Machine and method for making coffee, and method and device for calibrating coffee millator

Abstract

A method for calibrating a coffee grinder comprises receiving a plurality of sets of extraction data (20) inputs generated by a coffee preparation machine relating to one or more coffee extraction process parameters, processing the extraction data (20) using a machine learning based artificial intelligence model (30) algorithm, generating an output data set (21), it includes information for calibrating the coffee grinder (18).

Description

Technical Field

[0001] The present invention relates to a machine and method for preparing coffee, using an artificial intelligence (AI) algorithm, in particular based on machine learning, as well as a method and device for calibrating a coffee grinder. The present invention can be used to provide instructions for calibrating, perfecting, or adjusting a coffee grinder or grinder-distributor capable of grinding coffee beans for preparing any type of coffee beverage, such as espresso, long coffee, Americano, or similar coffee-based beverages. Background Art

[0002] In the world of coffee machines, the grind size of the coffee powder plays a fundamental role. To achieve a perfect extraction, the grind size must be kept within a certain particle size range.

[0003] In a coffee shop setting, the particle size of the coffee powder cannot be directly observed because this is a laboratory test. Instead, the barista will use his experience to judge whether the grinding effect is optimal, for example, by preparing a cup of espresso and measuring the extraction time; then, based on the measurement results, the barista calibrates the grinder or grinder-distributor. In fact, this calibration is done by manually adjusting the distance between the grinding wheels, tightening or loosening them, that is, bringing them closer or farther away accordingly, to produce finer or coarser powder, respectively. In combination with the above-mentioned distance adjustment, another parameter of the coffee grinder that the barista can manually adjust is the start-up time of the grinding wheels, which naturally affects the amount of coffee ground.

[0004] In fact, this manual method works because the hydraulic resistance of the coffee cake in the extraction chamber depends not only on the amount of coffee and the degree of compression, but also on the grind size of the coffee powder: the finer the coffee powder, the higher the hydraulic resistance, and thus the longer the extraction time.

[0005] This manual calibration method also has some issues, such as the variables when tamping the coffee in the filter cup, the need for the barista to be present (for example, when measuring extraction time), the need for the barista to understand the results, and the need to remove the grinder if drift is observed.

[0006] Mesin, L. et al., in their article (doi: 10.1109 / IJCNN.2012.6252493), addresses continuous control and product quality assessment in the food industry, in accordance with European standards. The article describes the use of neural networks to control two industrial grinders used in a factory producing ground coffee. According to the article, food product quality must be maintained at a high level throughout the entire production chain in every production plant. Various external factors, such as raw materials, mashing, fermentation, ripening, and mixing conditions, can influence final product quality. Therefore, automated control is necessary to ensure high and stable production quality. Furthermore, because different food products may require similar processing, adaptive control is often required. The article describes an example of an adaptive system for the food industry using an artificial neural network (ANN). The article describes an analysis of various relevant characteristics associated with coffee production in an industrial plant. The goal was to use ANNs to study the time series of coffee characteristics and their mutual influence. This approach allows the behavior of key variables in the ground coffee production process to be controlled and predicted, thereby maximizing operator assistance with the safest and most likely regulatory assessments. Therefore, the purpose of this article is to prevent the undesirable disruption of the entire production chain that can occur when certain critical parameters fall outside of expected ranges. This article records a dataset of coffee production variables from a production line. The ground coffee particles are subjected to particle size and density measurements to verify product quality. The resulting dataset consists of time series sampled at different instants in time. Product characteristics are represented by variables derived from particle size and density, based on which the operator decides how to control the grinder. To support the operator's decision-making, a neural control system is considered: two supervised artificial neural networks (ANNs) are used to control the first and second grinders, respectively, with the output being the inter-wheel distance of the first and second grinders, respectively. Possible input variables are the particle size and density data, measured at the current time or delayed by up to two sampling intervals, and the output is delayed by up to two sampling intervals. According to the article, selecting optimal input features for the ANN is crucial to reduce measurement noise, cope with the difficulty of handling large problems, and improve performance.

[0007] WO 2022 / 207953 A1 describes a method for monitoring a coffee grinder comprising a main hopper or inlet drum, multiple grinders, and one or more intermediate hoppers for ground coffee. The method includes measuring the grinder usage time for each use and in total, measuring the ambient temperature and humidity, measuring the height of the accumulated ground coffee in the intermediate hopper, calculating the expected height of the coffee in the intermediate hopper based on the aforementioned parameters and comparing it with the actual height achieved, providing a value for modifying the grinder startup time for each use, and adjusting the spacing or pressure between the grinders to ensure that the delivered dose weight is as close as possible to the set dose weight. For a given grinder state and a given particle size setting, a mathematical relationship is established, for example through machine learning, between the grinding time and the resulting grams of coffee. The relationship between the motor usage data and the intermediate hopper volume data provides information including the grinder lifespan, the ground particle size value, and the resulting quality of the service. The desired information (weight, particle size, and grinder state) is derived from the monitored variables (time, volume, humidity, and temperature) through appropriate calibration and training of an algorithm stored in a control unit.

[0008] US Pat. No. 5,645,230 A describes a device for controlling coffee grinding, comprising a pair of opposing grinding plates with an adjustable spacing to vary the size of the coffee beans obtained during grinding. The distance between the grinding plates is adjusted based on the humidity value detected by an ambient humidity sensor.

[0009] Therefore, there is a need to improve a machine for preparing coffee, a method for preparing coffee, a method and a device for calibrating a coffee grinder, which can overcome at least one disadvantage of the prior art.

[0010] In particular, an object of the present invention is to provide a machine for preparing coffee, a method for preparing coffee, a method for calibrating a coffee grinder and related devices, which are repeatable, standardized and controllable and can be automated.

[0011] More specifically, the present invention aims to make the method of preparing coffee and the operating steps of adjusting a coffee grinder automated and / or repeatable so as to provide as output information that can be used directly or indirectly for manual or automatic calibration of the coffee grinder.

[0012] Another object of the present invention is to provide a machine for preparing coffee, a method for preparing coffee, a method for calibrating a coffee grinder and related devices, which integrate the professional skills and expertise of a barista who usually performs manual adjustments of the coffee grinder. Summary of the Invention

[0013] The present invention is set forth and defined in the independent claims, while the dependent claims describe other characteristics of the invention or variants to the main inventive idea.

[0014] In accordance with the above-mentioned purposes, some embodiments described in the present specification relate to a machine for preparing coffee, the machine comprising a coffee extraction chamber capable of accommodating a certain amount of coffee powder, the coffee powder being obtained by grinding coffee beans by at least one coffee grinder, the machine further comprising a control unit configured to receive multiple sets of extraction data inputs regarding coffee extraction performed in the machine, the extraction data relating to one or more process parameters, the one or more process parameters being collected over time during at least one coffee extraction operation performed with a certain amount of coffee powder in the coffee extraction chamber.

[0015] According to one embodiment, the control unit may be associated with at least one processor to process the extraction data using an artificial intelligence model algorithm based on machine learning.

[0016] The algorithm generates an output data set comprising information for calibrating the coffee grinders by at least adjusting a distance between grinding wheels in at least one of the coffee grinders.

[0017] According to some embodiments, the at least one processor may be connected to the control unit locally or remotely (eg, via “cloud computing”).

[0018] According to one embodiment, the one or more process parameters include one or more of the following: a pressure measured in the extraction chamber, a flow rate of water fed into the extraction chamber and / or a temperature of water fed into the extraction chamber.

[0019] According to other embodiments, a method for preparing coffee in a coffee preparation machine having a coffee extraction chamber capable of containing a quantity of coffee powder obtained by grinding coffee beans with at least one coffee grinder is provided.

[0020] According to one embodiment, the method comprises receiving a plurality of sets of extraction data inputs relating to coffee extraction performed in the machine or a control unit of the machine, the extraction data relating to one or more process parameters collected over time during at least one coffee extraction operation performed with a quantity of ground coffee present in the coffee extraction chamber; By associating a control unit with at least one processor to process the extraction data using an artificial intelligence model algorithm based on machine learning; The algorithm generates an output data set comprising information for calibrating the coffee grinders, the calibration being achieved at least by adjusting a distance between grinding wheels in at least one of the coffee grinders; The one or more process parameters include one or more of the following: the pressure measured in the extraction chamber, the flow rate of water fed into the extraction chamber and / or the temperature of water fed into the extraction chamber.

[0021] According to an embodiment, which can be combined with all embodiments described herein, all three process parameters are used, in particular two of them are detected and fixed to desired preset values, while the third process parameter is allowed to vary freely and is detected.

