A temperature control method of a coffee machine temperature control self-learning algorithm

Abstract

The present application relates to coffee machine intelligent temperature control technology field, specifically to a kind of coffee machine temperature control self-learning algorithm temperature control method, comprising: by collecting the switch signal and temperature sequence of each heating unit, identify and according to the temperature inertia slip segment attenuation form after heating stops to the dynamic clustering grouping of heating unit.In homogeneous temperature control group, with contrast unit as benchmark to build attenuation template, calculate the form deviation sequence of learning unit, and interweave and fuse it with its own heating switch signal, generate timing compensation signal. Comprehensive group compensation signal forms collaborative control strategy, realizes adaptive temperature control.The method realizes self-organizing grouping according to actual thermal characteristics, and converts the characteristic difference into timing level accurate compensation, improves the temperature consistency and control accuracy between multiple heating units.

Description

Technical Field

[0001] This invention relates to the field of intelligent temperature control technology for coffee machines, and in particular to a temperature control method based on a self-learning algorithm for temperature control in coffee machines. Background Technology

[0002] Existing coffee machine temperature control technologies mostly employ fixed-parameter temperature control strategies, such as preset heating power curves or PID feedback regulation based on a single temperature sensor. These solutions typically treat multiple heating units as an ideal and uniform thermal system, using unified or pre-grouped control logic. When faced with variations in actual thermal characteristics due to manufacturing tolerances, component aging, or differences in usage environments, the system struggles to effectively differentiate and compensate for these variations. Some advanced solutions attempt to improve performance by calibrating or learning the parameters of individual heating units, but their learning targets are often static heating efficiency or thermal equilibrium temperature, failing to deeply analyze and utilize the unit characteristic information inherent in the dynamic temperature decay process after heating stops.

[0003] The main drawback of existing technologies lies in the rigidity of their grouping and compensation mechanisms. Grouping methods based on physical location or factory presets cannot reflect the dynamic thermal characteristics differences of heating units during actual operation, resulting in deviations in temperature control behavior among different units within the same group. Conventional compensation methods mostly focus on correcting offsets in target temperature or heating power; this compensation is static or scalar and fails to deeply couple with the core timing sequence of heating control—the switching signal sequence. Therefore, in scenarios requiring multiple heating units to work collaboratively to achieve a uniform and precise temperature field, such as coffee brewing, existing technologies struggle to achieve highly consistent adaptive temperature control, impacting final product quality and energy efficiency. This invention aims to address two core issues: how to self-organize grouping based on the actual dynamic thermal response characteristics of heating units, and how to transform identified characteristic differences into precise dynamic compensation for heating control timing. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a temperature control method based on a self-learning algorithm for temperature control in coffee machines.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a temperature control method for a coffee machine using a self-learning temperature control algorithm, comprising: A time-temperature sample set for the heating process of a coffee machine is established, the time-temperature sample set including the heater switching signal sequence of each of the multiple heating units and the corresponding temperature sensor sampling sequence; For each heating unit, the temperature change rate is extracted from its temperature sensor sampling sequence, and combined with the heater switching signal sequence to mark the temperature inertial slip segment after the heater is turned off; Based on the temperature decay pattern of all temperature inertial slip segments marked by each heating unit, the heating units are clustered, and heating units belonging to the same cluster are classified into homogeneous temperature control groups. Within each homogeneous temperature control group, any one heating unit is designated as the control heating unit, and the other heating units in the same group are designated as the learning heating units; A control attenuation template is constructed based on the temperature attenuation pattern of the temperature inertia slip segment of the control heating unit. The temperature inertia slip segment of each learning heating unit is morphologically compared with the control attenuation template to output a morphological deviation sequence. The morphological deviation sequence of each learning heating unit is interleaved and fused with its own heater switching signal sequence to generate a switching timing compensation signal; Within the homogeneous temperature control group, the collaborative heating control strategy of the homogeneous temperature control group is generated by integrating the switching timing compensation signals of all the learning heating units. The aforementioned collaborative heating control strategy is applied to perform adaptive temperature control on the target coffee machine during the heating process.

[0006] Preferably, the clustering of heating units based on the temperature decay pattern of all temperature inertial slip segments marked for each heating unit specifically includes: For each heating unit, the initial rate of decrease and the final stable temperature of the temperature decay curve for each temperature inertial slip segment are extracted; The initial descent rate and the final steady-state temperature of the same inertial slip segment are combined into a morphological feature vector. Calculate the average Mahalanobis distance between all morphological feature vectors of any two heating units, and classify the two heating units whose average Mahalanobis distance is less than a set threshold into the same homogeneous temperature control group.

