Intelligent cooking control system based on multi-sensor fusion

Abstract

The invention discloses an intelligent cooking control system based on multi-sensor fusion, which relates to the technical field of cooking control and comprises a temperature matrix analysis unit, a temperature estimation unit, a cookware identification unit, a target temperature calculation unit, a vector analysis unit and a single-head stove control unit. According to the invention, high-precision temperature field distribution is obtained, each pixel point is processed, and on this basis, the temperature value acquired in real time is subjected to form matching with the optimal cooking curve library, so that the current cooking stage is accurately identified, the power is adjusted in cooperation with a control strategy, each cooking can be smoothly carried out, and the cooking efficiency is improved. According to the method, the materials of the cookware are automatically identified through standard power pulses before cooking, corresponding parameters are loaded, self-adaption to the cookware with different physical characteristics is achieved, meanwhile, the influence of the environment parameters on cooking is compensated in real time through the environment sensor, regional and seasonal differences are eliminated, and it is ensured that users in all regions can obtain the consistent cooking effect.

Description

Technical Field

[0001] This invention relates to the field of cooking control technology, specifically to an intelligent cooking control system based on multi-sensor fusion. Background Technology

[0002] This invention relates to an intelligent cooking control system based on multi-sensor fusion, which achieves precise perception, dynamic decision-making, and automatic control of the cooking process. It digitizes cooking experience and ensures the accuracy, safety, and consistency of heat control in complex environments. Patent application number 202510882915.8 discloses a "big data-based intelligent cooking control system, including a cooking control data acquisition module, a data optimization processing module, a healthy dish intelligent recommendation module, a cooking control parameter optimization module, and an intelligent cooking execution control module." This invention relates to the field of information data processing technology, specifically to a big data-based intelligent cooking control system. This solution innovatively proposes a mechanism for personalized dish recommendations, dual-layered health control, and practical cooking process implementation, significantly improving the health guidance capability, recommendation adaptability, and operational effectiveness of intelligent cooking. It also innovatively proposes constructing a multi-dimensional feature analysis system, designing and improving a multi-layered graph neural structure, and introducing an interactive attention mechanism, achieving a breakthrough in the accuracy of personalized and health-oriented recommendations. Furthermore, it innovatively employs a piecewise sine and cosine chaotic mapping function, a reduction factor, and an improved inertia factor particle swarm optimization algorithm, improving the intelligence level of the cooking control process.

[0003] The aforementioned existing technologies have solved problems such as the lack of personalization in cooking dishes and the mismatch between the nutritional structure of dishes and users' health needs. However, during use, the system cannot sense the actual temperature field of the pot bottom, resulting in large temperature fluctuations, making it easy to burn or undercook. The quality of the dishes heavily depends on the user's experience, and it cannot identify the physical properties of different materials such as iron pots and ceramic pots. Using the same set of parameters to control all pots can easily lead to temperature control imbalance. There is a lack of environmental compensation, and the same recipe may fail completely in different seasons or in different families, failing to meet personalized needs. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent cooking control system based on multi-sensor fusion to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent cooking control system based on multi-sensor fusion, comprising: The temperature matrix analysis unit is used to acquire the raw data vector collected in the current window, align the temperature matrix in each vector to the time axis, clean it, and analyze the corresponding attribute characteristics. The temperature estimation unit is used to read the echo signal sequence at each sampling time in the original data vector and analyze the corresponding ultrasonic temperature estimate according to the signal sequence. The cookware identification unit is used to apply a standard heating power pulse before cooking begins and simultaneously acquire a sequence of actual temperature measurements at the bottom of the cookware, analyze the data, and thus determine the type of cookware. The target temperature calculation unit is used to extract the attribute representation of each temperature matrix and the center point temperature value, and match the corresponding cooking curve segment according to the center point temperature value to determine the target temperature value at each sampling time. The vector analysis unit is used to obtain the temperature matrix attribute representation, ultrasonic temperature estimate, target temperature value, and current power at each sampling moment within the current window, and then determine the cooking state feature vector. The single-burner control unit is used to generate control commands after statistically analyzing the cooking state feature vector at the current moment and the target temperature sequence at future moments, and then adjusts the heating power of the single-burner induction cooker according to the control commands.

