Coffee machine beverage self-adaptive adjustment method and system based on big data analysis

Abstract

The present application relates to the technical field of coffee beverage adjustment, and particularly relates to a coffee beverage self-adaptive adjustment method and system based on big data analysis, which establishes a nonlinear mapping relationship between control parameters and beverage quality, solves and implements optimal target parameters in an actuator. The method monitors beverage quality in real time, triggers re-optimization when parameters deviate, and controls sensory characteristics of irregular fluctuations. At the same time, big data analysis is combined with user multi-modal portrait data to mine the relationship between parameters to dynamically correct the mapping model. In addition, an optimized knowledge base and case base are established using historical data to form a standardized adjustment scheme, and adaptive control is realized based on knowledge matching, thereby improving the stability of beverage quality and personalized experience.

Description

Technical Field

[0001] This invention relates to the field of coffee machine beverage adjustment technology, and in particular to a coffee machine beverage adaptive adjustment method and system based on big data analysis. Background Technology

[0002] With the rapid development of smart home appliance technology and the increasing demand from consumers for personalized beverage experiences, traditional coffee machines, based on fixed programs or simple parameter adjustments, can no longer meet the diverse and high-quality drinking needs of users. Most existing coffee machines rely on preset parameters such as water temperature, pressure, grind size, and extraction time for standardized brewing, lacking real-time feedback and dynamic adjustment capabilities regarding the actual output quality. This results in inconsistent beverage taste, especially when raw materials (such as coffee bean batches) or environmental conditions change, making it difficult to guarantee consistent flavor. Furthermore, most machines do not fully consider individual user differences, such as taste preferences, drinking habits, and health needs, resulting in limited levels of personalized service.

[0003] In recent years, the integration of big data analytics, artificial intelligence, and the Internet of Things (IoT) has provided new solutions for the intelligent upgrading of coffee machines. Some studies have attempted to collect parameters during the beverage preparation process using sensors and combine them with machine learning models to achieve quality prediction or parameter optimization. However, these studies generally suffer from problems such as insufficient data utilization, poor model adaptability, and weak closed-loop control capabilities. Especially when faced with multi-source heterogeneous data (such as equipment operation data, sensory evaluation data, and user behavior data), there is a lack of effective fusion and analysis mechanisms, making it difficult to construct a precise nonlinear mapping relationship between control parameters and beverage quality. Furthermore, existing systems are still in the early stages of user profile construction, reuse of historical experience knowledge, and dynamic adaptive adjustment, and have not yet formed an intelligent control system with a closed loop from perception, analysis, decision-making, execution, and feedback.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a coffee machine beverage adaptive adjustment method and system based on big data analysis, aiming to solve the technical problem that existing coffee machines lack the ability to achieve stable beverage quality assurance and personalized adaptive adjustment based on deep fusion of multi-source heterogeneous data and user profile-driven dynamic closed-loop control.

[0006] To achieve the above objectives, the present invention provides a method for adaptive adjustment of coffee machine beverages based on big data analysis, the method comprising: Obtain the control parameters of the coffee machine and the beverage quality parameters, and establish a nonlinear mapping relationship between the control parameters and the beverage quality parameters; Based on the nonlinear mapping relationship between the constructed control parameters and the beverage quality parameters, the optimal combination of control parameters is optimized to obtain the target control parameters for achieving the target beverage quality. The optimized target control parameters are sent to the actuator of the coffee machine to collect beverage quality parameters in real time and monitor the process. When the beverage quality parameters deviate, re-optimization is triggered. During the beverage preparation process, beverage quality parameters are monitored, and sensory characteristics are controlled for beverages with irregular or uncontrollable fluctuations in quality parameters. Acquire user multimodal profile data, conduct big data analysis on beverage quality parameters, control parameters and user multimodal profile data in the beverage preparation process, and continuously correct and improve the nonlinear mapping relationship between the control parameters and the beverage quality parameters based on the inherent relationship between each parameter; Historical beverage data under different user profiles is obtained, including historical target control parameter combinations. A knowledge base and case library for control parameter optimization are established to form a standardized beverage control scheme. Based on the standardized beverage control scheme, the control parameters of the coffee machine are adaptively matched and adjusted.

[0007] Optionally, the control parameters include water temperature, pressure, grind size, and extraction time; the beverage quality parameters include taste score and aroma intensity; and the acquisition of the coffee machine's control parameters and beverage quality parameters, and the establishment of a non-linear mapping relationship between the control parameters and the beverage quality parameters, includes: Real-time data of water temperature, pressure, grinding degree and extraction time are collected, and the taste score and aroma intensity are measured. The collected raw data are then normalized. The moving average filtering method is used to remove noise from the normalized data, and outliers are identified by box plot method. Based on the filtered data, a nonlinear mapping relationship between the control parameters and the beverage quality parameters is established using a support vector machine algorithm.

[0008] Optionally, the control parameters include water temperature, pressure, grind size, and extraction time; the beverage quality parameters include taste score and aroma intensity; and the nonlinear mapping relationship between the constructed control parameters and the beverage quality parameters is intelligently optimized using a particle swarm optimization algorithm, wherein the optimization steps include: The four parameters—water temperature, pressure, grinding degree, and extraction time—are encoded into a real number vector according to a preset particle encoding method, and the vector dimension is set according to the parameter accuracy requirements. Obtain the predefined real number vector, set the fitness function of the particle swarm optimization algorithm, and use a normalization method to convert the taste score and aroma intensity into dimensionless indicators; After the fitness function is set, the particle velocity is adjusted using an inertia weight strategy, and the individual learning factor of the particle swarm optimization algorithm is set to a preset value. The position is iteratively updated using a global search method. Based on the iterative process of the particle swarm optimization algorithm, the population size and the maximum number of iterations are set to a preset value. Once the particle swarm optimization algorithm reaches the convergence condition or the maximum number of iterations, the position parameters corresponding to the optimal particle are decoded and used as target control parameters for the quality of the target beverage.

[0009] Optionally, the step of sending the optimized target control parameters to the actuator of the coffee machine, collecting beverage quality parameters in real time and monitoring the process, and triggering re-optimization when the beverage quality parameters deviate includes: The optimized target control parameters are obtained and sent to the automatic actuator of the coffee machine through a preset communication protocol. The coffee machine's water temperature, pressure, grind size, and extraction time are controlled according to the target control parameters, and the water temperature, pressure, grind size, and extraction time are monitored in real time. If the monitored actual parameters deviate from the target control parameters, the input signal of the actuator is dynamically adjusted using a fuzzy PID control algorithm. Sensors are used to continuously collect taste scores and aroma intensity data, and Kalman filtering algorithm is used to obtain the optimal estimated beverage quality parameters. The trend of the optimally estimated beverage quality parameters is calculated using the exponentially weighted moving average method, and anomaly detection is performed by combining it with control chart methods. If the taste score or aroma intensity of three consecutive samples exceeds the preset threshold of the target value, the control parameter re-optimization program is triggered, and the optimal combination of control parameters is recalculated.

[0010] Optionally, the sensory characteristics include body and acidity balance. Monitoring beverage quality parameters during the beverage preparation process, and controlling the sensory characteristics of beverages with irregular or uncontrollable fluctuations in quality parameters, includes: Statistical process control charts are used to monitor the changing trends of beverage quality parameters. The statistical process control charts apply the eight outlier criteria of sigma control charts. Determine whether the discrimination conditions are met based on the eight discrimination criteria described above; If the criteria for discrimination are not met, it is determined that the quality parameters of the beverage do not fluctuate irregularly or are uncontrollable, and the subsequent beverage production process continues. If the anomaly detection criteria are met, it is determined that the beverage quality parameters have irregular fluctuations or are uncontrollable, triggering a sensory characteristic detection and analysis program, and feeding the detection and analysis results back to the coffee machine's control system for adaptive compensation.

[0011] Optionally, the user multimodal profile data includes taste preferences, drinking habits, and health needs indicators. The process of acquiring user multimodal profile data, performing big data analysis on beverage quality parameters, control parameters, and user multimodal profile data during the beverage preparation process, and continuously refining the nonlinear mapping relationship between the control parameters and the beverage quality parameters based on the inherent connections between these parameters includes: Acquire user taste preference data, drinking habit data, and health needs indicators, and obtain control parameters and beverage quality parameters from the coffee machine; Outlier detection is performed on the taste preference data, drinking habit data, health requirement indicators, control parameters, and beverage quality parameters. Outliers are identified using box plots, and data exceeding a preset multiple of the interquartile range are smoothed. Based on the smoothed data, principal component analysis was used to reduce dimensionality and extract the main feature variables. Using the main feature variables, the FP-growth algorithm is used to perform association rule analysis to mine the association rules between the control parameters, the beverage quality parameters and the user multimodal profile, and to obtain the rule that if the water temperature is within a specified range and the pressure is within a specified range, the taste score is above a specified value. The adaptive response surface methodology was used to correct the model. New sampling points were generated through a central composite experimental design. The weighted least squares method was used to fit the nonlinear mapping relationship between the new control parameters and the beverage quality parameters.

[0012] Optionally, the step of acquiring historical beverage data under different user profiles, wherein the historical beverage data includes historical target control parameter combinations, establishing a knowledge base and case base for control parameter optimization, forming a standardized beverage control scheme, and adaptively matching and adjusting the control parameters of the coffee machine based on the standardized beverage control scheme, includes: Historical beverage data under different user profiles is obtained, including historical target control parameter combinations, and cluster analysis is performed using the K-means algorithm. Outlier detection is performed on the data based on the cluster analysis results. Outliers are identified by box plot method, and outliers exceeding the preset multiple interquartile range are replaced with the mean. Establish a rule-based expert knowledge base, which includes rules in the form of if-then; A case library is constructed using a case reasoning algorithm. The case library includes user profiles, combinations of control parameters, and actual beverage effects. The similarity between cases is calculated using Euclidean distance. The system receives a target user profile input, performs rule-based reasoning based on the expert knowledge base, retrieves similar cases from the case library, and obtains target control parameters using a fuzzy comprehensive evaluation method. Based on these target control parameters, the system adaptively matches and adjusts the coffee machine.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a coffee machine beverage adaptive adjustment system based on big data analysis, the system comprising: The parameter modeling module is used to acquire the control parameters of the coffee machine and the beverage quality parameters, and to establish a nonlinear mapping relationship between the control parameters and the beverage quality parameters; The optimization solution module is used to optimize the solution of the optimal combination of the control parameters based on the nonlinear mapping relationship between the control parameters and the beverage quality parameters, so as to obtain the target control parameters to achieve the target beverage quality. The execution monitoring module is used to send the optimized target control parameters to the actuator of the coffee machine, collect beverage quality parameters in real time and monitor the process, and trigger re-optimization when the beverage quality parameters deviate. The anomaly control module is used to monitor beverage quality parameters during the beverage preparation process and to control the sensory characteristics of beverages with irregular or uncontrollable fluctuations in quality parameters. The profile optimization module is used to acquire user multimodal profile data, perform big data analysis on beverage quality parameters, control parameters and user multimodal profile data in the beverage preparation process, and continuously correct and improve the nonlinear mapping relationship between the control parameters and the beverage quality parameters based on the inherent relationship between the parameters. The knowledge matching module is used to acquire historical beverage data under different user profiles. The historical beverage data includes historical target control parameter combinations. A knowledge base and case base for optimizing control parameters are established to form a standardized beverage control scheme. Based on the standardized beverage control scheme, the control parameters of the coffee machine are adaptively matched and adjusted.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a coffee machine beverage adaptive adjustment device based on big data analysis. The device includes: a memory, a processor, and a coffee machine beverage adaptive adjustment program based on big data analysis stored in the memory and executable on the processor. The coffee machine beverage adaptive adjustment program based on big data analysis is configured to implement the steps of the coffee machine beverage adaptive adjustment method based on big data analysis as described above.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a coffee machine beverage adaptive adjustment program based on big data analysis, wherein when the coffee machine beverage adaptive adjustment program based on big data analysis is executed by a processor, it implements the steps of the coffee machine beverage adaptive adjustment method based on big data analysis as described above.