[0007] Preferably, the construction of the control attenuation template specifically includes: Obtain the temperature inertial slip segment in the temperature sensor sampling sequence of the control heating unit, and extract the starting temperature point and ending temperature point of these slip segments; With the starting temperature point as the zero moment and the ending temperature point as the end moment, the temperature sampling sequences of all temperature inertial sliding segments are normalized and aligned on the time axis. Calculate the mean and variance of the temperature values ​​of all temperature inertial slip segments at each sampling time after normalization and alignment; Using the mean as the baseline and the variance as the tolerance band, a baseline for the temperature change of the control attenuation template and its upper and lower fluctuation ranges are generated.

[0008] Preferably, the output morphological deviation sequence specifically includes: For any temperature inertia slip segment of the learning heating unit, the temperature value at each sampling moment is obtained under the same normalized time scale; Calculate the difference between the temperature value of the learning heating unit at each sampling time and the temperature change baseline of the corresponding sampling time of the control attenuation template; Divide the difference by the floating range boundary value of the corresponding sampling time of the reference attenuation template to obtain the instantaneous morphological deviation of the learning heating unit at each sampling time in the temperature inertial slip segment; The instantaneous morphological deviations are arranged in chronological order to form a morphological deviation sequence of the learning heating unit corresponding to this temperature inertial slip segment.

[0009] Preferably, the step of interleaving and fusing the morphological deviation sequence of each learned heating unit with its own heater switching signal sequence specifically includes: Identify specific shutdown signals associated with the start point of the temperature inertial slip segment in the heater switching signal sequence of the learning heating unit; A one-to-one mapping relationship is established between the specific shutdown signal and the morphological deviation sequence of the corresponding temperature inertial slip segment in the time dimension; For each specific shutdown signal, extract the maximum shape deviation value and the rising edge duration of the shape deviation from its corresponding shape deviation sequence; The absolute occurrence time of the specific shutdown signal is adjusted according to the maximum morphological deviation value, and the duration of the specific shutdown signal is adjusted according to the rising edge duration of the morphological deviation, thereby generating a switching timing compensation signal.

[0010] Preferably, the method of generating a collaborative heating control strategy for the homogeneous temperature control group by integrating the switching timing compensation signals of all learning heating units specifically includes: Collect the switching timing compensation signals of all learning heating units within the homogeneous temperature control group for each specific shut-off signal type; Statistical analysis was performed on the time adjustment amount in the switching timing compensation signal of all learning heating units under the same specific shutdown signal type to obtain the consistency index between the median of the time adjustment amount and the adjustment direction. When the consistency index of the adjustment direction exceeds the preset consistency threshold, the median of the time adjustment amount is used as the collaborative correction amount of the homogeneous temperature control group for this specific shut-off signal type. The coordinated correction values ​​corresponding to all specific shutdown signal types are aggregated to form the instruction set of the coordinated heating control strategy.

[0011] Preferably, the adaptive temperature control operation for the heating process of the target coffee machine specifically includes: Identify the specific type of shutdown signal that occurs in the target coffee machine during the current heating cycle; The coordinated correction amount matching the specific shutdown signal type is queried from the instruction set of the coordinated heating control strategy; Based on the retrieved collaborative correction amount, the current heater switch control logic of the target coffee machine is corrected in real time using feedforward. The modified heater switching control logic is executed, and the actual temperature decay pattern is recorded after the next temperature inertia slip segment ends.

[0012] Preferably, the method further includes the following steps: The actual temperature decay pattern recorded by the target coffee machine is fed back to the homogeneous temperature control group to which it belongs; Based on the new actual temperature decay pattern, update the control decay template and the morphological deviation sequence of the learning heating unit in the homogeneous temperature control group. Based on the updated morphological deviation sequence, the interleaving and fusion steps are re-executed to iteratively update the collaborative heating control strategy.

[0013] Preferably, for each heating unit, extracting the temperature change rate from its temperature sensor sampling sequence and combining it with the heater switching signal sequence to mark the temperature inertial slip segment after the heater is turned off specifically includes: Monitor the heater switching signal sequence. When a signal edge transitioning from the on state to the off state is detected, record this signal edge as the sliding segment trigger point. Extract the starting sampling point from the temperature sensor sampling sequence after the sliding segment trigger point, where the temperature value begins to decrease continuously; Calculate the temperature difference between multiple consecutive sampling points starting from the initial sampling point. When the absolute value of the temperature difference is continuously lower than a set small change threshold, determine this moment as the end point of the sliding segment. The temperature sampling sequence between the starting sampling point and the end point of the slip segment is defined as a temperature inertial slip segment.

