Nutritional constraint adaptive control method for a commercial cooking appliance

Abstract

The present application relates to the technical field of intelligent cooking control, and discloses a kind of nutrition constraint adaptive control method of commercial cooking equipment, the method includes using data acquisition module to obtain cloud device-independent recipe semantics, food nutrition constraint parameters and local device capability profile;Using adaptive compilation module combines the local profile with nutrition constraint parameters, converts cloud recipe semantics into bottom execution sequence;Using cooking execution and perception module issues execution sequence, and collects temperature, current and other real-time state data to dynamically correct the running state;Finally, using learning iteration module based on state data and evaluation data to correct execution sequence, generate exclusive cooking parameter set.The present application realizes the accurate mapping of standardized recipe across devices, while automatically landing nutrition intervention indicators, through multi-dimensional closed-loop perception, ensures the stability of the cooking process and the quality of continuous meal service.

Description

Technical Field

[0001] This invention relates to the field of intelligent cooking control technology, specifically to a nutritional constraint adaptive control method for commercial cooking equipment. Background Technology

[0002] With the development of catering standardization, commercial automated cooking equipment has been widely used. Existing commercial cooking equipment typically relies on pre-programmed control programs to execute recipe actions. However, existing control programs are often deeply bound to the underlying hardware parameters of specific models. When the same recipe data is sent to cooking equipment with different thermodynamic properties or mechanical structural characteristics, the lack of a mechanism for reading and adaptively compiling local device capabilities often leads to serious deviations between the actual output heating power and the stir-frying action, making it difficult to achieve cross-device compatibility and accurate implementation of standardized recipes.

[0003] Meanwhile, existing commercial cooking control strategies mainly focus on basic cooking control and the sequential execution of mechanical actions, lacking a fundamental intervention link for the nutritional indicators of ingredients. When faced with scenarios with specific nutritional restrictions such as calories, salt, or allergens, the equipment cannot automatically reduce the feeding of corresponding materials based on nutritional constraints, nor can it provide reverse compensation correction for process deviations caused by material reduction.

[0004] Furthermore, in actual continuous commercial cooking processes, the initial temperature, texture, and environmental conditions of ingredients often vary. Existing equipment often employs open-loop control or single temperature threshold control when executing instructions, failing to integrate multi-dimensional state data during the cooking process. Especially when faced with changes in the heat absorption rate of materials in the pot or increased viscosity leading to variations in mechanical load, the equipment struggles to combine underlying parameters such as temperature and current for real-time dynamic correction, easily resulting in inconsistent doneness or triggering localized dry-burning, leading to poor consistency in the quality of continuously served dishes. Moreover, after cooking, existing systems typically cannot utilize process execution characteristics and actual finished product evaluation data for closed-loop parameter correction calculations, lacking the ability for self-iteration and the generation of customized cooking parameter sets. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a nutritional constraint adaptive control method for commercial cooking equipment, which solves the problems of poor cross-device compatibility, lack of underlying nutritional constraint intervention mechanism, and inconsistent meal quality due to lack of multi-dimensional closed-loop correction when existing commercial cooking equipment executes standardized recipes.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for adaptive control of nutritional constraints in commercial cooking equipment includes the following steps: The data acquisition module is used to obtain cloud-based device-independent recipe semantics, ingredient nutritional constraint parameters, and local device capability profiles. The adaptive compilation module is used to convert the semantics of the cloud-based recipes that are irrelevant to the local device capability profile and the nutritional constraint parameters of the ingredients, thereby generating an execution sequence adapted to the current device. The cooking execution and sensing module controls the operation of the underlying hardware device according to the execution sequence, collects real-time status data during the cooking execution process, and performs dynamic correction based on the real-time status data. The learning iteration module uses the real-time status data and finished product evaluation data to correct and calculate the equipment execution parameters for similar recipes, generating a set of cooking parameters specific to the target equipment.

[0007] Preferably, the step of using the data acquisition module to obtain cloud-device-independent recipe semantics includes: The data acquisition module is used to parse and process the initial recipe data package to extract recipe semantics that are irrelevant to the cloud device; The data acquisition module uses a tree-structured data structure to parse the cloud device-independent recipe semantics and read the action semantic instructions. The data acquisition module uses temporal logic to vectorize and reconstruct the action semantic instructions, outputting a state space vector sequence. The action semantic instruction consists of an action identifier, the characteristic value of the operation object, and the relative state constraint variable.