[0016] This invention provides a coffee machine beverage adaptive adjustment method based on big data analysis. The method constructs a nonlinear mapping relationship between control parameters and beverage quality parameters, and combines this with an intelligent optimization algorithm to solve for the optimal combination of control parameters. This achieves a shift from experience-driven to data-driven precise control, significantly improving the stability and consistency of coffee output. By monitoring beverage quality parameters in real time and triggering a re-optimization mechanism when deviations occur, a closed-loop feedback control is formed, effectively addressing the impact of uncertainties such as raw material differences and environmental fluctuations on beverage quality. The invention introduces anomaly detection for abnormal fluctuations in quality parameters such as taste scores and aroma intensity, as well as a sensory characteristic (e.g., body and acidity balance) control mechanism, enhancing the system's adaptability to complex operating conditions and improving quality control accuracy. Furthermore, by integrating multimodal user profile data and utilizing big data analysis to mine the intrinsic correlation between control parameters, quality performance, and user preferences, dynamic correction and continuous optimization of the model are achieved, improving the system's intelligence level. By establishing a knowledge base and case library containing historical target control parameter combinations, standardized control schemes are formed. Combined with case reasoning and fuzzy evaluation, rapid adaptive matching in new user scenarios is achieved, significantly improving personalized service efficiency and user experience satisfaction. The overall solution realizes a closed-loop intelligent system encompassing perception, analysis, decision-making, execution, feedback, and evolution throughout the coffee brewing process, providing efficient and reliable technical support for the personalized, high-quality, and adaptive operation of intelligent beverage equipment. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the coffee machine beverage adaptive adjustment method based on big data analysis according to the present invention. Figure 2 This is a structural block diagram of an embodiment of the coffee machine beverage adaptive adjustment system based on big data analysis according to the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] Reference Figure 1 , Figure 1This is a flowchart illustrating an embodiment of the coffee machine beverage adaptive adjustment method based on big data analysis according to the present invention.

[0021] In one embodiment, the coffee machine beverage adaptive adjustment method based on big data analysis includes: Step S100: Obtain the control parameters of the coffee machine and the beverage quality parameters, and establish a nonlinear mapping relationship between the control parameters and the beverage quality parameters.

[0022] The coffee machine's control parameters can be a set of adjustable variables used to control the coffee-making process, directly influencing the extraction process and final quality of the beverage. In an exemplary embodiment, the control parameters may include, but are not limited to, one or more of the following: water temperature setpoint, pump pressure control value, grind size parameter, and extraction time. Beverage quality parameters can be observable indicators characterizing the sensory or physicochemical properties of the coffee beverage, serving as a basis for evaluating product quality and for feedback adjustment and model training. For example, beverage quality parameters may include, but are not limited to, one or more of the following: taste score, aroma intensity, body, and acidity balance. The nonlinear mapping relationship between the control parameters and beverage quality parameters can be a mathematical or data-driven model describing the complex, nonlinear dependency between the combination of control parameters and beverage quality parameters, supporting the decision-making basis for deriving optimal control parameters from the target quality. In this embodiment, the nonlinear mapping relationship between the control parameters and beverage quality parameters can be constructed using machine learning methods by fitting historical production data, sensor-collected data, and user feedback data. Furthermore, the nonlinear mapping relationship between the control parameters and beverage quality parameters can be used in conjunction with user multimodal profile data for personalized correction, or in conjunction with a knowledge base for initial modeling.

[0023] Step S200: Based on the nonlinear mapping relationship between the constructed control parameters and beverage quality parameters, optimize the solution of the optimal combination of control parameters to obtain the target control parameters for achieving the target beverage quality.

[0024] The target control parameters can be the optimal combination of control parameters calculated to achieve a specific beverage quality target. These parameters can be used as input instructions to the actuator to achieve precise brewing control. In one specific embodiment, the target control parameters can be obtained by solving based on a nonlinear mapping relationship and an optimization algorithm. Optimizing the optimal combination of control parameters can be achieved by searching for the optimal combination of control parameters under the constraint of a nonlinear mapping relationship, with the target quality as the optimization objective. Furthermore, optimizing the optimal combination of control parameters can be achieved through gradient-based continuous optimization algorithms or population-based intelligent optimization algorithms such as genetic algorithms or particle swarm optimization, thereby generating precise control instructions that meet personalized quality requirements. Obtaining the target control parameters to achieve the target beverage quality can be achieved by formatting the optimization results into a parameter vector executable by the device, thus completing the conversion from an abstract target to a specific execution instruction.

[0025] Step S300: The optimized target control parameters are sent to the actuator of the coffee machine to collect beverage quality parameters in real time and monitor the process. When the beverage quality parameters deviate, re-optimization is triggered.

[0026] The actuator of the coffee machine can be a hardware module that receives control commands and performs physical operations, translating target control parameters into actual brewing actions. For example, the actuator can include a water pump control system, heating element, grinding motor, and flow control valve. The optimized target control parameters are sent to the actuator via a device control bus, transmitting parameter commands to each execution module to drive the hardware to execute the brewing process according to the optimal strategy. Real-time acquisition and monitoring of beverage quality parameters can be achieved by continuously acquiring quality-related sensor data and comparing it to expected ranges during the brewing process. Furthermore, real-time acquisition and monitoring of beverage quality parameters can be achieved through online near-infrared spectroscopy analysis of the extract composition or by indirectly inferring the extraction state using pressure-flow curves, enabling dynamic tracking and early warning of anomalies in the production process. Triggering re-optimization when beverage quality parameters deviate can be achieved by restarting the optimization process when monitoring detects that quality parameters exceed the allowable deviation, thus forming a closed-loop feedback mechanism to address raw material or environmental disturbances.

[0027] Step S400: Monitor the quality parameters of the beverage during the beverage preparation process, and control the sensory characteristics of beverages with irregular fluctuations or uncontrollable quality parameters.

[0028] Sensory characteristics can be a set of subjective or semi-objective attributes describing the flavor dimensions of a beverage. These characteristics can be used for anomaly detection and quality control intervention to improve flavor consistency. In an exemplary embodiment, sensory characteristics may include, but are not limited to, one or more of body, acidity balance, and bitterness harmony. Controlling the sensory characteristics of beverages with irregular or uncontrollable quality parameters can involve identifying abnormal patterns in quality parameters and intervening or terminating the process based on preset sensory rules. Furthermore, controlling the sensory characteristics of beverages with irregular or uncontrollable quality parameters can be achieved by triggering intervention based on statistical process control (SPC) anomaly detection rules, or by using an expert rule engine to match anomaly patterns and execute correction strategies, thereby preventing unacceptable flavor deviations from reaching the user.

[0029] Step S500: Obtain user multimodal profile data, perform big data analysis on beverage quality parameters, control parameters and user multimodal profile data in the beverage production process, and continuously correct and improve the nonlinear mapping relationship between control parameters and beverage quality parameters based on the inherent relationship between each parameter.

[0030] The user multimodal profile data can be a structured user feature representation formed by fusing heterogeneous information from multiple sources such as user behavior, preferences, and health. This data can be used to support personalized parameter recommendations and dynamic model correction. In this embodiment, the user multimodal profile data can be collected and fused through user interaction logs, evaluation feedback, user-provided health-related information, or questionnaires. For example, the user multimodal profile data may include, but is not limited to, one or more of the following: taste preference tags, drinking time habits, and health-related restrictions. Obtaining user multimodal profile data can involve extracting and fusing user features from multiple sources such as user accounts, interaction logs, and external health platforms, thereby providing contextual information for personalized modeling.

[0031] Step S600: Obtain historical beverage data under different user profiles. The historical beverage data includes historical target control parameter combinations. Establish a knowledge base and case library for control parameter optimization, form a standardized beverage control scheme, and adaptively match and adjust the control parameters of the coffee machine based on the standardized beverage control scheme.

[0032] Historical beverage data can be a complete set of parameters and results recorded during past production processes, which can be used to build a knowledge base, train mapping models, and perform case-based reasoning. For example, historical beverage data can include, but is not limited to, one or more of the following: successful production records, failed production records, and user-confirmed satisfaction records. Historical target control parameter combinations can be control parameter configurations verified in historical beverage data to effectively achieve target quality, and can be used as the basic unit of standardized solutions for rapid matching in new scenarios. In a specific embodiment, historical target control parameter combinations can constitute the core content of the knowledge base and case library for control parameter optimization. The knowledge base for control parameter optimization can be a structured knowledge system that stores verified and effective control parameter combinations and their corresponding quality performance, and can be used to support parameter initialization and experience reuse under new tasks. In this embodiment, the knowledge base for control parameter optimization can be built by continuously accumulating historical target control parameter combinations and labeling their applicable conditions. The case library for control parameter optimization can be a collection of historical successful cases organized by specific user scenarios, and can be used to support rapid adaptive matching based on similar scenarios. Furthermore, the case library for control parameter optimization can collaborate with a fuzzy evaluation mechanism to achieve case suitability assessment. Standardized beverage control schemes can be general control parameter templates extracted from typical user profiles or beverage types, which can be used to improve the initial adaptation efficiency for new users or new beverage types. In an exemplary embodiment, standardized beverage control schemes can be generated by clustering, summarizing, and abstracting knowledge bases and case bases.

[0033] Taking a new user's first use of personalized coffee service as an example, the adaptive adjustment method for coffee machine beverages based on big data analysis in this embodiment can be as follows: When user A uses the equipment for the first time, the system constructs a multimodal profile of the user (preferring low acidity, medium body, and no caffeine restriction) through questionnaires and health data. The system retrieves historical successful cases of users with similar profiles from the case library, and initially selects a set of parameters in combination with a standardized control scheme; at the same time, it calls a nonlinear mapping relationship model for fine-tuning, generates target control parameters, and issues them for execution. During the preparation process, a near-infrared sensor monitors the composition of the extract in real time. If the acidity is found to be slightly high, a re-optimization mechanism is triggered, and the water temperature and extraction time are adjusted for a second preparation. After completion, the user scores 8 points, and the system stores the complete data (including profile, parameters, quality, and feedback) in the knowledge base for subsequent model correction.