[0022] According to other embodiments, a computer-implemented method for calibrating a coffee grinder is provided, comprising: receiving a plurality of sets of extraction data inputs generated by a coffee preparation machine and relating to one or more coffee extraction process parameters, such as pressure, flow rate and / or temperature, wherein the process parameters are collected over time during at least one coffee extraction operation performed in the coffee extraction chamber, wherein a certain amount of coffee grounds obtained by grinding coffee beans by a coffee grinder is stored in the coffee extraction chamber; Processing the extracted data using a machine learning-based artificial intelligence (AI) model algorithm; An output data set is generated that includes information for calibrating a coffee grinder by at least adjusting a distance between grinding wheels in the coffee grinder.

[0023] Other embodiments described herein relate to a device for calibrating a coffee grinder, comprising: a control unit configured to receive a plurality of sets of extraction data inputs regarding coffee extraction in a coffee preparation machine having a coffee extraction chamber capable of containing a certain amount of ground coffee obtained by grinding coffee beans by a coffee grinder, the extraction data relating to one or more process parameters collected over time during at least one coffee extraction operation performed with the certain amount of ground coffee in the coffee extraction chamber; a control unit associated with at least one processor for processing the extraction data using an artificial intelligence model algorithm based on machine learning; The algorithm generates an output data set that includes information for calibrating the coffee grinder by at least adjusting the distance between grinding wheels in the coffee grinder.

[0024] Other embodiments relate to a data processing apparatus comprising means for executing the method according to the embodiments described in this description.

[0025] Other embodiments relate to a computer program comprising instructions that, when executed by a computer, cause the computer to implement the method according to the embodiments described in this specification.

[0026] Other embodiments relate to a computer including instructions that, when executed by the computer, cause the computer to implement the method according to the embodiments described in this specification.

[0027] The computer may be local, i.e., locally associated, such as being contained in or locally connected to the coffee making machine, or located near it, or remotely available, such as via a "cloud computing" architecture. The computer includes a processor as described above, which may be local or remote. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The above and other aspects, features and advantages of the present invention will become apparent from the following description of several embodiments, given by way of non-limiting examples with reference to the accompanying drawings, in which: Figure 1 A framework diagram illustrating how to use the method according to some embodiments of the present invention; Figure 2 is a partial schematic diagram of an apparatus according to several embodiments of the present invention; Figure 3 is a schematic diagram of an apparatus according to several embodiments of the present invention; Figure 4 is a schematic diagram of a neural network that can be used in the embodiments of the present invention; Figure 5 is a schematic diagram of another neural network that can be used in the embodiments of the present invention; Figure 6 Schematic diagram of a coffee grinder according to several embodiments of the present invention.

[0029] It must be clarified that since the scope of protection is limited by the claims, the words and terms used in this specification, as well as the figures in the drawings and their relationship with the description, are only used to better illustrate and explain the present invention, and their purpose is to provide non-limiting examples of the present invention itself.

[0030] To facilitate understanding, the same reference numerals have been used in the drawings to identify the same and common elements as much as possible. It should be understood that elements and features in one embodiment may be conveniently combined or incorporated in other embodiments without further explanation. DETAILED DESCRIPTION

[0031] Some embodiments described in the drawings of this specification relate to a machine 12 and method for preparing coffee.

[0032] According to some embodiments, the machine 12 has a coffee extraction chamber 14 capable of containing a certain amount of coffee powder 16 obtained by grinding coffee beans through at least one coffee grinder 18.

[0033] The machine 12 comprises a control unit 22 configured to receive input of a plurality of sets of extraction data 20 relating to coffee extraction in the machine 12, the extraction data relating to one or more process parameters.

[0034] The one or more process parameters are collected over time during at least one coffee extraction operation performed with a certain amount of coffee powder 20 in the coffee extraction chamber 14 .

[0035] The control unit 22 may be associated with at least one processor 24 to process the extracted data 20 using a machine learning-based artificial intelligence model 30 algorithm. For example, the at least one processor 24 may be communicatively coupled to the control unit 22. For example, the at least one processor 24 may be local, i.e., directly included in or locally coupled to the control unit 22, or remote, i.e., remotely coupled (e.g., "cloud computing"), e.g., via the internet.

[0036] The algorithm generates an output data set 21 comprising information for calibrating the coffee grinders 18 by at least adjusting the distance between the grinding wheels in at least one of the coffee grinders 18 .

[0037] The one or more process parameters include one or more of the following: the pressure measured in the coffee extraction chamber 14 , the flow rate of water fed into the coffee extraction chamber 14 and / or the temperature of the water fed into the coffee extraction chamber 14 .

[0038] Other embodiments relate to a method for preparing coffee, which is performed in a machine 12 for preparing coffee and having a coffee extraction chamber 14 capable of containing a quantity of coffee powder 16 obtained by grinding coffee beans with at least one coffee grinder 18 .

[0039] The method includes receiving a plurality of sets of extraction data 20 for performing coffee extraction in a control unit 22 of the machine 12 , the extraction data relating to one or more process parameters for performing coffee extraction in the machine 12 .

[0040] The one or more process parameters are collected over time during at least one coffee extraction operation performed with a certain amount of coffee powder 16 present in the coffee extraction chamber 14.

[0041] The method includes processing the extraction data 20 using a machine learning based artificial intelligence model 30 algorithm via at least one processor 24 connected locally or remotely to the control unit 22 .

[0042] The algorithm generates an output data set 21 comprising information for calibrating the coffee grinders 18 by at least adjusting the distance between the grinding wheels in at least one of the coffee grinders 18 .

[0043] The one or more process parameters include one or more of the following: the pressure measured in the coffee extraction chamber 14 , the flow rate of water fed into the coffee extraction chamber 14 and / or the temperature of the water fed into the coffee extraction chamber 14 .

[0044] According to some embodiments, which can be combined with all embodiments described in this specification, all three of the above process parameters can be used, in particular two of the process parameters are detected and fixed to desired preset values, while the third process parameter is allowed to vary freely and be detected.

[0045] Other embodiments relate to a computer-implemented method for calibrating a coffee grinder, comprising: receiving a plurality of sets of extraction data 20 inputs generated by the coffee preparation machine 12 and related to one or more coffee extraction process parameters, such as pressure, flow rate, and temperature associated with the extraction process in the coffee extraction chamber 14, wherein the process parameters are collected over time during at least one coffee extraction operation performed in the coffee extraction chamber 14, wherein a certain amount of coffee grounds 16 obtained by grinding coffee beans by a coffee grinder 18 is stored in the coffee extraction chamber; Processing the extracted data 20 using a machine learning-based artificial intelligence (AI) model 30 or an algorithm of an AI model; An output data set 21 is generated that includes information for calibrating the coffee grinder 18 by at least adjusting the distance between the grinding wheels in the coffee grinder 18 .

[0046] Throughout this specification, reference to coffee prepared by the machine 12 always refers to liquid coffee or a coffee-based beverage, particularly one prepared by a process of extracting water and coffee grounds 16. For example, the coffee may be espresso, espresso, Americano, cold brew, or other types.

[0047] Furthermore, in this specification, references to process parameters such as pressure, flow rate, and / or temperature refer to the pressure measured in the coffee extraction chamber 14, the flow rate of water supplied to the coffee extraction chamber 14, and the temperature of the water supplied to the coffee extraction chamber 14. According to some embodiments, which can be combined with all embodiments described herein, the machine 12 receives a temporally serialized automatic sequence input 19, which results in performing coffee extraction, i.e., preparing a quantity of coffee, in the coffee extraction chamber 14. The coffee extraction in the coffee extraction chamber 14 generates extraction data 20, which is provided to the AI ​​model 30.

[0048] Some embodiments described herein also include grinding coffee beans into coffee powder 16 using a coffee grinder 18 and performing an extraction operation using the coffee powder 16 in the machine 12 to obtain a coffee-based beverage, wherein the operation is performed in accordance with the aforementioned automatic sequence input 19 and the extraction data 20 is generated by the input.

[0049] As an example, according to some embodiments, the apparatus 10 for calibrating the coffee grinder 18 includes: A control unit 22 is configured to receive a plurality of sets of extraction data 20 inputs regarding coffee extraction performed in a coffee preparation machine 12, wherein the coffee preparation machine 12 has a coffee extraction chamber 14 capable of containing a certain amount of coffee powder 16 obtained by grinding coffee beans by a coffee grinder 18, the extraction data relating to one or more process parameters collected over time during at least one coffee extraction operation performed with a certain amount of coffee powder 16 in the coffee extraction chamber 14.