[0008] Preferably, the step of using the data acquisition module to acquire the nutritional constraint parameters of the ingredients and the local device capability profile includes: The data acquisition module reads local underlying parameters, performs classification and structuring processing, generates the local device capability profile, and converts the local device capability profile into a set of device capability vectors. The data acquisition module is used to convert the raw data of the food nutrition constraint parameters into numerical constraint variables and output a set of food nutrition constraint vectors, which includes quantitative restriction variables and logical masking variables.

[0009] Preferably, the step of using the adaptive compilation module to convert the semantics of the cloud-based, device-independent recipes according to the local device capability profile and the nutritional constraint parameters of the ingredients, and generating an execution sequence adapted to the current device, includes: The adaptive compilation module is used to decouple the state vectors in the state space vector sequence from the instructions, and to extract independent thermodynamic control factors, mechanical motion control factors, and material delivery control factors. The execution sequence is generated by using the adaptive compilation module to combine the thermodynamic parameter vectors in the equipment capability vector set with the mechanical motion control factor and the material delivery control factor for synchronous mapping calculation.

[0010] Preferably, the cooking execution and sensing module includes a sensing unit and a microcomputer numerical control unit; The method of collecting real-time status data during the cooking execution process using the cooking execution and sensing module includes: The surface temperature value is collected using an infrared temperature sensor in the sensing unit, and the substrate temperature value is collected using a K-type temperature sensor in the sensing unit. The actual comprehensive temperature is extracted by using the microcomputer numerical control unit to perform a weighted calculation based on the surface temperature value, the substrate temperature value, and the temperature measurement weighting coefficient. The actual comprehensive temperature is used as the basic data for subsequent splicing to generate a multi-dimensional real-time status data matrix.

[0011] Preferably, the step of collecting real-time status data during the cooking execution process using the cooking execution and sensing module further includes: The instantaneous measured operating current is monitored by the motor current sensor in the sensing unit, and the instantaneous measured operating current is converted into mechanical load parameters by the microcomputer numerical control unit. The actual cumulative mass of feed is obtained by using the flow meter device in the sensing unit, and the actual feed deviation variable is extracted by comparing the target feed mass with the actual outflow mass. The mechanical load parameters and the actual feeding deviation variables are concatenated into the real-time status data matrix.

[0012] Preferably, the dynamic correction based on the real-time status data includes: The microcomputer numerical control unit is used to perform a subtraction operation between the target temperature setpoint in the execution sequence and the actual comprehensive temperature in the real-time status data matrix to generate a temperature deviation variable, and the temperature deviation variable is used to obtain the subdivided firepower compensation amount. The microcomputer numerical control unit reads the mechanical load parameters in the real-time status data matrix. When the mechanical load parameters deviate from the load reference threshold, reverse intervention compensation is performed on the target operating speed in the execution sequence.

[0013] Preferably, the step of using the learning iteration module to correct the equipment execution parameters of similar recipes based on the real-time status data and finished product evaluation data includes: The learning iteration module is used to extract the execution feature vector of the underlying execution record file construction process; The learning iteration module is used to obtain the raw data for finished product evaluation, and the raw data for finished product evaluation is converted into an evaluation quantification feature vector. The learning iteration module is used to align and concatenate the process execution feature vector and the evaluation quantification feature vector to construct a comprehensive feedback data matrix.

[0014] Preferred options also include: The learning iteration module uses the process execution feature vector and evaluation quantization feature vector in the comprehensive feedback data matrix as input node data to calculate the correction increment of the underlying control parameters; The learning iteration module is used to superimpose the correction increment onto the original local execution parameter sequence of the corresponding recipe, and the local device capability file is called to perform a forced safety limit on the superimposed new parameter sequence to generate the target device-specific cooking parameter set; Preferably, after generating the target device-specific cooking parameter set, it also includes: The learning iteration module encapsulates the target device's specific cooking parameter set and the comprehensive feedback data matrix into an encrypted synchronization message and uploads it to the cloud server. The cloud server is used to retrieve a group of isomorphic commercial cooking equipment with the same thermodynamic boundary parameters and mechanical motion parameters from the cloud database, and the exclusive cooking parameter sets uploaded by multiple isomorphic devices are extracted to perform parameter group aggregation calculation.

