discharge the separated phase / precipitate that exits the decanter bowl, such that each separated precipitate is discharged from the decanter.

[0019] The decanter scroll differential speed can be understood as the absolute relative speed of the decanter scroll (e.g., revolutions per minute (RPM)) and the decanter bowl (e.g., RPM). A higher scroll differential speed can shorten the retention time of the separated phases / precipitates in the bowl (e.g., the period of time that the separated phases / precipitates may remain in the bowl), whereas a lower differential speed can lengthen the retention time of the separated phases / precipitates in the bowl.

[0020] Sludge can be understood as the treatment fluid injected into the decanter bowl. The sludge may consist of a liquid (preferably water) and one or more different precipitates that are separated from the liquid by the decanter. In some cases, the sludge is wastewater.

[0021] The polymer feed rate can be understood as the rate at which the polymer is added to the sludge (e.g., in kg / s or l / s). The polymer is adapted to aid in the aggregation of sediment particles into larger, heavier compounds, allowing better separation of the formed compounds from other sediments and / or liquids.

[0022] This can advantageously contribute to the optimization of material parameters, which is facilitated by adjusting operating parameters so that the material parameters converge to a preferred parameter range (e.g., a predefined parameter range for a particular material parameter that is deemed optimized).

[0023] The step of determining the plurality of material parameters may further include determining, by one or more sensors, the dryness of the cake, the purity of the centrate, and the loading of polymer in the centrate.

[0024] The dryness of the cake can be understood as a measure of the liquid content of the cake. In other words, the dryness of the cake can indicate the moisture content of the cake. A higher dryness of the cake indicates a lower moisture content of the cake, while a lower dryness of the cake indicates a higher moisture content of the cake.

[0025] Centrate purity can be understood as the cleanliness of the process fluid (e.g., water).

[0026] The amount of polymer added refers to the amount of polymer added to the sludge.

[0027] Determining the plurality of material parameters allows for optimizing one or more of the plurality of material parameters based at least in part on the enhanced AI engine, thereby facilitating closed-loop operation of the decanter by determining a plurality of operating parameters such that the plurality of material parameters are optimized.

[0028] The step of determining a quality value for each of the plurality of material parameters by the enhanced AI engine may further include the steps of: setting the cake dryness quality value to low if the cake dryness is in a range of 10 to 19.99% dry matter equivalent (%DS) and setting the cake dryness quality value to high if the cake dryness is in a range of 20 to 35%DS; setting the centrate purity quality value to low if the purity of the centrate is in a range of 300 to 1000 nephelometric turbidity units (NTU) and setting the centrate purity quality value to high if the purity of the centrate is in a range of 50 to 299.99 NTU; and setting the polymer dosage quality value to high if the amount of polymer added in the centrate is in a range of 2 to 9.99 kg / tDS (tons dry matter equivalent) and setting the polymer dosage quality value to low if the amount of polymer added in the centrate is in a range of 10 to 20 kg / tDS.

[0029] A high cake dryness quality value may indicate that the cake dryness is acceptable or may at least be slightly reduced. A low cake dryness quality value may indicate that the cake is not dry enough and that the dryness needs to be increased. The cake dryness quality value is considered low when the cake dryness is preferably within the range of 10-19.99% DS, more preferably within the range of 12-18% DS, and most preferably within the range of 14-16% DS, while the cake dryness quality value is considered high when the cake dryness is within the range of 20-35% DS, more preferably within the range of 25-30% DS, and most preferably within the range of 26-28% DS.

[0030] A high centrate purity quality value may indicate that the purity of the centrate is within an acceptable range or may at least be slightly reduced. A low centrate purity quality value may indicate that the purity of the centrate needs to be increased. A low centrate purity may be understood as a centrate with higher turbidity or more sediment than a centrate with higher purity. The centrate purity quality value is considered high when the purity of the centrate is preferably within the range of 50 to 299.99 NTU, more preferably within the range of 100 to 250 NTU, and most preferably within the range of 150 to 200 NTU. On the other hand, the centrate purity quality value is considered low when the purity of the centrate is preferably within the range of 300 to 1000 NTU, more preferably within the range of 500 to 800 NTU, and most preferably within the range of 600 to 700 NTU.

[0031] A high polymer dosage quality value may indicate that the polymer dosage is acceptable or may be slightly reduced. A low polymer dosage may be understood as a smaller amount of polymer being delivered to the sludge in the decanter centrifuge compared to a high polymer dosage. The polymer dosage quality value is considered high when the polymer dosage is preferably within the range of 2 to 9.99 kg / tDS, more preferably within the range of 4 to 8 kg / tDS, and most preferably within the range of 6 to 7 kg / tDS. On the other hand, the polymer dosage quality value is considered low when the polymer dosage is preferably within the range of 10 to 20 kg / tDS, more preferably within the range of 12 to 18 kg / tDS, and most preferably within the range of 14 to 16 kg / tDS.

[0032] Providing the respective quality values ​​for cake dryness, centrate purity, and polymer dosage facilitates quantitative estimation of the degree of optimization of material parameters related to decanter output. The quantitative measures can then be provided to an enhanced AI engine, which can then perform predictive adaptation of the multiple operating parameters so that the resulting multiple material parameters converge to their respective optimal values.

[0033] The step of predicting the plurality of adjusted operating parameters by the enhanced AI engine may further include the steps of determining a total energy consumption of the decanter based on the plurality of operating parameters, and determining optimal values ​​for each of the decanter scroll differential speed, the decanter bowl speed, the sludge feed flow rate, and the polymer feed flow rate, taking into account the balance between the dryness of the cake, the purity of the centrate, the total energy consumption of the decanter, and the amount of polymer added in the centrate.

[0034] The balance between cake dryness, centrate purity, total energy consumption, and polymer addition can be understood as the interrelationship of these parameters: as an example, changing cake dryness may necessarily change centrate purity, total decanter energy consumption, and polymer addition.

[0035] Predicting multiple adjusted operating parameters may further contribute to convergence of one or more material parameters to an optimal setting.

[0036] The step of determining the optimum value may further include, if the cake dryness quality value is high, decreasing the optimum value for the polymer feed flow rate, decreasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or increasing the optimum value for the sludge feed flow rate. The step of determining may also include, if the cake dryness quality value is low, increasing the optimum value for the polymer feed flow rate, increasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or decreasing the optimum value for the sludge feed flow rate.

[0037] If the cake dryness quality value is essentially high, at least two of the above-mentioned operating parameters of the decanter centrifuge may be adapted to reduce the cake dryness.

[0038] A reduced polymer feed rate may reduce the flocculation of the sludge injected into the decanter bowl. This results in less flocculation of the sludge particles and a lighter weight of sediment in the sludge (considering the original polymer feed rate). This reduces the degree of phase / sediment separation in the sludge. This results in a less dry cake and a higher cake moisture content compared to the exemplary situation with a higher polymer feed rate.

[0039] Decreasing the optimum differential velocity may result in a longer retention time for the separated phase / precipitate. Longer retention time allows more time for phase / precipitate separation to occur. This may result in a higher cake dryness, which may meet a higher cake dryness quality value.

[0040] As the decanter bowl speed decreases, the decanter differential speed may also decrease.

[0041] Increasing the sludge feed rate can shorten the time the sludge passes through the decanter bowl, resulting in less time for centrifugal force to act on the sediment in the sludge. This reduces the efficiency of phase / sediment separation. As a result, the cake may have a higher moisture content than would be achieved with a higher sludge feed rate.

[0042] The adjustment of at least two operating parameters of the decanter centrifuge described above can also be applied conversely to increase the dryness of the cake.

[0043] Adjustment of at least two operating parameters of the decanter bowl described above facilitates optimization of the material parameters to desired optimum values.

[0044] The determining of the optimum value may further include decreasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed and / or increasing the optimum value for the sludge feed flow rate when the centrate purity quality value is high, and the determining of the optimum value may further include increasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed and / or decreasing the optimum value for the sludge feed flow rate when the centrate purity quality value is low.

[0045] As mentioned above, reducing the decanter differential velocity increases the retention time of the phase / sediment that separates from the sludge liquid, thus increasing the degree to which the sediment separates from the liquid and potentially increasing the purity of the centrate.

[0046] Decreasing the optimum differential velocity may result in a longer retention time for the separated phase / precipitate. Longer retention time allows more time for phase / precipitate separation to occur and may result in a purer centrate.

[0047] Increasing the sludge feed rate can shorten the time it takes for the sludge to pass through the decanter bowl, which reduces the time that centrifugal force acts on sediments in the sludge, which can reduce the purity of the centrate.

[0048] Adjustment of one or more operating parameters of the decanter centrifuge described above can further assist in convergence of one or more material parameters to their respective optimum values.

[0049] The adjustment of at least two of the operating parameters described above can also be applied conversely to reverse the purity quality of the centrate.