[0034] In one embodiment, the control parameters include water temperature, pressure, grind size, and extraction time; the beverage quality parameters include taste score and aroma intensity. The control parameters of the coffee machine and the beverage quality parameters are acquired, and a non-linear mapping relationship between the control parameters and the beverage quality parameters is established, including: Real-time data on water temperature, pressure, grind size, and extraction time were collected, and taste scores and aroma intensity were measured. The collected raw data were then normalized. The moving average filtering method is used to remove noise from the normalized data, and outliers are identified by box plot method. Based on the filtered data, a nonlinear mapping relationship between the control parameters and the beverage quality parameters is established using the support vector machine algorithm.

[0035] The water temperature can be a setpoint or measured value of the hot water during coffee extraction, which can affect the dissolution rate of coffee components and the extraction ratio of flavor substances. Pressure can be the pressure applied to the water flow through the coffee puck during extraction, which can be used to control the extraction speed and the amount of crema generated. Grind size can be the particle size of the coffee beans after grinding, which can determine the water flow penetration resistance and effective extraction surface area. Extraction time can be the duration from the water flow contacting the coffee grounds to the end of extraction, which can be used to control the total extraction volume and flavor balance. Taste score can be a subjective or semi-structured evaluation indicator of the overall flavor pleasantness of the beverage, which can be used as one of the key dimensions of quality objectives for model training and optimization feedback. Aroma intensity can be the perceived concentration or sensor detection value of volatile aromatic substances in the beverage, which can be used to reflect the freshness and roasting characteristics of the coffee. Real-time data can be the raw parameter stream collected in real time by sensors or the control system during coffee making, which can be used as the original input source for modeling and monitoring. In an exemplary embodiment, real-time data can include, but is not limited to, one or more of the following: equipment runtime sequence data, online sensory evaluation data, and synchronous recording of ambient temperature and humidity.

[0036] Normalized data can be data formatted by scaling original data of different dimensions or magnitudes to a uniform numerical range, used to eliminate feature scale differences and improve model training stability and convergence speed. In this embodiment, normalized data can be obtained by processing the original data using methods such as min-max scaling, Z-score normalization, or fractional scaling. Noise can be random interference or measurement errors superimposed on the real signal, used to reduce data quality and affect model fitting accuracy. For example, noise can include, but is not limited to, one or more of the following: sensor thermal noise, discrete jumps introduced by communication jitter, and instantaneous bias in human scoring. Outliers can be observation samples that significantly deviate from the normal data distribution range, which can be used to prevent model learning bias if not removed. Further, outliers can include, but are not limited to, one or more of the following: extreme pressure readings caused by equipment failure, invalid scores generated by user misoperation, and outlier extraction results caused by sudden environmental changes. Filtered data can be a high-quality dataset after denoising and outlier removal, which can be used as reliable input for modeling nonlinear mapping relationships. In this embodiment, the filtered data can be formed by smoothing the sequence through moving average filtering and removing outliers using box plot rules.

[0037] Support Vector Machine (SVM) algorithm is a supervised learning algorithm based on the principle of minimizing structural risk. It is suitable for high-dimensional nonlinear classification and regression tasks and can be used to construct a nonlinear mapping model with high generalization ability between control parameters and quality parameters. In a specific embodiment, the SVM algorithm can map the input space to a high-dimensional feature space through a kernel function, and find the optimal regression hyperplane in this space. Moving average filtering method is a signal processing technique that smooths fluctuations by calculating the mean of data within a local window of a time series. It can be used to suppress high-frequency noise and preserve data trend characteristics. Box plot method is a data visualization and analysis method based on quartile statistics to identify outliers. It can be used to objectively determine and remove abnormal observations, ensuring the purity of training data. In this embodiment, the box plot method can be implemented by calculating the lower quartile (Q1), upper quartile (Q3), and interquartile range (IQR = Q3 − Q1), and marking points less than Q1 − 1.5 × IQR or greater than Q3 + 1.5 × IQR as outliers.

[0038] The nonlinear mapping relationship between control parameters and beverage quality parameters, in this embodiment, specifically refers to a regression model built using a support vector machine algorithm and filtered data as input. This model can be used to accurately predict taste scores and aroma intensity under a given combination of control parameters, supporting inverse optimization. In this embodiment, the nonlinear mapping relationship between control parameters and beverage quality parameters can rely on filtered data as a training set; its output is used for subsequent optimization to solve for the target control parameters. Real-time data on water temperature, pressure, grind size, and extraction time can be collected by synchronously recording the values ​​of the four control parameters throughout the entire process using the coffee machine's built-in temperature sensor, pressure sensor, grinder encoder, and timing controller. Furthermore, the real-time data on water temperature, pressure, grind size, and extraction time can be collected through multi-sensor fusion and high-frequency sampling, thereby obtaining quantifiable and traceable process execution records to provide input features for modeling. The indicators for measuring taste scores and aroma intensity can be obtained by collecting subjective scores through a user interface or by acquiring objective aroma intensity data through sensing devices such as electronic noses and near-infrared spectrometers. Furthermore, the metrics for measuring taste scores and aroma intensity can be achieved through human-machine collaborative scoring mechanisms or automated sensor arrays, thereby generating high-quality output labels corresponding to the control parameters to support supervised learning.

[0039] Normalizing the collected raw data can be achieved by scaling the raw data of each dimension to [0, 1] or a standard normal distribution range according to a uniform rule. Further, normalization can be achieved by using Min-Max normalization for sensor readings and Z-score normalization for cross-user rating data, thereby eliminating dimensional differences and preventing model training from being dominated by feature scale. Noise removal from the normalized data can be achieved by applying a sliding window average to the normalized data in time series form to smooth instantaneous fluctuations. Further, noise removal from the normalized data can be achieved by using a 5-point simple moving average for pressure signals and an exponentially weighted moving average for aroma intensity time series to preserve trend response, thereby improving the signal-to-noise ratio and enhancing the model's ability to learn the real process-quality relationship. Outliers can be identified using box plots, which can be based on interquartile range statistical rules to automatically mark and remove sample points that deviate from the normal range. Furthermore, outlier identification using the box plot method can be achieved by constructing box plot thresholds for each user profile and using Mahalanobis distance to assist in outlier detection for multidimensional joint distributions. This can eliminate outlier samples caused by equipment malfunctions or human error, thus ensuring the reliability of the training set.

[0040] Based on the filtered data, a nonlinear mapping relationship between control parameters and beverage quality parameters can be established using the support vector machine (SVM) algorithm. This can be achieved by using the filtered data as the training set, configuring a support vector regression model (such as an RBF kernel), and fitting a mapping function from four-dimensional control parameters to two-dimensional quality indicators. Furthermore, establishing this nonlinear mapping relationship based on the filtered data can be achieved by using grid search and cross-validation to select the optimal regularization parameter C and kernel parameter γ, and introducing a multi-output SVR to simultaneously predict taste scores and aroma intensity. This allows for the construction of a nonlinear prediction model with strong generalization capabilities, providing high-fidelity decision-making support for subsequent optimization and closed-loop control.

[0041] Taking adaptive modeling after batch coffee bean replacement as an example, the coffee machine beverage adaptive adjustment method based on big data analysis in this embodiment can be implemented after the user starts the machine with a new batch of Ethiopian coffee beans. The system automatically collects multiple rounds of brewing data under the new bean variety, including real-time trajectories of water temperature, pressure, grind size, and extraction time, as well as the user's taste rating for each cup and the aroma intensity measured by the electronic nose. After normalization, the raw data is filtered using a 7-point moving average to smooth pressure and temperature fluctuations, and an abnormally low score record caused by grinder blockage is removed using a box plot. Finally, based on the cleaned dataset, the system calls the support vector regression model to retrain the nonlinear mapping relationship. The updated model can more accurately predict the optimal parameter combination under this bean variety, ensuring flavor consistency.

[0042] In one embodiment, the control parameters include water temperature, pressure, grind size, and extraction time; the beverage quality parameters include taste score and aroma intensity; a nonlinear mapping relationship between the control parameters and the beverage quality parameters is constructed; and a particle swarm optimization algorithm is used for intelligent optimization, wherein the optimization steps include: The four parameters of water temperature, pressure, grinding degree and extraction time are encoded into real number vectors according to the preset particle encoding method, and the vector dimension is set according to the parameter accuracy requirements. In this context, water temperature can be the setpoint for hot water during coffee extraction, influencing the dissolution rate of coffee components and the extraction ratio of flavor compounds. Pressure can be the pressure applied to the coffee puck by the pumped water flow during extraction, controlling the water flow penetration speed and extraction uniformity. Grind size refers to the particle size of the ground coffee beans, determining the specific surface area and water flow resistance, directly affecting extraction efficiency and the risk of channeling effects. Extraction time is the duration from water contact with the coffee grounds to the end of extraction, controlling the total extraction volume and flavor balance. The real number vector can be an ordered array of continuous real values, used as a particle position representation in particle swarm optimization algorithms. In this embodiment, the real number vector is generated by mapping the four parameters—water temperature, pressure, grind size, and extraction time—to the required precision.

[0043] The vector dimension can be the number of independent variables in a real-number vector, and can be used to determine the search space complexity of the optimization problem. Furthermore, the vector dimension can be determined based on the required numerical precision of each control parameter to determine whether it needs to be split into multiple sub-dimensions. Encoding the four parameters—water temperature, pressure, grinding degree, and extraction time—as real-number vectors can be achieved by mapping each control parameter to one or more dimensions in the real-number domain according to its numerical range and precision. Furthermore, this operation can be achieved by directly mapping continuous parameters to their corresponding dimensions, thereby constructing a continuous search space representation suitable for particle swarm optimization algorithms. Setting the vector dimension according to parameter precision requirements means that if a parameter requires high precision (e.g., 0.1℃), multiple dimensions may be allocated or high-resolution encoding may be used. Furthermore, this operation can be achieved by dynamically adjusting the encoding granularity, thereby achieving a balance between computational efficiency and parameter representation precision.