[0015] This invention provides a method for adaptive control of nutritional constraints in commercial cooking equipment. It offers the following advantages: 1. This invention obtains cloud-based device-independent recipe semantics and local device capability profiles, and uses an adaptive compilation module to combine the two into a low-level execution sequence. This breaks the strong dependence of traditional commercial recipe programs on specific hardware models, enabling the same set of cloud-based recipe data to undergo adaptive parameter mapping and dimensionality reduction analysis based on the specific thermodynamic and mechanical motion parameters of different cooking devices. This improves the cross-device compatibility and execution accuracy of standardized recipes in complex hardware ecosystems.

[0016] 2. This invention introduces nutritional constraint parameters for ingredients and uses an adaptive compilation module to call quantitative constraint variables and logical masking variables. It performs channel filtering and proportional reduction of duration on the basic hardware execution sequence and simultaneously implements reverse compensation correction. This establishes a control link from macro-nutritional indicators to micro-hardware actions. The equipment can automatically reduce the opening amount of the corresponding seasoning solenoid valve based on the set calorie or allergen threshold, and automatically compensate by adjusting the heating duty cycle and stir-frying speed. While achieving automated health intervention, it ensures the basic cooking quality of the dishes.

[0017] 3. This invention utilizes a sensing unit to collect surface radiation temperature, substrate conduction temperature, and stator winding operating current in real time. These data are then converted into actual comprehensive temperature and mechanical load parameters by a microcomputer numerical control unit for dynamic correction. This multi-dimensional closed-loop sensing mechanism enables the equipment to calculate detailed heat compensation amounts in real time and perform reverse intervention on the operating speed based on changes in the actual heat absorption rate or viscosity of the materials during the cooking process. This solves the interference caused by differences in the initial state of ingredients and environmental temperature differences, ensuring the consistency of finished product quality and the reliability of equipment operation in continuous meal serving scenarios. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system architecture diagram of the present invention; Figure 3 This is a sub-flowchart of the present invention for acquiring and parsing cloud device-independent recipe semantics; Figure 4 This is a sub-flowchart of the present invention for constructing a local device capability profile; Figure 5 This is a flowchart illustrating the process of obtaining and converting nutritional constraint parameters of food ingredients according to the present invention. Figure 6 This is a sub-flowchart of the basic hardware execution sequence generated in this invention; Figure 7 This is a flowchart of the parameter quadratic mapping and adjustment sub-process based on nutrient constraints of the present invention; Figure 8 This is a flowchart of the security boundary check and final encapsulation sub-process for the execution parameters of this invention; Figure 9 This is a flowchart of the cooking execution, multi-dimensional state perception, and dynamic correction sub-process of the present invention. Detailed Implementation

[0019] 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.

[0020] See Figure 1 and Figure 2 This invention provides a nutrient constraint adaptive control system for commercial cooking equipment, the nutrient constraint adaptive control system for commercial cooking equipment comprising: The module includes a data acquisition module, an adaptive compilation module, a cooking execution and perception module, and a learning iteration module.

[0021] The data acquisition module is configured to acquire recipe semantics, ingredient nutritional constraint parameters, and local device capability profiles that are independent of cloud devices.

[0022] The adaptive compilation module is connected to the data acquisition module and is configured to convert cloud device-independent recipe semantics to be processed based on the local device capability profile and ingredient nutritional constraints, generating an execution sequence adapted to the current device.

[0023] The cooking execution and perception module is connected to the adaptive compilation module and configured to control the operation of the underlying hardware devices according to the execution sequence, while collecting real-time status data during the cooking execution process and performing dynamic correction based on the real-time status data.

[0024] The learning iteration module is connected to the cooking execution and perception module and is configured to correct and calculate the equipment execution parameters of similar recipes based on real-time status data and finished product evaluation data, and generate a set of cooking parameters specific to the target equipment.

[0025] In the nutritional constraint adaptive control system of the aforementioned commercial cooking equipment, the cooking execution and sensing module further includes a microcomputer numerical control unit, a sensing unit, and an execution unit.

[0026] The sensing unit includes an infrared temperature sensor, a K-type temperature sensor, a motor current sensor, and a flow meter device.

[0027] The execution unit includes an electromagnetic heating assembly, a drum motor, a planetary stirring motor, a multi-way solenoid valve spray nozzle, an independent material box dispensing mechanism, and an electric tilting mechanism. A microcomputer numerical control unit (MCU) is electrically connected to both the sensing unit and the execution unit. The MCU receives data collected by the sensing unit and sends control commands to the execution unit.

[0028] See Figure 2 This invention provides a nutrient constraint adaptive control method for commercial cooking equipment, applied to the nutrient constraint adaptive control system of the aforementioned commercial cooking equipment. The nutrient constraint adaptive control method for commercial cooking equipment includes: Step S100: Use the data acquisition module to obtain cloud device-independent recipe semantics, ingredient nutritional constraint parameters, and local device capability profiles.