[0050] The step of determining the optimum value may further include, if the polymer dosage quality value is high, increasing the optimum value for the polymer feed rate, and / or increasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or decreasing the optimum value for the sludge feed rate. The step of determining the optimum value may further include, if the polymer dosage quality value is low, decreasing the optimum value for the polymer feed rate, and / or decreasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or increasing the optimum value for the sludge feed rate.

[0051] Adjustment of one or more operating parameters of the decanter centrifuge described above can advantageously assist in the convergence of one or more material parameters to their respective optimum values.

[0052] The step of predicting the plurality of adjusted operating parameters by the enhanced AI engine may further include deriving the plurality of adjusted operating parameters based on optimal values ​​for each of the decanter scroll differential speed, the decanter bowl speed, the sludge feed flow rate, and the polymer feed flow rate.

[0053] In some cases, adjustment of an operating parameter may affect two or more material parameters due to their interrelationships.

[0054] In some scenarios, multiple operating parameters are derived based on two, three, or all four of the optimum values ​​for each of the decanter scroll differential speed, decanter bowl speed, sludge feed rate, and polymer feed rate.

[0055] The prediction may be based at least in part on quality values ​​assigned to one or more of the plurality of material parameters.

[0056] By deriving multiple adjusted operating parameters for one or more operating parameters, interrelationships between material parameters can be taken into account, improving the optimization of the material parameters (e.g., not limited to only individual material parameters, as ideally material parameters would all be optimized to converge toward a desired optimal value).

[0057] The computer-implemented method may further include adjusting, by the agent, the enhanced AI engine based on the plurality of material parameters and the plurality of operating parameters.

[0058] The agent may be adapted to determine, based at least in part on the one or more quality values, whether to increase or decrease a particular one of the one or more operating parameters of the decanter to ensure that the decanter is operating such that the one or more material parameters are at a desired optimum value.

[0059] The agent may be configured to adjust the reinforcement AI engines to maximize the sum of the assigned quality values.

[0060] By using an agent to adjust the enhanced AI engine, the AI ​​engine can be trained while the decanter is in operation. Thus, predictions of one or more operating parameters of the decanter can be continuously further improved while the decanter is in operation. That is, the AI ​​engine can be further trained while the decanter is in operation, thereby making it easier to adapt to the actual operating conditions of the decanter, and, for example, can be limited to training based only on simulations of the decanter. This can at least contribute to the decanter accurately operating with multiple optimized operating parameters.

[0061] The computer-implemented method may further include performing steps bg above multiple times.

[0062] Steps b-g may be performed continuously, repeatedly, and autonomously in a closed loop. Alternatively, steps b-g may be repeated once per minute, once per hour, or once per day. Additionally or alternatively, steps b-g may be performed only in response to user input.

[0063] Performing the above steps multiple times facilitates closed-loop operation of the decanter, allowing the decanter to be stably operated at multiple optimized operating parameters to ensure optimized decanter output.

[0064] Another aspect of the invention relates to a computer-implemented method for training an enhanced artificial intelligence (AI) engine usable for operating a decanter centrifuge. The method may include: a. simulating operation of the decanter centrifuge to generate a plurality of operating parameters; b. determining a plurality of material parameters for the simulated decanter centrifuge. The method may also include: c. passing the plurality of operating parameters and the plurality of material parameters to the enhanced AI engine; and d. tuning, by an agent, the enhanced AI engine based on the plurality of material parameters and the plurality of operating parameters.

[0065] The operating parameters obtained from the simulation may be the same as the operating parameters described above with respect to the computer-implemented method for operating a decanter. The operating parameters obtained from the simulation may be understood as training data for an enhanced AI engine.

[0066] The material parameters determined from the simulated decanter may be the same as the material parameters described above with respect to the computer-implemented method for operating a decanter. The material parameters obtained from the simulation may be used as training data for an enhanced AI engine.

[0067] In some cases, only simulated data may be passed to the enhanced AI engine to train it, however, data (e.g., multiple operating parameters) obtained from one or more physically operated decanter centrifuges may additionally or alternatively be passed to the enhanced AI engine for training.

[0068] The computer-implemented method for training the enhanced AI engine described above facilitates, for example, improving the operation of a decanter centrifuge using the computer-implemented method. This can be done at least in part based on pairs of simulation data of multiple operating parameters and multiple material parameters (related to the operating parameters). Because training data can be obtained from simulations, it is generally easier to provide the enhanced AI engine with larger training data sets in a more cost-effective manner than is possible with training data obtained from physically operated decanter centrifuges. This allows for more accurate and faster training of the enhanced AI engine and, based thereon, more accurate and reliable predictions of multiple operating parameters by the enhanced AI engine when the trained enhanced AI engine is used to operate a decanter centrifuge. This can assist in closing the decanter centrifuge at desired optimized operating parameters.

[0069] Another aspect of the invention relates to an enhanced artificial intelligence (AI) engine that can be trained according to the above-described computer-implemented method for training an enhanced AI engine. The enhanced AI engine can be used in the method as described above.

[0070] The enhanced AI engine can be located in one or more decanter centrifuges (e.g., in their respective processing units). Additionally or alternatively, the AI ​​engine may be located in a control unit that communicates with one or more decanters. In some applications, the AI ​​engine may be located in a remote entity (e.g., a server (cloud), etc.).

[0071] The AI ​​engine can be adapted to be used to control the operation of multiple decanter centrifuges.

[0072] By providing the above-described AI engine, it is possible to provide an AI engine for optimizing decanter operation that is detached from a specific decanter centrifuge and adaptable to different decanter centrifuges. Therefore, it is possible to provide highly versatile AI-based control for decanter centrifuge operation. Furthermore, if the AI ​​engine is in communication with multiple decanters, the AI ​​engine can receive training data from multiple decanter centrifuges. This increases the amount of training data, thereby improving the accuracy of AI-based predictions for multiple centrifuge operating parameters and further improving decanter centrifuge operation.

[0073] Another aspect of the invention relates to an apparatus that may include means for carrying out the above-described method.

[0074] In some cases, the device may be a physically separate entity from the decanter. Alternatively, the device may be built into the decanter. The device may be a computer. In some cases, the device may be part of the decanter (e.g., a microcontroller, integrated circuit (IC), etc.).

[0075] In some cases, the device may be a remote entity, which may be adapted to communicate with the decanter, for example by means of a network connection.

[0076] The device may be adapted to control and / or train only a single decanter, or the device may be adapted to control and / or train multiple decanters, and in some cases the device may be adapted to control a grid of multiple decanters.

[0077] By providing an apparatus comprising means for carrying out the above-described method, a control unit adapted to control and / or train an enhanced AI engine can be readily provided.

[0078] Another aspect of the invention relates to a computer program that may include instructions that, when executed by a processing system, cause the processing system to perform the above-described method.

[0079] The processing system may be located in the decanter, or may be located in an entity physically separate from the decanter, in which case the processing system may be a remote entity (e.g., a remote server operating a cloud system) that communicates with the decanter.

[0080] Another aspect of the present invention relates to a computer-implemented method for optimizing the output of an operating decanter centrifuge using a supervised artificial intelligence (AI) engine. The method may include one or more of: (a) operating the decanter centrifuge according to a plurality of operating parameters; (b) processing a physical input including sludge and polymers through the decanter to produce a physical output including centrate and cake; (c) determining a plurality of material parameters based on the physical output; and (d) passing the plurality of material parameters and the plurality of operating parameters to a supervised learning AI engine. The method may further include one or more of: (e) determining a quality value for each of the plurality of material parameters by the supervised learning AI engine; (f) predicting a plurality of adjusted operating parameters by the supervised learning AI engine; and (g) further operating the decanter based on the plurality of adjusted operating parameters.

[0081] The AI-based operation of Decanter based on a supervised learning AI engine can be considered as a second option (other than implementing an enhanced AI engine) to improve the AI-based operation of Decanter.

[0082] The plurality of operating parameters and the plurality of adjusted operating parameters may include one or more of a decanter scroll differential speed, a decanter bowl speed, a sludge feed rate, and / or a polymer feed rate.

[0083] The step of determining the plurality of material parameters may further include determining, by one or more sensors, the dryness of the cake, the purity of the centrate, and the loading of polymer in the centrate.

[0084] The step of determining a quality value for each of the plurality of material parameters by the supervised learning AI engine may further include the steps of: setting the cake dryness quality value to be low when the cake dryness is in a range of 10 to 19.99% dry matter equivalent, and setting the cake dryness quality value to be high when the cake dryness is in a range of 20 to 35% DS; setting the centrate purity quality value to be low when the purity of the centrate is in a range of 300 to 1000 nephelometric turbidity units, and setting the centrate purity quality value to be high when the purity of the centrate is in a range of 50 to 299.99 NTU; and setting the polymer dosage quality value to be high when the amount of polymer added in the centrate is in a range of 2 to 9.99 kg / ton dry matter equivalent (tDS), and setting the polymer dosage quality value to be low when the amount of polymer added in the centrate is in a range of 10 to 20 kg / tDS.