[0050] According to some embodiments, which can be combined with all the embodiments described herein, the control unit 22 is associated locally or remotely with at least one processor 24, possibly two or more, for processing the extracted data 20 using an artificial intelligence (AI) model 30 based on machine learning or an algorithm of an AI model. Figure 3 The processor 24 is shown as being locally associated, specifically being included in the control unit 22 , but such representation is only a limited example, and in fact, the processor 24 can also be connected remotely (eg “cloud computing”).

[0051] According to some embodiments, which can be combined with all the embodiments described herein, the control unit 22 can be local, ie associated with or located near the coffee preparation machine, or it can be remote, for example via a "cloud computing" architecture.

[0052] The algorithm generates an output data set 21 that includes information for calibrating the coffee grinder 18 by at least adjusting the distance between the grinding wheels in the coffee grinder 18 .

[0053] The coffee grinder 18 is used to grind coffee beans and produce coffee powder 16 with different grinding degrees. Figure 6 Provide a description.

[0054] Some embodiments further relate to a data processing device, comprising a tool for executing the method according to the embodiments described herein, such as but not limited to the control unit 22 described above.

[0055] According to some embodiments, which can be combined with all the embodiments described herein, the device 10 can include a machine 12. In other embodiments, the device 10 can include a coffee grinder 18. Furthermore, in some embodiments, the device 10 can also include both the machine 12 and the coffee grinder 18.

[0056] In some embodiments, the coffee grinder 18 and the machine 12 can be integrated into a single coffee preparation machine. In this case, the coffee grinder 18 is incorporated into the machine 12 and automatically adjusted according to the output of the AI ​​model 30.

[0057] In some embodiments, the coffee grinder 18 may be external to the machine 12 .

[0058] In some embodiments, whether the coffee grinder 18 is integrated into the machine 12 or the coffee grinder 18 is disposed outside the machine 12, both can communicate, and the communication method can be wired or wireless.

[0059] In some embodiments, the machine 12 may be connected to a network, such as the Internet, which allows for remote monitoring and control of the machine 12. This enables the AI ​​model 30 to be continuously updated and trained on new data, ensuring that the accuracy of the prediction of the appropriate grind size remains high.

[0060] In another embodiment, the machine 12 may include a user interface 23, such as a touchscreen display, that allows an operator to input various parameters and preferences, such as the type of coffee beans and the desired strength of the coffee. The AI ​​model 30 processes this input data and the raw material supply and provides information that can be used to adjust the grind size of the coffee grinder 18 to ensure that the resulting coffee meets the user's preferences.

[0061] In another embodiment, the machine 12 may include or be associated with multiple coffee grinders 18 with different grind sizes. The AI ​​model 30 processes the brewing data 20 generated during the brewing process and generates an output indicating the appropriate coffee grinder 18 to use for the current brewing process. For example, the machine 12 may automatically select a recommended coffee grinder.

[0062] In another embodiment, the output of the AI ​​model 30 can be displayed to the operator, for example, through a mechanical, audio, visual, color, or lighting display, or a combination thereof. The operator can manually adjust the coffee grinder 18 based on the output. Alternatively, the output of the AI ​​model 30 can be automatically sent to the coffee grinder 18 to adjust the grind size.

[0063] The output indication of the AI ​​model 30 may also be displayed in the form of a numerical value, numerical value category, or numerical value range. In certain implementations, these numerical values, numerical value categories, or numerical value ranges may be associated with a preferred extraction time for preparing a desired type of coffee-based beverage for reference by the operator.

[0064] According to a possible embodiment that can be combined with all the embodiments described herein, coffee extraction chamber 14 includes a fixed portion 14a and a removable portion 14c. Removable portion 14c, also known as a filter cup holder, can be temporarily coupled to fixed portion 14a. Coffee extraction chamber 14 is provided with at least one coffee liquid outlet conduit 14b for discharging the coffee liquid from coffee extraction chamber 14.

[0065] The removable member 14c is adapted to receive coffee grounds 16 of a selected particle size and in a desired amount.

[0066] According to some possible technical solutions, the removable component 14c has a geometric shape and size suitable for defining the volume of the coffee extraction chamber 14 when the removable component 14c is temporarily combined with the fixed component 14a.

[0067] Thus, the removable part 14c forms a receiving chamber into which coffee powder of a selected particle size can be placed. The receiving chamber is in fluid communication with the outlet conduit 14b.

[0068] The water necessary for preparing the coffee is fed into the receiving chamber, for example by combining the following Figure 3 The pump 42 delivers the coffee powder 16 contained in the holding chamber and flows out through the outlet conduit 14b.

[0069] According to a possible embodiment, the outlet conduit 14b is integrally formed with the removable component 14c or is fixedly provided therewith.

[0070] When the removable component 14c is coupled to the fixed component 14a, the coffee extraction chamber 14 has a fixed volume.

[0071] In some use Figure 2 and Figure 3 In the embodiment described herein, which can be combined with all embodiments described herein, the machine 12 includes a sensor 49 for measuring or detecting the pressure within the coffee extraction chamber 14. In some embodiments, the AI ​​model 30 can process the pressure value generated by the sensor 49 as input and generate an output data set indicating the appropriate grind size of the coffee grinder 18.

[0072] In some embodiments, which can be combined with all embodiments described herein, the machine 12 includes a sensor 47 for measuring or detecting the flow of water supplied to the coffee extraction chamber 14. In some embodiments, the AI ​​model 30 can process the flow rate value generated by the sensor 47 as input and generate an output data set indicating an appropriate grind size of the coffee grinder 18.

[0073] In some embodiments, which can be combined with all embodiments described herein, the machine 12 includes a sensor 48 for measuring or detecting the temperature of the heated water introduced into the coffee extraction chamber 14. In some embodiments, the AI ​​model 30 can process the temperature value generated by the sensor 48 as input and generate an output data set indicating an appropriate grind size for the coffee grinder 18.

[0074] In some embodiments, which can be combined with all the embodiments described herein, the machine 12 includes a sensor for measuring or detecting the humidity inside the coffee extraction chamber 14. The AI ​​model 30 can process the humidity value generated by the sensor as input and generate an output data set indicating the appropriate grind size of the coffee grinder 18.

[0075] In some embodiments, which may be combined with all embodiments described herein, the machine 12 includes a combination of two or more of the above-mentioned sensors, such as a pressure sensor 49 and a flow sensor 47, a pressure sensor 49 and a temperature sensor 48, a flow sensor 47 and a temperature sensor 48, or all three sensors 47, 48, 49, or a combination of one, two or more of the sensors 47, 48, 49 and a humidity sensor, or all four sensors.

[0076] In some embodiments, which can be combined with all the embodiments described herein, the machine 12 is capable of preparing a coffee-based beverage according to one or more coffee extraction profiles having typical extraction process parameters, such as the pressure in the coffee extraction chamber 14, the flow rate of water fed into the coffee extraction chamber 14 and / or the temperature of the water fed into the coffee extraction chamber 14.

[0077] The characteristic curve may be stored in a storage device 25 associated with the machine 12, which may be local or remote (e.g., via a cloud computing architecture), and may be accessed by a control unit 22 also associated with the machine 12, which may also be local or remote as described above. The control unit 22 is used to operate the extraction components of the machine, such as pumps, heaters, delivery valves, etc., to perform the desired extraction and prepare a certain amount of coffee-based beverage.

[0078] The extraction profile allows the determination of nominal operating parameters for the machine 12 to obtain a coffee liquid having desired characteristics.

[0079] Therefore, the extraction characteristic curve can determine the temporal variation trend of at least the pressure, temperature and flow rate of the water introduced into the coffee extraction chamber 14 at each delivery time point or interval of the coffee liquid.

[0080] The control unit 22 may be a computer system, or include a computer system, or be locally or remotely associated with a computer system, which includes a central processing unit or CPU, electronic memory (which may be storage device 25 or other memory), an electronic database, and auxiliary circuitry (or input / output circuitry) (not shown).

[0081] For example, the CPU can be any form of computer processor that can be used for processing artificial intelligence algorithms. Memory can be connected to the CPU and can be any commercially available memory type or types, such as random access memory (RAM), read-only memory (ROM), floppy disk, hard disk, mass storage, or any other form of digital storage, whether local or remote. For example, software instructions and data can be encoded and stored in the memory to command the CPU. Auxiliary circuits can also be connected to the CPU to assist the processor in conventional manner. Auxiliary circuits can include, for example, at least one of the following: cache circuits, power supply circuits, clock circuits, input / output circuits, subsystems, etc. A program (or computer instructions) readable by a computer system can define tasks that can be implemented according to the methods of the present disclosure. In some embodiments, the program is software readable by a computer system. The computer system includes code for generating and storing information and data introduced or generated during the process according to the methods of the present disclosure.