[0044] Obtain the predefined real number vector, set the fitness function of the particle swarm optimization algorithm, and use the normalization method to convert the taste score and aroma intensity into dimensionless indicators. Among these, the taste score can be a subjective or semi-structured evaluation index of the overall flavor experience of the beverage, and can be used as one of the quality objectives in the optimization process. Aroma intensity can be the perceived concentration or sensor detection value of volatile aromatic substances in the beverage, and can be used to reflect extraction efficiency and freshness, serving as a key quality dimension. The fitness function can be an objective function used to evaluate the merits of particle positions, and can be used to guide the particle swarm towards a high-quality parameter combination. In this embodiment, the fitness function can be a weighted combination constructed based on the normalized taste score and aroma intensity. The normalization method can be a mathematical transformation that converts quality parameters of different dimensions or scales into a unified dimensionless range, which can be used to eliminate dimensional differences and enable multi-objective fusion and comparison. Furthermore, the normalization method can employ Min-Max normalization, Z-score standardization, etc. The dimensionless index can be a standardized quality evaluation value after normalization and removal of physical units, which can be used as input to the fitness function, supporting cross-batch and cross-device consistency optimization. Setting the fitness function for the particle swarm optimization algorithm can be achieved by defining a weighted objective function with normalized taste scores and aroma intensity as inputs. Furthermore, this operation can be implemented through linear weighting or multi-objective Pareto front construction, thereby quantifying multi-dimensional quality objectives and guiding the optimization direction. Normalization methods are used to convert taste scores and aroma intensity into dimensionless indicators, which can be achieved by applying a standardization transformation to the [0, 1] or [-1, 1] interval to the original quality data. Further, this operation can be implemented through Min-Max linear scaling, quantile normalization based on historical distribution, etc., thereby eliminating dimensional and scale differences and supporting multi-objective fusion.

[0045] After the fitness function is set, the particle velocity is adjusted using an inertia weight strategy, and the individual learning factor of the particle swarm optimization algorithm is set to a preset value. The position is iteratively updated using a global search method. The particle swarm optimization (PSO) algorithm can be a metaheuristic optimization algorithm based on swarm intelligence, capable of efficiently solving for optimal combinations of control parameters in a nonlinear, non-convex parameter space. PSO can approximate the global optimum by initializing the particle swarm and iteratively updating its velocity and position based on the historical best positions of individuals and the entire population. The inertia weight strategy can be a control mechanism that dynamically adjusts the degree to which particle velocity inherits historical motion trends, balancing the algorithm's global exploration and local exploitation capabilities. Particle velocity can be the direction and step size vector of particle movement in the search space, used to determine the increment of the next position update. In this embodiment, particle velocity can be influenced by the inertia weight strategy, individual learning factors, and the population's optimal position. The individual learning factor can be a weight coefficient controlling the particle's learning towards its own historical best position, used to adjust the intensity of the influence of individual experience on the current position update. Furthermore, the individual learning factor can be set to a preset constant value to simplify parameter tuning and ensure convergence stability. The global search method can be a strategy where particles update by referencing the historical best position of the entire population in each iteration, used to accelerate convergence and increase the probability of finding the global optimum.

[0046] Position iteration update can be a process of progressively correcting the current position of a particle based on its velocity vector, which can be used to gradually optimize parameter combinations in the search space. Adjusting particle velocity using an inertia weight strategy can involve dynamically reducing the inertia weight during iteration, emphasizing exploration in the early stages and focusing on development in the later stages. Furthermore, this operation can be implemented using a preset decay function, thereby improving the algorithm's convergence speed and solution accuracy. Setting the individual learning factor of the particle swarm optimization algorithm to a preset value can be achieved by fixing hyperparameters such as c1 to empirical values ​​(e.g., 1.5 or 2.0). Furthermore, this operation can be implemented through configuration files or firmware constants, thereby simplifying configuration and enhancing the algorithm's stability in embedded environments. Using a global search method for position iteration update can involve each particle referencing the historical best position of the entire population during update. Furthermore, this operation can be achieved by sharing globally optimal position information, thereby accelerating convergence and avoiding getting trapped in local suboptimal conditions.

[0047] Based on the iterative process of the particle swarm optimization algorithm, the population size and the maximum number of iterations are set to a preset value. The population size can be the total number of parallel search individuals in the particle swarm, and can be used to influence algorithm diversity and computational overhead. In this embodiment, the population size can be set to a preset value to achieve a balance between search breadth and resource consumption. The maximum number of iterations can be the upper limit of the maximum number of generations the algorithm is allowed to run, and can be used to prevent infinite loops and control optimization time. The maximum number of iterations can be set to a preset value to adapt to the real-time requirements of embedded devices. The convergence condition can be a termination criterion for determining whether the algorithm has found a satisfactory solution, and can be used to terminate invalid iterations early to improve efficiency. Setting the population size to a preset value can be achieved by setting the number of particles (e.g., 20–50) according to the device's computing power limitations. Furthermore, this operation can be automatically matched by the system resource detection module, thereby controlling memory and computing resource usage. Setting the maximum number of iterations to a preset value can limit the algorithm to run a maximum of N generations (e.g., 100 generations). Furthermore, this operation can be implemented through hard limiting by the task scheduler, thereby ensuring real-time response capability and preventing timeouts.

[0048] Once the particle swarm optimization algorithm reaches the convergence condition or the maximum number of iterations, the position parameters corresponding to the optimal particle are decoded and used as the target control parameters for the quality of the target beverage.

[0049] The optimal particle can be the individual particle with the highest fitness value in the current or historical iterations, and its position can represent the best combination of control parameters found so far. The position parameter can be the coordinates of the optimal particle in the search space, which can be directly used as the target control parameter after decoding and executed. Decoding the position parameter corresponding to the optimal particle can be achieved by inversely mapping the real vector of the optimal particle back to the original control parameter value according to the encoding rules. Furthermore, this operation can be implemented through table lookup or linear inverse transformation, thereby restoring the executable physical control command. Using the position parameter as the target control parameter for the target beverage quality can be achieved by formatting the decoding result and transmitting it to the actuator control module. Furthermore, this operation can be implemented through a communication interface or internal message queue, thereby completing the closed-loop connection from optimization results to equipment action.

[0050] Taking the extraction deviation caused by a new batch of coffee beans as an example, the adaptive adjustment method for coffee machine beverages based on big data analysis in this embodiment can be as follows: When a user uses newly arrived Ethiopian light roast beans, the system detects that the aroma intensity is lower than the historical average. After triggering the re-optimization mechanism, the particle swarm optimization algorithm is started: water temperature (88–94℃), pressure (7–10 bar), grind size (digital setting 3–7), and extraction time (20–35s) are encoded as 4-dimensional real vectors; the taste score and aroma intensity are normalized by Min-Max to form a fitness function; a linearly decreasing inertial weight is used, the individual learning factor is set to 1.8, the population size is 30, and the maximum number of iterations is 80. After the algorithm converges, the optimal parameters are decoded as water temperature 91.2℃, pressure 8.5 bar, grind size 5.3, and extraction time 28s, and the algorithm is executed, successfully restoring the expected flavor performance.

[0051] In one embodiment, the optimized target control parameters are sent to the actuator of the coffee machine, beverage quality parameters are collected in real time and the process is monitored, and re-optimization is triggered when the beverage quality parameters deviate, including: The optimized target control parameters are obtained and sent to the automatic actuator of the coffee machine through a preset communication protocol. The preset communication protocol can be a predefined data transmission format and interaction rules between devices, used to ensure that the target control parameters are accurately and reliably sent to the automatic actuator. In this embodiment, the preset communication protocol may include, but is not limited to, one or more of the Modbus protocol, CAN bus protocol, and MQTT protocol. The automatic actuator of the coffee machine can be an electromechanical integrated execution module with autonomous response to control commands, used to receive the target control parameters and automatically adjust the water temperature, pressure, grind size, and extraction time. For example, the automatic actuator of the coffee machine may include an intelligent temperature control heating module, a closed-loop pressure pump system, a stepper motor-driven grinder, and an electronic flow meter timing unit. Sending the target control parameters to the automatic actuator of the coffee machine through the preset communication protocol can be achieved by encapsulating the target control parameters according to the protocol specification and transmitting them to the actuator controller through the communication interface. Furthermore, this operation can ensure the integrity and real-time performance of the parameter commands, avoid transmission distortion, and thus achieve the technical effect of high-fidelity command transmission.

[0052] The coffee machine's water temperature, pressure, grind size, and extraction time are controlled according to the target control parameters, and the water temperature, pressure, grind size, and extraction time are monitored in real time. The actual parameters can be physical quantity measurements fed back in real time by sensors during execution, which can be compared with the target control parameters to determine execution deviations. In an exemplary embodiment, the actual parameters may include measured water temperature, actual pump pressure, current grind size, and elapsed extraction time. Controlling the coffee machine's water temperature, pressure, grind size, and extraction time according to the target control parameters can be achieved by an automatic actuator parsing the target parameters and driving the corresponding hardware module to execute the set actions. Furthermore, this operation can achieve the technical effect of synchronous and precise control of multi-dimensional process parameters. Real-time monitoring of water temperature, pressure, grind size, and extraction time can be achieved by continuously collecting the actual values ​​of each physical quantity through built-in sensors and feeding them back to the control unit. Furthermore, this operation can provide a technical effect of providing real-time basis for deviation detection and dynamic correction.

[0053] If the actual parameters monitored deviate from the target control parameters, the input signal of the actuator is dynamically adjusted using a fuzzy PID control algorithm. The fuzzy PID control algorithm can be a composite control strategy combining fuzzy logic and traditional PID control, and can be used to handle the dynamic adjustment of nonlinear, time-varying systems. In a specific embodiment, the fuzzy PID control algorithm can be used to dynamically adjust the actuator input signal to reduce steady-state error and suppress overshoot when the actual parameter deviates from the target control parameter. Furthermore, the fuzzy PID control algorithm can construct a fuzzy rule base based on the deviation and its rate of change, and output the PID parameter correction amount in real time. The actuator input signal can be an electrical signal or digital command sent to the actuator to drive its action, and can be used to correct the execution deviation after dynamic adjustment by the fuzzy PID algorithm. For example, the actuator input signal can include a PWM duty cycle signal, an analog voltage control signal, a digital setpoint, etc. If there is a deviation between the monitored actual parameter and the target control parameter, the fuzzy PID control algorithm is used to dynamically adjust the actuator input signal. This can involve calculating the deviation and its rate of change, adjusting the PID gain online through fuzzy inference, and outputting the corrected control signal. Furthermore, this operation can be achieved through parameter self-tuning based on Mamdani-type fuzzy rules or through a continuous output fuzzy controller based on the TSK model, thereby improving the tracking accuracy and anti-interference capability under nonlinear conditions.

[0054] Sensors are used to continuously collect taste scores and aroma intensity data, and Kalman filtering algorithm is used to obtain the optimal estimated beverage quality parameters. The sensor can be a sensing device used to continuously acquire physical or chemical signals related to taste scores and aroma intensity, providing raw quality observation data for Kalman filtering. In an exemplary embodiment, the sensor may include an electronic nose sensor array, a near-infrared spectral sensor, an electrochemical taste sensor, etc. The Kalman filtering algorithm can be a recursive state estimation algorithm used to estimate the true state of the system from a noisy observation sequence. Furthermore, the Kalman filtering algorithm can be used to denoise the taste scores and aroma intensity acquired by the sensor, generating optimally estimated beverage quality parameters. In this embodiment, the Kalman filtering algorithm can minimize the estimation error covariance by iteratively predicting and updating the system state equation and the observation equation. The optimally estimated beverage quality parameters can be denoised and fused quality parameter estimates obtained after Kalman filtering, which can be used as input for trend analysis and anomaly detection. Furthermore, the optimally estimated beverage quality parameters can be used as input for the exponentially weighted moving average method to support the anomaly detection logic of the control chart method.