[0085] The step of predicting the plurality of adjusted operating parameters by the supervised learning AI engine may further include the steps of determining a total energy consumption of the decanter based on the plurality of operating parameters, and determining optimal values ​​for each of the decanter scroll differential speed, the decanter bowl speed, the sludge feed flow rate, and the polymer feed flow rate, taking into account the balance between the dryness of the cake, the purity of the centrate, the total energy consumption of the decanter, and the amount of polymer added in the centrate.

[0086] The step of determining the optimum value may further include the steps of: decreasing the optimum value for the polymer feed flow rate, decreasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or increasing the optimum value for the sludge feed flow rate when the cake dryness quality value is high; and increasing the optimum value for the polymer feed flow rate, increasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or decreasing the optimum value for the sludge feed flow rate when the cake dryness quality value is low.

[0087] The step of determining the optimum value may further include the steps of decreasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed and / or increasing the optimum value for the sludge feed flow rate when the centrate purity quality value is high, and increasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed and / or decreasing the optimum value for the sludge feed flow rate when the centrate purity quality value is low.

[0088] The step of determining the optimum value may further include the steps of increasing the optimum value for the polymer feed rate and / or increasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed and / or decreasing the optimum value for the sludge feed rate when the polymer dosage quality value is high, and decreasing the optimum value for the polymer feed rate and / or decreasing at least one of the optimum values ​​for the decanter scroll differential speed and the decanter bowl speed and / or increasing the optimum value for the sludge feed rate when the polymer dosage quality value is low.

[0089] The step of predicting the plurality of adjusted operating parameters by the supervised learning AI engine may further include deriving the plurality of adjusted operating parameters based on optimal values ​​for each of the decanter scroll differential speed, the decanter bowl speed, the sludge feed rate, and the polymer feed rate.

[0090] In some cases, the above steps b to g may be performed multiple times.

[0091] Another aspect of the invention relates to a computer-implemented method for training a supervised learning artificial intelligence (AI) engine that can be used to operate a decanter, which may include: a. simulating operation of the decanter to generate a plurality of operating parameters, b. determining a plurality of material parameters for the simulated decanter, c. passing the plurality of operating parameters and the plurality of material parameters to the supervised learning AI engine, and d. tuning the supervised learning AI engine based on the plurality of material parameters and the plurality of operating parameters.

[0092] Another aspect of the invention relates to a supervised learning artificial intelligence (AI) engine trained according to and usable in any of the above methods.

[0093] Another aspect of the invention relates to an apparatus which may include means for carrying out any of the methods described above.

[0094] Another aspect of the invention relates to a computer program that may include instructions that, when executed by a processing system, cause the processing system to perform any of the methods described above.

[0095] Aspects of the present invention will now be described in more detail with reference to the accompanying drawings. [Brief explanation of the drawings]

[0096] [Figure 1] FIG. 1 is a diagram illustrating an example of a decanter centrifuge. [Figure 2] FIG. 1 illustrates a decanter and exemplary inputs and outputs. [Figure 3A] FIG. 1 illustrates a decanter and potential sensors used to monitor inputs and outputs associated with the decanter. [Figure 3B] FIG. 1 illustrates a decanter and potential sensors used to monitor inputs and outputs associated with the decanter. [Figure 3C] FIG. 1 illustrates a decanter and potential sensors used to monitor inputs and outputs associated with the decanter. [Figure 4A] This is an explanatory diagram of the training and prediction cycle of the AI ​​engine for decanter operation. [Figure 4B] This is an explanatory diagram of the training and prediction cycle of the AI ​​engine for decanter operation. [Figure 5A] This is an explanatory diagram of reinforcement training for the AI ​​engine for decanter operation. [Figure 5B] This is an explanatory diagram of reinforcement training for the AI ​​engine for decanter operation. [Figure 6] FIG. 1 shows the performance over time of a decanter operated in different operating modes. [Figure 7A] Figure 1 shows the performance change over time when the decanter is operated manually and when it is operated based on a trained AI engine (supervised learning). [Figure 7B] Figure 1 shows the performance change over time when the decanter is operated manually and when it is operated based on a trained AI engine (supervised learning). [Figure 8] This figure shows the performance change over time when the decanter is operated manually and when it is operated based on a trained AI engine (reinforcement learning). [Figure 9] FIG. 1 illustrates an exemplary supervised learning infrastructure. [Figure 10] This is an example of training data that can be used to train the AI ​​engine. DETAILED DESCRIPTION OF THE INVENTION

[0097] Hereinafter, the embodiments and modifications of the present invention will be described in further detail.

[0098] FIG. 1 illustrates an exemplary decanter centrifuge 100 according to an embodiment of the present invention.

[0099] The exemplary decanter centrifuge 100 comprises a decanter bowl 110 having an inner volume 110a and an outer volume 110b (the latter surrounding the former). The inner volume 110a can be filled with sludge through an inlet 120.

[0100] Rotation of decanter bowl 110 about its longitudinal axis, driven by motor M, separates sediment from the liquid in the sludge. The sediment is transported radially along decanter bowl 110 toward the inner wall of inner volume 110a of decanter bowl 110. The separated sediment is discharged from inner volume 110a to outer volume 110b via outlet 130. Inner volume 110a is further surrounded by scroll 140, which is spirally wound around the outer wall of inner volume 110a. Scroll 140 is adapted to transport the sediment that leaves inner volume 110a via outlet 130 from decanter 100 toward outlet 150, which discards the sediment. The liquid in inner volume 110a can be discharged from decanter bowl 110 via centrate outlet 160.

[0101] More specifically, decanter bowl 110 may have a generally cylindrical volume (e.g., having a longer extension along the length of the decanter than a radial extension perpendicular to the length). The generally cylindrical volume may taper toward at least one end (e.g., toward the right side of decanter 100 as shown in FIG. 1). Decanter bowl 110 may have an inner volume 110a and an outer volume 110b. Inner volume 110a may be adapted to contain sludge to be treated by decanter 100. Outer volume 110b may be adapted to surround and contain inner volume 110a.

[0102] Decanter bowl 110 may be adapted to be rotatable about its longitudinal axis at a bowl speed (e.g., rotational speed in revolutions per minute (RPM)) driven by motor M. Rotation of decanter bowl 110 may generate a centrifugal force that acts on liquid and sediment within interior volume 110a of decanter bowl 110. Centrifugal force F Z The larger the mass m of different sediments in the sludge, the stronger the centrifugal force F Z receive.

number

[0103] Decanter bowl 110 is in fluid communication with inlet 120. Inlet 120 may be adapted as a feedthrough for sludge into interior volume 110a of decanter bowl 110. In some exemplary embodiments, inlet 120 includes a valve to control the sludge feed rate (e.g., when the valve is at least partially closed, the sludge feed rate is reduced, and vice versa).

[0104] The wall surrounding the interior volume 110a of the decanter bowl 110 may further include a sediment outlet 130 that provides fluid communication between the interior volume 110a of the decanter bowl 110 and the exterior volume 110b of the decanter bowl 110. As sediments separate from the sludge liquid, the separated sediments accumulate on the interior walls of the interior volume 110a of the decanter bowl 110, as described above. These sediments flow from the interior volume 110a of the decanter bowl 110 to the exterior volume 110b via the outlet 130.

[0105] Inner volume 110a may be surrounded by scroll 140. Scroll 140 may be disposed within outer volume 110b of decanter bowl 110 and rotatably relative to inner volume 110a. Scroll 140 may include a blade that spirals relative to inner volume 110a along the longitudinal direction of decanter bowl 110. The height of the blade may be equal to the radial distance between the outer wall of inner volume 110a and the inner wall of outer volume 110b. Scroll 140 may be adapted to rotate in the same direction relative to decanter bowl 110. In other words, if decanter bowl 110 is adapted to rotate clockwise, scroll 140 is adapted to rotate clockwise, and vice versa.

[0106] The scroll 140 can be adapted to transport sediment that leaves the interior volume 110a of the decanter bowl 110 towards the inlet 120 to the outlet 150, from which the separated sediment exits the decanter 100 and is transported to final waste disposal.

[0107] At the other end of the decanter 100 (relative to its longitudinal extension) opposite the inlet 120, the internal volume 110a and the decanter 100 itself may be provided with a centrate outlet 160. The sediment contained in the initially injected sludge is separated from the liquid in the sludge, so that the final treated sludge consists mainly of liquid, which can be discharged from the internal volume 110a and the decanter 100 as essentially clean liquid via the centrate outlet 160.