[0082] Some embodiments may perform various steps, processes, and operations according to the embodiments described herein. These steps, processes, and operations may be performed by instructions executed by a machine, with the machine comprising a general-purpose or special-purpose processor performing certain specific steps. Alternatively, these steps, processes, and operations may be performed by specific hardware components containing hardware logic to perform the steps, or by any combination of programmable computer components and custom hardware components.

[0083] Some method embodiments according to the present disclosure may be included in a computer program that can be stored in a computer-readable medium, wherein the computer program contains instructions that, once executed by an apparatus according to the present disclosure, will cause the performance of the method.

[0084] Specifically, some components according to the present invention may be provided as machine-readable means for storing machine-executable instructions. Such machine-readable means may include, but are not limited to, floppy disks, optical disks, CD-ROMs and magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, optical or magnetic cards, wired and / or wireless communication means, or other machine-readable media types suitable for storing electronic information. For example, certain embodiments described herein may be downloaded as a computer program that may be transmitted from a remote computer (e.g., a server) to a requesting computer (e.g., a client) via a data signal created by a wave carrier or other communication means via a communication link (e.g., a wired and / or wireless modem or network connection).

[0085] In some embodiments, the machine 12 may be, for example, a machine as described in international application WO-A-2019 / 102509, which is incorporated herein by reference in its entirety.

[0086] For example, in some embodiments that can be combined with all the embodiments described in this specification, the machine 12 ( Figure 3 ) may include an electrical circuit 40 including at least a pump 42 connected to a water source 41 and configured to deliver a certain amount of pressurized water, a heating device 43 configured to heat the water supplied by the pump 12, a coffee extraction chamber 14 disposed downstream of the heating device 43 and configured to contain a certain amount of coffee grounds 16, and a selectively adjustable delivery valve 44 for controlling the delivery amount of coffee liquid when it is discharged from the coffee extraction chamber 14. The delivery valve 44 may be, for example, a proportional type.

[0087] In some embodiments, which can be combined with all embodiments described herein, the machine 12 includes sensors 45, 46, 47, 48, and 49 configured to detect at least one operating parameter of the circuit 40, and further includes a user interface 23 connected to the control unit 22, through which a user can select one of a plurality of liquid coffee recipes, and a storage device 25 for storing a list of characteristic curves for extraction of the liquid coffee, each curve being associated with one of the recipes. The sensors 45, 46, 47, 48, and 49 are configured to repeatedly detect the one or more operating parameters of the circuit 40 during the delivery time.

[0088] In some embodiments, which can be combined with all the embodiments described herein, the sensors 45, 46, 47, 48, and 49 typically include one or two temperature sensors 46, 48, one or two pressure sensors 45, 49, and a flow sensor 47 or flow meter located downstream of the pump 42; for example, the first temperature sensor 46 is located upstream of the heating device 43 and / or the second temperature sensor 48 is located downstream of the heating device 43; or, for another example, the first pressure sensor 45 is located between the pump 42 and the heating device 43 and / or the second pressure sensor 49 is located within the coffee extraction chamber 14. The sensors 45, 46, 47, 48, and 49 are configured to repeatedly detect respective operating parameters of the circuit 40 during the delivery time, including the pressure and flow downstream of the pump 42, the temperature upstream and downstream of the heating device 43, and the pressure within the coffee extraction chamber 12. A humidity sensor may also be provided, if desired.

[0089] By way of example only, the recipes for the various types of coffee prepared by the machine 12 may be related to the type of liquid coffee delivered, such as espresso, espresso, Americano, cold brew, and / or to the type or origin of coffee to be used.

[0090] One of the above-described liquid coffee extraction profiles can be associated with each recipe.

[0091] According to its characteristic curve, the machine 12 can thus prepare a coffee-based beverage.During the extraction process, an automated sequence input 19 is generated which specifies the set points of various physical quantities such as temperature, pressure and flow that the machine 12 will follow during delivery.

[0092] For example, the pressure set point generates a pressure curve within the coffee extraction chamber 14 (as well as temperature and flow), which, in some embodiments, can then be used as an input to the algorithm of the AI ​​model 30 to determine the appropriate grind size of the coffee grinder 18.

[0093] In some embodiments, which may be combined with all embodiments described herein, the operating mode of machine 12 includes delivering coffee and detecting the evolution of at least one process parameter, such as pressure, over time. Machine 12 receives an automated sequence input 19, as a result of which it extracts coffee and, over time, collects corresponding values ​​of an extraction process parameter, such as pressure, and provides these values ​​as a set of extraction data 20 to an AI model 30 to determine an appropriate grind size for coffee grinder 18, such as pressure. During the extraction process, as a result of receiving the automated sequence input 19, one or more desired process parameters, such as pressure within coffee extraction chamber 14, are continuously sampled, generating a set of extraction data 20 that is provided as input to AI model 30. The AI ​​model processes the received data and generates an output data set containing information indicating an appropriate grind size for coffee grinder 18. The output of AI model 30 can thus be used to automatically adjust coffee grinder 18 or displayed to an operator for manual adjustment.

[0094] In another embodiment, the operation of calibrating the coffee grinder 18 may include adjusting the distance between the grinding wheels of the coffee grinder 18, which changes the setting (often referred to as "calibration") of the coffee grinder 18 itself, as will be described below in conjunction with Figure 6 In other embodiments, calibration of the coffee grinder 18 may include adjusting the activation time of the grinding wheels of the coffee grinder 18, possibly in conjunction with adjusting the distance between the grinding wheels.

[0095] The output of the AI ​​model 30 provides an indication of the degree of such adjustment, which can be manual, for example, indicating how many scale steps should be moved, or automatic. In particular, adjustment of the coffee grinder 18 can be performed manually by an operator based on the output of the AI ​​model 30, or automatically by the machine 12 using the control unit 22, which sends appropriate signals to the coffee grinder 18. This allows for precise and accurate calibration of the coffee grinder 18 to achieve the optimal grind size for the ground coffee.

[0096] The AI ​​model 30 can be trained using a training dataset obtained during one or more data collection campaigns, where a large number (e.g., hundreds) of coffee extractions are performed, with varying grind or particle sizes and roast levels. Each extraction is associated with a desired output, thus following a supervised learning approach. Generally speaking, any regression model can be used with properly defined input data.

[0097] Applicants have experimented with various architectures of machine learning-based AI models 30 that can be used in the embodiments described herein.

[0098] Therefore, the algorithm can be a machine learning-based algorithm, in particular a model-based machine learning algorithm, and more particularly a supervised learning algorithm (supervised machine learning).

[0099] Specifically, the machine learning algorithm may be or may include, for example, a neural network or a combination of neural networks, linear regression, deep learning or a decision tree, or any other model falling within the scope of the above definition.

[0100] For example, the neural network model takes as input an extracted dataset 20 generated from the entire automatic sequence input 19, while simple models (such as linear regression, decision tree, random forest, gradient boosting tree, bagging regressor, support vector regression, etc.) take as input a set of statistics (such as mean, standard deviation, skewness, etc.) calculated from the extracted dataset 20 generated from the entire automatic sequence input 19.

[0101] In some embodiments, AI model 30 may be based on a neural network.

[0102] In this case, the data generated by the machine 12 is a multivariate time series sampled at a desired frequency, for example, between 2 Hz and 10 Hz, such as 3 Hz, 4 Hz, 5 Hz, or 6 Hz. Conversely, the data used for training the neural network consists of a multivariate or univariate time series (values ​​of process parameters recorded during the coffee extraction process).

[0103] In some embodiments, the AI ​​model 30 may be based on a convolutional neural network (CNN) and / or a recurrent neural network (RNN).

[0104] According to some embodiments, the CNN and RNN networks have the same input extracted data 20 and the same type of output 21 (although the numerical values ​​of the outputs may be different since the outputs come from two different models). Therefore, the CNN and RNN networks can be used separately or interchangeably, and they can also work in parallel with the view of combining them into an integrated model as explained in more detail below.

[0105] CNN is used to extract features from the data, while RNN is used to capture the temporal dependencies between data points. The combination of these two neural network architectures makes it possible to accurately predict the appropriate grinding degree.

[0106] In some embodiments, an ensemble of neural networks can be used. A neural network ensemble is a group of multiple neural networks that work together to achieve better performance than a single network alone. Networks can be combined in a variety of ways, such as averaging their predictions or selecting predictions from specific networks based on their performance. Using an ensemble of neural networks can help reduce overfitting and improve model robustness.