[0029] Cloud-based, device-agnostic recipe semantics include semantic instructions for preheating, adding oil, adding ingredients, stir-frying, reducing sauce, and keeping warm. Nutritional constraints on ingredients include energy, fat, salt, sugar, protein ratio, and allergen restriction parameters.

[0030] The local equipment capability profile includes the target equipment's pot capacity, maximum heating power, number of spray channels, stirring capacity, drum capacity, temperature sensor configuration information, and safe temperature range limits.

[0031] Step S200: Use the adaptive compilation module to transform the cloud device-independent recipe semantics into an execution sequence adapted to the current device.

[0032] During the conversion process, the adaptive compilation module establishes the hardware operating boundaries of the target device based on the local device capability profile. Combining the nutritional constraints of the ingredients, the adaptive compilation module parses the cloud-based device-independent recipe semantics into corresponding temperature control curves, oil dosage, ingredient addition order, spraying amount, stirring speed, and cooking time parameters.

[0033] Through the above analytical operations, the adaptive compilation module generates an execution sequence suitable for the target device. The system ensures that all operating parameters in this execution sequence are within the hardware operating boundaries and safe temperature range, avoiding action conflicts caused by device capability mismatch.

[0034] Step S300: The cooking execution and sensing module controls the underlying execution mechanism to run according to the execution sequence, and collects the status data during the execution process through the sensing unit, and performs closed-loop correction on the current execution status based on the status data.

[0035] The microcomputer numerical control unit issues execution sequence instructions to the execution unit in a time-axis order. During the operation of the electromagnetic heating component, the multi-way solenoid valve injection port, and the stirring drum mechanism, the sensing unit monitors and outputs real-time data on actual temperature, actual operating power, feeding deviation, and pot movement status.

[0036] The microcomputer numerical control unit (MCU) compares the actual temperature and operating power with the preset target values ​​in the execution sequence in real time. When the comparison deviation exceeds the system's set allowable threshold, the MCU issues a correction command to adjust the output power level of the electromagnetic heating component or adjust the on / off duration of the multi-way solenoid valve's nozzle. The MCU strictly limits the execution state within the safe temperature range and implements anti-dry-burning protection.

[0037] In step S400, after the bottom-level execution unit completes a single cooking instruction sequence and produces a finished product, the learning iteration module updates the device execution parameters of similar recipes based on the status data during the execution process and the finished product evaluation data obtained after the execution is completed, generating a new local parameter version or cloud parameter version.

[0038] The learning iteration module receives comprehensive feedback data after a single cooking process is completed. This comprehensive feedback data consists of a state dataset during the cooking process and quantitative evaluation data of the finished product after the process is completed. The learning iteration module uses its internal learning model to perform reverse error calculation on the comprehensive feedback data.

[0039] The learning iteration module corrects and assigns values ​​to the heating curve, oil dosage, and spray amount parameters in the execution sequence. After correction, the system generates a target device-specific cooking parameter set. The learning iteration module writes this target device-specific cooking parameter set to local storage for offline caching and uploads it to the cloud server via a wireless communication interface, establishing a parameter canary release mechanism and version rollback mechanism among devices of the same model.

[0040] See Figure 3 The data acquisition module is configured to perform the operation of acquiring recipe semantics independent of cloud devices. This operation mainly includes data request, structure parsing, and semantic vectorization extraction steps: In step S110, the data acquisition module establishes a connection with the cloud server through the communication network and sends a data request instruction containing the target dish identifier to the cloud server. The cloud server responds to the data request instruction and sends an initial recipe data packet. The data acquisition module receives the initial recipe data packet and stores it in the local storage medium.

[0041] For the data transmission protocol and encryption authentication process of the communication network, those skilled in the art can make conventional selections and configurations according to the actual application environment. The network communication mechanism is a well-known technology in this field and will not be described in detail here.

[0042] In step S120, the data acquisition module parses and processes the initial recipe data package, extracts the cloud device-independent recipe semantics, and encapsulates and parses the cloud device-independent recipe semantics using a standardized tree data structure. This tree data structure includes basic information nodes for the dishes and cooking stage nodes arranged in time sequence.

[0043] The basic information nodes for dishes record the dish category, a list of main ingredient types, and a list of auxiliary ingredients. The cooking stage nodes unfold according to the process sequence, specifically including the preheating stage node, the ingredient adding stage node, the stir-frying stage node, and the sauce reducing stage node.