[0006] Preferably, the temperature matrix analysis unit includes a matrix alignment module, a pixel determination module, and a matrix purification module; The matrix alignment module obtains the current window. Raw data vectors acquired internally ,in , Indicates the first The original data vector at each sampling time point, This indicates the total number of sampling points within the current window. They represent the first The temperature matrix, echo signal sequence, environmental parameter vector, current power state, timestamp, and current heating power at each sampling time are used to align all temperature matrices to the standard time axis using nearest neighbor interpolation, resulting in a temperature matrix sequence. ,in Indicates the first after alignment A temperature matrix; The pixel determination module determines Process noise variance of each pixel and measurement noise variance , obtain the Temperature matrix The Middle Temperature measurement value of each pixel Then, according to and Analysis of the purification process Optimal state estimate at each sampling time point and error covariance ,in ， , Indicates the sequence number, Deposit the first A purified temperature matrix At the corresponding position, after traversing all pixels, output the first pixel. A purified temperature matrix ; The matrix purification module utilizes Determine the first State prediction value at each sampling time ,in ,pass and Calculate the error prediction value at the corresponding time. ,in Obtain the temperature matrix The Middle Temperature measurement value of each pixel Then, the gain coefficient was calculated. and deviation value ,in , ,use , , and Analyze the optimal estimate at the current moment. and error covariance ,in , ,Will Deposit the first A purified temperature matrix At the corresponding position, after traversing all pixels, output the first pixel. A purified temperature matrix .

[0007] Preferably, the temperature matrix analysis unit further includes a sequence output module and an attribute characterization combination module; The sequence output module repeats the process for each subsequent sampling time. The same operation at each sampling time point yields the purified temperature matrix sequence. , Indicates the first A purified temperature matrix This indicates the total number of sampling points within the current window; The attribute characterization and analysis module is based on the purified temperature matrix. After calculating the corresponding average temperature and maximum temperature, and analyzing... Horizontal partial derivative and vertical partial derivative ,in , , , Represent the horizontal and vertical deflection kernels, respectively, and determine... The Middle gradient magnitude of each pixel ,in , Indicates the first Horizontal bias kernel of each pixel Indicates the first The vertical partial derivative kernel of each pixel is used to calculate the average gradient after traversing each pixel. ,in , Indicates the serial number. Indicates the first Gradient magnitude of each pixel This represents the total number of pixels. The average temperature, maximum temperature, gradient magnitude, and average gradient are combined to construct... The attribute representation is repeated until the attribute representation of all temperature matrices is obtained.

[0008] Preferably, the temperature estimation unit includes a signal processing module, a temperature determination module, and a signal traversal module; The signal processing module obtains the transmission frequency. and the Echo signal sequence at each sampling time Extract it to obtain the main frequency peak. ,in , express The spectrum, Representing frequency, using and Calculate the frequency shift coefficient ,in ,pass Calculate the current speed of sound ,in , Indicates the distance of the sound wave. Representing fuzzy numbers, using Estimate the ultrasonic temperature value ,in , Indicates the molar mass of the propagation medium. This indicates the specific heat capacity ratio. Represents the molar gas constant; The temperature determination module obtains the temperature matrix at the corresponding time. Calculate If the variance of all pixels is less than a preset threshold, it is determined to be oil fume interference, and the ultrasonic temperature estimation value is used. As the actual temperature measurement value, otherwise... The average value is used as the actual temperature measurement value; The signal traversal module traverses the echo signal sequence at each sampling time to obtain a temperature estimation sequence.

[0009] Preferably, the cookware identification unit includes a constant calculation module and a type determination module; Before cooking begins, the constant calculation module applies a standard heating power pulse for 5 seconds and simultaneously acquires a sequence of actual temperature measurements at the bottom of the pot according to a preset sampling frequency. A thermodynamic model is used to perform nonlinear least-squares fitting on the temperature measurement sequence. The optimal thermal time constant is calculated by minimizing the sum of squared residuals between the model's predicted values ​​and the temperature measurement sequence. The core formula within the thermodynamic model is as follows:

[0010] in, Indicates room temperature. Indicates the degree of temperature rise. Represents the natural constant. Represents a time variable. Represents the thermal time constant. This represents the temperature value predicted by the model; The type determination module sets a low-degree threshold and a high-degree threshold. If the optimal heat time constant is less than the low-degree threshold, the cookware type is determined to be an iron pot. If the heat time constant is greater than or equal to the low-degree threshold and less than or equal to the high-degree threshold, the type is determined to be a stainless steel pot. If the heat time constant is greater than the high-degree threshold, the type is determined to be a ceramic pot. The corresponding control parameters are retrieved according to the cookware type.