[0108] FIG. 2 illustrates a decanter 200 (which may be identical to the decanter 100 described with reference to FIG. 1 above) and exemplary inputs relating to input polymer 210a and input polymer 210b fed to decanter 200, as well as exemplary outputs 220a and 220b obtained as outputs from decanter 200.

[0109] More specifically, the input 210a may be sludge as described above with reference to Figure 1A. The input 210a may be a sludge having a solids content (e.g., the amount of solid particles in the fluid, expressed in units of %DS or kg / h) and a density (e.g., mass per volume (e.g., g / cm 3 , kg / m 3 , t / m 3 It can be parameterized by the mass of the input (expressed in units of ).

[0110] The polymer input 210b can be parameterized by the amount of polymer added (e.g., in kg / tDS or l / h) and the amount of water with which the polymer is input into the decanter, the latter allowing for easy dilution of the polymer, thereby allowing for precise adaptation of the polymer concentration in the decanter according to the situational needs (e.g., desired material parameters).

[0111] Decanter 200 is provided with a feed 210a and a feed polymer 210b, and the feed is processed within decanter 200 to produce a decanter output. The decanter output includes centrate 220a and cake 220b.

[0112] The centrate 220a can be understood as the residue of the sludge (which was put into the decanter 200) after the sediment contained in the sludge has been separated from the liquid contained in the sludge. In other words, the centrate 220a is a washed liquid that typically contains only a small amount of sediment compared to the amount of sediment originally contained in the sludge.

[0113] The centrate quality can be parameterized in NTU, which relates to the quality / purity of the liquid after processing in the decanter 200. Furthermore, the centrate 220a discharged from the decanter 200 can preferably be recycled to the environment as clean or purified liquid.

[0114] The cake 200b can be understood as a sediment originally contained in the sludge as solid particles and separated from the liquid contained in the sludge. The cake 220b can be parameterized by its solids content (e.g., in DS% units). The cake 200b can be understood as a waste product resulting from the clarification of the sludge by the decanter 200b and can be discarded (i.e., no further use is intended within the scope of the present invention).

[0115] The above-described inputs and outputs are provided regardless of whether decanter 200 is operated manually or based on an AI engine. If decanter 200 is operated manually, an operator is required to collect centrate and / or cake samples, determine their quality and dryness, respectively, and infer potential operating settings for the decanter so that the output of decanter 200 (particularly the respective material parameters) is optimized.

[0116] On the other hand, if decanter 200 is controlled by an AI engine, decanter 200 is controlled as described elsewhere herein. The AI ​​engine is provided with multiple parameters related to decanter 200, such as one or more of the differential speed of scroll 140 of decanter 200, the speed of decanter bowl 110, the torque of scroll 140 of decanter 200, information regarding the vibration of the liquid side of decanter 200, information regarding the vibration of the solid side of decanter 200, current information for the main motor M of decanter 200, current information for the secondary motor of decanter 200 used to submerge scroll 140, and / or bearing temperature (bearings may be located between motor M and decanter bowl 110) to predict one or more potential operating parameters of decanter 200.

[0117] 3A-3C show an exemplary arrangement of sensors placed at or near a decanter (FIG. 3A) and exemplary sensors used to determine multiple material parameters and / or multiple operating parameters (FIGS. 3B and 3C).

[0118] The sensors may be configured to provide quantitative information regarding multiple material parameters and / or multiple operating parameters, as described elsewhere herein.

[0119] More specifically, Figure 3A illustrates a decanter 300 (which may be identical to decanter 100 or 200 described with reference to Figures 1 and 2 above) and exemplary locations of one or more sensors for providing, for example, the above-described quantitative information regarding multiple materials and inputs. In particular, the exemplary decanter 100 or 200 includes a feed control unit 310, a centrate monitor unit 320, and / or a cake monitor unit 330.

[0120] Decanter 300 may include a feed control 310 located at or near inlet 120, as described with reference to FIG. 1 above. Feed control 310 may include a valve for controlling the flux of sludge into decanter bowl 110 (as described with reference to FIG. 1 above). Feed control 310 may control the sludge throughput (e.g., l / h, m 3 / h, in kg / h).

[0121] The sensor adapted to monitor the supply may be an infrared duo-scattered light sensor. In this case, a light source (e.g., a laser light source) is positioned on one side of the tube of the decanter 300 to be monitored, and a corresponding light sensor is positioned, for example, on the opposite side of the tube of the decanter 300. Light emitted from the light source impinges on one or more sediments in the injected sludge and is scattered from the one or more sediments therein. The corresponding light sensor is adapted to detect the light scattered from the sediments. This makes it easy to derive, for example, the amount and size distribution of sediment particles in the injected sludge. This method of detecting the injected sludge is called turbidity measurement in accordance with DIN EN ISO 7027 and DIN 38414.

[0122] Additionally or alternatively, decanter 300 may include a centrate monitor 320. The centrate monitor 320 may include one or more sensors located near decanter 300, in the lines carrying the liquid being processed in decanter 300, within decanter 300, or in a chute connected to decanter 300. Monitoring of centrate quality may be based on an infrared duo-scattered light sensor, as described above with reference to supply control unit 310.

[0123] Additionally or alternatively, centrate quality can be measured based on a transmission inspection. This method also uses a light source (e.g., a laser light source) located on one side of the tube and a corresponding light sensor located on the opposite diametric side of the tube, adapted to transmit light emitted from the light source through the tube of the decanter 300 in a manner similar to a light barrier and monitor the intensity of the light reaching the light sensor. Any residual particles overlapping the light barrier may temporarily reduce the intensity of the light detected by the corresponding light sensor. This allows the centrate quality (i.e., the amount of residual sediment) to be derived from the difference between the intensity of the light emitted by the light source and the intensity of the light detected by the corresponding light sensor.

[0124] The decanter 300 may further comprise a cake monitor 330. The cake monitor 330 may include, for example, a sensor for determining the moisture content of the cake (e.g., the dryness of the cake). The measurement principle may be based on time domain reflectometry. Time domain reflectometry is based on the idea of ​​inserting one or more electrodes into the cake, sending an electric pulse along the one or more electrodes, and causing it to reflect off the edges of the one or more electrodes. Since the propagation speed of the pulse along the one or more electrodes depends on the moisture content (because the dielectric constant of the cake changes depending on the moisture content), the moisture content of the cake surrounding the one or more electrodes can be determined.

[0125] 3B shows an exemplary sensor 340 that can be used to sense the amount of sediment in the sludge being added to the decanter bowl 110 through the inlet 120, which may be based on the infrared duo-scattering light concept as described above with reference to FIG. 3A. The sensor 340 may include a light source 340a and a corresponding light sensor 340b.

[0126] 3C shows an exemplary sensor 350 for measuring moisture content in a cake. The sensor 350 may include a first electrode 350a through which an electric pulse propagates (up and down in the illustrated illustration). The electric pulse is reflected off the edge of the first electrode 350a. As a result of the reflection, the electric pulse propagates along a second electrode 350b. Because the propagation speed depends on the moisture in the surrounding environment, the moisture content can be derived from the length of time required for the electric pulse to propagate along the first electrode 350a and the second electrode 350b.

[0127] Some of the sensors described above may be located only in the decanter 300, and additionally or alternatively, some sensors may be located in the control cabinet.

[0128] Additional sensors (not shown) used in the decanter may include a magnetic induction sensor (MID) configured to measure the polymer concentration in the decanter, and / or a variable frequency converter (VFC) adapted to measure the torque of the scroll and / or motor.

[0129] FIG. 4A exemplarily illustrates a basic training concept of an AI engine according to an embodiment of the present invention, and FIG. 4B illustrates an operating concept based on the AI ​​engine of a decanter according to an embodiment of the present invention (FIG. 4B).

[0130] FIG. 4A exemplarily illustrates an iterative training method 400 for an AI engine (e.g., represented as a trained model). The training method 400 can be understood as a looping process that begins by providing input data 420 to an AI engine 410, which is trained based on the concept of supervised learning. The input data 420 may include a labeled training dataset. The labeled training dataset includes one or more parameters for controlling a decanter (e.g., decanter bowl speed), and the labels may be output data (e.g., cake dryness) corresponding to the decanter output obtained when the decanter is operated according to one or more parameters of the labeled training dataset. After the training cycle, in process step 440, the output data 430 is compared with the labels of the training dataset. The results of this comparison are provided to the AI ​​engine 410 being trained. If the comparison indicates that the output data corresponds to the labels, no adjustments to the AI ​​engine 410 may be made. On the other hand, if the results of the comparison indicate that the output data 430 deviate from the labels, the model is adjusted with the goal of producing output data 430 that corresponds to the labels of a further training data set used for training in a subsequent training cycle. As indicated by the dashed arrow, the training data may be modified or augmented with a further training data set, for example, by an operator during training, to improve the training.

[0131] The training data may be based on a pre-prepared data set (e.g., based on data obtained by simulating a decanter and / or based on data collected from one or more physically operated decanters), and labels for each set of training data may be provided (by a human operator).

