Automatic coffee roasting device and method

Abstract

The present invention relates to an automatic coffee roasting device. The automatic coffee roasting device may comprise: an automatic preheating unit that preheats a roasting machine; and an automatic roasting unit that roasts green coffee beans fed into the roasting machine.

Description

Coffee roasting automation device and method

[0001] The present invention relates to an automated coffee roasting device and method.

[0002] The coffee market continues to grow both domestically and internationally. Demand for whole bean coffee has increased due to the diversification of coffee consumption methods following the COVID-19 pandemic. Furthermore, the market continues to grow due to the growing popularity of premium coffee and diverse consumption patterns, including home cafes. This growth has led to a rise in the establishment of roastery cafes, which offer store branding, cost savings, and additional revenue streams. A roastery is a place that sells roasted coffee.

[0003] Roasting refers to the process of heating and roasting green coffee beans to transform them into coffee beans, a process that significantly impacts a roastery's sales. Each roastery's unique roasting style determines the perfection of the coffee, significantly impacting sales through beverage sales and coffee bean delivery. While roastery cafes offer an additional revenue stream, opening or transitioning to a roastery-style business is difficult without a dedicated roaster.

[0004] Every year, various coffees (i.e., more than 10,000 types of green coffee beans from different countries, varieties, and processes) are imported into Korea. Even Ethiopian coffee has different tastes depending on the roasting, and if you roast Colombian coffee like Brazilian coffee, it will have a green taste and be bitter, so roasting can be said to be that important. Coffee varieties include Caturra, Pacamara, Orange Bourbon, Typica, Bourbon, and Pacas. The taste you feel from the coffee (beans) varies depending on the roasting style. If the roasting style is light, you can feel flavors such as floral aromas, lemon candy, lime, and herbs. If the roasting style is medium, you can feel flavors such as black tea, peach, plum, and tropical fruits. If the roasting style is dark, you can feel flavors such as chocolate, grape, and cinnamon.

[0005] The competitiveness of the roastery can be said to be the taste of the coffee, and the taste of the coffee is determined by roasting and QC (Quality Control). Roasting cannot be taught in a simple way like a cooking recipe, and without a professional roaster, it is not easy to start a roastery.

[0006] Currently, roasteries of varying sizes are facing challenges related to human and time investment. Specifically, roasteries of varying sizes are experiencing difficulties in sales and operations, and operational issues arise due to the need to continuously invest and manage roasting personnel. Specifically, small-scale roasteries with in-house roasting spaces typically employ one or two roasters. While they seek to minimize the time spent roasting and invest in cafe operations and business growth, the increased sales channels require extensive tasks such as operations, logistics, and packaging, making it difficult to secure time for the cafe business. Furthermore, small-capacity roasters require extensive roasting time for deliveries, and sales challenges make it difficult to hire additional personnel. In medium to large-scale roasteries with production facilities such as warehouse-type factories, there are three or more roasters, and they need to strengthen coffee price competitiveness by reducing fixed costs and have a production system that is not affected by changes in roaster personnel. However, there are problems such as an increase in operational tasks such as HACCP management and order collection, problems with hiring / managing auxiliary personnel for packaging / logistics, disruption or strain in overall coffee production when roaster personnel change, and difficulty in increasing the production capacity of facilities and machines.

[0007] Furthermore, because traditional roasting involves direct human intervention, prolonged roasting next to the roasting machine exposes workers to hazardous chemicals, high temperatures, and loud noise. Due to the harsh working conditions associated with direct roasting, roasters are increasingly being categorized as a desirable occupation. Consequently, there is a pressing need to improve the roasting work environment.

[0008] Typically, roasters understand the characteristics of coffee and roasting machines and roast them according to various standards and methods to achieve the desired flavor of coffee. However, using only conventional mechanical algorithms, it is impossible to achieve the roasting style desired by the roaster. Therefore, there are limitations in automating roasting while implementing the roaster's unique style.

[0009] The technology underlying the present invention is disclosed in Korean Patent Publication No. 10-2022-0150683.

[0010] The present invention is intended to solve the problems of the prior art as described above, and to solve the problem that the roaster was exposed to harmful chemicals and had difficulty working due to exposure to high temperatures and loud noises when roasting next to the roasting machine for a long time during roasting, and to provide an automatic coffee roasting device and method capable of automatically roasting, which allows each roaster to automatically roast to achieve the coffee flavor that each roaster pursues by implementing a unique roasting style.

[0011] The present invention is intended to solve the problems of the prior art as described above, and to provide an automated coffee roasting device and method that effectively reduces the workload of conventional coffee bean production by automating the preheating process and roasting process, which previously required the input of roaster personnel in the process of producing coffee beans, and enables the roasting machine to be operated in a consistent preheating state, thereby enabling the production of consistent results (i.e., providing consistently roasted beans), thereby improving coffee quality and enabling the roaster to consistently implement the desired coffee flavor every time.

[0012] The present invention is intended to solve the problems of the prior art described above, and aims to provide an automated coffee roasting device and method that can automate the entire process of producing coffee (beans) by automatically preheating a roasting machine and automatically performing roasting using a calorie prediction model and two control decision engines, thereby efficiently producing and roasting coffee beans.

[0013] However, the technical problems to be solved by the embodiments of the present invention are not limited to the technical problems described above, and other technical problems may exist.

[0014] As a technical means for achieving the above-mentioned technical task, an automated coffee roasting device according to one embodiment of the present invention may include an automatic preheating unit for preheating a roasting machine; and an automatic roasting unit for roasting green coffee beans fed into the roasting machine.

[0015] In addition, the automatic preheating unit automatically preheats the roasting machine to a target calorie corresponding to a target calorie numerical value input by the user, and the automatic roasting unit can automatically proceed with the roasting when the roasting machine reaches the target calorie.

[0016] In addition, the automatic roasting unit performs monitoring of the roasting machine while the roasting is in progress, and when it is determined from the results of the monitoring that control of the roasting machine is necessary, the unit derives control information for the roasting machine using a predefined roasting decision model, and controls the operation of the roasting machine according to the derived control information.

[0017] In addition, when the automatic roasting unit detects that a roasting control event has occurred due to user input as a result of the monitoring, the control information can be derived using a first artificial intelligence model that has been learned to predict a control device and a control value based on a control record for the roasting machine among the pre-defined roasting decision models.

[0018] In addition, the automatic roasting unit may compare the sensor chart shape information determined from the monitoring results with the reference sensor chart shape information, and if a difference is detected, the control information may be derived using a second artificial intelligence model that has been previously learned to predict a control device and a control value based on the sensor chart shape information corresponding to the roasting record of the roasting machine among the pre-defined roasting decision models.

[0019] In addition, the automatic preheating unit measures the temperature of each machine part of the roasting machine to quantify the heat amount of the roasting machine, generates a temperature distribution data set for each machine part according to a heat source, and generates an overall temperature distribution model that predicts the overall temperature distribution of the entire roasting machine using the temperature distribution data set for each machine part, quantifies the heat amount of the roasting machine using the overall temperature distribution model, and automatically preheats the roasting machine based on information of the quantified heat amount.

[0020] Meanwhile, an automated coffee roasting method using an automated coffee roasting device according to one embodiment of the present invention may include a step of preheating a roasting machine in an automatic preheating unit; and a step of roasting green coffee beans fed into the roasting machine in an automatic roasting unit.

[0021] The above-described problem-solving methods are merely exemplary and should not be construed as limiting the present invention. In addition to the exemplary embodiments described above, additional embodiments may be included in the drawings and detailed description of the invention.

[0022] According to the above-described problem-solving means of the present invention, by providing an automated coffee roasting device and method, the problem that the roaster was exposed to hazardous chemicals and had difficulty working due to exposure to high temperatures and loud noises when roasting for a long time next to the roasting machine can be resolved, and each roaster can implement a unique roasting style so that roasting can be automatically performed according to the coffee flavor that each roaster pursues. That is, the present invention provides an automated coffee roasting device and method so that the user can be prevented from being exposed to hazardous chemicals, high temperatures, and loud noises when roasting, and can enable roasting to be performed more conveniently.

[0023] According to the problem solving means of the present invention described above, by providing an automated coffee roasting device and method, the preheating process and roasting process, which had to involve roaster personnel in the conventional coffee bean production process, can be automated, thereby effectively reducing the conventional coffee bean production workload, and the roasting machine can be operated in a consistent preheating state, thereby enabling the production of consistent results (i.e., providing consistently roasted coffee beans), thereby improving coffee quality and enabling the roaster to implement the desired coffee taste equally every time.

[0024] According to the problem solving means of the present invention described above, by providing an automated coffee roasting device and method, the entire process of producing coffee (beans) can be automated by automatically preheating the roasting machine and automatically performing roasting using a calorie prediction model and two control decision engines, thereby efficiently producing and roasting coffee beans.

[0025] However, the effects that can be obtained from the present invention are not limited to the effects described above, and other effects may exist.

[0026] FIG. 1 is a schematic diagram showing the configuration of a coffee roasting automation system including a coffee roasting automation device according to one embodiment of the present invention.

[0027] Figure 2 is a drawing showing an example of a conventional roasting machine.

[0028] FIG. 3 is a drawing for explaining a heat quantification process performed by an automatic preheating unit in an automated coffee roasting device according to one embodiment of the present invention.

[0029] FIGS. 4 to 6 are drawings for explaining a process of automatically controlling roasting by an automatic roasting unit of an automated coffee roasting device according to one embodiment of the present invention.

[0030] FIGS. 7 to 10 are drawings for explaining a temperature prediction engine constructed by an automatic roasting unit in an automated coffee roasting device according to one embodiment of the present invention.

[0031] FIG. 11 is a drawing for explaining a process in which an automatic roasting unit performs automatic roasting using a model reflecting a user's roasting style in an automatic coffee roasting device according to one embodiment of the present invention.

[0032] Figure 12 is a drawing for explaining a method of reproducing operation records among existing automatic roasting methods.