[0055] Continuously collecting taste scores and aroma intensity data using sensors can be achieved by deploying a multimodal sensor array for online analysis of the extract, outputting raw sensory signals. Furthermore, this operation enables the objective and continuous perception of subjective quality dimensions. Obtaining the optimal estimated beverage quality parameters using a Kalman filter algorithm can be achieved by establishing a state-space model of the quality parameters and recursively estimating them by fusing observations from multiple sensors. Further, this operation can be achieved by extending the Kalman filter to handle nonlinear observations or using unscented Kalman filtering to improve higher-order accuracy, thereby effectively suppressing measurement noise and improving the stability of quality assessment.

[0056] The trend of the optimally estimated beverage quality parameters is calculated using the exponentially weighted moving average method, and anomaly detection is performed by combining it with control chart methods. The exponentially weighted moving average (TWMA) method can be a smoothing algorithm that assigns higher weights to recent observations in time series data. It can be used to capture short-term trend changes in beverage quality parameters, providing a dynamic benchmark for anomaly detection. In a specific embodiment, the TWMA method can include single exponential smoothing, double exponential smoothing, triple exponential smoothing, etc. The trend calculated by the TWMA method can be a smoothed sequence reflecting the recent direction and rate of change of beverage quality parameters, which can be used to provide dynamic center lines and control limits for control chart methods. Furthermore, the trend calculated by the TWMA method can be accumulated over the optimally estimated beverage quality parameters with exponential decay weights. The control chart method can be a graphical tool used in statistical process control to monitor process stability and identify outliers. It can be used to combine the TWMA trend to determine whether beverage quality parameters have shifted significantly. For example, control chart methods can include X-bar control charts, CUSUM control charts, EWMA control charts, etc.

[0057] The trend of the optimally estimated beverage quality parameters is calculated using the exponentially weighted moving average method. This involves applying an exponentially weighted moving average to the filtered quality parameter sequence to generate a smooth trend line. Furthermore, this operation can highlight recent changes and support the effectiveness of dynamic anomaly detection. Anomaly detection can be performed by combining control chart methods, comparing trend values ​​with control limits calculated based on historical data to identify statistically significant deviations. This operation can be further achieved by using an EWMA control chart adapted to the exponentially weighted trend or by using CUSUM to detect small, persistent drifts, thus enabling early and reliable identification of quality degradation.

[0058] If the taste score or aroma intensity of three consecutive samples exceeds the preset threshold of the target value, the control parameter re-optimization program is triggered, and the optimal combination of control parameters is recalculated.

[0059] The preset threshold can be used to determine whether the taste score or aroma intensity significantly deviates from the target quality tolerance boundary, and can be used as one of the criteria for triggering re-optimization. In an exemplary embodiment, the preset threshold may include an upper control limit, a lower control limit, a tolerance bandwidth, etc. The taste score or aroma intensity sampled three times consecutively can be three consecutive quality parameter observation points acquired in a time series, which can be used to avoid accidental triggering by instantaneous disturbances and improve the robustness of re-optimization decisions. The control parameter re-optimization program can be an automated subroutine that restarts the parameter optimization process after detecting a quality anomaly, and can be used to trigger a new round of solving for the optimal combination of control parameters to achieve closed-loop adaptive adjustment. Furthermore, the control parameter re-optimization program can receive the anomaly detection result as input and call the nonlinear mapping relationship model and optimization algorithm module.

[0060] If the taste score or aroma intensity of three consecutive samples exceeds a preset threshold for the target value, a re-optimization procedure for the control parameters is triggered. This can be achieved by setting a counter, which activates the re-optimization process when three consecutive quality estimates exceed the limit. Furthermore, this operation balances sensitivity and robustness, preventing false triggering that could lead to system oscillations. The optimal combination of control parameters is recalculated, which can be done by using a nonlinear mapping model and optimization algorithm to solve for the new target parameters under the current constraints. Furthermore, this operation achieves a closed-loop response from anomaly detection to adaptive parameter adjustment.

[0061] Taking the example of excessive acidity in coffee bean extraction due to batch change, the adaptive adjustment method for coffee machine beverages based on big data analysis in this embodiment can be as follows: A user uses a new batch of lightly roasted beans to make an Americano. After the initial target control parameters are issued, the automatic actuator operates at the set water temperature of 92℃ and pressure of 9 bar. However, due to differences in bean density, the actual extraction flow rate is too fast, leading to increased acidity. The electronic nose sensor continuously collects aroma intensity, and the Kalman filter outputs the optimal estimate. The exponentially weighted moving average shows an upward trend in the acidity index, and the control chart determines that it exceeds the upper control limit three times consecutively. The system triggers a re-optimization program for the control parameters, and the fuzzy PID simultaneously fine-tunes the water pump pressure to slow down the flow rate. The optimization module recalculates based on a nonlinear mapping relationship, suggesting lowering the water temperature to 89℃ and extending the extraction time by 3 seconds. After the new parameters are issued, the acidity of the finished product returns to the normal range.

[0062] In one embodiment, sensory characteristics include body and acidity balance. Monitoring beverage quality parameters during the beverage preparation process and controlling sensory characteristics of beverages with irregular or uncontrollable fluctuations in quality parameters includes: Statistical process control charts were used to monitor the changing trends of beverage quality parameters. The statistical process control charts applied the eight criteria for identifying discrepancies in sigma control charts.

[0063] The statistical process control chart (SPC) can be a statistical tool used to monitor process stability and identify abnormal patterns. With time series as the horizontal axis and quality parameters as the vertical axis, and including a center line and control limits, it can visualize the changing trends of beverage quality parameters and support anomaly judgment based on statistical rules. In this embodiment, the SPC can calculate the mean and standard deviation of beverage quality parameters under historical stable operating conditions to construct upper and lower control limits (typically ±3σ). The eight outlier criteria of the sigma control chart can be a set of standardized statistical rules used to determine whether the process is out of control. Based on the distribution pattern of data points within the control limits, it can be used to structurally identify non-random fluctuations, improving the sensitivity and accuracy of anomaly detection. Monitoring the changing trends of beverage quality parameters using the SPC can be achieved by plotting the real-time collected beverage quality parameters in a time series on a chart with control limits. Furthermore, this operation can be achieved by real-time acquisition of parameters such as pressure, flow rate, and conductivity during the extraction stage using sensors, thereby enabling visualization and quantitative assessment of process stability. Applying the eight outlier criteria of the sigma control chart can be achieved by matching each of the eight statistical rules to the data point sequence on the control chart. Furthermore, this operation can be achieved by using an embedded algorithm module to perform sliding window analysis on the data stream, thereby systematically identifying potential runaway signals and going beyond simple threshold judgment.

[0064] Determine whether the conditions for discrimination are met based on eight criteria for discrimination.

[0065] The anomaly detection condition can be any state where one of the eight anomaly detection criteria of the sigma control chart is met, and can be used as the decision-making basis for whether to initiate the sensory characteristic detection and analysis program. Determining whether the anomaly detection condition is met based on the eight criteria can be done by declaring that the anomaly detection condition is met if any one of the criteria is true. Furthermore, this judgment logic can be executed by a state machine or a rule engine, thereby providing explicit anomaly triggering logic.

[0066] If the criteria for discrepancy are not met, it is determined that the beverage quality parameters do not exhibit irregular fluctuations or are uncontrollable, and the subsequent beverage preparation process continues.

[0067] Irregular or uncontrollable fluctuations in beverage quality parameters can be identified as batches that violate statistical steady-state assumptions and cannot be attributed to normal random variation during the production process. These batches can serve as indicators to trigger in-depth sensory analysis and intervention mechanisms. Determining that beverage quality parameters do not exhibit irregular or uncontrollable fluctuations can be made when all discrepancy criteria have not been triggered. Furthermore, this determination can be recorded by internal status flags within the control system, thus maintaining the current production process and avoiding unnecessary intervention. Continuing subsequent beverage production can be achieved by adjusting parameters according to the original target to complete the remaining brewing stage. Furthermore, this operation can be achieved by maintaining preset water temperature, pressure, and extraction time, thereby ensuring normal process efficiency.

[0068] If the anomaly detection criteria are met, it is determined that the beverage quality parameters have irregular fluctuations or are uncontrollable, triggering a sensory characteristic detection and analysis program, and feeding the detection and analysis results back to the coffee machine's control system for adaptive compensation.

[0069] The sensory characteristic detection and analysis program can be a software module that performs specialized evaluation and analysis of sensory dimensions that are difficult to quantify directly, such as body and acidity balance. This can be used to transform abstract sensory characteristics into controllable parameters. In one specific embodiment, the sensory characteristic detection and analysis program can combine multi-sensor fusion (such as flow rate, pressure, conductivity, and spectrum) with a pre-trained sensory mapping model for inference. Furthermore, the sensory characteristic detection and analysis program can be activated upon receiving an anomaly trigger signal, and its output can be used to drive adaptive compensation. The coffee machine's control system can be a central processing unit that coordinates the operation of various functional modules of the coffee machine. It can be used to receive feedback information and dynamically adjust actuator parameters to achieve adaptive compensation. For example, the coffee machine's control system can include a main control MCU, an embedded Linux system, a real-time operating system (RTOS), etc. Adaptive compensation can be the behavior of dynamically correcting subsequent stage control parameters based on real-time perceived quality deviations during the brewing process. This can be used to suppress sensory characteristic degradation and maintain the target flavor performance. In an exemplary embodiment, adaptive compensation can, based on the sensory characteristic detection and analysis results, call a nonlinear mapping model to back-calculate the correction amount and update the remaining process parameters.

[0070] Determining irregular or uncontrollable fluctuations in beverage quality parameters can be made when at least one discrepancy criterion is met. Furthermore, this determination can activate an anomaly handling path, thereby initiating subsequent intervention mechanisms. Triggering the sensory characteristic detection and analysis program can be achieved by sending a start command to the sensory characteristic analysis module and inputting the current quality parameter sequence. Further, this operation can be implemented through message queues or function callback mechanisms, thereby initiating specific assessments of key flavor dimensions such as body and acidity balance. Feeding the detection and analysis results back to the coffee machine's control system for adaptive compensation can be achieved by inputting the quantified sensory characteristic deviation results into the control system, which then recalculates and issues corrected control parameters. Further, this operation can be implemented by adjusting only the water temperature and flow rate in the remaining extraction stage, or by pausing the current process and replanning the complete parameter combination, thereby achieving closed-loop intervention in sensory quality and improving product consistency.