[0132] The training data may include label data, which can be understood as a number of known material parameters obtained by a number of known operating parameters, the labeling being performed, for example, by the operator of the decanter.

[0133] In some cases, the input data 420 may be pre-processed (eg, the input data 420 may be scaled, for example, by a robust scaler).

[0134] In one example, the training data includes a data set of one or more operating parameters used to operate the decanter and one or more corresponding material parameters resulting from setting the decanter to the one or more operating parameters. The input data 420 may further include a label (e.g., assigned to one or more of the plurality of operating parameters and / or plurality of material parameters) indicating whether a particular operating parameter is considered positive (e.g., the respective operating parameter may result in a change in one or more material parameters to a desired optimum value) or negative (e.g., the respective operating parameter does not result in a change in one or more material parameters to a desired optimum value).

[0135] Training of the AI ​​engine 410 may be based on adjusting an internal training model of the AI ​​engine 410 such that a mathematical mapping between the plurality of operating parameters and the plurality of material parameters is achieved, i.e., such that the trained AI engine 410 is able to predict one or more operating parameters that may realize one or more of the plurality of material parameters of the decanter. This may be achieved, for example, by minimizing an error function.

[0136] Successful training enables the AI ​​engine 410 to derive a plurality of operating parameters for the decanter such that the plurality of material parameters are optimized by providing one or more of the plurality of material parameters to the AI ​​engine 410. If the value of the calculated error function is less than a predefined threshold, it can be assumed that the training was successful.

[0137] The above-described supervised learning concept in the AI ​​engine 410 may further include maximizing arbitrarily assigned feedback values ​​(e.g., where feedback values ​​ranging from 0 (worst) to 1 (best) are assigned, similar to quality values ​​in reinforcement learning, as described further below with reference to FIGS. 5A and 5B). In some embodiments, the assigned feedback values ​​include an aggregation of individual quality values ​​assigned for cake dryness, centrate quality, and polymer addition amount. This aggregation can be understood as a quantitative measure for determining how closely the decanter is operating at its optimized settings. This allows monitoring the running costs of the decanter, such as the total cost of ownership (TCO), among other things. In general, to obtain an accurate mapping of the current decanter operation, it is beneficial to collect as many parameter values ​​as possible for multiple material parameters and multiple operating parameters. An accurate mapping of the decanter can advantageously contribute to the cost-efficient operation of the decanter.

[0138] After providing the decanter with one or more operating parameters (output data 430), feedback 440 may be determined to indicate whether the predicted one or more operating parameters have led to an improvement in one or more material parameters (i.e., whether the current values ​​of the one or more material parameters are close to the desired optimal values ​​of the one or more material parameters, such as in a previous iteration step). In some cases, the predicted output data 430 may not be used directly to operate the decanter for training purposes, but may be used to manually assign feedback 440, for example, by an operator of the decanter. In some cases, the feedback is provided as an error function. The error function may be configured to calculate, for one or more of the plurality of operating parameters, an error between one or more of the plurality of material parameters included in the training data set (assigned to the corresponding one or more of the plurality of operating parameters) and one or more of the predicted plurality of material parameters that would be obtained if the decanter were operated according to the predicted one or more of the predicted plurality of operating parameters. Training may be configured to minimize the error function, also commonly referred to as an error or error metric. The error function may be based on, for example, mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), Hubert loss, other suitable measures, and / or combinations thereof. The AI ​​Engine model may be based on random forest regression alone or in combination with adaptive boosted regression, k-nearest neighbor regression, Gaussian naive Bayes, or logistic regression. Stacking regression may be required when multiple regressions are used.

[0139] The predicted output 430 is fed back to the AI ​​engine 410 being trained, and the determined feedback 440 is fed back to the AI ​​engine 410, optionally in combination with the predicted output 430, to further train the AI ​​engine 410 in subsequent iteration steps of the training method 400.

[0140] In some embodiments, a data pair is defined that includes a parameter from the plurality of material parameters and a parameter from the plurality of operating parameters. Such a data pair, also referred to as a regressor, can be trained according to the supervised learning method described above. Once a first regressor is successfully trained, a second regressor is similarly trained. Once at least two regressors have been trained, a combination of the at least two regressors is trained.

[0141] As an example, supervised learning in AI engine 410 obtains one or more of a plurality of operating parameters (e.g., included in input data set 420) including decanter differential speed, decanter bowl speed, decanter scroll torque, liquid side vibration, solid side vibration, main motor current, secondary motor current, bearing temperature, feed flow rate, feed dry matter and / or polymer flow rate. As label data (a plurality of material parameters), one or more of cake dryness, polymer loading, and / or centrate quality are provided to AI engine 410.

[0142] During training, the AI ​​engine 410 can apply its internal calculations to facilitate mapping of the operating parameters. In other words, when one or more sets of operating parameters are provided to the AI ​​engine 410, the AI ​​engine 410 can reproduce the associated material parameters. During training, the AI ​​engine 410 predicts certain material parameters (output data 430) and sends them to feedback 440, which calculates the current deviation between the predicted values ​​and the labeled data. The feedback 440 is provided to the AI ​​engine 410. In subsequent iterations of the training method 400, the AI ​​engine 410 predicts subsequent material parameters (output data 430), which are again evaluated by the feedback 440. This process is repeated until the difference between the predicted material parameters and the labeled data falls below a predefined threshold.

[0143] 4B exemplarily illustrates a method 450 for operating a decanter based on a trained AI engine 470. This method is preferably performed iteratively. This method may be identical to the training method 400 described with reference to FIG. 4A above, except that the AI ​​engine 470 may, in some cases, not be further trained and may only be used to predict output data 480 (although further training of the AI ​​engine 470 is not precluded).

[0144] The operating method 450 may be based on an input data set 460, which preferably includes a plurality of material parameters obtained by analysis of the decanter output (e.g., by the cake monitor 330 and / or the centrate monitor 320). Additionally, the input data set 460 may include a plurality of operating parameters (output data 480).

[0145] The input dataset 460 is optimized by the trained AI engine 470. The trained AI engine 470 (e.g., a trained AI engine such as that described with reference to FIG. 4A above, or any other suitable AI engine, or any suitable AI engine) is provided with the input dataset 470 and, based on its training, predicts the output dataset 480. In some cases, the input dataset 460 may be preprocessed before being provided to the trained AI engine 470.

[0146] The output dataset 480 may include a plurality of operating parameters provided to the decanter. When the decanter is operated according to the plurality of operating parameters, the decanter generates an output (e.g., centrate 220a and / or cake 220b and associated plurality of material parameters) that is analyzed to determine whether the output dataset 480 predicted by the trained AI engine 470 resulted in optimization of the plurality of material parameters.

[0147] The output produced by the decanter when operated according to the predicted output data 480 may result in a change in the input data 490. This new input data 490 may be used as input data 460 and provided to the trained AI engine 470 to obtain new output data 480 in a subsequent iteration step.

[0148] 5A and 5B show an illustration of another exemplary embodiment of a method 500 for training an AI engine 510 based on reinforcement learning according to an aspect of the present invention (FIG. 5A), and a detailed illustration of an exemplary agent interacting with an environment as part of the reinforcement learning concept shown in FIG. 5A (FIG. 5B).

[0149] The training method 500 shown in FIG. 5A may be implemented generally similarly to the (supervised learning) training method 400 described with reference to FIG. 4A above, except that the actual training operations for training the AI ​​engine 510 are based on reinforcement learning, which is described in more detail below.

[0150] 5A illustrates an enhanced training method 500. The training method 500 may include an input data set 520 containing training data. The training data may include a plurality of operating parameters of the decanter and, optionally, a plurality of material parameters associated with the plurality of operating parameters. The training data may also include one or more quality values ​​assigned to each of the plurality of operating parameters, as further described below.

[0151] An input dataset 520 is provided to the AI ​​engine 510 to be trained. Based on the input dataset 520, i.e., the plurality of operating parameters (and their respective quality values), the AI ​​engine 510 can adapt itself (e.g., based on a reinforcement learning algorithm) such that, when presented with one or more parameter values ​​of the plurality of material parameters, the AI ​​engine 510 predicts one or more parameter values ​​of the plurality of operating parameters (i.e., determines new plurality of operating parameters for the decanter) with the goal of ideally improving one or more parameters included in the plurality of material parameters. An improvement in this case can be understood as one or more parameter values ​​of the plurality of material parameters in a current iteration of the method 500 being closer to a desired optimal value for the one or more parameters compared to one or more parameter values ​​of the plurality of material parameters in a previous iteration step of the method 500. That is, the plurality of operating parameters are predicted such that, when the decanter is operated according to the predicted plurality of operating parameters, the plurality of material parameters will converge toward a desired optimal value. Training of the AI ​​engine 510 may include interaction with an agent, as further described below with reference to FIG. 5B .