[0011] Preferably, the target temperature calculation unit includes a characterization acquisition module, a candidate position determination module, and a target temperature compensation module; The characterization acquisition module extracts the current window. The temperature matrix attributes and center point temperature values ​​at each sampling time are represented, and the set of pre-stored cooking curves for the current dish are retrieved. The candidate position determination module determines the first candidate position within the cooking curve set. Target curve and all undetermined starting positions ,in Indicates the first An undetermined starting position, express Given the total number of undetermined starting positions, select the undetermined starting position. ,exist From Start by capturing the current window. For curve segments of the same size, calculate the temperature value at the center point of each temperature matrix. Correlation coefficient between curve segments ,in , Indicates the cross-correlation value. express Inner Temperature values ​​at the center point of each temperature matrix Indicates the first The first target curve Temperature values ​​at each time point, , Indicates the index of the target curve. The index represents the sequence number. Each undetermined starting position within all target curves is selected sequentially. After calculating the corresponding correlation coefficient, the 10 undetermined starting positions with the highest correlation coefficients are selected as candidate positions, and the rest are deleted. The target temperature compensation module calculates the distance value of the candidate positions, determines the curve with the highest matching degree and the undetermined starting position, and then calculates the target temperature value corresponding to each sampling moment during the cooking process according to the curve. Combining the environmental parameter vector and the corresponding compensation coefficient, it calculates the deviation value corresponding to each parameter in the environmental vector, calculates the total compensation amount using the deviation value and the compensation coefficient, and adds it to the target temperature value to obtain the compensated target temperature value.

[0012] Preferably, the vector analysis unit includes a vector generation module, a feature vector extraction module, and a sensor monitoring module; The vector generation module extracts the temperature matrix attribute representation, ultrasonic temperature estimate, compensated target temperature value and current power for each sampling moment in the current window, and then splices them together to form a multidimensional vector sequence. The feature vector extraction module maps the multidimensional vector sequence to a 128-dimensional embedding space through linear projection, adds position encoding, and obtains the embedding sequence. This embedding sequence is then input into a stacked 4-layer Transformer encoder. The average output of all time steps in the last layer is taken, and the cooking state feature vector is output through a fully connected layer. The sensor monitoring module acquires the output status of each sensor at the current moment. The temperature sensor provides a basic probability allocation based on infrared temperature measurement data, the ultrasonic sensor provides a basic probability allocation based on ultrasonic temperature measurement data, and the power sensor provides a basic probability allocation based on feedback data from the power controller. After calculating the normalization factor, the product of the points of each sensor pointing to the same state is summed and divided by the normalization factor to obtain the fused confidence level. The state with the highest confidence level is taken as the decision result. If the decision result is abnormal, a safety protection mechanism is triggered. If it is a deviation, the sensor control parameters are adjusted. If it is normal, normal control continues.

[0013] Preferably, the single-head furnace control unit includes a temperature prediction module, an instruction generation module, an operation execution module, and a cycle end module; The temperature prediction module obtains the cooking state feature vector at the current moment and the compensated target temperature sequence for each future moment. It then determines the dynamic relationship between the pot bottom temperature and the heating power. It inputs the temperature value after the most recent purification and the power sequence to be optimized, and it iteratively calculates the predicted temperature values ​​for each future moment. The dynamic relationship is specifically as follows:

[0014] in, These all indicate the corresponding parameters loaded according to the cookware type. Indicates at time The predicted temperature value, Indicates at time The predicted temperature value, Indicates at time Heating power; After constructing the optimization objective function and constraints, the instruction generation module uses an embedded solver to analyze them and obtain the optimal power sequence for each future time. The control instruction for the current time is then generated by combining the first value in the sequence. After receiving the control command, the operation execution module retrieves the power compensation coefficient according to the type of cookware. ,according to For the optimal power value Adjustments were made to obtain the compensated power value. ,in Adjust the heating power of the single-burner induction cooker according to the compensated power value; After the current control cycle ends, the cycle termination module acquires a new temperature matrix and recalculates the cooking state feature vector and target temperature sequence before entering the next control cycle.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention obtains a high-precision temperature field distribution and processes each pixel. Based on this, the real-time collected temperature value is matched with the optimal cooking curve library to accurately identify the current cooking stage. The power is adjusted in conjunction with the control strategy to ensure that every cooking process is successful. 2. This invention designs a dual-modal redundant temperature measurement mechanism. When the infrared sensor detects oil fumes causing a sudden drop in the variance of the temperature matrix, it instantly switches to the medium temperature derived by the piezoelectric ultrasonic sensor. This cross-validation based on physical principles ensures the continuity of temperature feedback under any harsh environment. 3. Before cooking, the present invention uses standard power pulses to automatically identify the material of the cookware and load the corresponding parameters, realizing the self-adaptation of cookware with different physical properties. At the same time, the invention uses environmental sensors to compensate for the influence of environmental parameters on cooking in real time, eliminating regional and seasonal differences and ensuring that users in different places can obtain consistent cooking results. 4. This invention extracts temperature attribute characteristics, state features, and future trend coefficients, enabling the system to adjust power in advance based on the prediction of future temperature trajectories, avoiding temperature overshoot or lag, and achieving a leap from passive response to active control, making the temperature control of automated cooking more precise and stable. Attached Figure Description