[0071] Taking batch differences in raw materials leading to extraction anomalies as an example, the adaptive adjustment method for coffee machine beverages based on big data analysis in this embodiment can be as follows: The user uses a new batch of coffee beans to make espresso. During the initial extraction phase, the system continuously collects data using flow and pressure sensors and plots a statistical process control chart. If six consecutive data points show an upward trend starting from the 7th data point, triggering the "six consecutive points of continuous increase" anomaly criterion, the system determines that uncontrollable fluctuations exist and immediately initiates a sensory characteristic detection and analysis program. Combining near-infrared spectroscopy and the extraction curve, it infers that the current acidity is too high and the body is insufficient. This analysis result is fed back to the control system, which dynamically reduces the water temperature by 2°C and extends the pre-soaking time by 3 seconds in subsequent stages. After the final product is confirmed by the user to have a balanced flavor, the system stores this intervention record in the knowledge base for model optimization.

[0072] In one embodiment, the user multimodal profile data includes taste preferences, drinking habits, and health needs indicators. By acquiring this user multimodal profile data, big data analysis is performed on beverage quality parameters, control parameters, and user multimodal profile data during the beverage preparation process. Furthermore, the nonlinear mapping relationship between the control parameters and beverage quality parameters is continuously corrected and improved based on the inherent connections between these parameters. This includes: Acquire user taste preference data, drinking habit data, and health needs indicators, and obtain control parameters and beverage quality parameters from the coffee machine; Taste preference data can be subjective descriptions or ratings of users' preferences for coffee flavor dimensions (such as acidity, bitterness, and body), which can be used for personalized parameter recommendations. Drinking habit data can be user coffee consumption behavior patterns across time, frequency, and scenarios, which can be used to help predict users' current needs and optimize beverage types and parameter presets. For example, drinking habit data may include, but is not limited to, one or more of the following: average daily drinking frequency, common drinking times, and weekend / weekday difference patterns. Health requirement indicators can be constraints or limiting parameters related to the user's health status, which can be used to filter or adjust inappropriate combinations of control parameters to ensure beverage safety and suitability. Furthermore, health requirement indicators may include, but are not limited to, one or more of the following: caffeine intake limits, sugar control requirements, and allergy labels for specific ingredients.

[0073] Acquiring user taste preference data, drinking habit data, and health need indicators can be achieved by collecting structured user characteristics through user registration information, interactive questionnaires, historical evaluations, or third-party health platform interfaces. Furthermore, this operation can be implemented by supporting the fusion of multi-source heterogeneous data, thereby providing fundamental data support for building fine-grained multimodal user profiles. Obtaining control parameters and beverage quality parameters from the coffee machine can be achieved by synchronously reading parameters and sensor feedback from the current or historical production tasks through the device communication interface. Furthermore, this operation can be achieved through real-time data streaming or batch historical record synchronization, thereby ensuring the spatiotemporal alignment of user profiles with actual production data.

[0074] Outlier detection was performed on taste preference data, drinking habit data, health demand indicators, control parameters, and beverage quality parameters. Outliers were identified using box plots, and data exceeding a preset multiple of the interquartile range were smoothed. Outliers can be data samples that significantly deviate from the normal distribution range in a multidimensional parameter space. They can be used as targets for outlier detection to identify data collection errors or extreme cases. In one specific embodiment, outliers can be determined using box plots based on interquartile ranges (IQRs). A preset multiple of the IQR can be a statistical multiple parameter used to define the outlier threshold, controlling the sensitivity of outlier detection and balancing noise removal with information preservation. Box plots are a non-parametric statistical method for outlier identification based on quartiles and IQRs, and can be used to perform robust anomaly detection on multi-source heterogeneous raw data. Furthermore, box plots can calculate the upper and lower quartiles and IQRs for each variable, setting boundaries to identify outliers.

[0075] Outlier detection for taste preference data, drinking habit data, health demand indicators, regulatory parameters, and beverage quality parameters can be achieved by standardizing multi-source data and then applying statistical methods to identify samples deviating from normal patterns. Furthermore, this operation can be implemented using a variable-by-variable or joint multivariate approach, thus preventing erroneous or extreme data from interfering with subsequent modeling and rule mining. Outlier identification using box plots can involve calculating the quartiles and IQRs of each variable and marking data points exceeding the interval [Q1−k×IQR, Q3+k×IQR]. This operation can be further implemented by fixing k=1.5 for general noise filtering or dynamically adjusting the k value to adapt to different data distribution characteristics, thus achieving robust anomaly detection for multidimensional heterogeneous data. Smoothing data exceeding a preset multiple of the interquartile range can involve numerical correction or replacement of identified outliers. This operation can be achieved by replacing with the historical mean of the same user, interpolating with neighboring normal samples, or truncating to a boundary threshold, thus improving the overall consistency of the dataset and the stability of modeling. The smoothed data can be a cleaned dataset after outlier correction or replacement, which can be used to improve the data quality and stability of subsequent modeling inputs. In an exemplary embodiment, the smoothed data can be obtained by processing outliers exceeding a preset multiple of the interquartile range using methods such as nearest neighbor interpolation, truncation, or moving average.

[0076] Based on the smoothed data, principal component analysis was used to reduce dimensionality and extract the main feature variables. Principal Component Analysis (PCA) can be a linear dimensionality reduction technique that transforms high-dimensional correlated variables into low-dimensional uncorrelated principal components through orthogonal transformation. This reduces data dimensionality redundancy and extracts the most explanatory comprehensive features for modeling mapping relationships. PCA can perform eigenvalue decomposition on the covariance matrix, selecting principal components with sufficient cumulative contribution. The principal feature variables can be low-dimensional, high-information comprehensive variables extracted by PCA, which can be used as inputs for association rule mining and model fitting, improving computational efficiency and generalization ability. In a specific embodiment, the principal feature variables can serve as the input data source for the FP-growth algorithm. Dimensionality reduction using PCA on smoothed data can involve constructing a covariance matrix from the cleaned data, extracting principal components, and retaining those with sufficient cumulative variance contribution. Furthermore, this operation can be achieved by setting a cumulative contribution threshold (e.g., 85%) to filter principal components, thereby eliminating redundancy and collinearity among variables and compressing the feature space. Extracting principal feature variables can involve selecting the top few principal components as new features to represent the original high-dimensional data. Furthermore, this operation can be achieved by retaining principal components whose cumulative variance contribution rate exceeds a preset threshold, thereby providing a low-dimensional and efficient input for subsequent correlation analysis.

[0077] Using the main feature variables, the FP-growth algorithm is used to perform association rule analysis to mine the association rules between control parameters, beverage quality parameters and user multimodal profiles, and to obtain the rule that if the water temperature is within a specified range and the pressure is within a specified range, the taste score is above a specified value. The FP-growth algorithm is an efficient candidate-free itemset generation algorithm for mining frequent itemsets and association rules. It can be used to discover interpretable rule patterns from high-dimensional user-parameter-quality data. In an exemplary embodiment, the FP-growth algorithm can recursively mine conditional pattern bases by constructing an FP-tree to compress the transaction database. Association rules can be logical expressions describing the condition-outcome relationship between control parameters, quality parameters, and user profiles. They can be used to provide interpretable personalized control basis, enhancing model transparency and credibility. Using the main feature variables, the FP-growth algorithm can be used for association rule analysis by discretizing the principal component scores into itemsets, constructing an FP-tree, and mining frequent patterns. Furthermore, this operation can be achieved by setting minimum support and confidence thresholds to filter effective rules, thereby achieving the technical effect of discovering potential patterns between user profiles, parameters, and quality. Mining association rules between control parameters, beverage quality parameters, and user multimodal profiles can be achieved by generating implication rules that satisfy minimum support and confidence from frequent itemsets. Furthermore, this operation can be achieved through rule pruning and semantic parsing, thereby obtaining interpretable and personalized control strategy basis. The rule that if the water temperature and pressure are within a set range, the taste score reaches or exceeds a set threshold can be obtained by instantiating the mined high-confidence rules into specific parameter-quality condition statements. Furthermore, this operation can be achieved through rule template matching and parameter binding, thereby providing clear rule-based guidance for the decision-making module.

[0078] The adaptive response surface methodology was used to correct the model. New sampling points were generated through a central composite experimental design. The weighted least squares method was used to fit the nonlinear mapping relationship between the new control parameters and the beverage quality parameters.

[0079] Central composite experimental design (CCD) is a commonly used experimental design strategy in response surface methodology. It includes factor points, axial points, and center points, and can be used to systematically generate representative new sampling points in the parameter space, supporting local fine-grained modeling. These new sampling points can be combinations of control parameters generated by CCD to supplement training data, enhancing the fitting accuracy of nonlinear mapping relationships in key regions. In one specific embodiment, new sampling points can be obtained by distributing points in the parameter space according to CCD rules based on current model uncertainty or the target region. Weighted least squares (WLS) is a regression fitting method that assigns different weights to different data points, prioritizing the fitting accuracy of high-quality or high-confidence samples when fitting nonlinear mapping relationships. Furthermore, WLS can dynamically allocate weights based on data source reliability, user feedback confidence, or sampling density. Adaptive response surface methodology (OSM) is a dynamic modeling method combining experimental design and weighted regression, used for iterative optimization of nonlinear mapping relationships, enabling local fine-grained model updates and online adaptive evolution. In an exemplary embodiment, OSM can rely on CCD to generate sampling points and use WLS to complete the fitting.

[0080] Adaptive response surface methodology (RSM) is used to correct the model. This involves locally updating the existing nonlinear mapping relationship through new sampling and weighted fitting. Furthermore, this operation can be achieved by triggering a model update mechanism in regions of high uncertainty or changing user preferences, thus achieving dynamic and refined model evolution. New sampling points are generated through central composite experimental design (CCD), which involves placing new experimental points around the current optimal parameter region or high uncertainty region according to CCD rules. This operation can be further achieved by collaboratively determining the sampling point locations through parameter space grid partitioning and response gradient estimation, thus achieving efficient exploration of key subdomains in the parameter space. Weighted least squares method is used to fit the nonlinear mapping relationship between new control parameters and beverage quality parameters. This involves jointly modeling the original data and new sampling points, assigning fitting weights based on data quality. Furthermore, this operation can be achieved by assigning high weights to user-satisfied data and high-confidence quality parameters from the sensor, thereby improving the prediction accuracy and robustness of the mapping model in key areas.