[0033] Figure 13 is a drawing for explaining the PID reproduction method among existing automatic roasting methods.

[0034] FIG. 14 is a drawing showing an example of a service provision screen that can be provided on a screen of a user terminal by a control unit of an automated coffee roasting device according to one embodiment of the present invention.

[0035] Figure 15 is a flowchart of an automated coffee roasting method according to one embodiment of the present invention.

[0036] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily implement them. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In the drawings, irrelevant parts have been omitted for clarity of description, and similar reference numerals have been used throughout the specification to indicate similar elements.

[0037] Throughout the specification of the present invention, when a part is said to be "connected" to another part, this includes not only cases where it is "directly connected" but also cases where it is "electrically connected" or "indirectly connected" with another element in between.

[0038] Throughout the specification of the present invention, when it is said that a member is located “on”, “above”, “upper”, “lower”, “lower” or “lower” another member, this includes not only cases where a member is in contact with another member, but also cases where another member exists between the two members.

[0039] Throughout the specification of the present invention, when a part is said to "include" a certain component, this does not mean that other components are excluded, but rather that other components may be included, unless specifically stated otherwise.

[0040] Throughout the specification of the present invention, some of the operations or functions described as being performed by a terminal, apparatus, or device may instead be performed by a server connected to the terminal, apparatus, or device. Similarly, some of the operations or functions described as being performed by a server may also be performed by a terminal, apparatus, or device connected to the server.

[0041] Throughout the specification of the present invention, the term "at least one" may be defined as a term including both singular and plural, and it will be apparent that even if the term "at least one" does not exist, each component may exist singularly or plurally and may mean singular or plural. Furthermore, whether each component is provided singularly or plurally may vary depending on the embodiment.

[0042] FIG. 1 is a drawing showing a schematic configuration of a coffee roasting automation system (100) including a coffee roasting automation device (10) according to one embodiment of the present invention.

[0043] Hereinafter, for the convenience of explanation, the coffee roasting automation device (10) according to one embodiment of the present invention will be referred to as the present device (10), and the coffee roasting automation system (100) according to one embodiment of the present invention will be referred to as the present system (100). In addition, even if the details depicted (described) in the drawings of the present application, including FIG. 1, are omitted below, they may be equally applied to the description of the present device (10) or the present system (100).

[0044] Referring to FIG. 1, the present system (100) may include the present device (10), a roasting machine (20), and a user terminal (30).

[0045] This device (10) relates to a coffee roasting automation device, and may be a device or server that provides at least one of a web page, an app page, a program, an application (app, app), a service, and a platform related to coffee roasting automation. In this case, the program, application, service, and platform related to coffee roasting automation provided by this device (10) will be referred to as this program, this app, this service, and this platform, respectively, for the convenience of the following description.

[0046] The roasting machine (20) may refer to a roasting machine whose operation is controlled by the present device (10). The roasting machine (20) is a machine (device) that provides roasted green coffee beans by roasting green coffee beans, and may be referred to by other terms such as a roaster. The roasting machine (20) may be, for example, a Giessen W1A toaster, an Easyster 1.8 roaster, etc., but is not limited thereto, and various roasting machines (roasters) that have been previously known or will be developed in the future may be applied. The roasting machine (20) illustrated in FIG. 1 may be, for example, a user-owned roasting machine that the user owns (possesses) by purchasing or renting it.

[0047] The user terminal (30) may refer to a terminal possessed by a user using the device (10), and may refer to a terminal possessed by a user who wishes to automatically preheat a roasting machine (20) using the device (10) and automatically perform roasting through the preheated roasting machine (20).

[0048] For example, a user may access the device (10) (i.e., the program or app provided by the device (10)) using a user terminal (30) that he or she possesses, and remotely control the operation of the roasting machine (20). After accessing the program or app via the user terminal (30), the user may use the service with or without registering as a member. The user may be a person who roasts purchased green coffee beans and provides, sells, or delivers the roasted coffee beans themselves, or may be a person who makes coffee from roasted coffee beans and provides, sells, or the like. That is, the user may be referred to by other terms such as a roaster, a coffee bean seller, a cafe operator, or a coffee seller.

[0049] The user terminal (30) may include all types of wired and wireless communication devices, such as, but not limited to, PCS (Personal Communication System), GSM (Global System for Mobile communication), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (WCode Division Multiple Access), Wibro (Wireless Broadband Internet) terminals, smartphones, SmartPads, tablet PCs, laptops, wearable devices, desktop PCs, etc.

[0050] In addition, although not shown in the drawing, the system (100) may include an administrator terminal (not shown) possessed by an administrator who develops (manufactures), distributes, operates, and manages the device (10). Here, the administrator terminal, like the user terminal (30), may include any type of wired or wireless communication device.

[0051] This device (10) can transmit and receive data by being linked to a roasting machine (20), a user terminal (30), and an administrator terminal (not shown) via a network (5), and can control the operation of each of the roasting machine (20), the user terminal (30), and the administrator terminal. For example, this device (10) can control the screen display operation of each terminal in the case of the user terminal (30) and the administrator terminal.

[0052] The network (5) may include, but is not limited to, a 3GPP (3rd Generation Partnership Project) network, an LTE (Long Term Evolution) network, a WIMAX (World Interoperability for Microwave Access) network, the Internet, a LAN (Local Area Network), a Wireless LAN (Wireless Local Area Network), a WAN (Wide Area Network), a PAN (Personal Area Network), a Bluetooth network, an NFC (Near Field Communication) network, a satellite broadcasting network, an analog broadcasting network, a DMB (Digital Multimedia Broadcasting) network, etc., and may include various wired / wireless communication networks.

[0053] In addition, in the example illustrated in FIG. 1, one roasting machine (20) and one user terminal (30) are included in the system (100), but this is merely an example to help understand the present invention and is not limited thereto. As another example, the system (100) may include multiple roasting machines and multiple user terminals owned by multiple users. In this case, even if the content is omitted below, the content described with respect to the roasting machine (20) may be equally applied to the description of each of the multiple roasting machines, and similarly, the content described with respect to the user terminal (30) may be equally applied to the description of each of the multiple user terminals. A more specific description of the device (10) is as follows.

[0054] This device (10) may include an automatic preheating unit (11), an automatic roasting unit (12), and a control unit (13).

[0055] The automatic preheating unit (11) can preheat the roasting machine (20). The automatic preheating unit (11) can automatically preheat the roasting machine (20) by heating it to a target calorie corresponding to a target calorie numerical value input by the user.

[0056] Fig. 2 is a drawing showing an example of a conventional roasting machine. Referring to Fig. 2, a temperature sensor is generally located in one area of ​​a roasting machine, and when preheating the roasting machine, it can be confirmed that a relatively wide area among the entire area of ​​the roasting machine is affected by the preheating due to the thermal power (heat) supplied by the heat source of the roasting machine (i.e., it can be confirmed that the area affected by the preheating is relatively large). However, when roasting with a roasting machine, the intensity of the thermal power generally needs to be adjusted in several stages, and since thermal energy is frequently exchanged inside and outside, there is a problem that it is difficult to determine the amount of heat (amount of thermal energy) accumulated in the roasting machine. That is, in the past, the amount of heat accumulated in such a roasting machine could not be predicted by a temperature sensor provided in one area of ​​the roasting machine, and because of this, there was a problem in which it was difficult for the user to maintain a constant preheating every time he or she roasted with the roasting machine (20) (a problem in which it was difficult to preheat the roasting machine to have a constant level of heat).

[0057] In order to solve this problem, the automatic preheating unit (11) performs a heat digitization process that digitizes the heat (heat level, heat quantity) of the roasting machine, which was previously difficult to determine with a temperature sensor, so that, based on the heat information digitized through the heat digitization process, when the user wants to preheat the roasting machine (20), the target preheating degree of the roasting machine (20) is set (input) (in other words, the target heat numerical value, which is a numerical value corresponding to the target heat, which is the heat of the roasting machine that the user aims for), and then the roasting machine (20) is heated until the heat (target heat) corresponding to the target preheating degree (i.e., the target heat numerical value) set (input) by the user is reached, thereby automatically preheating the roasting machine (20). The automatic preheating unit (11) can control the operation of the roasting machine (20) so that heat can be supplied from the heat source of the roasting machine (20) until the heat of the roasting machine (20) reaches the target heat, and thereby automatically preheat the roasting machine (20) until the temperature corresponding to the target heat is reached.

[0058] The automatic preheating unit (11) can solve the problem of difficulty in maintaining constant preheating for each roasting through the above-mentioned calorie digitization process, and by performing calorie digitization, the user can set target preheating information (target calorie digitization value) and, based on this, enable automatic preheating of the roasting machine (20) for complete automation in the roastery production process. The automatic preheating unit (11) can perform a calorie digitization process that digitizes the degree of calorie accumulation of the roasting machine (20), thereby enabling automatic preheating to be performed so as to heat up to the target calorie.

[0059] At this time, the automatic preheating unit (11) can enable automatic preheating by digitizing the heat accumulation of the roasting machine (20). At this time, in order to digitize the comprehensive heat accumulation of the roasting machine (20) (to perform the heat digitization process), for example, the temperature of each machine part with respect to the thermal power supplied to the roasting machine (20) is measured, and the information on the measured temperature of each machine part is used to perform optimal load ensemble (i.e., ensemble learning) to predict the overall temperature (overall temperature distribution) for the roasting machine (20), and then digitization can be performed based on this. At this time, the automatic preheating unit (11), for example, when performing the heat digitization process, can utilize the implementation method of a conventional study that predicts the temperature of building elements by ensembling them to predict the temperature inside a building.

[0060] Here, ensemble learning refers to a technique that generates multiple classifiers (estimators) and combines their predictions to obtain better prediction results than a single classifier. In a broader sense, ensemble learning also refers to combining different models. Types of ensemble learning include voting, bagging, boosting, and stacking, and since descriptions of these are well known in the art, a detailed description thereof will be omitted in the present invention. The description of the calorie quantification process can be more easily understood with reference to Figure 3.