[0016] Figure 1 A schematic diagram of the overall system flow is provided for embodiments of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 This invention provides a technical solution: an intelligent cooking control system based on multi-sensor fusion, comprising: The temperature matrix analysis unit is used to acquire the raw data vector collected in the current window, align the temperature matrix in each vector to the time axis, clean it, and analyze the corresponding attribute characteristics. The temperature estimation unit is used to read the echo signal sequence at each sampling time in the original data vector and analyze the corresponding ultrasonic temperature estimate according to the signal sequence. The cookware identification unit is used to apply a standard heating power pulse before cooking begins and simultaneously acquire a sequence of actual temperature measurements at the bottom of the cookware, analyze the data, and thus determine the type of cookware. The target temperature calculation unit is used to extract the attribute representation of each temperature matrix and the center point temperature value, and match the corresponding cooking curve segment according to the center point temperature value to determine the target temperature value at each sampling time. The vector analysis unit is used to obtain the temperature matrix attribute representation, ultrasonic temperature estimate, target temperature value, and current power at each sampling moment within the current window, and then determine the cooking state feature vector. The single-burner control unit is used to generate control commands after statistically analyzing the cooking state feature vector at the current moment and the target temperature sequence at future moments, and then adjusts the heating power of the single-burner induction cooker according to the control commands.

[0019] The temperature matrix analysis unit includes a matrix alignment module, a pixel determination module, and a matrix purification module; The matrix alignment module retrieves the current window. Raw data vectors acquired internally ,in , Indicates the first The original data vector at each sampling time point, This indicates the total number of sampling points within the current window. They represent the first The temperature matrix, echo signal sequence, environmental parameter vector, current power state, timestamp, and current heating power at each sampling time are used to align all temperature matrices to the standard time axis using nearest neighbor interpolation, resulting in a temperature matrix sequence. ,in Indicates the first after alignment A temperature matrix; Pixel determination module Process noise variance of each pixel and measurement noise variance , obtain the Temperature matrix The Middle Temperature measurement value of each pixel Then, according to and Analysis of the purification process Optimal state estimate at each sampling time point and error covariance ,in ， , Indicates the sequence number, Deposit the first A purified temperature matrix At the corresponding position, after traversing all pixels, output the first pixel. A purified temperature matrix ; Matrix purification module utilizes Determine the first State prediction value at each sampling time ,in ,pass and Calculate the error prediction value at the corresponding time. ,in Obtain the temperature matrix The Middle Temperature measurement value of each pixel Then, the gain coefficient was calculated. and deviation value ,in , ,use , , and Analyze the optimal estimate at the current moment. and error covariance ,in , ,Will Deposit the first A purified temperature matrix At the corresponding position, after traversing all pixels, output the first pixel. A purified temperature matrix ; The temperature matrix analysis unit also includes a sequence output module and an attribute characterization combination module; The sequence output module repeats the process for each subsequent sampling time. The same operation at each sampling time point yields the purified temperature matrix sequence. , Indicates the first A purified temperature matrix This indicates the total number of sampling points within the current window; The attribute characterization analysis module is based on the purified temperature matrix. After calculating the corresponding average temperature and maximum temperature, and analyzing... Horizontal partial derivative and vertical partial derivative ,in , , , Represent the horizontal and vertical deflection kernels, respectively, and determine... The Middle gradient magnitude of each pixel ,in , Indicates the first Horizontal bias kernel of each pixel Indicates the first The vertical partial derivative kernel of each pixel is used to calculate the average gradient after traversing each pixel. ,in , Indicates the serial number. Indicates the first Gradient magnitude of each pixel This represents the total number of pixels. The average temperature, maximum temperature, gradient magnitude, and average gradient are combined to construct... The attribute representation is repeated until the attribute representation of all temperature matrices is obtained. The temperature estimation unit includes a signal processing module, a temperature determination module, and a signal traversal module; The signal processing module obtains the transmission frequency. and the Echo signal sequence at each sampling time Extract it to obtain the main frequency peak. ,in , express The spectrum, Representing frequency, using and Calculate the frequency shift coefficient ,in ,pass Calculate the current speed of sound ,in , Indicates the distance of the sound wave. Representing fuzzy numbers, using Estimate the ultrasonic temperature value ,in , Indicates the molar mass of the propagation medium. This indicates the specific heat capacity ratio. Represents the molar gas constant; The temperature determination module obtains the temperature matrix at the corresponding time. Calculate If the variance of all pixels is less than a preset threshold, it is determined to be oil fume interference, and the ultrasonic temperature estimation value is used. As the actual temperature measurement value, otherwise... The average value is used as the actual temperature measurement value; After the signal traversal module traverses the echo signal sequence at each sampling time, it obtains a sequence of temperature estimates. The cookware recognition unit includes a constant calculation module and a type determination module; Before cooking begins, the constant calculation module applies a standard heating power pulse for 5 seconds and simultaneously acquires a sequence of actual temperature measurements at the bottom of the pot according to a preset sampling frequency. A thermodynamic model is used to perform nonlinear least-squares fitting on the temperature measurement sequence. The optimal thermal time constant is calculated by minimizing the sum of squared residuals between the model's predicted values ​​and the temperature measurement sequence. The core formula within the thermodynamic model is as follows:

[0020] in, Indicates room temperature. Indicates the degree of temperature rise. Represents the natural constant. Represents a time variable. Represents the thermal time constant. This represents the temperature value predicted by the model; The type determination module sets a low threshold and a high threshold. If the optimal heat time constant is less than the low threshold, the cookware type is determined to be an iron pot. If the heat time constant is greater than or equal to the low threshold and less than or equal to the high threshold, the type is determined to be a stainless steel pot. If the heat time constant is greater than the high threshold, the type is determined to be a ceramic pot. The corresponding control parameters are retrieved according to the cookware type. The target temperature calculation unit includes a characterization acquisition module, a candidate location determination module, and a target temperature compensation module; The representation acquisition module extracts the current window. The temperature matrix attributes and center point temperature values ​​at each sampling time are represented, and the set of pre-stored cooking curves for the current dish are retrieved. The candidate position determination module determines the first position within the set of cooking curves. Target curve and all undetermined starting positions ,in Indicates the first An undetermined starting position, express Given the total number of undetermined starting positions, select the undetermined starting position. ,exist From Start by capturing the current window. For curve segments of the same size, calculate the temperature value at the center point of each temperature matrix. Correlation coefficient between curve segments ,in , Indicates the cross-correlation value. express Inner Temperature values ​​at the center point of each temperature matrix Indicates the first The first target curve Temperature values ​​at each time point, , Indicates the index of the target curve. The index represents the sequence number. Each undetermined starting position within all target curves is selected sequentially. After calculating the corresponding correlation coefficient, the 10 undetermined starting positions with the highest correlation coefficients are selected as candidate positions, and the rest are deleted. The target temperature compensation module calculates the distance value of the candidate positions, determines the curve with the highest matching degree and the undetermined starting position, and then calculates the target temperature value corresponding to each sampling time during the cooking process according to the curve. Combining the environmental parameter vector and the corresponding compensation coefficient, it calculates the deviation value corresponding to each parameter in the environmental vector, calculates the total compensation amount using the deviation value and the compensation coefficient, and adds it to the target temperature value to obtain the compensated target temperature value. The specific process for calculating the distance value of candidate locations is as follows: For a selected candidate matching pair, the number of current temperature matrices and the length of the target curve segment are both determined. Then, a matrix is ​​constructed based on the absolute difference between the temperature value at the center point of each temperature matrix and the temperature values ​​at each point in the target curve segment. The difference matrix, and initialized with a The cumulative distance matrix has the same first element as the first element of the difference matrix, and the remaining elements are equal to the data at the corresponding position in the difference matrix plus the minimum cumulative distance before reaching that position. The cumulative distance formula is as follows:

[0021] in, Represents the position in the cumulative distance matrix The value on, Represents the position in the cumulative distance matrix The value on, Represents the position in the cumulative distance matrix The value on, Represents the position in the cumulative distance matrix The value on, Represents the position in the difference matrix The value on, Indicates parameters; After filling the entire cumulative distance matrix, the first... Line number The column value is used as the actual distance value of the candidate matching pair. The same calculation is repeated for all candidate positions to obtain the corresponding distance value. All distance values ​​are compared, and the target curve index and starting position corresponding to the smallest distance value are selected as the global best matching result. The vector analysis unit includes a vector generation module, a feature vector extraction module, and a sensor monitoring module; The vector generation module extracts the temperature matrix attribute representation, ultrasonic temperature estimate, compensated target temperature value and current power for each sampling moment in the current window, and then concatenates them to form a multidimensional vector sequence. The feature vector extraction module maps the multidimensional vector sequence to a 128-dimensional embedding space through linear projection, adds positional encoding, and obtains the embedding sequence. This is then input into a stacked 4-layer Transformer encoder. Each layer contains a multi-head self-attention mechanism (8 attention heads, each with 64 dimensions) and a feedforward network. Layer normalization and residual connections are applied. After taking the average of the outputs of all time steps of the last layer, the cooking state feature vector is output through a fully connected layer. The sensor monitoring module acquires the output status of each sensor at the current moment. The temperature sensor provides a basic probability allocation based on infrared temperature measurement data, the ultrasonic sensor provides a basic probability allocation based on ultrasonic temperature measurement data, and the power sensor provides a basic probability allocation based on feedback data from the power controller. After calculating the normalization factor, the product of the points of each sensor pointing to the same state is summed and divided by the normalization factor to obtain the fused confidence level. The state with the highest confidence level is taken as the decision result. If the decision result is abnormal, the safety protection mechanism is triggered. If it is a deviation, the sensor control parameters are adjusted. If it is normal, the normal control continues. The single-head furnace control unit includes a temperature prediction module, an instruction generation module, an operation execution module, and a cycle end module; The temperature prediction module obtains the cooking state feature vector at the current moment and the compensated target temperature sequence for each future moment. It then determines the dynamic relationship between the pot bottom temperature and the heating power. Inputting the temperature value after the most recent purification and the power sequence to be optimized, the module iteratively calculates the predicted temperature values ​​for each future moment. The specific dynamic relationship is as follows:

[0022] in, These all indicate the corresponding parameters loaded according to the cookware type. Indicates at time The predicted temperature value, Indicates at time The predicted temperature value, Indicates at time Heating power; After the instruction generation module constructs the optimization objective function and constraints, it uses an embedded solver to analyze them and obtain the optimal power sequence for each future time step. The control instruction for the current time step is then generated by combining the first value in the sequence. The specific objective function is as follows:

[0023] in, This represents the cost function value. Indicates the prediction time domain, Indicates step index, Indicates the temperature tracking weight. Indicates power smoothing weights, This represents the change in power. This indicates the predicted temperature value. This indicates the target temperature after compensation. Indicates the serial number; The specific constraints are as follows:

[0024]

[0025]

[0026] in, Indicates at time The predicted temperature value, Indicates at time The predicted temperature value, Indicates at time Heating power, Indicates at time The change in power, This indicates the upper limit of the power change. Indicates the serial number. Represent a linear equation; After receiving the control command, the operation execution module retrieves the power compensation coefficient according to the type of cookware. ,according to For the optimal power value Adjustments were made to obtain the compensated power value. ,in The heating power of the single-head furnace is adjusted according to the compensated power value. This single-burner uses a multispectral infrared array sensor to obtain a high-precision temperature field at the bottom of the pot, supplemented by a piezoelectric ultrasonic sensor as a redundant backup for infrared temperature measurement, used for medium temperature inversion under the interference of oil fumes. At the same time, it has a built-in environmental sensor to monitor environmental parameters to eliminate their impact on cooking, and a power sensor to provide real-time feedback on heating power, forming a complete control closed loop. After the current control cycle ends, the cycle end module collects a new temperature matrix and recalculates the cooking state feature vector and target temperature sequence before entering the next control cycle.

[0027] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0028] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent cooking control system based on multi-sensor fusion, characterized in that, include: The temperature matrix analysis unit is used to acquire the raw data vector collected in the current window, align the temperature matrix in each vector to the time axis, clean it, and analyze the corresponding attribute characteristics. The temperature estimation unit is used to read the echo signal sequence at each sampling time in the original data vector and analyze the corresponding ultrasonic temperature estimate according to the signal sequence. The cookware identification unit is used to apply a standard heating power pulse before cooking begins and simultaneously acquire a sequence of actual temperature measurements at the bottom of the cookware, analyze the data, and thus determine the type of cookware. The target temperature calculation unit is used to extract the attribute representation of each temperature matrix and the center point temperature value, and match the corresponding cooking curve segment according to the center point temperature value to determine the target temperature value at each sampling time. The vector analysis unit is used to obtain the temperature matrix attribute representation, ultrasonic temperature estimate, target temperature value, and current power at each sampling moment within the current window, and then determine the cooking state feature vector. The single-burner control unit is used to generate control commands after statistically analyzing the cooking state feature vector at the current moment and the target temperature sequence at future moments, and then adjusts the heating power of the single-burner induction cooker according to the control commands.