[0107] Specifically, the AI ​​model 30 can be an integrated model that combines the prediction results of a CNN model and one or more recurrent neural network (RNN) models. There are several ways to create an integration of a CNN and RNN, some of which include stacking, bagging, boosting, and combining.

[0108] As mentioned above, in possible instances, CNN and RNN networks can be combined to work in parallel in an integrated model.

[0109] For example, in a possible example that can be used with the present invention, the ensemble model can use the average of the predictions of the individual models to generate the final output. This approach can provide greater accuracy and robustness in predicting the appropriate grind size of the coffee grinder 18.

[0110] In another embodiment, the CNN architecture can be a WaveNet-inspired model (CNN WaveNetInspired). WaveNet is a specific neural network architecture for text and audio generation, which was introduced by Google DeepMind in 2016. Its architecture is based on a series of dilated convolutional layers, each of which expands the convolution window to capture long-range relationships between input data. The convolutional neural network inspired by WaveNet (CNN WaveNet Inspired) combines the WaveNet architecture and CNN, with the goal of using the WaveNet architecture to capture long-range relationships between input data while using CNN to process the input data. Specifically, the CNN WaveNet Inspired neural network that can be used in the embodiments described herein can be a network with a structure similar to WaveNet (convolutional layers with increasing dilation rates), but with fewer layers and parameters, and modified for the regression action to which it is applied.

[0111] For example, a combination of CNN WaveNetInspired and RNN can be implemented by making them work in parallel, as described above.

[0112] Furthermore, according to some embodiments, a CNN WaveNet Inspired can be used as a processing layer, followed by an RNN: in this approach, a CNN WaveNet Inspired is trained to process input data, and its output is used as input to the RNN, which is then used to analyze the temporal relationships between the processed data.

[0113] Figure 4The following describes an embodiment of a CNN WaveNet-inspired network that can be used in some embodiments described herein. It receives an extracted dataset 20 as input. In some embodiments, the CNN includes an input layer, multiple one-dimensional convolutional layers (temporal convolution, Conv1D), dropout layers, pooling layers (e.g., MaxPooling 1D layers and GlobalAveragePooling 1D layers), and multiple dense layers. Each convolutional layer has its own activation function (e.g., ReLU, i.e., rectified linear unit or rectifier), except for the last convolutional layer. Each dense layer has its own activation function (e.g., ReLU, i.e., rectified linear unit or rectifier), except for the last dense layer, which provides the output of the CNN network. The Conv1D layers preferably include a first set of layers and a second set of layers. Each layer in the first set has a larger dilation rate than the previous layer (e.g., a multiplication of the dilation rate), and the layers in the second set have the same dilation rate as the first set of layers, i.e., an increasing dilation rate. For example, the first group of Conv1D layers can have dilation rates of 1, 2, 4, 8, etc., respectively, and the second group of Conv1D layers can also have dilation rates of 1, 2, 4, 8, etc., respectively. Provide a specific example: an input layer, eight Conv1D layers, a Dropout layer, a MaxPooling 1D layer, a GlobalAveragePooling1D layer, and two Dense layers.

[0114] Figure 5 Used to describe some embodiments of RNN networks that can be used in the embodiments described in this specification. In some embodiments, the RNN includes: an input layer, one or more one-dimensional convolutional layers (temporal convolution, Conv1D), a Drop layer, one or more pooling layers (e.g., MaxPooling 1D layer), one or more recurrent layers (e.g., Gated Recurrent Unit, Gru), another Dropout layer, and one or more dense layers. Each convolutional layer has its own activation function (e.g., Selu, i.e., scaled exponential linear unit), and each recurrent layer has its own activation function (e.g., TanH, i.e., hyperbolic tangent function). A specific example is provided: an input layer, a Conv1D layer, a Dropout layer, a MaxPooling 1D layer, another Dropout layer, and a dense layer.

[0115] Therefore, according to some embodiments, the AI ​​model 30 may be based on a neural network, which is specifically a model used in the calibration method described herein, and the neural network is specifically a CNN and / or RNN, or a combination of CNN and RNN, more specifically an integrated model of CNN and RNN. For example, the CNN may be WaveNet Inspired. The CNN and RNN networks may respectively conform to Figure 4 and Figure 5 The embodiments described in .

[0116] In other embodiments, the AI ​​model 30 may be based on linear regression.

[0117] Linear regression is a supervised learning technique used to establish a linear relationship between one or more independent variables (also called features or inputs) and a dependent variable (also called target or output). The linear relationship is given by a mathematical function, for example, the function has the form y = a + b*x, where y is the dependent variable, x is the independent variable, and a and b are the parameters or coefficients of the model. Linear regression attempts to find the optimal values ​​of parameters a and b to best describe the linear relationship between x and y. The model thus obtained can be used to predict the value of y when the value of x is known. Of course, the number of independent variables can be greater than 1. In general, the predicted value can be expressed as a function given by the relationship y = w0+ w1*x1+ w2*x2… + w p *x p The values ​​given, where p is the number of features, w0, w1, w2…w p is the coefficient.

[0118] In this case, a data preprocessing step and a data extraction step are provided before the data is fed into the linear regression model. The extracted dataset 20 generated from the automated sequence input 19 can be processed to extract useful features, which are then used as inputs in the regression model. Some examples of these features include, but are not limited to, the mean, standard deviation, and maximum value of the process parameters used in the extraction process (e.g., pressure, flow rate, or temperature, as defined herein), as well as the first discrete difference of the process parameters used. In this case, it is important to extract useful features, i.e., features that correlate with the target variable.

[0119] Once the features are selected, a linear regression model can be constructed by fitting the above linear model with coefficients in order to minimize the sum of squared residuals between the targets observed in the dataset and the targets predicted by the linear approximation.

[0120] In some embodiments that can be combined with all embodiments described herein, the AI ​​model 30 may include a general structure generally consisting of the following components: Data preprocessing to extract statistical features; Data standardization, for example using the StandardScaler function; Introducing polynomial features, through which the original features are converted into polynomial features of a certain degree, and then applying linear regression; Regularization of linear regression models, such as ridge regression (a statistical regularization technique used to correct overfitting in machine learning models).

[0121] The algorithm of AI model 30 typically considers three process variables: the pressure generated in the extraction chamber, the flow rate of water entering the extraction chamber, and the temperature of the water. As described above, two of these process variables can be fixed at set values ​​and monitored, while the third process variable is allowed to vary freely and be monitored.

[0122] In one embodiment, due to the control system, the water flow and temperature follow preset set points, while the pressure is allowed to vary freely in order to observe the hydraulic response of the coffee cake in the coffee extraction chamber 14.

[0123] In another embodiment, observations such as changes in flow rate may be provided so that the coffee cake is subjected to a constant water pressure.

[0124] During the data collection, data analysis, and system development process, the applicant discovered that the grinding control process, which is the subject of the embodiments described herein, generally includes grinding coffee, preparing a portion, and extracting the coffee in a machine to prepare a coffee beverage. The grinding control process is highly uncertain and can cause interference. For example, the following reasons may be involved: a) Intrinsic factors of the process: The particle size of a single shot of coffee follows a certain probability distribution; for example, two shots of the same coffee prepared using the same grinder, settings, and type of coffee may have different particle size distributions and exhibit different macroscopic behaviors. Individual beans being ground may inherently exhibit slightly different roast levels, which introduces variability and uncertainty into the process; Preparation of the serving itself.

[0125] b) Uncontrollable environmental conditions: Ambient humidity; The heads of coffee beans get stuck on the grinding wheel of the coffee grinder.

[0126] Regarding the generation of strong uncertainty and interference, in model training and selection, the applicant found that although neural network models are more accurate on training data, they may not be accurate enough in the testing step due to overfitting. Overfitting is a situation in which a model overfits to the training data so that it cannot generalize to new data and make correct predictions.

[0127] Therefore, the applicants found that by using simpler models, in particular regularized linear regression, such as ridge regression, possibly combined with more in-depth feature selection and extraction steps, better results can be obtained in terms of robustness against overfitting; models with fewer parameters are able to better generalize the problem and can mitigate the interference of the process, so that only useful information is obtained from the learning data.

[0128] Therefore, the applicant has found that choosing to use linear regression as a machine learning algorithm, especially regularized linear regression, such as ridge regression, is also advantageous for the uncertainty and interference problems of the above-mentioned system.