[0061] FIG. 3 is a drawing for explaining a heat digitization process performed by an automatic preheating unit (11) in an automated coffee roasting device (10) according to one embodiment of the present invention.

[0062] Referring to FIG. 3, the automatic preheating unit (11) can perform the first to fourth processes illustrated in FIG. 3 when performing the heat digitization process for digitizing the heat of the roasting machine (20). That is, the automatic preheating unit (11) can perform the heat digitization process including the first to fourth processes in order to digitize the heat of the roasting machine (20).

[0063] Specifically, the automatic preheating unit (11) can measure the temperature of each machine part of the roasting machine (20) using a thermal imaging camera or an attached thermometer in the first step of performing the above-described heat quantity digitization process, thereby generating a temperature distribution data set for each machine part according to the heat source (according to the change in the heat source, i.e., the heat power supplied by the heat source). At this time, the machine part of the roasting machine (20) where the temperature is measured means a part (main part) among the parts (components) within the roasting machine (20) that is affected by preheating, and may include, for example, a drum shaft, a drum, a casting front plate, a heat source, etc.

[0064] After performing the above first process, the automatic preheating unit (11) can perform the second and third processes of generating an overall temperature distribution model that predicts the overall temperature distribution of the entire roasting machine (20) using the temperature distribution data set for each machine part generated above.

[0065] Specifically, after performing the first process, the automatic preheating unit (11) trains the temperature distribution data set for each machine part generated in the second process using an artificial intelligence model, for example, the LightTS model, which is a multivariate time series prediction deep learning model known in the art, thereby constructing a temperature prediction model for each machine part of the roasting machine (20) according to variables of the temperature of the heat source, time, and exhaust flow (exhaust temperature) based on the results of the training.

[0066] Here, the LightTS model may refer to a model that extracts local / global temporal information for an input sequence through, for example, continuous / discrete sampling, concatenates the results from each sampling, and then performs prediction through an information exchange block.

[0067] In addition, the temperature prediction model for each machine part being constructed and considered in the second process may include a drum shaft temperature prediction model that predicts the temperature of the drum shaft, a drum temperature prediction model that predicts the temperature of the drum, and a casting front plate temperature prediction model that predicts the temperature of the casting front plate.

[0068] After performing the above second process, the automatic preheating unit (11) can generate an overall temperature distribution model that predicts the overall temperature distribution of the entire machine of the roasting machine (20) by combining the temperature prediction models for each machine part constructed in the third process by optimal load ensemble (i.e., performing ensemble learning). At this time, the entire machine may mean all parts including parts affected by preheating (i.e., drum shaft, drum, casting front plate, heat source). Accordingly, as an example, the overall temperature distribution model generated in the third process may be a model that is designed to predict (estimate) the temperature distribution (overall temperature distribution) of the entire parts including the temperature of the drum shaft, the drum temperature, and the temperature of the casting front plate according to the variable value change of the variable of the temperature, time, and exhaust flow (exhaust temperature) of the heat source. This overall temperature distribution model may be referred to as a heat prediction model that predicts the heat (accumulated heat) of, for example, a roasting machine (20).

[0069] After performing the above third process, the automatic preheating unit (11) can quantify the heat (particularly, the heat amount, the accumulated heat amount, the heat amount level, and the overheating degree) of the roasting machine (20) using the overall temperature distribution model in the fourth process.

[0070] At this time, in the fourth process, the automatic preheating unit (11) can quantify the heat amount of the roasting machine (20) by designating the case where the roasting machine (20) is overheated as the maximum, designating the thermal equilibrium state at room temperature as the minimum, and then comparing the designated data with the overall temperature distribution model. In other words, for example, in the fourth process, the automatic preheating unit (11) designates the overall temperature distribution information of the entire roasting machine (20) as the maximum value when the roasting machine (20) is in an overheated (maximum overheated) state, and designates the overall temperature distribution information of the entire roasting machine (20) as the minimum value when the roasting machine (20) is in a thermal equilibrium state of room temperature (i.e., a state of the same temperature as the room temperature, neither overheated nor cold), and thereafter, by comparing the overall temperature distribution information of the roasting machine (20) predicted (estimated) using the overall temperature distribution model based on the designated maximum and minimum values, the heat amount of the roasting machine (20) can be quantified. For example, the automatic preheating unit (11) can express the heat of the roasting machine (20) as one of the values ​​from 0% to 100% by performing the heat digitization process including the first to fourth processes.

[0071] Accordingly, for example, the automatic preheating unit (11) can preheat the roasting machine (20) based on the digitized calorie information after digitizing the calorie amount of the roasting machine (20). In particular, the automatic preheating unit (11) can receive information on a target preheating level (i.e., a target calorie numerical value) from a user after digitizing the calorie amount of the roasting machine (20), and preheat the roasting machine (20) to the target calorie amount corresponding to the input target preheating level information. This target preheating level may be referred to in different ways by terms such as a target calorie numerical value, a target preheating numerical value, a preheating value, etc. For example, when a user inputs a target preheating level (target calorie value) of, for example, 85%, the automatic preheating unit (11) can automatically heat and preheat the roasting machine (20) until the roasting machine (20) reaches a calorie value (target calorie value) corresponding to 85% (i.e., until it is preheated to a level having the target calorie value).

[0072] Accordingly, the automatic preheating unit (11) can model the internal heat distribution of the roasting machine (20) (particularly, the overall temperature distribution, which is the heat distribution for the entire roasting machine) by performing the above-mentioned heat digitization process, and based on this, the temperature of each machine part and the heat of the entire machine can be digitized using only the supplied heat source, and the heat of the roasting machine (20) can be predicted using the overall temperature distribution model.

[0073] The automatic preheating unit (11) can automatically preheat the roasting machine (20) based on the digitized heat information by performing the calorie digitization process, thereby preventing the occurrence of a case where roasting fails due to the roasting machine (20) overheating or not being sufficiently preheated. In addition, the automatic preheating unit (11) can enable a user to share the target preheating level (target calorie numerical value, preheating value, etc.), which is the digitized heat information, with other users (other roasters), thereby enabling an accurate roasting recipe to be transmitted to other users.

[0074] After the roasting machine (20) is preheated to the target calorie corresponding to the target calorie value (target preheating level) by the automatic preheating unit (11), the automatic roasting unit (12) can then roast the green coffee beans fed into the roasting machine (20). At this time, the automatic roasting unit (12) can automatically proceed (start) with roasting (i.e., perform automatic roasting) when the roasting machine (20) reaches the target calorie. The green coffee beans that have been roasted can be referred to by the term “coffee beans.” In addition, the green coffee beans can be referred to by other terms such as “coffee beans.”

[0075] The automatic roasting unit (12) monitors the roasting machine (20) during the roasting (automatic roasting) and, if it is determined that control of the roasting machine (20) is necessary based on the monitoring results (monitoring results, analysis results), it can derive (predict) control information for the roasting machine (20) using a predefined roasting decision model and automatically control the operation of the roasting machine (20) according to the derived control information.

[0076] Here, the roasting decision-making model may be a model designed to assist the roaster (user) in making decisions during roasting. In addition, the control information for the roasting machine (20) may include control information regarding the control devices of the roasting machine (20) and information regarding control values ​​(control fluctuation ranges) regarding the control devices. That is, the control information for the roasting machine (20) may mean information regarding which of the plurality of control devices should be controlled and at what control value. Here, the control devices may include a burner, an airflow speed, a drum speed, a damper, etc.

[0077] Here, airflow is something that is necessarily formed during the coffee roasting process, and the airflow changes depending on the firepower, exhaust, batch size, etc. Since all actions of the roaster throughout the entire roasting process from preheating to exhaust are related to changes in the airflow, it can be said that understanding the airflow during coffee roasting is very important. The fan speed control of the blower (sirocco fan) installed in the cyclone dust collector or the damper that controls the cross-sectional area of ​​the exhaust pipe by controlling the opening and closing of the gate in the exhaust pipe play a role in forming the airflow during coffee roasting. However, the fundamental element that forms the airflow is the temperature increase of the air around the heater from the gas or electricity supply. Hereinafter, with reference to FIGS. 4 to 6, the process in which the automatic roasting unit (12) automatically controls the roasting machine (20) during roasting (i.e., the automatic roasting control process) will be described in detail.

[0078] FIGS. 4 to 6 are drawings for explaining a process (automatic roasting process) of automatically controlling roasting by an automatic roasting unit (12) of an automatic coffee roasting device (10) according to one embodiment of the present invention.

[0079] Referring to FIGS. 4 to 6, when the automatic roasting unit (12) detects that the roasting machine (20) has been automatically preheated to the target heat value in the automatic preheating unit (11), roasting can be automatically performed (started) when the heat value (target heat value) is reached (S11).

[0080] The automatic roasting unit (12) can monitor the temperature of the roasting machine (20) and the sensors within the roasting machine (20) in real time during the roasting process (S12) when the roasting process is in progress (started). Here, the sensors may include a sensor for measuring the temperature of the green beans within the drum of the roasting machine (20), a sensor for measuring the exhaust temperature, a sensor for measuring the firepower (the intensity of the firepower), a sensor for measuring the fan speed, etc. Here, the temperature of the roasting machine (20) may refer to, for example, the temperature during roasting (input temperature) measured by a temperature sensor provided in one area of ​​the roasting machine (20). That is, the automatic roasting unit (12) can monitor predefined monitoring indices (for example, the temperature of the green beans within the drum, the exhaust temperature, the temperature increase rate, etc.) by monitoring the sensors within the roasting machine (20) in real time during the roasting process.

[0081] The automatic roasting unit (12) can perform two types of monitoring when monitoring the roasting machine (20). The two types of monitoring may include roasting control event monitoring (particularly, roasting control event monitoring including roasting events) as a first type of monitoring, and sensor chart modification monitoring as a second type of monitoring.