[0152] Based on the initial training step of the AI ​​engine 510 and the input data 520 provided to the AI ​​engine 510, the AI ​​engine 510 can predict an output data set 530. The output data set 530 can include a plurality of operating parameters provided to the decanter, based on which the decanter can generate an output (as described above) from which new subsequent plurality of material parameters can be derived. The output data set 530 can include, for example, one or more of the differential speeds of the scrolls.

[0153] In subsequent method steps of method 500, the plurality of operating parameters may be assigned respective quality values ​​540 based on the obtained plurality of material parameters (i.e., in a preferred embodiment, each parameter value of the plurality of operating parameters is assigned a respective quality value 540 depending on one or more of the parameter values ​​of the plurality of material parameters).

[0154] The quality value may indicate whether one or more parameter values ​​of the plurality of operating parameters may be associated with an improvement of one or more parameter values ​​of the plurality of material parameters toward a respective desired optimized material parameter, or whether one or more parameter values ​​of the plurality of operating parameters may be associated with a deterioration of one or more parameter values ​​of the plurality of material parameters (e.g., whether the predicted plurality of operating parameters caused the plurality of material parameters to have values ​​further away from the respective optimized material parameter compared to a previous iteration of method 500), or evaluate one or more parameter values ​​of the plurality of material parameters as neutral (e.g., one or more parameter values ​​of the plurality of material parameters are considered neither an improvement nor a deterioration compared to a previous iteration of method 500). In some cases, the quality value may be passed to an agent of AI engine 510. The agent, further described below with reference to agent model 550, may be adapted to prioritize achieving a relatively high cake dryness (e.g., a cake dryness of 20% DS or greater). If optimized cake dryness is achieved, the AI ​​engine 510 attempts to optimize centrate quality (e.g., below 300 NTU), followed by keeping the polymer loading low (e.g., below 10 kg / tDS), with the goal of minimizing decanter energy consumption. From an energy consumption perspective (e.g., when energy consumption is a priority), decanter energy consumption can be reduced by 25-35%, preferably 30%, by reducing cake dryness by 0.1-1.2%, preferably 0.2-0.8%. Such cake dryness reduction can be achieved, for example, by reducing the decanter bowl speed.

[0155] The assigned quality value 540, and optionally the output data 530 and the plurality of material parameters obtained by providing the output data 530 to the decanter, can be integrated as input data 520 for a subsequent iteration step of the method 500. Based on this new subsequent input data set 520, a subsequent training step of the AI ​​engine 510 can be performed.

[0156] With reinforcement learning, the AI ​​engine 510 learns to maximize the sum of all assigned quality values, thereby becoming more efficient with respect to the assigned quality values. Furthermore, reinforcement learning offers greater flexibility with respect to the training data used (compared to supervised learning). For example, supervised learning requires data pairs of operating parameters and material parameters, whereas reinforcement learning may require only the operating parameters. This allows for further continuous training of the reinforcement AI engine 510, even while the decanter is in operation.

[0157] Further in this regard, the training data used to train the enhanced AI engine 510 may not require labeling.

[0158] In some cases, the AI ​​engine 510 includes a neural network including one or more layers and one or more nodes in each layer. The neural network may be a feedforward neural network, a convolutional neural network, or any combination thereof. Training the AI ​​engine 510 may include determining one or more weighting coefficients associated with the connection of nodes between two adjacent layers.

[0159] Additionally, training may be based on double deep Q-network (DDQN) or deep deterministic policy gradient (DDPG) architectures.

[0160] During training, training data (e.g., including inflow, defined time steps) may be provided to the AI ​​engine 510. Additionally, control parameters for time step t+1, such as decanter bowl velocity and differential velocity, may be included as part of the training data. The model then predicts the remaining parameters for time step t+1 that are not part of the training data. The DDQN-based agent outputs all possible actions, such as increasing or decreasing the decanter bowl velocity, from which the action with the highest value is selected. The DDPG-based agent outputs the magnitude of each parameter, where t represents the time step.

[0161] Figure 5B is an illustration of an exemplary method 550 of interaction between an agent 560 and an environment 570, further illustrating the concepts of the reinforcement training method 500, and in particular the impact of the training method 500 on an AI engine 510 (the latter described with reference to Figure 5A above). Figure 5B and the corresponding description use terminology commonly used in the art for reinforcement learning, but do not introduce any new elements compared to the description with reference to Figure 5A above.

[0162] 5A above, the AI ​​engine 510 may include an agent 560. The agent 560 may provide an action A to an environment 570 in which the agent 560 resides.

[0163] Action A may be the same as output data 530 described with reference to FIG. 5A above.

[0164] The environment 570 may include one or more decanters operated according to the action A. The environment 570 may therefore provide state information S (which may be the same as the plurality of operating parameters as described with reference to FIG. 5A above) and a quality value R (which may be the same as the quality value as described with reference to FIG. 5A above). The quality value R and the state information S may be understood to be included in the input data set 520, as described with reference to FIG. 5A above.

[0165] The state information S and the quality value R are provided to the agent 560 based on which the agent 560 can predict a new action A in a subsequent iteration step of the interaction method 550 .

[0166] The interaction between the assignment of quality values ​​R and the prediction of an action A by agent 560 based at least in part on the assigned quality values ​​R and state information S is described below.

[0167] Described below are exemplary assignments of quality values ​​R to a number of operating parameters based on one or more parameter values ​​of the material parameters. As outlined elsewhere herein, the number of material parameters may include polymer loading, cake dryness, and centrate quality.

[0168] Generally, the assignment of quality values ​​R is configured such that quality values ​​R between 0 and 1 are assigned (although other suitable intervals are also applicable). As an example, if one or more parameter values ​​of the predicted plurality of operating parameters lead to an improvement in one or more parameter values ​​of the plurality of material parameters, a quality value of 1 is assigned. Quality values ​​R can be assigned continuously within the interval between 0 and 1.

[0169] Regarding centrate quality, a quality value R is assigned when the centrate quality is in the range of 150 to 250, with a maximum quality value R being assigned when the centrate quality is 200. In this case, a maximum quality value of 1 is assigned. This can be understood as indicating that 200 is the desired, optimized centrate quality. On the other hand, if the centrate quality is 225, a quality value R of 0.5 can be assigned. This can be understood as indicating that higher centrate quality values ​​(i.e., essentially less pure centrate) are less desirable than a centrate quality of 200. This is due to the fact that higher centrate quality values ​​can affect, for example, cake dryness and / or polymer addition amount, leading to only partial optimization of several material parameters (e.g., related to centrate quality) during decanter operation, while potentially resulting in a deterioration of material parameters. If the centrate quality is below 150 or above 250, a quality value of 0 is assigned. This indicates that these material parameters should be avoided.

[0170] A similar allocation of quality values ​​R can be applied to the polymer dosage, which can range from 8 to 12 kg / tDS. In this example, if the polymer dosage is 10 kg / tDS (the desired optimum), a quality value of 1 is assigned, and if the polymer dosage is below 9 kg / tDS or above 11 kg / tDS, a lower quality value, say 0.5, is assigned. If the polymer dosage is below 8 kg / tDS or above 12 kg / tDS, an even lower quality value R of 0 is assigned.

[0171] A similar assignment can be made for cake dryness. As an example, the maximum (theoretical) cake dryness is 25% DS (however, this exemplary theoretical assumption does not limit the cake dryness that can be empirically achieved, and the cake dryness may exceed the theoretical dryness, up to 35%. In some examples, the maximum (theoretical) cake dryness is set to a maximum of 35%). In this case, a quality value of 1 is assigned. If the cake dryness is 20% DS or less or 30% DS or more, a quality value of 0.5 is assigned. Furthermore, if the cake dryness is 15% DS or less or 35% DS or more, a quality value of 0 is assigned.

[0172] It will be understood that the above-described assignment of quality values ​​to a plurality of material parameters is merely exemplary, and any other suitable assignment is also possible. Furthermore, rather than agent 560 being adapted to maximize the sum of the assigned quality values ​​R, agent 560 may also be adapted to minimize the sum of the assigned quality values ​​R.

[0173] The agent may be adapted to maximize the sum of the individual quality values ​​R assigned to each parameter of the plurality of material parameters. More specifically, as an example, based on the assignment of a respective quality value R to each individual parameter included in the plurality of operating parameters (action A), the agent 560 is notified of which of the operating parameter values ​​is considered to be an improvement in the operation of the decanter (i.e., one or more of the parameters included in the plurality of material parameters approaching the desired optimized parameter value) and which of the operating parameter values ​​is considered to be a deterioration in the operation of the decanter (e.g., one or more of the parameters included in the plurality of material parameters moving away from the desired optimized parameter value). Based thereon, the agent 560 may set a winning function including a function for calculating the sum of the assigned quality values ​​with the goal of optimizing the sum. Based on the goal of maximizing the sum of the assigned quality values, the agent 560 predicts one or more of the plurality of operating parameters such that the determined one or more of the plurality of operating parameters (when provided to the decanter) may lead to an improvement in the plurality of material parameters (action A) based on experience the agent 560 has gained over previous iterations of the method 550.