[0129] Generally, regarding the algorithm of AI Model 30, Applicants adjusted the above general structure for each roast degree (classic, rich, strong, decaffeinated) to achieve better results. Specifically: Classic Coffee: Statistical characteristics: Consider the maximum, mean, and standard deviation of pressure, as well as the pressure at a specific volume; Normalized scaler; Quadratic polynomial characteristics; Ridge regression, with a regularization coefficient α equal to 0.1; Strong coffee: Statistical characteristics: consider the maximum value, average value and standard deviation of pressure; Normalized scaler; Quadratic polynomial features; Ridge regression, where the regularization coefficient α is equal to 1; Strong coffee: Statistical characteristics: Consider the maximum, mean, and standard deviation of pressure, as well as the pressure at a specific volume; Normalized scaler; First-order polynomial features; Ridge regression, with a regularization coefficient α equal to 0.1; Decaffeinated coffee: Statistical characteristics: consider the maximum value, mean value and standard deviation of pressure, as well as the maximum value, mean value and standard deviation of its second-order derivative; Normalized scaler; First-order polynomial features; Ridge regression, with a regularization coefficient α equal to 1.

[0130] The applicant believes that the above architecture can also be used to create models for different grinding targets for each mixture and roasting situation mentioned above in the future, such as mocha or drip.

[0131] Figure 6The present invention describes some embodiments of a coffee grinder 18, which can be combined with all embodiments described herein, in which the spacing of the grinding wheels is adjustable to calibrate the grind size based on the indications obtained according to the present invention. As a non-limiting example, the coffee grinder can be manufactured in the manner described in the applicant's international application WO-A-2016 / 166216, which is incorporated herein by reference in its entirety.

[0132] In some embodiments, the coffee grinder 18 includes a grinding portion 50 configured to grind coffee beans and produce coffee powder 16, a transition chamber 51 configured to receive the coffee powder 16 ground by the grinding portion 50, and a discharge portion 52 configured to discharge the coffee powder 16 from the transition chamber 51. The coffee grinder 18 also includes an inlet 53 and a discharge port 54, wherein the inlet is used to feed coffee beans from the supply portion 71 and the discharge port is used to discharge the coffee powder 16 to the outside.

[0133] In some embodiments, which can be combined with all the embodiments described herein, the grinding unit 50 includes two grinding wheels 55 and 56 that move relative to each other, and the grinding unit is configured to perform grinding using the relative rotational motion of the grinding wheels. This rotational motion generates a centrifugal force acting on the coffee grounds 16, causing them to move toward the transition chamber 51 and then toward the discharge portion 52.

[0134] Typically, the grinding wheels 55, 56 are coaxially arranged with respect to a common central axis Z. The grinding wheels 55, 56 are typically provided with grinding teeth on their respective grinding surfaces.

[0135] Furthermore, the grinding wheels 55, 56 are capable of moving relative to one another, at least for the purpose of performing the grinding operation. Specifically, a grinding wheel 55, or rotor, can be provided that moves during the grinding process, and a grinding wheel 56, or stator, can be provided that is stationary during the grinding process. The terms "moving / rotor" and "stationary / stator" refer to the respective states of the grinding wheels 55, 56 during the grinding operation.

[0136] The relative movement of the grinding wheels 55, 56 for grinding may be determined by a drive unit 57. According to some embodiments, the drive unit 57 may include a motor 58 with a drive shaft 59, which is configured to determine the required relative movement of the grinding wheels 55, 56. A base 60 may also be provided for supporting the drive unit 57, in particular the motor 58.

[0137] In some embodiments, the grinding wheels 55 and 56 are configured to rotate relative to each other about a common central axis Z. Specifically, the grinding wheel 55 can rotate about the central axis Z while the fixed grinding wheel 56 remains stationary. To this end, the grinding wheel 55 can be connected to a drive shaft 59 driven by the motor 32.

[0138] In a possible embodiment, the grinding wheels 55 and 56 are configured as a male-female fit. Because their shapes match, the grinding wheels 55 and 56 nest within one another. For example, the grinding wheels 55 and 56 may have substantially interfitting truncated cone shapes. In this configuration, the truncated cone-shaped grinding wheels have grinding surfaces that are inclined at a certain angle. For example, the angle may be between 12° and 22°, more specifically between 15° and 20°, and more specifically between 16° and 18°.

[0139] In an embodiment where the grinding wheels 55 and 56 are nested with each other, coaxial, and relatively rotatable about the central axis Z for grinding, the grinding wheel 55 can be disposed internally, that is, disposed inside the grinding wheel 56 and thus completely surrounded by the grinding wheel 56 .

[0140] The grinding operation performed by the grinding section 50 is generally subject to various construction and operating parameters. Construction parameters are generally fixed and determined by the grinding wheel manufacturer, such as the geometry of the grinding body and grinding teeth, the friction and surface hardness of the grinding wheel related to the material and process. Operating parameters, such as the spacing between the grinding wheels 55, 56 and / or the rotational speed of the grinding wheels 55, 56 and / or the start-up time of the grinding section 50, can be varied as a function of the raw materials used, environmental conditions, and the desired final product.

[0141] In some references Figure 6 In an embodiment, which can be combined with all embodiments described herein, the relative distance between the grinding wheels 55 and 56 is adjustable to vary the particle size of the ground coffee powder 16. Advantageously, the adjustability of this distance is on the order of microns. This adjustability can be manual, automatic, or semi-automatic. Thus, the grinding wheels 55 and 56 can be moved relative to each other, moving closer to or further away from each other along the central axis Z, to adjust their relative distance.

[0142] For example, for adjustment purposes, the grinding wheel 56 can be moved axially away from / closer to the grinding wheel 55. Here, the fact that the grinding wheel 56 moves relative to the grinding wheel 55 to adjust their relative distance should not be understood as contradicting the fact that the grinding wheel 56 remains stationary during the grinding operation while the grinding wheel 55 moves during the operation. Furthermore, the operation of adjusting the distance between the grinding wheels 55, 56 is performed prior to the grinding operation, and once initially set, the distance does not change during the grinding operation.

[0143] Specifically, adjustment can be performed by rotating the grinding wheel 56 about the central axis Z while the grinding wheel 55 remains stationary.

[0144] In one possible embodiment, each time the grinding wheel 56 completes a full rotation, it descends a certain distance and then moves a certain distance toward the grinding wheel 55 based on a sinusoidal function of the inclination angle and the descending distance. Depending on the direction of rotation, the grinding wheel 56 moves closer to or farther from the grinding wheel 55 per degree of rotation, specifically between 0.5 and 2 microns, more specifically between 0.75 and 1.5 microns, for example, 1 micron. This movement of the grinding wheels 55 and 56 toward or away from each other can be controlled by a micron-level control system. Thus, the grinding wheels 55 and 56 can be positioned at a variable distance from each other to selectively define a grinding gap or spacing associated with the desired particle size of the ground coffee 16.

[0145] In some references Figure 6 In an embodiment that can be combined with all embodiments described in this specification, the coffee grinder 18 can include an automatic adjustment unit 61 for adjusting the relative distance between the grinding wheels 55 and 56.

[0146] In some embodiments, which can be combined with all embodiments described herein, the automatic adjustment unit 61 can include an adjustment member 62 associated with the grinding wheel 56. The adjustment member 62 is movable to position the grinding wheel 56 relative to the grinding wheel 55 for adjustment purposes.

[0147] In some embodiments, which may be combined with all embodiments described herein, the adjusting member 62 is axially coupled to the support body 63 along the central axis Z via a threaded coupler 64. Specifically, for example, the adjusting member 62 is inserted into the interior of the support body 63. Because the adjusting member 62 is threadably coupled to the support body 63, it is movable relative to the support body 63, allowing rotation about the central axis Z and axial movement along the same central axis Z. Furthermore, the adjusting member 62 is integrally connected to the grinding wheel 56, for example, via one or more screws 65. In this manner, the adjusting member 62 moves the grinding wheel 56 along with it, causing it to advance or retract along the central axis Z. Due to the threaded coupler 64, rotation of the adjusting member 62 also produces axial movement, causing the adjusting member 62 and the grinding wheel 56 to advance or retract, depending on the direction of rotation. This displacement along the central axis Z can thus change the gap width between the grinding wheels 55 and 56.