[0082] <Type 1 Monitoring (Roasting Control Event Monitoring)>

[0083] The automatic roasting unit (12) performs a first type of monitoring (roasting control event monitoring) that monitors whether a roasting control event occurs in the roasting machine (20) during the roasting process, for example, in step S12, and if it is detected as a result of the monitoring that a roasting control event has occurred due to user input, it can determine that control of the roasting machine (20) is necessary (S13-Y). Here, the roasting control event may include i) an event in which the user controls (changes, adjusts) the temperature (e.g., input temperature) of the roasting machine (20) to a specific temperature during the roasting process, ii) an event in which the user controls (changes, adjusts) the roasting time of the roasting machine (20) to a specific time during the roasting process, and iii) an event in which the user generates a specific roasting event during the roasting process. At this time, a specific roasting event may include, for example, an input event in which a first crack occurs when green coffee beans are input into the roasting machine (20), a turning point event in which a second crack occurs due to a turning point, and an exhaust event in which exhaust occurs. Here, the turning point refers to the temperature (for example, air temperature) of the roasting machine (20) dropping for a certain period of time after the green coffee beans are input and then rising again, and the section in which the temperature drops and then rises is called a turning point. The automatic roasting unit (12) may be able to determine the timing before and after the occurrence of the roasting control event through the first type of monitoring.

[0084] In this way, if the automatic roasting unit (12) detects that a roasting control event by user input has occurred in the monitoring result for the first type of monitoring in step S12, it can determine that there is control performed by the user (roaster) for the temperature, time, and roasting event at the update point (the point in time when the sensor value is updated by monitoring) in the monitoring result (i.e., it can determine that the roasting control event has occurred by the user as the user directly performs control for a specific temperature, a specific time, and a specific roasting event) (S12-Y), and thereafter automatically control the roasting machine (20) (S14). Specifically, the automatic roasting unit (12), when performing step S14 after step S12-Y, can derive (predict) control information for the roasting machine (20) by using the first artificial intelligence model that has been previously learned to predict the control device and the control value based on the control record for the roasting machine (20) among the predefined roasting decision-making models. Here, the control device may include, as mentioned above, firepower, airflow speed, drum speed, etc., and the control value may mean information about the control variation range (the variation range of the control numerical value for a specific control device that is desired to be controlled).

[0085] Here, the first artificial intelligence model is a model used to derive (predict) control information of a roasting machine (20) for automatically controlling the roasting machine (20) when a roasting control event occurs (i.e., when a control with a clear point in time, such as a specific temperature, a specific time, or a specific roasting event, is detected), and may be, for example, a Random Forest model, which is one of the machine learning models.

[0086] That is, the automatic roasting unit (12) can build (create) the first artificial intelligence model in advance to perform roasting control decision-making based on a specific point in time (a specific point in time when a roasting control event occurs).

[0087] At this time, the automatic roasting unit (12) can reconstruct a data table as shown in (a) of FIG. 6 based on the control record of the roasting machine (20) (for example, this may mean a control record of controlled roasting for the control devices of the roasting machine), in order to build a first artificial intelligence model. Thereafter, the automatic roasting unit (12) can train the first artificial intelligence model with the information of the reconstructed data table so that the first artificial intelligence model can predict the control information of the roasting machine (20) (i.e., information on the control device and the control fluctuation range). At this time, the information of the reconstructed data table may include information on temperature, time, proximity event, event elapsed time, control device, control value before control, control value after control, value (fluctuation value) of the control fluctuation range, etc. At this time, the automatic roasting unit (12) can learn to learn and predict three items corresponding to the values ​​of the control device, the control value after control, and the control fluctuation range based on the information of the reconstructed table when training the first artificial intelligence model, for example. Thereafter, the automatic roasting unit (12) can simulate by comparing the learning result of the trained first artificial intelligence model with a similar control record of a user (or a similar roasting record corresponding to a similar control record) similar to the control record used for training, and then optimize the hyper parameters of the first artificial intelligence model, thereby completing the construction of the first artificial intelligence model.

[0088] The automatic roasting unit (12) can derive (predict) control information of the roasting machine (20) using the first artificial intelligence model that has been constructed (learned) in this way. In particular, when a roasting control event is detected to have occurred during roasting, the automatic roasting unit (12) can apply the control record of the roasting machine (20) corresponding to the information of the detected roasting control event as an input value of the first artificial intelligence model, thereby obtaining a result value derived (predicted) from the first artificial intelligence model (i.e., a result value corresponding to the input value) as control information of the roasting machine (20), which is information necessary for controlling the roasting machine (20). Thereafter, the automatic roasting unit (12) can automatically control the operation of the roasting machine (20) (automatically control the roasting process) according to the control information of the roasting machine (20) obtained from the first artificial intelligence model. That is, the automatic roasting unit (12) can use the first artificial intelligence model to derive control criteria for control devices (i.e., derive control information) based on the roasting control event.

[0089] <Type 2 Monitoring (Sensor Chart Modification Monitoring)>

[0090] In addition, the automatic roasting unit (12) can perform a second type of monitoring (sensor chart deformation monitoring) that monitors chart deformation (chart shape, appearance) by sensors in real time based on sensor values ​​of sensors in the roasting machine (20) during roasting, for example, at step S12, and when a difference is detected by comparing the sensor chart deformation information determined from the monitoring result with the reference sensor chart deformation information, it can be determined that control of the roasting machine (20) is necessary (S15-Y). After step S15-Y, the automatic roasting unit (12) can derive (predict) control information of the roasting machine (20) using the second artificial intelligence model that has been previously learned to predict the control device and control value based on the sensor chart modification information corresponding to the roasting record of the roasting machine (20) among the above-definition roasting decision-making models (S16), and can then automatically control the roasting machine (20) based on the derived control information (S14). The automatic roasting unit (12) can derive information (control information) on which control device should be controlled and with what control value in step S16, and can automatically control the operation of the roasting machine (20) based on this (S14).

[0091] At this time, the sensor chart modification information may mean, for example, modification information of a chart (graph) for sensing values ​​related to predefined monitoring indicators (e.g., temperature of green beans in a drum, exhaust temperature, temperature increase rate, etc.).

[0092] For example, in the example illustrated in FIG. 4, the process of step S15 is exemplified as being performed after the roasting control event is detected not to have occurred in step S13 (when S13-N), but this is only one example to help understanding of the present invention and is not limited thereto. The automatic roasting unit (12) may perform step S15 after step S13, may perform step S13 after performing step S15, or may perform the process of step S13 and the process of step S15 simultaneously.

[0093] In the above description, steps S11 to S16 may be further divided into additional steps or combined into fewer steps, depending on the implementation of the present invention. Furthermore, some steps may be omitted as needed, and the order of steps may be changed.

[0094] Specifically, the automatic roasting unit (12) can obtain sensor chart deformation information by performing the second type of monitoring, thereby determining (confirming) deformation of a temperature chart, deformation of a temperature rise rate chart, etc. The sensor chart deformation information may include deformation information of a temperature chart, deformation information of a temperature rise rate chart, etc. Thereafter, the automatic roasting unit (12) can compare the sensor chart deformation information obtained through the second type of monitoring with the reference sensor chart deformation information, and if a difference exists, determine that control of the roasting machine (20) is necessary. The automatic roasting unit (12) can analyze (compare and analyze) the pattern of the two sensor chart deformation information (i.e., the sensor chart deformation information obtained through the second type of monitoring and the reference sensor chart deformation information), compare the temperature rise rate, and compare temperature difference information for the monitoring index through the comparison between the two pieces of sensor chart deformation information.

[0095] Here, the reference sensor chart modification information may be information obtained using a pre-built roasting machine temperature prediction engine, and the roasting machine temperature prediction engine may be referred to by other terms such as a roasting machine temperature modification prediction model, which will be described in detail later.

[0096] The second artificial intelligence model is a model used to derive (predict) control information of a roasting machine (20) for automatically controlling the roasting machine (20) by comparing two sensor chart modification information, and may be, for example, a VGG-Net16 model, which is a CNN-based deep learning model.

[0097] That is, the automatic roasting unit (12) can build (create) the second artificial intelligence model in advance to perform roasting control decision-making based on sensor chart modification.

[0098] At this time, the automatic roasting unit (12) can label the sensor chart deformation information corresponding to the roasting record of the roasting machine (20) obtained by monitoring the sensors of the roasting machine (20), the reference sensor chart deformation information for reference, the pattern information of the temperature rise rate, and the control records (control data, control record data) of the control devices corresponding to the roasting record in order to build a second artificial intelligence model, and thereafter, can train the chart deformation by applying the labeled data as input to the second artificial intelligence model. At this time, the automatic roasting unit (12) can complete the construction of the second artificial intelligence model by training the second artificial intelligence model with the labeled data so that the second artificial intelligence model learns the chart deformation from the labeled data and predicts the control information of the roasting machine (20) (i.e., information on the control device and the control fluctuation range).

[0099] The automatic roasting unit (12) can derive (predict) control information of the roasting machine (20) using the second artificial intelligence model that has been previously constructed (learned) in this way. In particular, when a difference is detected between the two pieces of sensor chart deformation information during roasting, the automatic roasting unit (12) can apply the sensor chart deformation information detected to have a difference (i.e., the sensor chart deformation information acquired through the second type of monitoring) as an input value of the second artificial intelligence model that has been previously learned, thereby obtaining a result value derived (predicted) from the second artificial intelligence model (i.e., a result value corresponding to the input value) as control information of the roasting machine (20), which is information necessary for controlling the roasting machine (20). Thereafter, the automatic roasting unit (12) can automatically control the operation of the roasting machine (20) (automatically control the roasting process) according to the control information of the roasting machine (20) acquired from the second artificial intelligence model.