[0014] Preferably, the calculation of the average Mahalanobis distance between all morphological feature vectors of any two heating units specifically includes: For any two heating units, obtain the set of all morphological feature vectors contained in each of them; Calculate the Mahalanobis distance from any morphological feature vector of the first heating unit to the set of all morphological feature vectors of the second heating unit, and take the minimum value as the shortest distance from the morphological feature vector to the second set. Repeat the steps for all morphological feature vectors of the first heating unit and calculate the arithmetic mean of all nearest distances as the average Mahalanobis distance from the first heating unit to the second heating unit.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: By analyzing the temperature inertia slip segment of each heating unit during its natural cooling phase after the heater is turned off, and clustering them based on their specific temperature decay curve shapes, heating units with similar thermal inertia and heat dissipation characteristics can be automatically grouped into the same cluster. This overcomes the limitations of relying on physical layout or preset parameters for grouping, achieving adaptive classification based on real-time, dynamic thermal behavior characteristics. Based on this, the system can more accurately identify groups of units with essentially similar temperature control responses, establishing an accurate object basis for subsequent collaborative control and improving the starting point for consistent thermal behavior among units within a group.

[0016] The morphological differences between the learning unit and the control unit in the temperature inertia slip segment are quantified into a temporal deviation sequence. This sequence is then interleaved and fused with the original heating switching signal sequence of the learning unit itself to directly generate a new switching timing compensation signal. This process means that the compensation behavior is no longer a simple addition or subtraction of temperature values ​​or adjustment of power percentage, but rather a fine-tuning and reshaping at the timing level of control commands. It can dynamically advance, delay, or fine-tune the timing and width of specific heating pulses based on differences in attenuation patterns, thereby accurately offsetting the differences in heating and cooling processes caused by different thermal characteristics in the time domain. This fusion generation mechanism allows the compensation action to be seamlessly integrated with the original control logic, achieving refined, timing-level intervention in the heating process and improving the synchronization and tracking accuracy of the temperature trajectories of all heating units within the group. Attached Figure Description

[0017] Figure 1 This is a flowchart of the temperature control method of the coffee machine temperature control self-learning algorithm described in this invention; Figure 2 This is a flowchart illustrating the clustering of heating units based on temperature decay patterns. Figure 3 A flowchart constructed for the control attenuation template; Figure 4 A trend chart showing the changes in temperature control error and consistency index under strategy iteration updates; Figure 5 Box plot of the iteration rounds and temperature stability of the self-learning algorithm for temperature control of coffee machine. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0020] See Figure 1 A time-temperature sample set for the coffee machine heating process is established, containing heater switching signal sequences and corresponding temperature sensor sampling sequences for each heating unit. For each heating unit, the temperature change rate is extracted from its temperature sensor sampling sequence, and the temperature inertia slip segment after the heater is turned off is marked by combining it with the heater switching signal sequence. Heating units are clustered based on the temperature decay patterns of all marked temperature inertia slip segments for each heating unit, and heating units belonging to the same cluster are classified into a homogeneous temperature control group. Within each homogeneous temperature control group, any heating unit is designated as a control heating unit, and the other heating units in the same group are designated as learning heating units. A control decay template is constructed based on the temperature decay pattern of the control heating unit's temperature inertia slip segment. The temperature inertia slip segment of each learning heating unit is compared with the control decay template to output a pattern deviation sequence. The pattern deviation sequence of each learning heating unit is interleaved and fused with its own heater switching signal sequence to generate a switching timing compensation signal. Within the homogeneous temperature control group, the switching timing compensation signals of all learning heating units are combined to generate a collaborative heating control strategy for that homogeneous temperature control group. A collaborative heating control strategy is applied to perform adaptive temperature control on the target coffee machine during the heating process.

[0021] In one embodiment of the present invention, see [reference] Figure 2 The clustering operation of the heating units is performed based on the morphological features of the temperature inertia slip segments. For each heating unit, the system extracts the initial rate of temperature decay and the final stable temperature of the temperature decay curve for each temperature inertia slip segment. The initial rate of temperature decay is obtained by linearly fitting the temperature values ​​of consecutive sampling points at the beginning of the temperature inertia slip segment, and the final stable temperature is the temperature value that remains stable after the end of the temperature inertia slip segment. The initial rate of temperature decay and the final stable temperature of the same temperature inertia slip segment are combined to form a morphological feature vector representing the decay pattern. For a heating unit, all temperature inertia slip segments marked in its historical heating processes will generate a set of morphological feature vectors.

[0022] In some embodiments, the average Mahalanobis distance of morphological similarity between two heating units is calculated. For any two heating units to be compared, such as heating unit A and heating unit B, it is necessary to obtain the sets of all morphological feature vectors contained in heating unit A and heating unit B, respectively. The Mahalanobis distance from any morphological feature vector of heating unit A to the set of all morphological feature vectors of heating unit B is calculated. This step involves calculating the Mahalanobis distance between a single morphological feature vector of heating unit A and every morphological feature vector in the set of heating unit B, and selecting the minimum value from all results. This minimum value is recorded as the closest distance from the morphological feature vector to the second set.