[0081] Taking the adaptive model after a gradual change in user taste preferences as an example, the adaptive adjustment method for coffee machine beverages based on big data analysis in this embodiment can be as follows: User B previously preferred a high-acidity, bright flavor, but recently lowered the acidity score multiple times. The system collects the taste preference data, drinking time (all in the evening), and health indicators (unchanged) of the user's three most recent drinks, and merges them with the water temperature, pressure, extraction time, and taste score in the corresponding production record. Box plot analysis reveals an abnormal pressure value (exceeding 2.0 times the IQR) caused by equipment malfunction, which the system smooths to a near-normal value. Principal component analysis is then performed on the 10-dimensional original variables, extracting three main feature variables with a cumulative contribution rate of 85%. The FP-growth algorithm extracts a new rule: "If the water temperature is ≤88℃ and the grind size is ≥ medium-coarse, then the acidity score is ≤3 (out of 5)." Based on this, the system initiates the adaptive response surface methodology, adds 5 new sampling points in the 86–90℃ water temperature range according to the central composite design for trial production, and, combined with user feedback, updates the nonlinear mapping model using the weighted least squares method, so that subsequent recommendations automatically favor the low-temperature extraction strategy.

[0082] In one embodiment, historical beverage data under different user profiles is acquired. This historical beverage data includes historical target control parameter combinations. A knowledge base and case library for control parameter optimization are established to form a standardized beverage control scheme. Based on this standardized beverage control scheme, the control parameters of the coffee machine are adaptively matched and adjusted, including: Historical beverage data was obtained under different user profiles. The historical beverage data included combinations of historical target control parameters. K-means algorithm was used for cluster analysis. Historical beverage data under different user profiles can be a collection of historical production records archived according to user multimodal profile tags. This data can be used as raw input for cluster analysis and knowledge base construction, supporting personalized modeling. In an exemplary embodiment, historical beverage data under different user profiles can include, but is not limited to, historical data from users with high acidity preferences, users with evening drinking habits, and users with health restrictions, among one or more of these. The K-means algorithm, an unsupervised clustering method based on distance metrics, divides the data into K clusters. It can be used to structurally group historical beverage data and identify typical user-parameter patterns. Cluster analysis using the K-means algorithm can be performed by using historical beverage data under different user profiles as input and executing unsupervised clustering. Furthermore, cluster analysis using the K-means algorithm can be achieved by clustering the joint vector of user profiles and control parameters, or by clustering only the combination of control parameters and then associating them with profile tags, thereby revealing typical user-parameter patterns and providing a foundation for standardized schemes.

[0083] Outlier detection is performed on the data based on the cluster analysis results. Outliers are identified by box plot method, and outliers exceeding the preset multiple interquartile range are replaced with the mean. The clustering analysis results can be the sample groups and their cluster center representations output by the K-means algorithm, which can be used for subsequent anomaly detection and standardization scheme refinement. Box plots, a method for visualizing and detecting outliers based on quartile statistics, can be used to identify outliers in historical data that deviate from the normal distribution. A preset multiple of the interquartile range (IQR) can be a threshold coefficient used to define outlier boundaries, typically 1.5 or 3 times the IQR, which can be used to control the sensitivity of anomaly detection and avoid excessive removal of valid data. Mean replacement for outliers involves replacing identified outliers with the mean of their respective clusters or variables, which can be used to preserve data integrity while reducing noise interference. Furthermore, mean replacement for outliers can be implemented by replacing outliers with the within-group mean based on cluster grouping.

[0084] Outlier detection based on cluster analysis results can be performed independently within each cluster to avoid cross-cluster misjudgments. This method improves outlier identification accuracy and prevents global statistics from masking local patterns. Box plots effectively identify extreme values ​​in non-normal distributions. Outliers exceeding a preset multiple of the interquartile range are replaced with the mean, which can be achieved by replacing the outlier value with the mean of the dimension corresponding to its cluster. Furthermore, replacing outliers exceeding the preset multiple of the interquartile range can be achieved by using the median for enhanced robustness, or by replacing with the value of the dimension corresponding to the cluster center, thus purifying the training data while preserving the sample size.

[0085] Establish a rule-based expert knowledge base, which includes rules in the form of if-then; The rule-based expert knowledge base can be a set of structured rules encoding domain expert experience in an if-then form, which can be used to provide interpretable prior guidance in scenarios with sparse data or cold start. In an exemplary embodiment, the rule-based expert knowledge base can include, but is not limited to, one or more of flavor control rules, health adaptation rules, and equipment safety constraint rules. If-then rules can be logical rule expressions consisting of preconditions and execution actions, which can be used to achieve a deterministic mapping from user characteristics to control strategies. For example, if-then rules can take the form of "if the user prefers low acidity, then the water temperature should not exceed 88°C". Establishing a rule-based expert knowledge base can involve converting the experience of baristas or nutrition experts into structured if-then rules and storing them. Furthermore, establishing a rule-based expert knowledge base allows for the introduction of domain knowledge, compensating for the shortcomings of purely data-driven methods.

[0086] A case library is constructed using a case-based reasoning algorithm. The case library includes user profiles, combinations of control parameters, and actual beverage effects. The similarity between cases is calculated using Euclidean distance. The case-based reasoning algorithm can be an analogical reasoning method for solving new problems based on historical similar cases, and can be used to support the transfer of solutions from existing successful experiences. In a specific embodiment, the case-based reasoning algorithm can complete the application of cases through four stages: retrieval, reuse, modification, and retention. The actual beverage effect can be the quality performance of the beverage after it has been prepared, confirmed by sensors or user feedback, and can be used as a basis for verifying the validity of cases, thereby improving the reliability of the case library. For example, the actual beverage effect can include, but is not limited to, one or more of user rating results, sensory evaluation labels, and measured values ​​of physicochemical indicators. Euclidean distance can be a measure of the straight-line distance between two points in a multi-dimensional space, and can be used to quantify the similarity between user profiles and historical cases. Furthermore, Euclidean distance can be implemented by calculating the L2 norm after standardizing the user profile feature vector. Building a case library using the case-based reasoning algorithm can be achieved by organizing the cleaned historical beverage data into triples of <user profile, control parameter combination, actual beverage effect>. Furthermore, building a case library using the case-based reasoning algorithm forms searchable and reusable experience units. Calculating the similarity between cases using Euclidean distance can be achieved by performing vector distance calculations between the target user profile and the profile portions of each case in the case library. Furthermore, calculating the similarity between cases using Euclidean distance can be achieved by using weighted Euclidean distance to highlight key features, or by first reducing dimensionality and then calculating the distance to eliminate redundancy, thereby enabling efficient and interpretable similar case retrieval.

[0087] The system receives the target user profile input, performs rule reasoning based on the expert knowledge base, retrieves similar cases from the case library, and obtains the target control parameters by combining the fuzzy comprehensive evaluation method. Based on the target control parameters, the system performs adaptive matching and adjustment of the coffee machine.

[0088] The target user profile input can be a multimodal feature representation of the current user to be served, which can be used as a query condition for adaptive matching. In an exemplary embodiment, the target user profile input can drive the expert rule reasoning and case retrieval process. Similar cases can be several historical cases in the case library with the smallest Euclidean distance to the target user profile, which can be used to provide a reusable set of control parameter candidates. Furthermore, similar cases can serve as input candidate schemes for fuzzy comprehensive evaluation. The fuzzy comprehensive evaluation method can be a decision-making method based on fuzzy mathematics theory to weighted evaluate multi-factor and multi-index schemes, which can be used to integrate preferences of multiple dimensions such as taste, health, and efficiency to generate comprehensive optimal target control parameters. In a specific embodiment, the fuzzy comprehensive evaluation method can operate by constructing a set of evaluation factors, a set of weights, and a membership function, and obtaining a comprehensive evaluation value through fuzzy synthesis operations.

[0089] Receiving target user profile input can involve obtaining multimodal profile features of the current session from the user's client. Furthermore, receiving target user profile input can trigger the personalized matching process. Rule reasoning based on an expert knowledge base can match target user profile features with if-then rule premises, activating rule conclusions that meet the conditions. Furthermore, rule reasoning based on the expert knowledge base can quickly narrow down the feasible parameter space, improving subsequent retrieval efficiency. Retrieving similar cases from the case library can involve selecting the Top-N cases with the smallest Euclidean distance as candidates. Furthermore, retrieving similar cases from the case library can obtain validated and effective historical solutions.

[0090] The target control parameters are derived by combining fuzzy comprehensive evaluation methods. This can be achieved by using rule-based reasoning results and control parameters from similar cases as candidate solutions, and then performing fuzzy evaluation based on multi-dimensional preference weights. Furthermore, the target control parameters can be derived by constructing a three-factor evaluation matrix (taste, health, and efficiency) or dynamically adjusting the weights to reflect the user's current situation, thereby generating comprehensive optimal control parameters that take multiple objectives into account. Adaptive matching and adjustment of the coffee machine based on these target control parameters can be achieved by issuing the finalized control parameters to the execution mechanism. Furthermore, adaptive matching and adjustment of the coffee machine based on these target control parameters completes a closed-loop adaptation from user profiling to physical brewing.

[0091] Taking a cold-start user's first request for low-acid coffee as an example, the coffee machine beverage adaptive adjustment method based on big data analysis in this embodiment can be as follows: New user B is using the machine for the first time, and the profile shows "prefers low acidity, has no caffeine restriction, and usually drinks in the morning". The system first matches the expert knowledge base rule "if you prefer low acidity, the water temperature should not be higher than 89℃ and the extraction time should not exceed 25 seconds", initially limiting the parameter range; then, it searches the case library for the 5 similar cases with the closest Euclidean distance, and finds that 3 of them use the 88℃ / 22-second combination and the user rating is ≥8; the system merges the rule suggestion and case parameters into a candidate set, and selects 88℃ / 23 seconds as the target control parameter after evaluation by fuzzy comprehensive judgment (taste weight 0.5, efficiency 0.3, health 0.2); after execution, the user feedback is good, and the record is stored in the case library for future use.

[0092] In addition, refer to Figure 2 To achieve the above objectives, the present invention also provides a coffee machine beverage adaptive adjustment system based on big data analysis, the system comprising: The parameter modeling module 10 is used to acquire the control parameters of the coffee machine and the beverage quality parameters, and to establish a nonlinear mapping relationship between the control parameters and the beverage quality parameters. The optimization and solution module 20 is used to optimize and solve the best combination of the control parameters based on the nonlinear mapping relationship between the control parameters and the beverage quality parameters, so as to obtain the target control parameters to achieve the target beverage quality. The execution monitoring module 30 is used to send the optimized target control parameters to the actuator of the coffee machine, collect beverage quality parameters in real time and monitor the process, and trigger re-optimization when the beverage quality parameters deviate. The anomaly control module 40 is used to monitor beverage quality parameters during the beverage preparation process and to control the sensory characteristics of beverages with irregular fluctuations or uncontrollable quality parameters. The profile optimization module 50 is used to acquire user multimodal profile data, perform big data analysis on beverage quality parameters, control parameters and user multimodal profile data in the beverage making process, and continuously correct and improve the nonlinear mapping relationship between the control parameters and the beverage quality parameters based on the inherent relationship between the parameters. The knowledge matching module 60 is used to acquire historical beverage data under different user profiles. The historical beverage data includes historical target control parameter combinations. A knowledge base and case base for optimizing control parameters are established to form a standardized beverage control scheme. Based on the standardized beverage control scheme, the control parameters of the coffee machine are adaptively matched and adjusted.