[0044] In step S130, the data acquisition module reads action semantic instructions from each cooking stage node. The action semantic instructions consist of action identifiers, operation object feature values, and relative state constraint variables. The action identifiers are used to define the process category of the execution unit action, which covers preheating, adding oil, adding ingredients, stir-frying, reducing sauce, and keeping warm. The operation object feature values ​​are used to point to specific ingredient list items or seasoning list items.

[0045] Relative state constraint variables are used to describe the conditions for the end of an action or the process requirements for the execution of an action. These relative state constraint variables do not contain absolute parameters bound to specific hardware devices. In the semantic instruction of the preheating action, the relative state constraint variable is represented as the percentage parameter that the target preheating medium temperature reaches the smoke point of the oil.

[0046] In the stir-frying action semantic command, the relative state constraint variable is represented as the threshold of food heating uniformity or the relative stir-frying intensity level. In the ingredient feeding action semantic command, the relative state constraint variable is represented as the mass ratio of seasoning to main ingredient. In the oil feeding action semantic command, the relative state constraint variable is represented as the percentage weight of the target oil volume to the total mass of the main ingredient. In the sauce reducing action semantic command, the relative state constraint variable is represented as the target sauce viscosity relative to the reference value or the target water evaporation percentage. In the heat preservation action semantic command, the relative state constraint variable is represented as the percentage of maintaining the core temperature relative to the cooking temperature. By extracting these relative state constraint variables, the data acquisition module removes the dependence of recipes on specific rated power, specific pot size, or specific motor model.

[0047] In step S140, the data acquisition module converts the parsed cloud device-independent recipe semantics into a state space vector sequence so that the adaptive compilation module can perform calculation mapping. The data acquisition module reconstructs each action semantic instruction into a vector according to the temporal logic.

[0048] The data acquisition module extracts the state vector corresponding to a single action semantic command into a multi-dimensional feature combination containing action identifiers, operation object feature values, relative state constraint variables, and relative time constraint weight factors. The relative time constraint weight factors are used to characterize the baseline proportion range of the current time sequence step in the entire cooking time cycle.

[0049] The data acquisition module outputs a state space vector sequence consisting of multiple state vectors arranged in order, and transmits it to the adaptive compilation module. This state vector sequence provides a unified data input basis for multi-source heterogeneous commercial cooking equipment.

[0050] See Figure 4 The data acquisition module is configured to perform operations to read and identify local device hardware parameters and build a local device capability profile while or before acquiring recipe semantics unrelated to cloud devices.

[0051] This process includes the following steps: underlying hardware status polling, parameter classification and structuring, and construction of the device capability matrix. In step S150, during the system power-on initialization or standby reset phase, the data acquisition module sends a status query message to each hardware execution unit through the underlying communication bus of the microcomputer numerical control unit. After receiving the message, the hardware execution unit returns a data frame containing the factory hardware configuration identifier and the current peripheral mounting status. The data acquisition module parses the data frame to read the local underlying parameters.

[0052] For the definition of the data frame structure and bus polling protocol of the underlying communication bus, those skilled in the art can make conventional settings according to the specific main control board architecture. Its microcontroller bus communication mechanism is a well-known technology in the field and will not be described in detail here.

[0053] In step S160, the data acquisition module classifies and structures the read local underlying parameters to generate a local equipment capability profile. This local equipment capability profile consists of four feature dimensions: thermodynamic boundary parameters, mechanical motion parameters, material delivery parameters, and safety protection boundary parameters.

[0054] Thermodynamic boundary parameters record the maximum rated output power of the electromagnetic heating assembly, the pot loading capacity, and the heating response time constant of the electromagnetic coil.

[0055] The mechanical motion parameters indicate the type of the bottom stirring mechanism. A specific identification code distinguishes whether the current equipment is equipped with a 360° roller mechanism or a planetary stirring and scraping mechanism. The mechanical motion parameters further include the lower and upper limits of the speed adjustment supported by the corresponding drive motor, as well as the maximum stroke tilt angle of the stepless electric stirring mechanism.

[0056] The material feeding parameter calibration equipment currently has the total number of multi-channel solenoid valve nozzles available, the upper limit of the liquid capacity of the corresponding independent material box, and the theoretical flow rate threshold per unit time of a single nozzle under the reference hydraulic pressure.