[0023] It is understandable that the above nearest distance calculation process is repeated for all morphological feature vectors of heating unit A. Assuming heating unit A has N morphological feature vectors, N nearest distance values ​​will be obtained. The arithmetic mean of these N values ​​is calculated, and this arithmetic mean is defined as the average Mahalanobis distance from heating unit A to heating unit B. The formula for calculating the average Mahalanobis distance can be expressed as: Where: symbol The symbol represents the average Mahalanobis distance from heating unit A to heating unit B. The symbol represents the total number of morphological feature vectors of heating unit A. Indicates the first heating unit A A morphological feature vector, symbol Indicates the first heating unit B A morphological feature vector, symbol The covariance matrix represents the set of morphological eigenvectors of heating unit B, and the operation... Represents the index of all morphological feature vectors for heating unit B. Take the minimum value, sign This represents the transpose operation of a vector. The average Mahalanobis distance from heating unit B to heating unit A is calculated using the same method described above.

[0024] Optionally, a distance threshold can be set as a clustering criterion. The calculated average Mahalanobis distance is compared with the set threshold. When the average Mahalanobis distance from heating unit A to heating unit B is less than the set threshold, and the average Mahalanobis distance from heating unit B to heating unit A is also less than the set threshold, heating unit A and heating unit B are determined to have similar decay patterns. The system groups all heating units whose average Mahalanobis distances to each other are less than the set threshold together to form a homogeneous temperature control group. Heating units within each homogeneous temperature control group are considered to have consistent thermodynamic response characteristics.

[0025] In one embodiment of the present invention, see [reference]Figure 3 The construction of the control attenuation template begins with acquiring all temperature inertia slip segments from the temperature sensor sampling sequence designated as the control heating unit. Each temperature inertia slip segment contains continuous temperature sampling data from the starting temperature point to the ending temperature point. The system extracts the starting and ending temperature points of these temperature inertia slip segments. Taking the sampling time corresponding to the starting temperature point as time zero and the sampling time corresponding to the ending temperature point as time end, the system performs normalization and alignment processing on the time axis for all temperature inertia slip segments. This processing maps temperature inertia slip segments with different absolute durations to a unified normalized time scale.

[0026] In some embodiments, after normalization and alignment, at each identical normalized sampling time, the system calculates the mean and variance of the temperature values ​​of all temperature inertia slip segments of the control heating unit at that time. The mean is used to characterize the typical trajectory of temperature decay, and the variance is used to describe the dispersion of the typical trajectory. Using the calculated mean as a benchmark and the calculated variance as a tolerance band, the system generates a temperature change baseline and its upper and lower fluctuation ranges for the control decay template. The temperature change baseline is a series of means connected in chronological order, and the upper and lower fluctuation ranges are formed by superimposing or subtracting the variance values ​​from the baseline. For any temperature inertia slip segment of a learning heating unit, the system first normalizes it according to the same rules to place it under the same normalized time scale as the control decay template, and then obtains the temperature value of the learning heating unit at each normalized sampling time of the temperature inertia slip segment.

[0027] It is understandable that the calculation of morphological deviation involves a point-by-point comparison between the temperature trajectory of the learning heating unit and the baseline of the control attenuation template. The difference between the temperature value of the learning heating unit at each normalized sampling moment and the corresponding temperature change baseline value of the control attenuation template at the same moment is calculated. This difference is then compared with the floating range boundary value of the control attenuation template at the same moment. The floating range boundary value can be either an upper or lower boundary value, depending on whether the temperature value of the learning heating unit is higher or lower than the baseline. The result is the instantaneous morphological deviation of the learning heating unit at each sampling moment in that temperature inertial slip segment. Arranging the instantaneous morphological deviations of all normalized sampling moments in the temperature inertial slip segment in chronological order forms the morphological deviation sequence of the learning heating unit for that specific temperature inertial slip segment.

[0028] Optionally, the instantaneous morphological deviation can be calculated using a formula: Where: symbol Indicates the learning heating unit at the normalized time index instantaneous morphological deviation at the location, symbol This indicates the temperature inertia slip segment of the learning heating unit in normalized time index. Temperature value at, symbol Indicates the normalized time index of the control decay template. Temperature change baseline value, sign Indicates the normalized time index of the control decay template. The floating range boundary value at that point, when hour, Take the upper boundary value, when hour, Take the lower boundary value. The morphological deviation sequence of a complete temperature inertial slip segment of the learning heating unit can be represented as: ,in It is the total number of sampling points after normalization.