[0100] That is, the automatic roasting unit (12) can use the second artificial intelligence model (CNN-based model) to determine (analyze, recognize) the sensor chart deformation or deformation pattern of the monitored sensors and derive (predict) the control information of the roasting machine (20) based on this. That is, the automatic roasting unit (12) can use the second artificial intelligence model to derive control criteria for the control devices (i.e., derive control information) based on the chart deformation. When the automatic roasting unit (12) intends to automatically control the roasting process of the roasting machine (20) by looking at the deformation of the temperature chart, the deformation of the temperature rise rate chart, etc., the first artificial intelligence model can recognize the sensor chart deformation and predict the control information, thereby automatically controlling the roasting.

[0101] In this way, the automatic roasting unit (12) monitors the roasting machine (20) during the roasting process (in particular, monitors roasting control events and sensor chart modifications), and derives control information (control reference information on which control device to control and with what control value) of the roasting machine (20) by using the first artificial intelligence model and the second artificial intelligence model based on this, and automatically controls the operation of the roasting machine (20) based on the derived control information (in particular, automatically performs control of the control devices). At this time, the automatic roasting unit (12), when controlling the roasting machine (20) according to the derived control information, may perform control with three types of control actions, such as, for example, one-time control, gradually increasing control, or gradually decreasing control.

[0102] The automatic roasting unit (12) can utilize a pre-defined (built) roasting decision-making model (particularly, two roasting decision-making models including a first artificial intelligence model and a second artificial intelligence model) to enable an AI model to make a control decision instead of the user (roaster) making a decision to monitor and control the roasting machine (20) during roasting. That is, the automatic roasting unit (12) can build a roasting decision-making model by AI modeling the decision of the user (roaster) to monitor and control the roasting machine, and can automatically predict information about the control timing and control criteria for the control devices during roasting (i.e., control information of the roasting machine (20)) without the need for the user's judgment (intervention) by using this roasting decision-making model, and can automatically control the operation of the roasting machine (20) during roasting (i.e., automatically control the roasting) based on the automatically predicted information (control information).

[0103] The automatic roasting unit (12) can generate a first artificial intelligence model and a second artificial intelligence model by performing AI modeling by dividing the method of making a judgment after monitoring during the user's decision-making process into two types (i.e., the first type of monitoring method, the roasting control event monitoring method, and the second type of monitoring method, the sensor chart modification monitoring method), and thereafter, when monitoring the first type, the first artificial intelligence model is used, and when monitoring the second type, the second artificial intelligence model is used to derive control information. That is, the automatic roasting unit (12) can construct and use two detailed models including the first artificial intelligence model and the second artificial intelligence model as roasting decision-making models, taking into account that the user (roaster) makes a decision on controlling the roasting machine by dividing it into two major characteristics when monitoring the roasting process of the roasting machine (20). At this time, as mentioned above, the automatic roasting unit (12) can use the first artificial intelligence model (random forest model) to predict control information when the control is at a specific time, such as a specific temperature, a specific time, or a specific roasting event, and can use the second artificial intelligence model (CNN-based deep learning model) to recognize the shape and predict control information when the control is based on the shape of the temperature chart, the shape of the temperature rise rate chart, etc.

[0104] Hereinafter, the process of constructing a roasting machine temperature prediction engine capable of providing reference sensor chart deformation information used for comparison when the automatic roasting unit (12) performs the second type of monitoring described above will be described in more detail. That is, the roasting machine temperature prediction engine may be a model provided to provide reference sensor chart deformation information, which is a criterion for judgment of the roasting decision-making model. At this time, the roasting machine temperature prediction engine may be referred to by other terms such as a roasting machine temperature deformation prediction model (or engine), a temperature prediction model (or engine), a roasting temperature prediction model, etc., and hereinafter, for the convenience of explanation, it will be referred to as a temperature prediction engine.

[0105] FIGS. 7 to 10 are drawings for explaining a temperature prediction engine constructed by an automatic roasting unit (12) in an automated coffee roasting device (10) according to one embodiment of the present invention.

[0106] Referring to FIGS. 7 to 10, the automatic roasting unit (12) can construct a temperature prediction engine. The temperature prediction engine can be a model designed to predict the value of a temperature sensor to be measured in a roasting machine (20) during coffee roasting (i.e., during the roasting process).

[0107] The automatic roasting unit (12) can control the roasting machine (20) by predicting the temperature development and the timing of physical changes in the coffee beans during the roasting process, and can build (create) the temperature prediction engine capable of predicting the coffee temperature during roasting with given roasting conditions to provide a judgment standard for the decision-making model of the roaster. In this case, the coffee temperature may mean the temperature of the green beans.

[0108] FIG. 7 shows an example of predicting the coffee temperature (green bean temperature) inside the roasting machine (20) when controlling the firepower of the roasting machine (20) using a temperature prediction engine built by the automatic roasting unit (12).

[0109] In Fig. 7, the graph at t0 shows an example of a sensor prediction model (i.e., sensor chart model information) that measures the coffee temperature when only the initial conditions are added (set). At this time, the initial conditions may mean the conditions of variables (initial variables) corresponding to coffee (type of green coffee beans), machine (type of roasting machine), and environment (input temperature). In Fig. 7, the graphs at t1 and t2 show examples of chart models in which control variables are added during the development (progress) of roasting and the coffee temperature is predicted. Here, the control variables may include firepower, damper, drum speed, green bean input amount, etc. According to this, the temperature prediction engine, t0~t nIf you input t, it will continue to n+1 ~ t drop can be predicted. That is, the temperature prediction engine can predict and provide a chart (graph) shape corresponding to the change information of the coffee temperature measured by controlling control variables such as firepower during the roasting process.

[0110] In addition, when constructing a temperature prediction engine, the automatic roasting unit (12) has multiple roasting records (e.g., about 800,000 roasting records held by the manager's company) already stored in the database (not shown) within the device (10), and since there is a temperature change due to the control of the roasting machine during roasting (because temperature change information is included), it is difficult to directly utilize them in the temperature prediction engine. Therefore, for example, a dataset with fixed initial variables without control during roasting is created to train an initial model (initial temperature prediction engine), and thereafter, the learned initial model is additionally trained using the multiple roasting records (i.e., using the company's roasting records), thereby completing the construction of the temperature prediction engine.

[0111] Specifically, in order to build a temperature prediction engine, the automatic roasting unit (12) can perform a first construction process for generating a dataset with fixed initial variables. When performing the first construction process, the automatic roasting unit (12) can directly roast and generate a temperature curve dataset that has been performed with the initial control variables of the roasting machine (20). Here, the temperature curve dataset may mean, for example, a time-series dataset of roasting results in a state where the initial variables are fixed, as illustrated in FIG. 8. In other words, it may mean a set of data for a time-series temperature change curve graph (temperature chart) of coffee temperature (green bean temperature) as roasting is performed with the initial control variables in a state where the initial variables are fixed. At this time, the temperature curve dataset may include, for example, information on changes in exhaust temperature as well as changes in coffee temperature.

[0112] Initial variables considered when performing the first construction process include, for example, Ethiopian Natural and Colombian Washed for coffee (type of green coffee beans), Giesen W1A and Easyster 1.8 for machine (type of roasting machine) (for example, facilities and equipment within a shared roastery can be used), and 140°C to 240°C (20°C interval) for environment (for example, input temperature).

[0113] In addition, the initial control variables considered when performing the first construction process may include, for example, firing power control, damper control, drum speed control, and green bean input amount control. Here, looking at the control range of each initial control variable, the firing power control can be controlled in three stages: high, medium, and low; the damper control can be controlled in three stages: high, medium, and low; the drum speed control can be controlled in 5 Hz intervals in the range of 50 Hz or more and 60 Hz or less; and the green bean input amount control can be controlled in two stages, including half size and full size.

[0114] The automatic roasting unit (12) can identify trends by using two types of green coffee beans (i.e., Ethiopian Natural and Colombia Washed) and two types of machines (i.e., Gysen W1A and Easyster 1.8) with different characteristics when performing the first construction process. In addition, the automatic roasting unit (12) can learn about detailed control steps and control cases by utilizing its own roasting records when performing the first construction process.

[0115] After a dataset (temperature curve dataset) with fixed initial variables is generated in the first construction process, the automatic roasting unit (12) can then perform a second construction process to construct (create) an initial model (initial temperature prediction engine) that predicts the temperature change of the roasting machine (20) with the given initial variables. At this time, the automatic roasting unit (12), when performing the second construction process, divides the temperature curve dataset generated in the first construction process into learning data and ground truth and trains the LightTS model, which is a multivariate time series prediction deep learning model, to construct the initial model. The automatic roasting unit (12) can learn and construct the initial model at a low cost and in a short time by constructing the initial model using the LightTS model. The LightTS model refers to a fast multivariate time series prediction model using a light sampling-oriented multilayer perceptron (MLP) structure.

[0116] After the second construction process is performed, the automatic roasting unit (12) can perform a third construction process to construct a temperature prediction engine capable of predicting temperature reform changes according to control variables by adding control variables to the initial model (initial temperature prediction engine) constructed in the second construction process. The automatic roasting unit (12) can construct the temperature prediction engine as a model capable of predicting temperature reform according to the addition of control variables during roasting by performing the third construction process. At this time, the automatic roasting unit (12), when constructing the temperature prediction engine, can use a plurality of roasting records (i.e., its own roasting records) previously stored in a database (not shown) to apply records with a smaller number of variable control counts among the plurality of roasting records to the initial model to perform additional training on the initial model to reduce the difference in temperature reform, and can construct the temperature prediction engine through the additional training.

[0117] The process of building a temperature prediction engine will be described again with reference to FIG. 9 as follows. Referring to FIG. 9, the automatic roasting unit (12) extracts (selects) initial variables and initial control variables for training an initial model in the first building process, for example, and generates a temperature curve dataset according to roasting progress based on the extracted initial variables and initial control variables, and then trains the temperature curve dataset with the LightTS model in the second building process, thereby building an initial model, and enables prediction of temperature deformation through the built initial model. Here, the extracted (selected) initial variables and initial control variables may be, for example, as follows: [Machine: Easyster 1.8, Green Beans: A Blend, Green Bean Input Amount: 1 kg, Input Temperature: 170 degrees, Firepower: 20% (lower), Damper: 120 Pa, Drum Speed: 56 Hz]. Thereafter, the automatic roasting unit (12) adds the control value of the roasting machine to the initial model in the third construction process (i.e., adds a control variable) so that the initial model predicts the temperature change according to the added control value, and then compares the information of the predicted temperature change (i.e., the information of the predicted temperature curve as the information of the temperature change according to the added control value) with the original record using a plurality of roasting records to additionally learn (re-learn) the initial model, thereby constructing the additionally learned initial model as the temperature prediction engine.