2. The intelligent cooking control system based on multi-sensor fusion according to claim 1, characterized in that, The temperature matrix analysis unit includes a matrix alignment module, a pixel determination module, and a matrix purification module; The matrix alignment module obtains the current window. Raw data vectors acquired internally ,in , Indicates the first The original data vector at each sampling time point, This indicates the total number of sampling points within the current window. They represent the first The temperature matrix, echo signal sequence, environmental parameter vector, current power state, timestamp, and current heating power at each sampling time are used to align all temperature matrices to the standard time axis using nearest neighbor interpolation, resulting in a temperature matrix sequence. ,in Indicates the first after alignment A temperature matrix; The pixel determination module determines Process noise variance of each pixel and measurement noise variance , obtain the Temperature matrix The Middle Temperature measurement value of each pixel Then, according to and Analysis of the purification process Optimal state estimate at each sampling time point and error covariance ,in ， , Indicates the sequence number, Deposit the first A purified temperature matrix At the corresponding position, after traversing all pixels, output the first pixel. A purified temperature matrix ; The matrix purification module utilizes Determine the first State prediction value at each sampling time ,in ,pass and Calculate the error prediction value at the corresponding time. ,in Obtain the temperature matrix The Middle Temperature measurement value of each pixel Then, the gain coefficient was calculated. and deviation value ,in , ,use , , and Analyze the optimal estimate at the current moment. and error covariance ,in , ,Will Deposit the first A purified temperature matrix At the corresponding position, after traversing all pixels, output the first pixel. A purified temperature matrix .

3. The intelligent cooking control system based on multi-sensor fusion according to claim 2, characterized in that, The temperature matrix analysis unit also includes a sequence output module and an attribute characterization combination module; The sequence output module repeats the process for each subsequent sampling time. The same operation at each sampling time point yields the purified temperature matrix sequence. , Indicates the first A purified temperature matrix This indicates the total number of sampling points within the current window; The attribute characterization and analysis module is based on the purified temperature matrix. After calculating the corresponding average temperature and maximum temperature, and analyzing... Horizontal partial derivative and vertical partial derivative ,in , , , Represent the horizontal and vertical deflection kernels, respectively, and determine... The Middle gradient magnitude of each pixel ,in , Indicates the first Horizontal bias kernel of each pixel Indicates the first The vertical partial derivative kernel of each pixel is used to calculate the average gradient after traversing each pixel. ,in , Indicates the serial number. Indicates the first Gradient magnitude of each pixel This represents the total number of pixels. The average temperature, maximum temperature, gradient magnitude, and average gradient are combined to construct... The attribute representation is repeated until the attribute representation of all temperature matrices is obtained.

4. The intelligent cooking control system based on multi-sensor fusion according to claim 1, characterized in that, The temperature estimation unit includes a signal processing module, a temperature determination module, and a signal traversal module; The signal processing module obtains the transmission frequency. and the Echo signal sequence at each sampling time Extract it to obtain the peak frequency. ,in , express The spectrum, Representing frequency, using and Calculate the frequency shift coefficient ,in ,pass Calculate the current speed of sound ,in , Indicates the distance of the sound wave. Representing fuzzy numbers, using Estimate the ultrasonic temperature value ,in , Indicates the molar mass of the propagation medium. This indicates the specific heat capacity ratio. Represents the molar gas constant; The temperature determination module obtains the temperature matrix at the corresponding time. Calculate If the variance of all pixels is less than a preset threshold, it is determined to be oil fume interference, and the ultrasonic temperature estimation value is used. As the actual temperature measurement value, otherwise... The average value is used as the actual temperature measurement value; The signal traversal module traverses the echo signal sequence at each sampling time to obtain a temperature estimation sequence.

5. The intelligent cooking control system based on multi-sensor fusion according to claim 1, characterized in that, The cookware identification unit includes a constant calculation module and a type determination module; Before cooking begins, the constant calculation module applies a standard heating power pulse for 5 seconds and simultaneously acquires a sequence of actual temperature measurements at the bottom of the pot according to a preset sampling frequency. A thermodynamic model is used to perform nonlinear least-squares fitting on the temperature measurement sequence. The optimal thermal time constant is calculated by minimizing the sum of squared residuals between the model's predicted values ​​and the temperature measurement sequence. The core formula within the thermodynamic model is as follows: in, Indicates room temperature. Indicates the degree of temperature rise. Represents the natural constant. Represents a time variable. Represents the thermal time constant. This represents the temperature value predicted by the model; The type determination module sets a low-degree threshold and a high-degree threshold. If the optimal heat time constant is less than the low-degree threshold, the cookware type is determined to be an iron pot. If the heat time constant is greater than or equal to the low-degree threshold and less than or equal to the high-degree threshold, the type is determined to be a stainless steel pot. If the heat time constant is greater than the high-degree threshold, the type is determined to be a ceramic pot. The corresponding control parameters are retrieved according to the cookware type.