[0029] In one embodiment of the present invention, the system identifies a specific shutdown signal in the heater switching signal sequence of the learning heating unit that is directly associated with the starting point of a temperature inertial slip segment. This specific shutdown signal refers to a control command event that causes the heater to stop heating during the operation of the heating unit, and this event precisely corresponds to the start of a temperature inertial slip segment on the time axis. A one-to-one mapping relationship is established between each specific shutdown signal and the corresponding morphological deviation sequence of the generated temperature inertial slip segment in the time dimension, forming a "signal-morphology" association pair. This mapping relationship allows each shutdown action to be associated with the morphological characteristics of the resulting temperature decay.

[0030] In some embodiments, for each specific shutdown signal in the mapping relationship, the system extracts key feature parameters from its corresponding shape deviation sequence. These key feature parameters include the maximum shape deviation value and the duration of the shape deviation rising edge. The maximum shape deviation value is the maximum absolute value of all instantaneous shape deviation values ​​in the shape deviation sequence, and the duration of the shape deviation rising edge is the time elapsed from the start of the shape deviation sequence until it first reaches the maximum shape deviation value. Based on the extracted maximum shape deviation value, the absolute occurrence time of the specific shutdown signal is adjusted; for example, a positive maximum shape deviation value may indicate that the shutdown signal needs to be issued earlier. Based on the extracted rise edge duration, the duration of the specific shutdown signal is adjusted; for example, a longer rise edge duration may indicate that the preset duration of the signal needs to be extended. After these two adjustments, the system generates a switching timing compensation signal for that specific shutdown signal.

[0031] It is understandable that generating a collaborative heating control strategy requires aggregating the individual experiences of all learning heating units within the same homogeneous temperature control group. The system collects the switching timing compensation signals generated by each learning heating unit within the homogeneous temperature control group for each specific shutdown signal type. Specific shutdown signal types can be categorized based on the temperature threshold triggering shutdown or the heating stage. Statistical analysis is performed on the time adjustment amounts contained in the switching timing compensation signals of all learning heating units under the same specific shutdown signal type. These time adjustment amounts include adjustments for the signal occurrence time and adjustments for the signal duration. The median of the time adjustment amounts is calculated through statistical analysis, and a consistency index for the adjustment direction is calculated. This consistency index quantifies the degree to which the adjustment recommendations of all learning heating units under this type tend to be earlier or later, longer or shorter in direction.

[0032] Optionally, the consistency index for adjusting the direction can be calculated using the following formula: Where: symbol This represents the consistency index of the adjustment amount of all learning heating units for a specific time period under a particular shut-off signal type. The symbol is... Indicates the number of learning heating units within the same temperature control group, symbol Indicates the first The suggested time adjustment value for each learning heating unit, symbol... It is a symbolic function used to extract... The sign of the calculated consistency index. When the absolute value exceeds the preset consistency threshold, it indicates that the adjustment recommendations of most learning heating units are consistent. In this case, all time adjustment values ​​are used. The median value is used as the collaborative correction amount for this specific shut-off signal type by the homogeneous temperature control group. The system summarizes the collaborative correction amounts corresponding to all different specific shut-off signal types to form an instruction set for the collaborative heating control strategy. The instruction set exists in the form of a data table or rule set, which clearly lists what kind of time correction should be applied under what signal type.

[0033] In one embodiment of the present invention, when performing adaptive temperature control on the target coffee machine during the heating process, the system first identifies a specific shutdown signal type occurring in the target coffee machine during the current heating cycle. The identification of the specific shutdown signal type is based on the temperature threshold that triggers heater shutdown or the program logic stage. The system queries the instruction set of the generated cooperative heating control strategy for a cooperative correction amount that matches the identified specific shutdown signal type. The cooperative correction amount is stored in the form of data records, including adjustment values ​​for the timing of the shutdown signal and adjustment values ​​for the duration of the signal. Based on the queried cooperative correction amount, the system performs real-time feedforward correction on the current heater switching control logic of the target coffee machine. The feedforward correction operation is to directly superimpose or apply a time offset earlier / later on the preset control logic output.

[0034] In some embodiments, the system executes modified heater switching control logic to drive the heater to operate according to a modified timing sequence. After completing a heating cycle and generating the next temperature inertia slip segment, the system records the actual temperature decay pattern generated by the target coffee machine during this heating process. The actual temperature decay pattern is stored in the form of a temperature sampling sequence under a normalized time scale. The system feeds back the actual temperature decay pattern recorded by the target coffee machine to its corresponding homogeneous temperature control group, using it as new sample data to participate in the group's data update. Based on the newly fed-back actual temperature decay pattern, the system updates the control decay template within the homogeneous temperature control group. The update operation involves incorporating the sampled data of the new pattern into the original temperature inertia slip segment set and recalculating the temperature mean and variance at each normalized time point.