[0118] Accordingly, the automatic roasting unit (12) compares the sensor chart deformation information obtained through the second type of monitoring with the reference sensor chart deformation information (which may be referred to as a reference record or the like) obtained from the temperature prediction engine by applying the obtained sensor chart deformation information to the temperature prediction engine, and if there is a difference, it can determine that control of the roasting machine (20) is necessary.

[0119] In addition, as an example, the automatic roasting unit (12) may, after performing the third construction process, perform a fourth construction process of fine-tuning the temperature prediction engine by applying the user's user roasting record to the temperature prediction engine constructed in the third construction process. That is, the automatic roasting unit (12), in the fourth construction process, may fine-tune the temperature prediction engine with the user's user roasting record as illustrated in FIG. 10. At this time, the fine-tuned temperature prediction model may be referred to differently as a user-customized temperature prediction model that is customized for the user so that the user's roasting style can be implemented, such as a roasting style reflection model that reflects the user's unique roasting style. For example, in FIG. 10, the Deep Learning Model may refer to a temperature prediction engine built in the third construction process, the Small Data Set may refer to data regarding user roasting records, and the Big Data Set may refer to multiple roasting records stored in a database (not shown).

[0120] In the fourth construction process, the automatic roasting unit (12) can be designed so that the temperature prediction engine constructed in the third construction process can be expanded to various roasting machines or environments.

[0121] In addition, in the fourth construction process, the automatic roasting unit (12) can apply the user's roasting record to the temperature prediction engine and learn it, thereby enabling the creation of a roasting style reflection model that reflects the user's unique roasting style, and the roasting style reflection model can improve the accuracy of temperature prediction of the roasting machine (20).

[0122] That is, the automatic roasting unit (12) can generate a roasting style reflection model that can implement the unique roasting style of the user (roaster). At this time, the user roasting record directly roasted by each user can be fine-tuned in the temperature prediction engine to generate a unique roasting style reflection model for each user. Afterwards, automatic roasting can be performed for the roasting machine (20) using the desired reproduction roasting record desired to be reproduced entered by the user of the user terminal (30) and the user's roasting style reflection model generated for the user. This can be more easily understood with reference to FIG. 11.

[0123] FIG. 11 is a drawing for explaining a process in which an automatic roasting unit (12) in an automatic coffee roasting device (10) according to one embodiment of the present invention performs automatic roasting using a model reflecting a user's roasting style.

[0124] Referring to FIG. 11, specifically, the automatic roasting unit (12) may perform the following process to generate a unique roasting style reflection model of a user (roaster, for example, Roaster A) of a user terminal (30). To generate a user's roasting style reflection model, the automatic roasting unit (12) may input (acquire) from the user a plurality (for example, about 5) of user roasting records, including roasting record 1 (first user roasting record) for Ethiopian Light of the user (Roaster A), roasting record 2 (second user roasting record) for Colombian Light of the user (Roaster A), etc. Thereafter, the automatic roasting unit (12) classifies the input multiple user roasting records according to coffee characteristics (type of green coffee beans, etc.) and machine characteristics (type of roasting machine), and then fine-tunes the pre-learned temperature prediction engine (which may mean a roasting temperature prediction model, which is a pre-learned temperature prediction engine built in the third construction process) with the classified records (classified user roasting records), and then generates a roasting decision-making model unique to the user (Roaster A) through an analysis process using a pre-defined roasting decision-making model. At this time, the fine-tuned temperature prediction engine fine-tuned with the classified records is, for example, a model generated (learned) based on the light roasting records of the user (Roaster A), and may be otherwise referred to as a light roasting style model of the user (Roaster A).

[0125] Thereafter, the automatic roasting unit (12) can perform automatic roasting for the roasting machine (20) by additionally inputting the desired reproduction roasting record that the user wants to reproduce into the fine-tuned temperature prediction engine. At this time, in order to perform automatic roasting using the user's roasting style reflection model (particularly, the light roasting style model), when the user inputs (additionally inputs) an 'Ethiopian roasting record' as the desired reproduction roasting record, for example, the automatic roasting unit (12) can apply the desired reproduction roasting record entered by the user as an input value of the fine-tuned temperature prediction engine (i.e., the light roasting style model of Roaster A). Thereafter, the automatic roasting unit (12), in response to the application of the input value, obtains sensor chart deformation information corresponding to the desired reproduction roasting record (i.e., development prediction information, which is chart deformation information that predicts a change in the temperature development of the roasting according to the desired reproduction roasting record) as a result value (predicted value) corresponding to the input value from the fine-tuned temperature prediction engine, and thereafter, compares the obtained sensor chart deformation information (i.e., development prediction information) with reference sensor chart deformation information (i.e., existing roasting record) to determine whether to control the roasting machine (20). At this time, if the automatic roasting unit (12) determines that control of the roasting machine (20) is necessary as a result of the above judgment, it can derive control information (control value) of the roasting machine (20) by querying a predefined roasting decision model (e.g., the second artificial intelligence model) (i.e., by performing the processes of steps S15 and S16 mentioned above), and can automatically control the operation of the roasting machine (20) according to the derived control information.

[0126] The control unit (13) can provide coffee beans roasted by the automatic roasting unit (12) to the user. In addition, the control unit (13) can control the operation of each part within the device (10). In addition, the control unit (13) can control the operation of each of the roasting machine (20), user terminal (30), and administrator terminal (not shown) linked to the device (10) via a network (5).

[0127] According to the above, the device (10) uses a heat prediction model (overall temperature distribution model), two roasting decision-making models (i.e., a first artificial intelligence model and a second artificial intelligence model), and a temperature prediction engine (roasting machine temperature prediction engine), thereby automating the entire roasting process, including the preheating process and the roasting process of the roasting machine (20), so that it is performed automatically without user intervention, and in particular, automatic roasting is possible while implementing the coffee taste desired by the user (roaster) in the same way every time.

[0128] For example, the control unit (13) can develop and distribute the device (10) (i.e., the program or the app) in a form that can be immediately used by existing customers by building the system (100) on the company's software previously created by the administrator. In addition, the control unit (13) can control the operation of each component by synchronizing with an artificial intelligence (AI) model (engine) based on the company's software that can communicate with the roasting machine (20) and providing control commands (control requests) to the AI ​​model and the roasting machine (20).

[0129] The control unit (13) can, for example, lighten and optimize the model (artificial intelligence model) so that it can operate on a PC (desktop PC) and an MCU board for Android, etc., when developing (producing, constructing) the program or the app. The control unit (13) can, for example, store and manage AI models considered in the device (10) (i.e., the first and second artificial intelligence models which are the heat prediction model, the roasting decision-making model, and the temperature prediction engine (roasting machine temperature prediction engine), etc.) in the form of HDF5 so that they can be used in Tensorflow lite and Tensorflow JS, and the AI ​​models can be AI models embedded in the device (10). The device (10) can be arranged to be linked with a monitoring and control system according to the platform used.

[0130] In addition, the control unit (13) can transmit and receive data using, for example, a remote communication protocol with an MCU board or PLC device installed in the roasting machine (20), and the remote communication protocol may include, for example, Modbus RTU, Modbus TCP, Siemens S7, Bluetooth, Phidget, Websocket, etc. Through this, the control unit (13) can obtain sensing values ​​of the temperature sensor or other sensors of the roasting machine (20), and can also control each control device by sensing the firepower, exhaust, drum speed, etc., and control the operation of the roasting machine (20). The control unit (13) can enable the user to use the service immediately by downloading and installing the program or app through the user terminal (30).

[0131] This device (10) can automate the entire roasting process by providing this service (i.e., coffee roasting automation related service provided by this device (10) including an automatic preheating unit (11), an automatic roasting unit (12), and a control unit (13). This device (10) can reduce 90% of the coffee bean production work compared to the conventional coffee bean production process through AI automation excluding the physical process (i.e., by automating the part that previously required manpower for preheating and roasting with AI).

[0132] For example, the existing coffee bean production process requires the user to perform a preheating operation, which takes about 5 minutes. After that, the green beans are prepared and placed in. The user then performs the roasting operation, which completes the roasting process (roasting for about 10 to 15 minutes). After the beans are discharged, they are cleaned up. As such, the existing coffee bean production process required essential human intervention, which lowered the efficiency of the coffee bean production work and made it difficult for the user. In contrast, the present device (10) can automatically preheat the roasting machine (20) through the automatic preheating unit (11), and when the prepared green coffee beans are fed in, the automatic roasting unit (12) automatically starts (proceeds with) roasting to perform the roasting through automatic roasting, and when the roasted coffee beans are discharged, they can be cleaned up, thereby reducing the workload of the conventional coffee bean production process by more than 90% and providing convenience to the user performing the roasting.

[0133] That is, this device (10) can solve the problem that the roastery production process could not be automated in the past, and in particular, it can increase work efficiency by automating preheating and roasting, which are problems that inevitably require the input of roaster personnel in the main production process. In addition, this device (10) can be integrated with a green bean distributor, a bean conveyor, etc., and through this, the entire process of the coffee bean production can be automated.

[0134] This device (10) can improve the quality of coffee by providing (producing) consistent results (beans with a consistent taste) through automation of preheating and roasting, and in particular, by digitizing preheating and automatically performing preheating based on the digitized information, it can enable roasting to proceed (operate) with a more consistent preheating state of the roasting machine (20) than when the user manually controls the preheating. In addition, this device (10) can provide (implement) reproducible performance equivalent to that of an actual roaster (user), and can enable response to unexpected situations during roasting that could not be handled with existing mechanical algorithms.