[0174] FIG. 6 is a diagram 600 that qualitatively illustrates the performance change over time of an exemplary decanter (eg, as described with reference to FIGS. 1 and 2 above).

[0175] The performance of a decanter can be understood as its ability to separate the sediment contained in the sludge from the liquid contained in the sludge. In this context, high performance can be understood as a relatively high degree of separation (e.g., most of the sediment is separated from the liquid in the sludge), whereas low performance can be understood as a relatively low degree of separation (e.g., little of the sediment is separated from the liquid in the sludge).

[0176] More specifically, FIG. 6 shows the performance over time of an exemplary decanter when manually operated in manual operation section 610, after stabilizing the operation of the decanter in stabilization section 620, and after optimizing the decanter in optimization section 630.

[0177] More specifically, when the decanter is started, a number of operating parameters are initially set by the decanter operator in manual operation section 610, e.g., based on experience. After the initial start-up of the decanter, the operator, for example, periodically collects and analyzes samples of the decanter output (e.g., centrate 220a, cake 220b) (e.g., to determine respective material parameters, such as cake dryness, centrate quality, polymer addition amount, etc.), and derives new adjusted operating parameters for the decanter therefrom to achieve desired optimum values ​​of the material parameters. Because this concept is essentially based on "trial and error" to find the operating parameters that lead to optimization of the material parameters, the decanter performance achieved by manual operation may fluctuate around a particular average performance baseline M.

[0178] In the potential (subsequent) stabilization section 620, the operating parameters of the decanter are found and the decanter is configured accordingly, thereby reducing the peak-to-peak variation in the performance progression over time of diagram 600. However, the reduced peak-to-peak variation still varies around an average performance baseline M (similar to that described with reference to manual operation section 610 above) that is below the desired optimum O of decanter performance. That is, even though the stabilization section 620 has reduced the peak-to-peak variation, the desired optimum O of performance has not yet been reached.

[0179] The optimization section 630 shows how the decanter's performance changes over time when several operating parameters are set to optimal values, where the peak-to-peak performance variation over time is minimized and the mean value around which the variation is centered is shifted to a performance optimum O that is slightly below the expected performance limit L.

[0180] It is an aspect of the present invention that the optimization section 630 is reached shortly after initial start-up of the decanter to minimize the period during which the decanter is not operating at an optimized performance level.

[0181] 7A and 7B show the performance change over time of an exemplary decanter for manual operation of the decanter (FIG. 7A) and operation of the decanter based on a trained AI engine (trained based on supervised learning, FIG. 7B).

[0182] As shown in FIG. 7A, the performance of the decanter, when operated manually, clearly fluctuates around a constant mean performance value M. Individual points where the slope of the performance curve changes abruptly may be associated with manual readjustment of one or more of the decanter's operating parameters, such as through operator sampling (e.g., every 45 minutes) followed by readjustment of one or more of the operating parameters. Performance fluctuations over time can cause the decanter to fail to consistently and effectively separate sediment from water in the sludge, adversely affecting centrate quality (variable centrate quality and / or overall poor centrate quality) and / or cake dryness (e.g., excessive cake moisture, resulting in increased cake waste disposal costs).

[0183] In contrast, Figure 7B shows the performance change over time of the decanter when it was operated based on the trained AI engine, which was trained based on supervised learning.

[0184] It can be seen that the peak-to-peak variation of the performance curve is significantly reduced compared to the performance curve shown in Figure 7A. Furthermore, the average performance baseline around which the performance variation is centered is higher than the performance curve shown in Figure 7A. Therefore, when a decanter is operated based on the trained AI engine, not only will the performance of the decanter be consistent over time, but it will also generally achieve high performance.

[0185] 7A and 7B, respectively, occur from the initial ascent of decanter bowl 110 and / or scroll 140 from a stationary state (e.g., a state in which the decanter is stopped) to an operating state (e.g., a state in which sludge processing, such as separation of centrate 220a and cake 220b, is occurring). During this ascent process, the decanter bowl velocity and / or differential speed increases to a desired value, and the separation capacity of the decanter increases with the increase in decanter bowl velocity and / or differential speed, preferably peaking at its desired (average) efficiency.

[0186] FIG. 8 shows an example simulated performance of a decanter over time when operated based on the trained enhanced AI engine.

[0187] As can be seen from Figure 8, when an enhanced AI engine is used to operate the decanter, the peak-to-peak variability of the decanter is also reduced compared to manual operation of the decanter, as shown in Figure 7 A. Additionally, operation of the decanter based on the trained enhanced AI engine also exhibits variability around the mean performance baseline M that is higher than the mean performance baseline M of Figure 7 A (decanter operated in manual mode).

[0188] 9 is a schematic diagram of an example system 900 (infrastructure) that can be used to operate a decanter. System 900 can include an environment 910 that can communicate with a human-machine interface 920, an IoT edge device 930, and a cloud service 940. Cloud service 940 can communicate with a computer 950, which can communicate with human-machine interface 920.

[0189] Environment 910 may be identical to environment 570 (described with reference to FIG. 5B above) and may include one or more decanters. Environment 910 may communicate (wirelessly and / or wired) with human machine interface (HMI) 920.

[0190] The HMI 920 may be physically separate from the environment 910 or may be part of the environment 910 (e.g., part of the decanter as a display). The HMI 920 may display decanter-specific settings to the operator and provide a means for receiving commands from the operator.

[0191] The HMI 920 can communicate with an IoT edge device 930. The IoT edge device 930 can be configured to provide access to a network and / or the Internet, based on which the IoT edge device 930 can communicate with a cloud service 940.

[0192] The cloud service 940 may be accessible via a (local) network and / or via the Internet. The cloud service 940 may at least partially implement a trained AI engine (based on supervised learning and / or reinforcement learning), providing material parameters to the AI ​​engine, which may then predict operating parameters of one or more decanters placed in the environment 910. In some cases, the cloud service 940 may be implemented as a database that stores material parameters and operating parameters (and optionally feedback and / or quality values) such that a potential training dataset is generated when one or more decanters in the environment 910 are operated. The training dataset can be used, for example, to train an AI engine that can be used to operate one or more other decanters.

[0193] Cloud service 940 can further be in communication with computer 950. Computer 950 can be used to monitor and control the prediction of multiple operating parameters of one or more decanters. Additionally, computer 950 can be in communication with HMI 920.

[0194] FIG. 10 illustrates an exemplary training data set that can be used to train an AI engine, as described elsewhere herein. This training data set is particularly applicable to reinforcement learning. Each column shown in the training data set of FIG. 10 is associated with a particular operating parameter of a plurality of operating parameters that are particularly relevant to the operation of a decanter and suitable for training. Each row shown in the training data set of FIG. 10 is associated with a distinct set of operating parameters that were determined simultaneously, with the operating parameters in different rows being determined at different times.

[0195] An exemplary training data set may include one or more of bearing liquid temperature (Bearing_Liquid_Temp), decanter bowl speed (Bowl_Speed), inlet XX (TS_Inlet), decanter differential speed (Diff_Speed), centrate TRB (TRB_Centrate), motor current (Current), solids inlet temperature (Ti_Solid), bearing solids temperature (Bearing_Solid_Temp), solids temperature (TS_Solid, typically in the range of 15-45°C, preferably in the range of 20-40°C, depending on process and weather conditions), feed flow rate (Feed_Flow), scroll torque (Torque, measured in %, typically in the range of 40-90%, preferably in the range of 45-85%, depending on each wastewater treatment plant's process conditions), bowl speed setting (Bowl_Speed_Setting), and differential speed setting (Diff_Speed_Setting).

[0196] As described above with reference to FIG. 5B, each of the operating parameters included in the training data set can be provided with a quality value that essentially scores whether a particular operating parameter of the set of operating parameters positively impacts, i.e., improves, one or more of the plurality of material parameters or causes a deterioration of one or more of the plurality of material parameters.

[0197] The following describes an exemplary training of an enhanced AI engine that includes an agent (such as agent 560 described with reference to FIG. 5B above).