[0035] Understandably, after the reference attenuation template is updated, the morphological deviation sequence of the learning heating units within the group also needs to be recalculated. Based on the updated reference attenuation template, the system recalculates the morphological deviation sequence corresponding to each historical and newly generated temperature inertia slip segment of each learning heating unit. According to the updated morphological deviation sequence, the system re-executes the interleaving and fusion step, that is, it again interleaves and fuses the new morphological deviation sequence of each learning heating unit with its own heater switching signal sequence to generate a new switching timing compensation signal. Based on the newly generated switching timing compensation signals of all learning heating units, the system re-performs statistical analysis and generates a new collaborative correction quantity, thereby completing one iterative update of the collaborative heating control strategy.

[0036] Optionally, the instruction set for the coordinated heating control strategy can be stored and queried in tabular form. See Table 1 for a simplified example of an instruction set.

[0037] Table 1: Command Table for Coordinated Heating Control Strategy During the strategy update process, the impact of the new sample on the control attenuation template can be incorporated in a weighted manner, and the normalized time after the update can be used to calculate the impact. reference temperature value It can be calculated using the following formula: Where: symbol This indicates the time after the update at the normalization moment. Temperature change baseline value, symbol This indicates the time before the update at the normalization moment. Temperature change baseline value, symbol This indicates the actual temperature decay pattern of the new feedback at the normalization time. Temperature value, symbol It is a forgetting factor between 0 and 1, used to control the weight ratio of historical data to new data.

[0038] See Figure 4 In the strategy iteration and update phase of the coffee machine temperature control self-learning algorithm, the system continuously optimizes the collaborative heating control strategy through multiple iterations. As clearly shown in the graph, the temperature control error (bar) exhibits a significant decreasing trend as the iteration count progresses from the initial version to the fourth iteration: the initial temperature control error was approximately 4.2℃, which decreased to approximately 0.9℃ after four iterations. This indicates that each strategy update effectively improves temperature control accuracy, reflecting the optimization effect of the collaborative correction on the heater switching timing. Simultaneously, the consistency index (line graph), representing the degree of uniformity in the adjustment direction of heating units within the group, gradually increases from an initial 0.65 to 0.95. This indicates that as iterations deepen, the interweaving and fusion effect of the morphological deviation sequence of each learned heating unit within the homogeneous temperature control group and the heater switching signal continuously improves, and the synergy of the adjustment direction is continuously enhanced. This change verifies the effectiveness of the "feedback-update-iteration" mechanism in the algorithm: new temperature decay patterns are continuously incorporated into the control decay template, making the calculation of morphological deviation more accurate, thereby generating more consistent switching timing compensation signals, ultimately driving the collaborative heating control strategy towards a better direction.

[0039] In one embodiment of the present invention, the process of marking the temperature inertial slip segment after the heater is turned off begins with continuous monitoring of the heater switching signal sequence. The system scans the state transitions of the heater switching signal sequence in real time. When a signal edge is detected in the heater switching signal sequence transitioning from the on state to the off state, the system immediately records the occurrence time of this signal edge and marks it as the slip segment trigger point. The system extracts the data segment after the slip segment trigger point from the temperature sensor sampling sequence. In this data segment, the system searches for the starting sampling point where the temperature value begins to show a continuous decreasing trend. The criterion for determining the starting sampling point is that the temperature values ​​of multiple consecutive sampling points show a monotonically decreasing trend.

[0040] In some embodiments, the system starts from a determined initial sampling point and calculates the temperature difference between multiple consecutive sampling points. The temperature difference is calculated by subtracting the temperature value of the previous sampling point from the temperature value of the next sampling point. The system continuously calculates and checks the absolute value of the temperature difference for each sampling interval. When the system detects that the absolute value of the temperature difference is consistently below a predefined small change threshold (a predefined positive constant), the system determines that the temperature decrease process is stalling and identifies the temperature sampling point at this moment as the end point of the sliding segment. The complete temperature sampling sequence between the initial sampling point and the end point of the sliding segment is defined as a temperature inertial sliding segment. This temperature sampling sequence contains complete data from the start of the inertial temperature decrease after the heater is turned off until the temperature reaches a stable state again.

[0041] Understandably, accurately defining the temperature inertia slip segment relies on real-time analysis of the temperature difference sequence between consecutive sampling points. The system maintains a sliding analysis window containing the temperature differences of multiple most recent consecutive sampling points. A consistency check is performed on the absolute values ​​of all temperature differences within the window to confirm whether the temperature change has consistently fallen below a small change threshold. The determination of the slip segment's end point is not based on a single temperature difference falling below the threshold, but rather requires that the absolute values ​​of all temperature differences within a continuous window length satisfy the condition of being below the threshold, thus ensuring the robustness of the determination that the temperature has entered a stable phase.