[0135] Furthermore, by providing this service, the device (10) can reproduce, for example, the level of coffee beans and roasting profiles directly roasted by the user (roaster), thereby ensuring a consistent coffee flavor. Furthermore, by providing this service, the device (10) can automate the coffee bean production process by allowing the head roaster (e.g., the user) to set a roasting style guide.

[0136] This device (10) can enable each user to implement the coffee taste they seek through roasting by providing this service through various indicator interpretation and control methods, and for this purpose, specifically, various indicators (e.g., coffee bean temperature, machine exhaust temperature, coffee bean condition, temperature rise rate, etc.) can be monitored to enable decision making, and the control device (firepower, damper, drum speed, etc.) of the roasting machine (20) can be controlled in timing (control timing) and adjustment range (control fluctuation range) according to each user's individual standards that suit their style, and even for the same user (roaster), different roasting styles can be applied to enable roasting depending on the characteristics of the green coffee beans (e.g., country, variety, process) or the characteristics of the roasting machine (e.g., type).

[0137] Meanwhile, existing automatic roasting methods include operation (control) record reproduction methods and PID reproduction methods (bean temperature PID reproduction methods). However, the operation record reproduction method has the limitation that it can only perform simple control without considering the monitoring indicators of the roasting machine (e.g., green bean temperature, exhaust temperature, temperature rise rate, etc.), and the PID reproduction method has the limitation that it cannot implement individual roasting styles for each user due to the firepower control to follow the temperature of the coffee beans (i.e., green bean temperature). In other words, the existing automatic roasting method has the problem that it cannot implement the roasting style at the level desired by the user with only the current mechanical algorithm. Below, the operation record reproduction method and the PID reproduction method are explained in more detail.

[0138] <How to reproduce the operation record>

[0139] Figure 12 is a drawing for explaining a method of reproducing operation records among existing automatic roasting methods.

[0140] Referring to Fig. 12, the operation record reproduction method is a simple method of recording and reproducing operation by time, as shown in (a) of Fig. 12, and is used by storing the record in a roasting machine and reproducing it.

[0141] However, when roasting coffee using this method of reproducing operation records, the user (roaster) faces the problem of not being able to apply the roasting process differently depending on environmental conditions such as weather, preheating conditions, and green bean condition. Currently, this problem is solved by creating a reproducibility profile that the user can use according to a number of environmental conditions, monitoring it, and replacing it when necessary, and using it for automatic operation for about 3-4 minutes, and then the user (roaster) directly manually completing the operation. In this way, since the roasting process is not fully automated in the past, the user (roaster) continuously monitors it directly and solves the problem of automatic reproduction using the method of reproducing operation records. However, this has the aspect of being cumbersome and difficult for the user because the user has to continuously monitor it.

[0142] Figure 12 (b) is a drawing for explaining an example in which the operation record reproduction method is carried out incorrectly. Referring to this, if the operation record reproduction method is carried out incorrectly, even though the user roasts with the same firepower operation, the difference between the ambient temperature and the preheating accumulates, resulting in a time difference of more than 1 minute, and this causes the resultant product (roasted coffee beans) to have a green smell.

[0143] <PID 재현 방식>

[0144] Figure 13 is a drawing for explaining the PID reproduction method among existing automatic roasting methods.

[0145] Referring to Fig. 13, in the PID reproduction method, PID (Proportional-Integral-Differential) control means a method of controlling by calculating the value of input operation (heating power) through differentiation / integration in order to match the target value (specific temperature) as shown in (a) of Fig. 13, and is used to maintain the same target value in industrial equipment such as constant temperature and humidity warehouses and boilers.

[0146] However, when roasting coffee using this PID reproducible method, the user (roaster) experiences the following problems. That is, if the roasting environment, such as the preheating state, is not manually adjusted, the PID method has the problem of making it difficult to roast consistently because the firepower is too high or too low. In addition, some users try to differentiate the taste of coffee by roasting according to their own roasting style, such as adjusting the firepower at a specific time for each roaster. However, the PID method has the problem of not being able to implement each roaster's unique roasting style. In other words, the PID reproducible method is generally only available for roasting inexpensive coffee, so the market adoption rate is low.

[0147] Figure 13 (b) is a drawing for explaining an example of an incorrectly performed PID reproduction method. Referring to this, it can be seen that although response according to the roasting style and coffee is essential for the taste of coffee, the PID method has a problem in that its use is limited due to the control of the firepower to meet the target temperature.

[0148] Accordingly, the present device (10) can solve the problems of the automatic roasting method that is conventionally performed by the operation record reproduction method or the PID reproduction method by providing this service, and can implement the roasting style at the level desired by the roaster, and based on this, the roasting machine (20) can be automatically controlled, but in particular, the roasting machine (20) can be efficiently controlled with optimized control that can implement the coffee taste that the user seeks by considering various monitoring indicators.

[0149] By providing this service, this device (10) can provide an environment in which users can easily perform roasting in a more comfortable and safe working environment, by eliminating the need for users to be exposed to high temperatures and noise for a long time during the roasting process and also eliminating the need for users to be exposed to hazardous chemicals.

[0150] FIG. 14 is a drawing showing an example of a service provision screen that can be provided on the screen of a user terminal (30) by a control unit (13) of a coffee roasting automation device (10) according to one embodiment of the present invention.

[0151] Referring to FIG. 14, the control unit (13) can provide a service provision screen, such as that shown in FIG. 14, to the screen of the user terminal (30) to enable the user to monitor the roasting process of the roasting machine (20) and operate and manage the roastery through linkage with the roasting machine (20) owned by the user.

[0152] The above service provision screen may refer to a screen that monitors roasting and enables management of data related to the roastery. Specifically, the service provision screen may be configured to monitor the temperature and sensors of the roasting machine (20) in conjunction with the roasting machine (20), and record, manage, and confirm key roasting events. In addition, the service provision screen may be configured to enable integrated management of coffee data (roasting-related data), result measurements, and green bean inventory before and after roasting, thereby assisting in roasting operations.

[0153] The control unit (13) can display the sensing values ​​obtained from the roasting machine (20) in real time on the screen of the user terminal (30), thereby enabling the user to monitor the sensing values ​​remotely in real time through a service provision screen such as that shown in FIG. 14 (in particular, a screen such as that shown in the upper drawing in FIG. 14) and check the roasting record (record of the progress) in real time, thereby enabling the current status of roasting to be checked and the roasting machine (20) to be controlled.

[0154] For example, referring to FIG. 14, the service provision screen displayed on the screen of the user terminal (30) by the control unit (13) may display information of the first graph (g1) to the fourth graph (g4). Here, the first graph (g1) refers to a change graph of ET (temperature inside the drum), the second graph (g2) refers to a change graph of BT (temperature of green coffee beans), the third graph (g3) refers to a BTRoR graph (i.e., a temperature rise rate graph of the temperature of the coffee beans), and the fourth graph (g4) refers to an ETRoR graph (i.e., a temperature rise rate graph of the temperature inside the drum). Here, RoR (Rate of Rise) represents a set temperature rise rate per hour.

[0155] For example, if the user sets 'RoR as 9 for 1 minute', the control unit (13) can control the operation of the roasting machine (20) in response to the setting so that the preset heat power is maintained for 1 minute during the roasting process, and thereby, after the 1 minute has elapsed, the temperature can be increased by 9 degrees (9 ℃) compared to before the 1 minute has elapsed. For example, if the user sets 'RoR as 9 for 1 minute', when 6 minutes have elapsed since the start of roasting, if the BTRoR at 6 minutes is 9 and the BT at 6 minutes is 190 degrees, the BT at 7 minutes thereafter can be 199 degrees (i.e., 199 degrees, which is an increase of 9 degrees from 190 degrees).

[0156] In addition, the control unit (13) may, for example, cause the first graph (g1) to be displayed on the screen of the user terminal (30), cause the underpressure information to be displayed together in conjunction with the first graph (g1). Here, underpressure refers to gas pressure, and the expression underpressure may be used because a weak gas pressure is used during roasting. The display of such underpressure information may mean that information recording the firepower adjustment point (i.e., information regarding how the firepower was changed at a specific point in time) is displayed. Such underpressure information may be information that the control unit (13) can obtain from the roasting machine (20) by automatically checking (measuring) the firepower in the roasting machine (20).

[0157] In addition, the control unit (13) can automatically determine information about the time point at which the green coffee beans were put into the roasting machine (20), the time point of the turning point, etc., through analysis of the second graph (g2). For example, if the control unit (13) analyzes that the preset temperature (e.g., 7 degrees in 10 seconds) has dropped for a certain period of time in the second graph (g2), it can detect that the green coffee beans have been put into the roasting machine (20).

[0158] In addition, the control unit (13) can analyze that there is a possibility (high probability) that baked or well-developed may not occur, for example, if the descending slope of the third graph (g3, BTRoR) graph after the first crack is steeper than 45 degrees. Here, baked refers to a phenomenon in which, when the input temperature of the green coffee beans is too low or roasting starts with a small amount of heat, the rate of heat transfer from input to the first crack slows down, causing the roasting time to become long, the expansion and color change of the green coffee beans are not smooth, and the organic acid is not properly separated due to the dehydration of the green coffee beans, and chemical reactions such as the Maillard reaction and caramelization do not occur properly.

[0159] In addition, the control unit (13) can receive sensing values ​​from the sensors of the roasting machine (20) in real time while the roasting is in progress, and display graphs (i.e., the first to fourth graphs) according to the sensing values ​​on the service provision screen. At this time, the control unit (13) can, for example, allow the user to directly input the setting values ​​of the current roasting information (e.g., setting values ​​for roasting machine information, green bean information, roasting machine information, green bean input amount, temperature, humidity, moisture content, density, batch name, etc.) through the user terminal (30) on the service provision screen while the roasting is in progress, or can automatically input the current roasting information by loading the values ​​of the previous roasting information that was performed previously.