[0198] As an example, if the agent determines that increasing Bowl_Speed ​​from 37.49 au (the first row of the column "Bowl_Speed" in the training dataset of FIG. 10) to 65.729 au will have a positive impact on the dryness of the cake, then, for example, a quality value of 1 is assigned to the "Bowl_Speed" parameter of 65.729 au. This indicates to the agent that increasing "Bowl_Speed" is advantageous to further increase the output value of the agent's win function. The agent can then predict an even higher "Bowl_Speed" (e.g., 97.789 au) to further increase the value of the win function. In a subsequent iteration step, it may be determined that the dryness of the cake has further increased. Therefore, a quality value of 1 is assigned to the "Bowl_Speed" of 97.789 au, indicating that further increasing "Bowl_Speed" is considered advantageous, for example, from the perspective of optimizing the dryness of the cake. Based on this, the agent can further increase the value of the win function by again predicting a higher "Bowl_Speed" of 124.325 au. If it is determined that further increases in "Bowl_Speed" will result in a deterioration of one or more of the material parameters, then a decrease in "Bowl_Speed" is initiated so that the material parameters as a whole are optimized, rather than just a single parameter of the material parameters. Training can be similarly performed for additional training parameters included in the training dataset shown in FIG. 10.

[0199] Based on the exemplary training method described above, the agent can gain knowledge of how any one or more of the plurality of operating parameters should be adjusted so that optimization of the plurality of material parameters is achieved.

[0200] The method according to the invention can be implemented as a computer program executable on any suitable data processing apparatus comprising means configured accordingly (e.g., a memory and one or more processors operatively coupled to the memory). The computer program can be stored as computer-executable instructions on a non-transitory computer-readable medium.

[0201] Embodiments of the present disclosure can be implemented in any of a variety of forms, for example, in some embodiments, the present invention is implemented as a computer-implemented method, a computer-readable storage medium, or a computer system.

[0202] In some embodiments, a non-transitory computer-readable storage medium is configured to store program instructions and / or data that, when executed by a computer system, cause the computer system to perform a method, such as any of the method embodiments described herein, or any combination of the method embodiments described herein, or any subset of the method embodiments described herein, or any combination of such subsets.

[0203] In some embodiments, a computing device is configured to include a processor (or set of processors) and a storage medium, the storage medium storing program instructions, the processor configured to read and execute the program instructions from the storage medium, the program instructions executable to perform any of the various method embodiments described herein (or any combination of the method embodiments described herein, or any subset of the method embodiments described herein, or any combination of such subsets). This device may be embodied in any of a variety of forms.

[0204] Although specific embodiments have been described above, these embodiments are not intended to limit the scope of the present disclosure, even if only a single embodiment is described with respect to a particular feature. Examples of features provided in this disclosure are intended to be illustrative rather than limiting, unless otherwise stated. The above description is intended to cover alternatives, modifications, and equivalents that will be apparent to those skilled in the art having the benefit of this disclosure.

[0205] The scope of the present disclosure includes any feature or combination of features disclosed herein (explicitly or implicitly), or generalization thereof, whether or not it alleviates any or all of the problems addressed herein. In particular, with reference to the appended claims, features of the dependent claims may be combined with features of the independent claims, and features of each independent claim may be combined in any suitable manner and not merely in the specific combinations recited in the appended claims.

Claims

1. A computer-aided method for optimizing the output of a decanter in operation using an enhanced artificial intelligence (AI) engine, a. A step of operating the decanter according to multiple operating parameters, b. The decanter processes the physical inputs, including sludge and polymer, to produce a physical output, including a centrifugal liquid and cake. c. A step of determining a plurality of material parameters based on the physical output, d. The step of passing the plurality of material parameters and the plurality of operating parameters to the enhanced AI engine, e. The steps of determining a quality value for each of the multiple material parameters using the enhanced AI engine, f. The steps of predicting multiple adjusted driving parameters using the enhanced AI engine, g. A step of further operating the decanter based on the plurality of adjusted operating parameters, A computer implementation method including

2. The plurality of operating parameters and the plurality of adjusted operating parameters are, The difference speed of the decanter's scroll, The speed of the decanter bowl, The supply flow rate of the sludge, and / or, The supply flow rate of the aforementioned polymer A computer implementation method according to claim 1, including the following:

3. The step of determining the aforementioned multiple material parameters is: One or more sensors, The dryness of the aforementioned cake, The purity of the centrifugation solution, and To determine the amount of the polymer added to the centrifugation solution. The computer implementation method according to claim 2, further comprising:

4. The step of determining the quality value for each of the multiple material parameters using the enhanced AI engine is as follows: The steps include setting the cake dryness quality value to low if the dryness of the cake is in the range of 10 to 19.99% dry weight (%DS), and setting the cake dryness quality value to high if the dryness of the cake is in the range of 20 to 35%DS. The steps include setting the purity quality value of the centrifugated liquid to be low if the purity of the centrifugated liquid is in the range of 300 to 1000 turbidity units (NTU), and setting the purity quality value of the centrifugated liquid to be high if the purity of the centrifugated liquid is in the range of 50 to 299.99 NTU, The steps include setting the polymer addition quality value to high when the amount of polymer added to the centrifugation solution is in the range of 2 to 9.99 kg / ton (tDS), and setting the polymer addition quality value to low when the amount of polymer added to the centrifugation solution is in the range of 10 to 20 kg / tDS, The computer implementation method according to claim 3, further comprising:

5. The step of predicting multiple adjusted driving parameters using the enhanced AI engine is as follows: The steps include determining the total energy consumption of the decanter based on the aforementioned multiple operating parameters, The steps include determining optimal values ​​for the decanter scroll differential velocity, the decanter bowl velocity, the sludge supply flow rate, and the polymer supply flow rate, taking into consideration the balance between the dryness of the cake, the purity of the centrifugal liquid, the total energy consumption of the decanter, and the amount of polymer added to the centrifugal liquid, A computer implementation method according to any one of claims 2 to 4, further comprising:

6. The step of determining the aforementioned optimal value is: If the cake dryness quality value is high, the steps include: reducing the optimal value for the polymer supply flow rate, reducing at least one of the optimal values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or increasing the optimal value for the sludge supply flow rate; If the cake dryness quality value is low, the steps include increasing the optimal value for the polymer supply flow rate, increasing at least one of the optimal values ​​for the decanter scroll differential speed and the decanter bowl speed, and / or decreasing the optimal value for the sludge supply flow rate, Computer implementation method according to claim 5, further comprising

7. The step of determining the aforementioned optimal value is: If the purity quality value of the centrifugal liquid is high, the steps include reducing at least one of the optimal values ​​for the differential speed of the decanter scroll and the speed of the decanter bowl, and / or increasing the optimal value for the sludge supply flow rate, If the purity quality value of the centrifugal liquid is low, the steps include increasing at least one of the optimal values ​​for the differential speed of the decanter scroll and the speed of the decanter bowl, and / or decreasing the optimal value for the sludge supply flow rate, The computer implementation method according to claim 5, further comprising:

8. The step of determining the aforementioned optimal value is: If the polymer additive quality value is high, the steps include increasing the optimal value for the polymer supply flow rate and / or increasing at least one of the optimal values ​​for the decanter scroll differential speed and the decanter bowl speed and / or decreasing the optimal value for the sludge supply flow rate, If the polymer additive quality value is low, the steps include: reducing the optimal value for the polymer supply flow rate and / or reducing at least one of the optimal values ​​for the decanter scroll differential speed and the decanter bowl speed and / or increasing the optimal value for the sludge supply flow rate; The computer implementation method according to claim 5, further comprising:

9. The step of predicting multiple adjusted driving parameters using the enhanced AI engine is as follows: The computer implementation method according to claim 5, further comprising the step of deriving the plurality of adjusted operating parameters based on the optimal values ​​for each of the decanter scroll differential speed, the decanter bowl speed, the sludge supply flow rate, and the polymer supply flow rate.

10. A computer-aided implementation method according to any one of claims 1 to 4, further comprising the step of adjusting the enhanced AI engine based on the plurality of material parameters and the plurality of operating parameters by an agent.

11. The computer implementation method according to any one of claims 1 to 4, further comprising performing steps b to g multiple times.

12. A computer-based method for training an enhanced artificial intelligence (AI) engine usable for operation of a decanter-type centrifuge, a. A step of simulating the operation of the decanter-type centrifuge and generating multiple operating parameters, b. The step of determining multiple material parameters of the simulated decanter-type centrifuge, c. The step of passing the plurality of operating parameters and the plurality of material parameters to the enhanced AI engine, d. The steps of adjusting the enhanced AI engine based on the plurality of material parameters and the plurality of operating parameters by an agent, A method that includes this.

13. An enhanced artificial intelligence (AI) engine trained according to the computer implementation method described in claim 12.

14. An enhanced artificial intelligence (AI) engine usable in a computer implementation method according to any one of claims 1 to 4.

15. An apparatus comprising means for carrying out the computer implementation method described in any one of claims 1 to 4.

16. An apparatus comprising means for carrying out the computer implementation method described in claim 12.

17. A computer program that, when executed by a processing system, includes an instruction that causes the processing system to execute the computer implementation method described in any one of claims 1 to 4.

18. A computer program that, when executed by a processing system, includes instructions that cause the processing system to execute the computer implementation method described in claim 12.

Images