6. The intelligent cooking control system based on multi-sensor fusion according to claim 1, characterized in that, The target temperature calculation unit includes a characterization acquisition module, a candidate position determination module, and a target temperature compensation module; The characterization acquisition module extracts the current window. The temperature matrix attributes and center point temperature values ​​at each sampling time are represented, and the set of pre-stored cooking curves for the current dish are retrieved. The candidate position determination module determines the first candidate position within the cooking curve set. Target curve and all undetermined starting positions ,in Indicates the first An undetermined starting position, express Given the total number of undetermined starting positions, select the undetermined starting position. ,exist From Start by capturing the current window. For curve segments of the same size, calculate the temperature value at the center point of each temperature matrix. Correlation coefficient between curve segments ,in , Indicates the cross-correlation value. express Inner Temperature values ​​at the center point of each temperature matrix Indicates the first The first target curve Temperature values ​​at each time point, , Indicates the index of the target curve. The index represents the sequence number. Each undetermined starting position within all target curves is selected sequentially. After calculating the corresponding correlation coefficient, the 10 undetermined starting positions with the highest correlation coefficients are selected as candidate positions, and the rest are deleted. The target temperature compensation module calculates the distance value of the candidate positions, determines the curve with the highest matching degree and the undetermined starting position, and then calculates the target temperature value corresponding to each sampling moment during the cooking process according to the curve. Combining the environmental parameter vector and the corresponding compensation coefficient, it calculates the deviation value corresponding to each parameter in the environmental vector, calculates the total compensation amount using the deviation value and the compensation coefficient, and adds it to the target temperature value to obtain the compensated target temperature value.

7. The intelligent cooking control system based on multi-sensor fusion according to claim 1, characterized in that, The vector analysis unit includes a vector generation module, a feature vector extraction module, and a sensor monitoring module; The vector generation module extracts the temperature matrix attribute representation, ultrasonic temperature estimate, compensated target temperature value and current power for each sampling moment in the current window, and then splices them together to form a multidimensional vector sequence. The feature vector extraction module maps the multidimensional vector sequence to a 128-dimensional embedding space through linear projection, adds position encoding, and obtains the embedding sequence. This embedding sequence is then input into a stacked 4-layer Transformer encoder. The average output of all time steps in the last layer is taken, and the cooking state feature vector is output through a fully connected layer. The sensor monitoring module acquires the output status of each sensor at the current moment. The temperature sensor provides a basic probability allocation based on infrared temperature measurement data, the ultrasonic sensor provides a basic probability allocation based on ultrasonic temperature measurement data, and the power sensor provides a basic probability allocation based on feedback data from the power controller. After calculating the normalization factor, the product of the points of each sensor pointing to the same state is summed and divided by the normalization factor to obtain the fused confidence level. The state with the highest confidence level is taken as the decision result. If the decision result is abnormal, a safety protection mechanism is triggered. If it is a deviation, the sensor control parameters are adjusted. If it is normal, normal control continues.

8. The intelligent cooking control system based on multi-sensor fusion according to claim 1, characterized in that, The single-head furnace control unit includes a temperature prediction module, an instruction generation module, an operation execution module, and a cycle end module; The temperature prediction module obtains the cooking state feature vector at the current moment and the compensated target temperature sequence for each future moment. It then determines the dynamic relationship between the pot bottom temperature and the heating power. It inputs the temperature value after the most recent purification and the power sequence to be optimized, and it iteratively calculates the predicted temperature values ​​for each future moment. The dynamic relationship is specifically as follows: in, These all indicate the corresponding parameters loaded according to the cookware type. Indicates at time The predicted temperature value, Indicates at time The predicted temperature value, Indicates at time Heating power; After constructing the optimization objective function and constraints, the instruction generation module uses an embedded solver to analyze them and obtain the optimal power sequence for each future time. The control instruction for the current time is then generated by combining the first value in the sequence. After receiving the control command, the operation execution module retrieves the power compensation coefficient according to the type of cookware. ,according to For the optimal power value Adjustments were made to obtain the compensated power value. ,in Adjust the heating power of the single-burner induction cooker according to the compensated power value; After the current control cycle ends, the cycle termination module acquires a new temperature matrix and recalculates the cooking state feature vector and target temperature sequence before entering the next control cycle.

Images