[0160] At this time, the control unit (13) can control the current temperature development graph (i.e., sensor chart modification information) according to the progress of the current roasting to be displayed in real time on the service provision screen while the roasting is in progress, and at the same time, the previous temperature development graph (i.e., the temperature development graph of the previous roasting performed previously) to be displayed and displayed together on the service provision screen. Based on this, the control unit (13) can display the difference in arrangement and graph aspect between the current temperature development graph and the previous temperature development graph in numbers, thereby enabling the user to intuitively confirm (check) the difference in real time. That is, the control unit (13) can display, for example, the current temperature development graph and the previous temperature development graph together on the screen of the user terminal (30), and can also display a value corresponding to the difference between the two temperature development graphs (in particular, the difference value for each of the first to fourth graphs, for example) as a number on the screen of the user terminal (30), thereby enabling the user (roster) to intuitively check the temperature difference.

[0161] The control unit (13) can enable the user to check the difference in RoR values ​​on the service provision screen, and can also enable the service provision screen to check the difference in heat by looking at the difference in how the temperature has changed from when the user initially started, and can also enable the user to determine in advance, before the first crack occurs, information on how quickly the desired discharge temperature can be raised when the first crack occurs. In this way, the control unit (13) can enable the user to implement the uniform coffee taste he or she wants without failure by providing the service provision screen to the user, and can conveniently perform attempts to implement a more delicious and better coffee taste.

[0162] In addition, the present device (10) can provide a function for standardizing the specifications of sensors within a roasting machine (20), and in relation to this, it is possible to standardize the difficulty in sharing accurate data due to sensors being installed in different locations and specifications for each roasting machine, and when installing automatic roasting hardware within a roasting machine, it is possible to standardize the specifications of sensors so that they can be installed (installed).

[0163] In addition, the present device (10) can provide a function to display a preheating calorie index in the roasting record, and in this regard, after developing a preheating calorie index as mentioned above, it can be shared in a form integrated into the roasting record, and in the case of a record without an index, the value of sharing is reduced, so that it can be established as a de facto standard.

[0164] In addition, the present device (10) can provide a function capable of integrating existing automatic roasting profiles, and in this regard, can create a new format of automatic roasting profiles and support integration of third-party automatic roasting profiles to induce dominant use of the company's automatic roasting profile format.

[0165] In addition, this device (10) can be provided to users by charging or free of charge for the use of this program and this app, for example. In this case, if the use fee is charged, it can be provided at a price lower than the labor cost of one roaster, so that when using this service in a medium or large roastery, convenience can be provided by automating 90% of the roasting time for about 2 million won.

[0166] In addition, the present device (10) can perform a performance evaluation on the results of the present program (or the present app), for example, after developing (producing) the present program (or the present app). At this time, the performance indicators (evaluation items) considered in the performance evaluation may include, for example, a first evaluation item corresponding to the consistency of the temperature graph of green coffee beans when reproducing roasting, a second evaluation item corresponding to the prediction of temperature change due to heat transfer for an arbitrary point in the roasting machine, a third evaluation item corresponding to the error range of the score obtained by sensory evaluation of the results of automatic roasting reproduction, and a fourth evaluation item corresponding to the color consistency rate of the results of coffee beans when reproducing roasting. At this time, the unit of the first evaluation item may be MSE (Mean Squared Error), the unit of the second evaluation item may be % (matching ratio of predicted temperature - measured temperature), the unit of the third evaluation item may be SCA Cupping Form Score, and the unit of the fourth evaluation item may be % (error ratio - 2σ).

[0167] At this time, when the device (10) performs the above performance evaluation, for example, when performing the evaluation for the first evaluation item, as an evaluation method, i) a curve is formed by recording the temperature sensor measurement value of the roasting machine every 0.5 seconds to measure the performance that combines accuracy and reproducibility, and ii) an evaluation method of calculating the Mean Squared Error of the sensor records measured during roasting after attempting 10 roasting reproductions with 3 types of green coffee beans and performing a target evaluation can be used.

[0168] In addition, when performing an evaluation for the second evaluation item, the present device (10) may use an evaluation method of: i) measuring the actual temperature of the front casting plate, drum temperature, and drum shaft of the roasting machine to evaluate whether it matches the expected temperature; ii) randomly performing the supply and cut-off of the heat source in the roasting machine for 10 minutes, measuring the actual temperature result, and comparing it with the expected temperature, and performing a target evaluation based on the result of repeating this process 30 or more times.

[0169] In addition, when performing an evaluation for the third evaluation item, this device (10) may, as an evaluation method, i) conduct a sensory evaluation based on the results of attempting to reproduce roasting 5 times with 4 types of green coffee beans, and ii) at this time, the sensory evaluation may be conducted according to the cupping protocol of the SCA (Specialty Coffee Association), and the score may be calculated using an evaluation method of calculating scores for aroma, acidity, body, etc. using an SCA cupping evaluation sheet. This device (10) may secure objectivity and standardization by using, for example, the SCA cupping form, which is a de facto global standard in coffee sensory evaluation.

[0170] In addition, when performing an evaluation for the fourth evaluation item, this device (10) can use an evaluation method that measures the numerical value of the Agtron Gourmet value, which is a spectrophotometric measurement, by attempting to reproduce roasting five times with four types of green coffee beans.

[0171] In addition, the device (10) can, for example, i) generate roasting data, ii) generate a roasting decision-making model (i.e., a roaster control decision-making model), iii) generate a roasting temperature prediction model (i.e., a temperature prediction engine), iv) generate a full temperature distribution model (i.e., a heat prediction model) which is a temperature distribution prediction algorithm, and v) enhance an AI engine to provide the service.

[0172] At this time, the present device (10) can, i) generate a dataset by performing roasting with fixed initial variables when generating roasting data, and can also generate a temperature dataset by heat source - machine part (i.e., by machine part according to heat source), and can measure temperature change data. In addition, the present device (10) can, ii) reconstruct and label existing roasting records so that they can be learned, and learn ML (Machine Learning) / DL (Deep Learning) models when generating a roasting decision-making model. In addition, the present device (10) can, iii) construct a model utilizing a LightTS time-series deep learning model when generating a roasting temperature prediction model, and supplement (re-learn) the constructed model so that it can be predicted when a control variable is added.

[0173] In addition, the present device (10) can construct a model utilizing the LightTS time series deep learning model when generating a full temperature distribution model (i.e., a heat prediction model) which is a temperature distribution prediction algorithm, research an optimal ensemble algorithm, develop a heat quantification index, and enable automatic preheating and automatic roasting by standardizing the preheating value through the temperature distribution prediction algorithm. In addition, the present device (10) can v) evaluate the AI ​​engine itself, check supplementary items, and take additional measures to enhance the AI ​​engine.

[0174] Below, based on the detailed description above, we will briefly examine the operational flow of the present invention.

[0175] Figure 15 is a flowchart of an automated coffee roasting method according to one embodiment of the present invention.

[0176] The coffee roasting automation method illustrated in Fig. 15 can be performed by the device (10) described above. Therefore, even if the details are omitted below, the description of the device (10) can be equally applied to the description of the coffee roasting automation method.

[0177] Referring to Fig. 15, in step S21, the automatic preheating unit can preheat the roasting machine. At this time, the automatic preheating unit can automatically preheat the roasting machine to a target calorie corresponding to a target calorie numerical value input by the user (roaster).

[0178] Next, in step S22, the automatic roasting unit can roast the green coffee beans (coffee beans) fed into the roasting machine. At this time, the automatic roasting unit can automatically proceed with roasting when the roasting machine reaches the target heat level.

[0179] In the above description, steps S21 and S22 may be further divided into additional steps or combined into fewer steps, depending on the implementation of the present invention. Furthermore, some steps may be omitted as needed, and the order of the steps may be changed.

[0180] A coffee roasting automation method according to an embodiment of the present invention may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., either singly or in combination. The program commands recorded on the medium may be those specially designed and configured for the present invention or may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. The above hardware devices may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

[0181] Additionally, the aforementioned coffee roasting automation method can also be implemented in the form of a computer program or application executed by a computer stored in a recording medium.

[0182] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will readily appreciate that the present invention can be readily modified into other specific forms without altering the technical spirit or essential characteristics of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single entity may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined manner.

[0183] The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention.

Claims

1. Automatic preheating unit for preheating the roasting machine; and An automatic roasting unit that roasts the green coffee beans fed into the roasting machine. An automated coffee roasting device comprising:

2. In paragraph 1, The above automatic preheating unit automatically preheats the roasting machine to the target calorie corresponding to the target calorie value input by the user, The above automatic roasting unit is an automated coffee roasting device that automatically performs the roasting when the roasting machine reaches the target heat.

3. In paragraph 1, The above automatic roasting unit, A coffee roasting automation device, wherein, during the roasting process, monitoring of the roasting machine is performed, and if it is determined from the results of the monitoring that control of the roasting machine is necessary, control information for the roasting machine is derived using a predefined roasting decision model, and the operation of the roasting machine is controlled according to the derived control information.

4. In paragraph 3, The above automatic roasting unit, A coffee roasting automation device, wherein, when a roasting control event caused by user input is detected as a result of the above monitoring, the control information is derived by using a first artificial intelligence model that has been learned based on a control record for the roasting machine among the above-defined roasting decision-making models to predict the control device and control value.

5. In paragraph 3, The above automatic roasting unit, A coffee roasting automation device, wherein, when a difference is detected by comparing the sensor chart deformation information judged from the results of the above monitoring with the reference sensor chart deformation information, the control information is derived by using a second artificial intelligence model that has been previously learned to predict the control device and the control value based on the sensor chart deformation information corresponding to the roasting record of the roasting machine among the above-definition roasting decision-making models.

6. A coffee roasting automation method using the coffee roasting automation device of Article 1, In the automatic preheating section, a step of preheating the roasting machine; and In the automatic roasting unit, a step of roasting the green coffee beans fed into the roasting machine. A coffee roasting automation method comprising:

Images