[0093] Other embodiments or specific implementations of the coffee machine beverage adaptive adjustment system based on big data analysis described in this invention can be referred to the above-mentioned method embodiments, and will not be repeated here.

[0094] Furthermore, to achieve the above objectives, the present invention also provides a coffee machine beverage adaptive adjustment device based on big data analysis. The device includes: a memory, a processor, and a coffee machine beverage adaptive adjustment program based on big data analysis stored in the memory and executable on the processor. The coffee machine beverage adaptive adjustment program based on big data analysis is configured to implement the steps of the coffee machine beverage adaptive adjustment method based on big data analysis as described above.

[0095] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a coffee machine beverage adaptive adjustment program based on big data analysis, wherein when the coffee machine beverage adaptive adjustment program based on big data analysis is executed by a processor, it implements the steps of the coffee machine beverage adaptive adjustment method based on big data analysis as described above.

[0096] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for adaptive adjustment of coffee machine beverages based on big data analysis, characterized in that, The method includes: Obtain the control parameters of the coffee machine and the beverage quality parameters, and establish a nonlinear mapping relationship between the control parameters and the beverage quality parameters; Based on the nonlinear mapping relationship between the constructed control parameters and the beverage quality parameters, the optimal combination of control parameters is optimized to obtain the target control parameters for achieving the target beverage quality. The optimized target control parameters are sent to the actuator of the coffee machine to collect beverage quality parameters in real time and monitor the process. When the beverage quality parameters deviate, re-optimization is triggered. During the beverage preparation process, beverage quality parameters are monitored, and sensory characteristics are controlled for beverages with irregular or uncontrollable fluctuations in quality parameters. Acquire user multimodal profile data, conduct big data analysis on beverage quality parameters, control parameters and user multimodal profile data in the beverage preparation process, and continuously correct and improve the nonlinear mapping relationship between the control parameters and the beverage quality parameters based on the inherent relationship between each parameter; Historical beverage data under different user profiles is obtained, including historical target control parameter combinations. A knowledge base and case library for control parameter optimization are established to form a standardized beverage control scheme. Based on the standardized beverage control scheme, the control parameters of the coffee machine are adaptively matched and adjusted.

2. The coffee machine beverage adaptive adjustment method based on big data analysis as described in claim 1, characterized in that, The control parameters include water temperature, pressure, grind size, and extraction time; the beverage quality parameters include taste score and aroma intensity; and the acquisition of the coffee machine's control parameters and beverage quality parameters, and the establishment of a non-linear mapping relationship between the control parameters and the beverage quality parameters, includes: Real-time data of water temperature, pressure, grinding degree and extraction time are collected, and the taste score and aroma intensity are measured. The collected raw data are then normalized. The moving average filtering method is used to remove noise from the normalized data, and outliers are identified by box plot method. Based on the filtered data, a nonlinear mapping relationship between the control parameters and the beverage quality parameters is established using a support vector machine algorithm.

3. The coffee machine beverage adaptive adjustment method based on big data analysis as described in claim 1, characterized in that, The control parameters include water temperature, pressure, grind size, and extraction time. The beverage quality parameters include taste score and aroma intensity. The nonlinear mapping relationship between the control parameters and the beverage quality parameters is intelligently optimized using a particle swarm optimization algorithm. The optimization steps include: The four parameters—water temperature, pressure, grinding degree, and extraction time—are encoded into a real number vector according to a preset particle encoding method, and the vector dimension is set according to the parameter accuracy requirements. Obtain the predefined real number vector, set the fitness function of the particle swarm optimization algorithm, and use a normalization method to convert the taste score and aroma intensity into dimensionless indicators; After the fitness function is set, the particle velocity is adjusted using an inertia weight strategy, and the individual learning factor of the particle swarm optimization algorithm is set to a preset value. The position is iteratively updated using a global search method. Based on the iterative process of the particle swarm optimization algorithm, the population size and the maximum number of iterations are set to a preset value. Once the particle swarm optimization algorithm reaches the convergence condition or the maximum number of iterations, the position parameters corresponding to the optimal particle are decoded and used as target control parameters for the quality of the target beverage.

4. The coffee machine beverage adaptive adjustment method based on big data analysis as described in claim 1, characterized in that, The process of sending the optimized target control parameters to the actuator of the coffee machine, collecting beverage quality parameters in real time and monitoring the process, and triggering re-optimization when the beverage quality parameters deviate includes: The optimized target control parameters are obtained and sent to the automatic actuator of the coffee machine through a preset communication protocol. The coffee machine's water temperature, pressure, grind size, and extraction time are controlled according to the target control parameters, and the water temperature, pressure, grind size, and extraction time are monitored in real time. If the monitored actual parameters deviate from the target control parameters, the input signal of the actuator is dynamically adjusted using a fuzzy PID control algorithm. Sensors are used to continuously collect taste scores and aroma intensity data, and Kalman filtering algorithm is used to obtain the optimal estimated beverage quality parameters. The trend of the optimally estimated beverage quality parameters is calculated using the exponentially weighted moving average method, and anomaly detection is performed by combining it with control chart methods. If the taste score or aroma intensity of three consecutive samples exceeds the preset threshold of the target value, the control parameter re-optimization program is triggered, and the optimal combination of control parameters is recalculated.

5. The coffee machine beverage adaptive adjustment method based on big data analysis as described in claim 1, characterized in that, The sensory characteristics include body and acidity balance. Monitoring beverage quality parameters during the beverage preparation process, and controlling the sensory characteristics of beverages with irregular or uncontrollable fluctuations in quality parameters, includes: Statistical process control charts are used to monitor the changing trends of beverage quality parameters. The statistical process control charts apply the eight outlier criteria of sigma control charts. Determine whether the discrimination conditions are met based on the eight discrimination criteria described above; If the criteria for discrimination are not met, it is determined that the quality parameters of the beverage do not fluctuate irregularly or are uncontrollable, and the subsequent beverage production process continues. If the anomaly detection criteria are met, it is determined that the beverage quality parameters have irregular fluctuations or are uncontrollable, triggering a sensory characteristic detection and analysis program, and feeding the detection and analysis results back to the coffee machine's control system for adaptive compensation.

6. The coffee machine beverage adaptive adjustment method based on big data analysis as described in claim 1, characterized in that, The user multimodal profile data includes taste preferences, drinking habits, and health needs indicators. The process of acquiring this user multimodal profile data involves big data analysis of beverage quality parameters, control parameters, and user multimodal profile data during the beverage preparation process. Furthermore, it involves continuously refining and improving the nonlinear mapping relationship between the control parameters and the beverage quality parameters based on the inherent connections between these parameters. This includes: Acquire user taste preference data, drinking habit data, and health needs indicators, and obtain control parameters and beverage quality parameters from the coffee machine; Outlier detection is performed on the taste preference data, drinking habit data, health requirement indicators, control parameters, and beverage quality parameters. Outliers are identified using box plots, and data exceeding a preset multiple of the interquartile range are smoothed. Based on the smoothed data, principal component analysis is used to reduce dimensionality and extract the main feature variables. The main feature variables are low-dimensional, high-information comprehensive variables extracted by principal component analysis, which are used as inputs for association rule mining and model fitting. Using the main feature variables, the FP-growth algorithm is used to perform association rule analysis to mine the association rules between the control parameters, the beverage quality parameters and the user multimodal profile, and to obtain the rule that if the water temperature is within a specified range and the pressure is within a specified range, the taste score is above a specified value. The adaptive response surface methodology is used to correct the model. New sampling points are generated through central composite experimental design. The weighted least squares method is used to fit the nonlinear mapping relationship between the new control parameters and the beverage quality parameters.

7. The coffee machine beverage adaptive adjustment method based on big data analysis as described in claim 1, characterized in that, The process involves acquiring historical beverage data under different user profiles, including historical target control parameter combinations. This data is used to establish a knowledge base and case library for control parameter optimization, forming standardized beverage control schemes. Based on these standardized schemes, the control parameters of the coffee machine are adaptively adjusted. Historical beverage data under different user profiles is obtained, including historical target control parameter combinations, and cluster analysis is performed using the K-means algorithm. Outlier detection is performed on the data based on the cluster analysis results. Outliers are identified by box plot method, and outliers exceeding the preset multiple interquartile range are replaced with the mean. Establish a rule-based expert knowledge base, which includes rules in the form of if-then; A case library is constructed using a case reasoning algorithm. The case library includes user profiles, combinations of control parameters, and actual beverage effects. The similarity between cases is calculated using Euclidean distance. The system receives a target user profile input, performs rule-based reasoning based on the expert knowledge base, retrieves similar cases from the case library, and obtains target control parameters using a fuzzy comprehensive evaluation method. Based on these target control parameters, the system adaptively matches and adjusts the coffee machine.

8. A coffee machine beverage adaptive adjustment system based on big data analysis, characterized in that, The system includes: The parameter modeling module is used to acquire the control parameters of the coffee machine and the beverage quality parameters, and to establish a nonlinear mapping relationship between the control parameters and the beverage quality parameters; The optimization solution module is used to optimize the solution of the optimal combination of the control parameters based on the nonlinear mapping relationship between the control parameters and the beverage quality parameters, so as to obtain the target control parameters to achieve the target beverage quality. The execution monitoring module is used to send the optimized target control parameters to the actuator of the coffee machine, collect beverage quality parameters in real time and monitor the process, and trigger re-optimization when the beverage quality parameters deviate. The anomaly control module is used to monitor beverage quality parameters during the beverage preparation process and to control the sensory characteristics of beverages with irregular or uncontrollable fluctuations in quality parameters. The profile optimization module is used to acquire user multimodal profile data, perform big data analysis on beverage quality parameters, control parameters and user multimodal profile data in the beverage preparation process, and continuously correct and improve the nonlinear mapping relationship between the control parameters and the beverage quality parameters based on the inherent relationship between the parameters. The knowledge matching module is used to acquire historical beverage data under different user profiles. The historical beverage data includes historical target control parameter combinations. A knowledge base and case base for optimizing control parameters are established to form a standardized beverage control scheme. Based on the standardized beverage control scheme, the control parameters of the coffee machine are adaptively matched and adjusted.

9. A coffee machine beverage adaptive adjustment device based on big data analysis, characterized in that, The device includes: a memory, a processor, and a coffee machine beverage adaptive adjustment program based on big data analysis stored in the memory and executable on the processor, the coffee machine beverage adaptive adjustment program based on big data analysis being configured to implement the steps of the coffee machine beverage adaptive adjustment method based on big data analysis as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a coffee machine beverage adaptive adjustment program based on big data analysis, which, when executed by a processor, implements the steps of the coffee machine beverage adaptive adjustment method based on big data analysis as described in any one of claims 1 to 7.

Images