[0148] In some embodiments, which can be combined with all embodiments described herein, the automatic adjustment unit 61 can include a driven wheel 66 connected to the adjustment member 62. A transmission member 67 is provided, which is wound around the driven wheel 66. A drive wheel 68 is also provided. The transmission member 67 is wound around the drive wheel 68. The drive wheel 68 is connected to a drive element 69, which is configured to drive the rotation of the drive wheel 68 clockwise or counterclockwise, as required. Advantageously, the automatic adjustment unit 61 is configured such that for every degree of rotation of the driven wheel 66, the adjustment member 62 rotates, corresponding to a micrometer-level adjustment movement of the grinding wheel 56 relative to the grinding wheel 55, specifically between 0.5 micrometers and 2 micrometers, more specifically between 0.75 micrometers and 1.5 micrometers, for example, approximately 1 micrometer. Advantageously, an angular position sensor 70 or encoder associated with the drive element 69 can be provided, by which the angular rotation applied to the drive wheel 68 can be precisely controlled and adjusted, thereby enabling reliable and precise control of the adjustment of the grinding grit size. Specifically, due to the angular position sensor 70 or encoder, the degree of rotation of the grinding wheel 56 can be precisely controlled, thereby enabling its movement away from / closer to the grinding wheel 55 to be controlled as described above.

[0149] In summary, the described embodiments provide a method and apparatus for automatically providing reliable information for calibrating a coffee grinder 18 using an algorithm based on a machine learning AI model 30 .

[0150] The AI ​​model 30 receives as input a plurality of extraction data 20 (e.g., pressure, flow, or temperature) automatically generated by the machine 12 when processing a quantity of ground coffee 16 to prepare coffee. Specifically, the machine 12 receives automated timed input 19 and executes the coffee extraction process based thereon, generating extraction data 20 that is processed by the AI ​​model 30. The output of the AI ​​model 30 is information, which may be an indication or even a signal, based on which a coffee grinder, for example, integrated into or external to the machine, may be automatically or manually adjusted, or a specific coffee grinder may be selected from a plurality of grinders with varying grind sizes.

[0151] The embodiments described herein are capable of accurately predicting the appropriate grind size and eliminate the need for manual intervention and specific operator skills.

[0152] In some embodiments, integrating the coffee grinder 18 and the machine 12 into one coffee preparation machine can further simplify the process and provide a convenient solution for the operator. However, this does not exclude that the embodiments described in this specification are also applicable to the case where the coffee grinder 18 and the machine 12 are not integrated into one machine.

[0153] In addition, connecting to the network enables remote monitoring and control of the machine 12, and the user interface allows the operator to input different preferences and parameters. One or more sensors for measuring various physical quantities provided in the coffee extraction chamber 14 allow further optimization of the extraction process and the final coffee quality.

[0154] Obviously, modifications and / or additions to steps and / or components may be made to the machine 12 for preparing coffee, as well as to the method and device 10 described above, without departing from the sphere and scope of the invention as defined by the claims.

[0155] It is also clear that, although the present invention has been described with reference to certain specific examples, a person skilled in the art will be able to implement other equivalent forms of machines, methods and devices for preparing coffee having the characteristics recited in the claims, all of which fall within the scope of protection defined thereby.

[0156] In the following claims, references in parentheses are provided for the sole purpose of ease of reading and shall not be construed as limiting factors as to the scope of protection defined by the claims.

Claims

1. A machine (12) for preparing coffee, comprising a coffee extraction chamber (14) capable of accommodating a quantity of coffee powder (16), the coffee powder being obtained by grinding coffee beans using at least one coffee grinder (18), the machine (12) further comprising a control unit (22) configured to receive an input of extraction data (20) regarding coffee extraction performed in the machine (12), the extraction data relating to one or more process parameters collected over time during at least one coffee extraction operation performed with the quantity of coffee powder (16) present in the coffee extraction chamber (14), The control unit (22) is associated with at least one processor (24) for processing the extraction data (20) using an artificial intelligence model (30) algorithm based on machine learning, The algorithm generates an output data set (21) comprising information for calibrating the coffee grinder (18), the calibration being achieved at least by adjusting the distance between the grinding wheels in the at least one coffee grinder (18), in, The one or more process parameters include one or more of the following: the pressure measured in the coffee extraction chamber (14), the flow rate of water fed into the coffee extraction chamber (14) and / or the temperature of the water fed into the coffee extraction chamber (14).

2. The machine according to claim 1, characterized in that The algorithm for implementing the artificial intelligence model (30) includes a model-based machine learning algorithm, specifically a supervised learning algorithm.

3. The machine according to claim 1 or 2, characterized in that The algorithm for implementing the artificial intelligence model (30) is based on a neural network, specifically, the neural network is selected from a convolutional neural network (CNN) and a recurrent neural network (RNN), or a combination of the two, more specifically, selected from an integrated combination of a convolutional neural network (CNN) and a recurrent neural network (RNN).

4. The machine according to claim 1 or 2, characterized in that The artificial intelligence model (30) is based on linear regression.

5. The machine according to claim 4, characterized in that The artificial intelligence model (30) is based on ridge regression regularized linear regression.

6. A machine according to any preceding claim, characterised in that Using all three of the process parameters: the pressure measured in the coffee extraction chamber (14), the flow rate of water fed into the coffee extraction chamber (14) and the temperature of the water fed into the coffee extraction chamber (14), two of the process parameters are detected and fixed to desired preset values, while the third process parameter is allowed to vary freely and is detected.

7. A machine according to any one of the preceding claims, characterised in that The machine (12) is configured to deliver at least one dose of coffee beverage and detect the evolution of at least one process parameter over time, the machine (12) being configured to receive an automated sequence input (19), as a result of which the machine performs coffee extraction and samples corresponding values ​​of coffee extraction process parameters over time, the extraction process parameters being provided as extraction data (20) to the artificial intelligence model (30) for determining an appropriate grind size for the coffee grinder (18), wherein during the execution of the extraction process, the required one or more process parameters are continuously sampled and provided as input to the artificial intelligence model (30), the artificial intelligence model being configured to process the received data and generate an output data set containing information indicating an appropriate grind size for the coffee grinder (18), the output of the artificial intelligence model (30) being used for adjusting the coffee grinder (18).

8. A machine according to any preceding claim, characterised in that The coffee extraction chamber (14) includes a fixed part (14a), a removable part (14c) that can be temporarily combined with the fixed part (14a), and at least one outlet conduit (14b), which is used to output coffee liquid from the coffee extraction chamber (14), the removable part (14c) forms a receiving chamber for placing coffee powder (16) of a selected particle size into the receiving chamber, the receiving chamber is fluidically connected to the outlet conduit (14), and water necessary for preparing coffee is fed into the receiving chamber by a pump (42), flows through the coffee powder (16) contained in the receiving chamber and flows out through the outlet conduit (14b), the machine (12) also includes a sensor (49) for measuring or detecting the pressure in the coffee extraction chamber (14), and the artificial intelligence model (30) is configured to process at least the pressure value generated by the sensor (49) as input and generate an output data set indicating the appropriate grinding degree of the coffee grinder (18).

9. A machine according to any preceding claim, characterised in that The machine (12) includes a sensor (47) for measuring or detecting the flow rate of water supplied to the coffee extraction chamber (14), and the artificial intelligence model (30) is configured to process the flow rate value generated by the sensor (47) as input and generate an output data set indicating the appropriate grind size of the coffee grinder (18).

10. A machine according to any preceding claim, characterised in that The machine (12) includes a sensor (48) for measuring or detecting the temperature of heated water introduced into the coffee extraction chamber (14), and the artificial intelligence model (30) is configured to process the temperature value generated by the sensor (48) as input and generate an output data set indicating an appropriate grind size for the coffee grinder (18).

11. A machine according to any preceding claim, characterised in that The machine (12) is configured to prepare a coffee-based beverage according to one or more coffee extraction characteristic curves having typical extraction process parameters, the typical extraction process parameters including one or more of the following: pressure in the coffee extraction chamber (14), water flow rate fed into the coffee extraction chamber (14) and / or temperature of water fed into the coffee extraction chamber (14), the characteristic curves being stored in a storage device (25) locally or remotely associated with the machine (12), the storage device being called by a control unit (22) associated with the machine (12) to operate the extraction components of the machine (12), perform the required extraction and prepare a certain amount of coffee-based beverage.

12. A machine according to any preceding claim, characterised in that The machine (12) comprises a circuit (40) equipped with at least a pump (42) connected to a water source (41) and configured to deliver a certain amount of pressurized water, a heating device (43) configured to heat the water supplied by the pump (42), a coffee extraction chamber (14) arranged downstream of the heating device (43) and configured to accommodate a certain amount of coffee powder (16), and a selectively adjustable delivery valve (44) for controlling the delivery amount of the coffee liquid when it is discharged from the coffee extraction chamber (14), and the machine (12) further comprises one or more sensors A sensor (45, 46, 47, 48, 49), a user interface (23) and a storage device (25), wherein the sensor is configured to detect at least one operating parameter of the circuit (40), the user interface is connected to the control unit (22), and a user can select one of a plurality of liquid coffee recipes through the user interface, and the storage device is used to store a list of characteristic curves of liquid coffee extraction, each curve being associated with one of the recipes, wherein the sensor (45, 46, 47, 48, 49) is configured to repeatedly detect at least one operating parameter of the circuit (40) within the delivery time.