[0042] Optionally, the logic for determining the threshold of minute changes can be expressed as a formula to calculate the temperature change assessment within the sliding window. : Where: symbol This represents the average absolute temperature change calculated within a specific sliding window, with the symbol... The fixed length of the sliding window, i.e., the number of consecutive sampling points, is indicated by the symbol. This indicates the index number of the sliding window's starting point in the temperature sensor's sampling sequence, symbol [symbol missing]. This indicates that the index number in the temperature sensor sampling sequence is The sampling points and index numbers are The temperature difference between the sampling points, i.e. The system will calculate the average absolute temperature change. Compared with the preset threshold of minute changes Compare, when satisfied When the condition is met, the system determines the index. The corresponding sampling points are used as candidate end points of the slip segment, and the stability conditions are further verified to finally confirm the end point of the slip segment.

[0043] See Figure 5 In the iterative optimization phase of the coffee machine's temperature control self-learning algorithm, temperature stability changes are quantified using a box plot of the algorithm iteration cycle and temperature standard deviation. The vertical axis represents the temperature standard deviation (°C), reflecting the dispersion of temperature fluctuations during heating; a smaller value indicates higher temperature stability. The horizontal axis represents the algorithm iteration cycle, covering the complete optimization period from iteration 1 to iteration 5. From iteration 1 to iteration 5, the median (red line) of the box plot shows a significant downward trend. The median in iteration 1 is approximately 1.55°C, while in iteration 5 it drops to approximately 0.82°C, indicating that as the number of algorithm iterations increases, the dispersion of temperature fluctuations continuously decreases, and temperature control stability is systematically improved. Simultaneously, the box height (interquartile range) of the box plot gradually narrows. The interquartile range in iteration 1 is approximately 0.2°C, while in iteration 5 it shrinks to approximately 0.08°C, indicating enhanced consistency in temperature fluctuations during heating, and a convergence in the temperature decay patterns of each heating unit within the homogeneous temperature control group. Furthermore, the number and degree of deviation of outliers in the figure decrease with the increase of iteration rounds. Iteration 1 has outliers that are significantly lower than the median, while iteration 5 has only one outlier with a small deviation, reflecting that the algorithm’s ability to suppress extreme temperature fluctuations gradually increases.

[0044] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A temperature control method using a self-learning algorithm for temperature control in a coffee machine, characterized in that, Includes the following steps: A time-temperature sample set for the heating process of a coffee machine is established, the time-temperature sample set including the heater switching signal sequence of each of the multiple heating units and the corresponding temperature sensor sampling sequence; For each heating unit, the temperature change rate is extracted from its temperature sensor sampling sequence, and combined with the heater switching signal sequence to mark the temperature inertial slip segment after the heater is turned off; Based on the temperature decay pattern of all temperature inertial slip segments marked by each heating unit, the heating units are clustered, and heating units belonging to the same cluster are classified into homogeneous temperature control groups. Within each homogeneous temperature control group, any one heating unit is designated as the control heating unit, and the other heating units in the same group are designated as the learning heating units; A control attenuation template is constructed based on the temperature attenuation pattern of the temperature inertia slip segment of the control heating unit. The temperature inertia slip segment of each learning heating unit is morphologically compared with the control attenuation template to output a morphological deviation sequence. The morphological deviation sequence of each learning heating unit is interleaved and fused with its own heater switching signal sequence to generate a switching timing compensation signal; Within the homogeneous temperature control group, the collaborative heating control strategy of the homogeneous temperature control group is generated by integrating the switching timing compensation signals of all the learning heating units. The aforementioned collaborative heating control strategy is applied to perform adaptive temperature control on the target coffee machine during the heating process.

2. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 1, characterized in that, The clustering of heating units based on the temperature decay patterns of all temperature inertial slip segments marked for each heating unit specifically includes: For each heating unit, the initial rate of decrease and the final stable temperature of the temperature decay curve for each temperature inertial slip segment are extracted; The initial descent rate and the final steady-state temperature of the same inertial slip segment are combined into a morphological feature vector. Calculate the average Mahalanobis distance between all morphological feature vectors of any two heating units, and classify the two heating units whose average Mahalanobis distance is less than a set threshold into the same homogeneous temperature control group.

3. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 2, characterized in that, The construction of the control attenuation template specifically includes: Obtain the temperature inertial slip segment in the temperature sensor sampling sequence of the control heating unit, and extract the starting temperature point and ending temperature point of these slip segments; With the starting temperature point as the zero moment and the ending temperature point as the end moment, the temperature sampling sequences of all temperature inertial sliding segments are normalized and aligned on the time axis. Calculate the mean and variance of the temperature values ​​of all temperature inertial slip segments at each sampling time after normalization and alignment; Using the mean as the baseline and the variance as the tolerance band, a baseline for the temperature change of the control attenuation template and its upper and lower fluctuation ranges are generated.

4. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 3, characterized in that, The output shape deviation sequence specifically includes: For any temperature inertia slip segment of the learning heating unit, the temperature value at each sampling moment is obtained under the same normalized time scale; Calculate the difference between the temperature value of the learning heating unit at each sampling time and the temperature change baseline of the corresponding sampling time of the control attenuation template; Divide the difference by the floating range boundary value of the corresponding sampling time of the reference attenuation template to obtain the instantaneous morphological deviation of the learning heating unit at each sampling time in the temperature inertial slip segment; The instantaneous morphological deviations are arranged in chronological order to form a morphological deviation sequence of the learning heating unit corresponding to this temperature inertial slip segment.

5. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 4, characterized in that, The process of interleaving and fusing the morphological deviation sequence of each learned heating unit with its own heater switching signal sequence specifically includes: Identify specific shutdown signals associated with the start point of the temperature inertial slip segment in the heater switching signal sequence of the learning heating unit; A one-to-one mapping relationship is established between the specific shutdown signal and the morphological deviation sequence of the corresponding temperature inertial slip segment in the time dimension; For each specific shutdown signal, extract the maximum shape deviation value and the rising edge duration of the shape deviation from its corresponding shape deviation sequence; The absolute occurrence time of the specific shutdown signal is adjusted according to the maximum morphological deviation value, and the duration of the specific shutdown signal is adjusted according to the rising edge duration of the morphological deviation, thereby generating a switching timing compensation signal.

6. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 5, characterized in that, The method of generating a collaborative heating control strategy for the homogeneous temperature control group by integrating the switching timing compensation signals of all the learning heating units specifically includes: Collect the switching timing compensation signals of all learning heating units within the homogeneous temperature control group for each specific shut-off signal type; Statistical analysis was performed on the time adjustment amount in the switching timing compensation signal of all learning heating units under the same specific shutdown signal type to obtain the consistency index between the median of the time adjustment amount and the adjustment direction. When the consistency index of the adjustment direction exceeds the preset consistency threshold, the median of the time adjustment amount is used as the collaborative correction amount of the homogeneous temperature control group for this specific shut-off signal type. The coordinated correction values ​​corresponding to all specific shutdown signal types are aggregated to form the instruction set of the coordinated heating control strategy.

7. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 6, characterized in that, The adaptive temperature control operation for the heating process of the target coffee machine specifically includes: Identify the specific type of shutdown signal that occurs in the target coffee machine during the current heating cycle; The coordinated correction amount matching the specific shutdown signal type is queried from the instruction set of the coordinated heating control strategy; Based on the retrieved collaborative correction amount, the current heater switch control logic of the target coffee machine is corrected in real time using feedforward. The modified heater switching control logic is executed, and the actual temperature decay pattern is recorded after the next temperature inertia slip segment ends.

8. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 7, characterized in that, It also includes the following steps: The actual temperature decay pattern recorded by the target coffee machine is fed back to the homogeneous temperature control group to which it belongs; Based on the new actual temperature decay pattern, update the control decay template and the morphological deviation sequence of the learning heating unit in the homogeneous temperature control group. Based on the updated morphological deviation sequence, the interleaving and fusion steps are re-executed to iteratively update the collaborative heating control strategy.

9. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 1, characterized in that, For each heating unit, the temperature change rate is extracted from its temperature sensor sampling sequence, and combined with the heater switching signal sequence to mark the temperature inertial slip segment after the heater is turned off. Specifically, this includes: Monitor the heater switching signal sequence. When a signal edge transitioning from the on state to the off state is detected, record this signal edge as the sliding segment trigger point. Extract the starting sampling point from the temperature sensor sampling sequence after the sliding segment trigger point, where the temperature value begins to decrease continuously; Calculate the temperature difference between multiple consecutive sampling points starting from the initial sampling point. When the absolute value of the temperature difference is continuously lower than a set small change threshold, determine this moment as the end point of the sliding segment. The temperature sampling sequence between the starting sampling point and the end point of the slip segment is defined as a temperature inertial slip segment.

10. The temperature control method of a coffee machine with a self-learning temperature control algorithm according to claim 2, characterized in that, The calculation of the average Mahalanobis distance between all morphological feature vectors of any two heating units specifically includes: For any two heating units, obtain the set of all morphological feature vectors contained in each of them; Calculate the Mahalanobis distance from any morphological feature vector of the first heating unit to the set of all morphological feature vectors of the second heating unit, and take the minimum value as the shortest distance from the morphological feature vector to the second set. Repeat the steps for all morphological feature vectors of the first heating unit and calculate the arithmetic mean of all nearest distances as the average Mahalanobis distance from the first heating unit to the second heating unit.

Images