13. The machine according to claim 12, characterized in that The sensors (45, 46, 47, 48, 49) include one or two temperature sensors (46, 48), one or two pressure sensors (45, 49), and a flow sensor (47) located downstream of the pump (42), the first temperature sensor (46) is located upstream of the heating device (43) and / or the second temperature sensor (48) is located downstream of the heating device (43), the first pressure sensor (45) is located between the pump (42) and the heating device (43) and / or the second pressure sensor (49) is located in the coffee extraction chamber (14), and the sensors (45, 46, 47, 48, 49) are configured to repeatedly detect the respective operating parameters of the circuit (40) during the delivery time, including the pressure and flow downstream of the pump (42), the temperature upstream and downstream of the heating device (43), and the pressure in the coffee extraction chamber (12).

14. A machine according to any preceding claim, characterised in that The machine (12) is associated with at least one coffee grinder (18), wherein the operation of calibrating the coffee grinder (18) includes adjusting the distance between the grinding wheels of the coffee grinder (18) and / or adjusting the start time of the grinding wheels of the coffee grinder (18), wherein the output of the artificial intelligence model 30 gives an indication of the degree of adjustment, which adjustment can be made manually by an operator or automatically by the machine (12) using a control unit (22) sending an appropriate signal to the coffee grinder (18).

15. A method for preparing coffee in a machine (12) for preparing coffee having a coffee extraction chamber (14), the coffee extraction chamber being capable of containing a quantity of coffee powder (16), the coffee powder being obtained by grinding coffee beans with at least one coffee grinder (18), the method comprising: A plurality of sets of extraction data (20) for coffee extraction are received as input in a control unit (22) of a machine (12), the extraction data relating to one or more process parameters of coffee extraction in the machine (12), the one or more process parameters being collected over time during at least one coffee extraction operation with a certain amount of coffee powder (16) present in the coffee extraction chamber (14), processing the extraction data (20) using a machine learning based artificial intelligence model (30) algorithm via at least one processor (24) associated with the control unit (22), The algorithm generates an output data set (21) comprising information for calibrating the coffee grinders (18), the calibration being achieved at least by adjusting the distance between the grinding wheels in at least one coffee grinder (18), in, The one or more process parameters include one or more of the following: the pressure measured in the coffee extraction chamber (14), the flow rate of water fed into the coffee extraction chamber (14) and / or the temperature of the water fed into the coffee extraction chamber (14).

16. The method according to claim 15, characterized in that The algorithm for implementing the artificial intelligence model (30) includes a model-based machine learning algorithm, specifically a supervised learning algorithm.

17. The method according to claim 15 or 16, characterized in that The algorithm for implementing the artificial intelligence model (30) is based on a neural network, specifically, the neural network is selected from a convolutional neural network (CNN) and a recurrent neural network (RNN), or a combination of the two, more specifically, selected from an integrated combination of a convolutional neural network (CNN) and a recurrent neural network (RNN).

18. The method according to claim 15 or 16, characterized in that The artificial intelligence model (30) is based on linear regression.

19. The method according to claim 18, characterized in that The artificial intelligence model (30) is based on ridge regression regularized linear regression.

20. The method according to any one of claims 15 to 19, characterized in that: Using all three of the process parameters: the pressure measured in the coffee extraction chamber (14), the flow rate of water fed into the coffee extraction chamber (14) and the temperature of the water fed into the coffee extraction chamber (14), two of the process parameters are detected and fixed to desired preset values, while the third process parameter is allowed to vary freely and is detected.

21. The method according to any one of claims 15 to 20, characterized in that The machine (12) delivers at least one dose of coffee beverage and detects the evolution of at least one process parameter over time, the machine (12) receiving an automated sequence input (19), as a result of which the machine performs coffee extraction and samples corresponding values ​​of coffee extraction process parameters over time, the extraction process parameters being provided as extraction data (20) to the artificial intelligence model (30) for determining an appropriate grind size for the coffee grinder (18), wherein during the execution of the extraction process, the required one or more process parameters are continuously sampled and provided as input to the artificial intelligence model (30), the artificial intelligence model being configured to process the received data and generate an output data set containing information indicating an appropriate grind size for the coffee grinder (18), the output of the artificial intelligence model (30) being used for adjusting the coffee grinder (18).

22. The method according to any one of claims 15 to 21, characterized in that The method includes measuring or detecting the pressure within the coffee extraction chamber (14) via a sensor (49) within the machine (12), the artificial intelligence model (30) processing the pressure value generated by the sensor (49) as input and generating an output data set indicating an appropriate grind size for the coffee grinder (18).

23. The method according to any one of claims 15 to 22, characterized in that The method includes measuring or detecting the flow rate of water supplied to the coffee extraction chamber (14) via a sensor (47) within the machine (12), the artificial intelligence model (30) processing the flow rate value generated by the sensor (47) as input and generating an output data set indicating the appropriate grind size of the coffee grinder (18).

24. The method according to any one of claims 15 to 23, characterized in that The method includes measuring or detecting the temperature of heated water introduced into the coffee extraction chamber (14) via a sensor (48) within the machine (12), the artificial intelligence model (30) processing the temperature value generated by the sensor (48) as input and generating an output data set indicating an appropriate grind size for the coffee grinder (18).

25. The method according to any one of claims 15 to 24, characterized in that The machine (12) prepares a coffee-based beverage according to one or more coffee extraction characteristic curves having typical extraction process parameters, the typical extraction process parameters including one or more of the following: the pressure in the coffee extraction chamber (14), the flow rate of water fed into the coffee extraction chamber (14) and / or the temperature of the water fed into the coffee extraction chamber (14), the characteristic curves being stored in a storage device (25) locally or remotely associated with the machine (12), the storage device being called by a control unit (22) associated with the machine (12) to operate the extraction components of the machine (12), perform the required extraction and prepare a certain amount of coffee-based beverage.

26. The method according to any one of claims 15 to 25, characterized in that The machine (12) comprises a circuit (40) equipped with at least a pump (42) connected to a water source (41) and configured to deliver a certain amount of pressurized water, a heating device (43) configured to heat the water supplied by the pump (42), a coffee extraction chamber (14) arranged downstream of the heating device (43) and configured to accommodate a certain amount of coffee powder (16), and a selectively adjustable delivery valve (44) for controlling the delivery amount of the coffee liquid when it is discharged from the coffee extraction chamber (14), and the machine (12) further comprises one or more sensors (45, 46, 47, 48, 49), a user interface (23) and a storage device (25), the sensor being configured to detect at least one operating parameter of the circuit (40), the user interface being connected to the control unit (22), the user being able to select one of a plurality of liquid coffee recipes through the user interface, the storage device being used to store a list of characteristic curves for extraction of liquid coffee, each curve being associated with one of the recipes, wherein the one or more sensors (45, 46, 47, 48, 49) are configured to repeatedly detect the at least one operating parameter of the circuit (40) within a delivery time.

27. The method according to claim 26, characterized in that The sensors (45, 46, 47, 48, 49) include one or two temperature sensors (46, 48), one or two pressure sensors (45, 49), and a flow sensor (47) located downstream of the pump (42), the first temperature sensor (46) is located upstream of the heating device (43) and / or the second temperature sensor (48) is located downstream of the heating device (43), the first pressure sensor (45) is located between the pump (42) and the heating device (43) and / or the second pressure sensor (49) is located in the coffee extraction chamber (14), and the sensors (45, 46, 47, 48, 49) are configured to repeatedly detect the respective operating parameters of the circuit (40) during the delivery time, including the pressure and flow downstream of the pump (42), the temperature upstream and downstream of the heating device (43), and the pressure in the coffee extraction chamber (12).

28. The method according to any one of claims 15 to 27, characterized in that The operation of calibrating the coffee grinder (18) includes adjusting the distance between the grinding wheels of the coffee grinder (18) and / or adjusting the start time of the grinding wheels of the coffee grinder (18), wherein the output of the artificial intelligence model 30 gives an indication of the degree of adjustment, which can be performed manually by an operator or automatically by the machine (12) using a control unit (22) that sends appropriate signals to the coffee grinder (18).

29. A computer program comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any one of claims 15 to 28.

30. A computer-readable medium comprising instructions which, when executed by the computer, cause the computer to implement the method according to any one of claims 15 to 28.

Images