[0057] The safety protection boundary parameter is extracted based on the range of the sensing unit and the material heat resistance limit of the mechanical structure. This parameter specifies the absolute high-temperature alarm temperature calibrated by the infrared temperature sensor or K-type temperature sensor, as well as the maximum heating duration triggered by the anti-dry-burn protection mechanism under no-load conditions.

[0058] In step S170, the data acquisition module converts the structured local device capability profile into a set of device capability vectors, which serves as the absolute hardware constraint boundary for the subsequent adaptive compilation and calculation module.

[0059] The data acquisition module integrates the extracted thermodynamic parameters, mechanical motion parameters, material delivery parameters, and safety protection parameters into a multi-dimensional hardware capability feature vector. The data acquisition module outputs a complete set of equipment capability vectors to the adaptive compilation module, establishing the absolute boundary of the target equipment's execution capability.

[0060] See Figure 5 The data acquisition module is configured to acquire the nutritional constraint parameters of ingredients for the target population or specific dietary needs, and convert these parameters into constraints that can be read by the underlying control system, which are then used as the limiting inputs for generating the subsequent device execution sequence.

[0061] In step S180, the data acquisition module receives the raw data of the nutritional constraint parameters of the ingredients. This raw data is acquired through methods including parsing user-defined commands received from the touchscreen display connected to the microcomputer numerical control unit, or receiving standard constraint messages from an external dietary health management system via a communication interface. For the message protocol parsing of the external communication interface and the input response mechanism of the touchscreen, those skilled in the art can program according to conventional terminal application systems; the data exchange process is well-known in the field and will not be elaborated upon here.

[0062] In step S190, the data acquisition module transforms the extracted raw data into numerical constraint variables that the underlying control system participates in the calculation. This process transforms the dietary concept instruction into direct constraints on the material delivery and heating process.

[0063] The data acquisition module constructs and outputs a set of food nutrition constraint vectors. In this set of food nutrition constraint vectors, the data acquisition module establishes quantitative limiting variables and logical masking variables. The quantitative limiting variables set upper and lower limit thresholds for the intake of energy, oil, salt and sugar, and limit the mass ratio range of specific macronutrients.

[0064] The logical masking variable masks the allergen restriction parameters and is used to forcibly cut off the control signals of specific spray channels during subsequent execution. The data acquisition module compares the seasoning ingredient labels in each spray channel with the input allergen restriction parameters. When there is an overlap in ingredients, it indicates that the action of that channel is forcibly masked.

[0065] When there is no overlap in ingredients, the characterization allows the adaptive compilation module to make normal instruction calls to the channel. The data acquisition module transmits the completed set of food nutrition constraint vectors to the adaptive compilation module, which combines the recipe semantics independent of the cloud device and the local device capability file as the boundary control conditions for generating device action parameters.

[0066] See Figure 6 The adaptive compilation module is configured to transform the state space vector sequence extracted by the data acquisition module into the specific execution control quantity of the corresponding underlying hardware peripheral. This transformation process establishes the baseline hardware working state of the target device, and the transformation result before the superposition of food nutrition constraint parameters forms the basic hardware execution sequence.

[0067] In step S210, the adaptive compilation module receives a state space vector sequence and a local device capability profile containing a set of device capability vectors.

[0068] The adaptive compilation module decouples the state vectors of each time step in the state space vector sequence from the underlying hardware execution logic, extracting independent thermodynamic control factors, mechanical motion control factors, and material feeding control factors. Specifically, the thermodynamic control factor is characterized by the target temperature rise range and the desired heating rate parameter; the mechanical motion control factor is characterized by the relative stirring intensity level parameter; and the material feeding control factor is characterized by the target feeding mass percentage parameter.

[0069] In step S220, the adaptive compilation module combines the thermodynamic parameter vectors in the device capability vector set and calls the decoupled thermodynamic control factors to perform time and power duty cycle mapping calculations on the heating action factors.

[0070] For preheating or reducing actions that include heat treatment requirements, the adaptive compilation module, based on the principle of thermodynamic work, calculates the reference heating time required for the target equipment to reach the current relative state constraint variables and the corresponding reference heating power duty cycle, according to the maximum rated output power of the equipment, the heating response time constant of the electromagnetic coil, the total loading mass of the material in the pot and its equivalent specific heat capacity.

[0071] The adaptive compilation module uses the baseline heating power duty cycle, calculated based on the desired heating rate, as the control parameter sent by the microcomputer numerical control unit to the electromagnetic heating component. The calculation of the baseline heating time is based on the formula for the work done by the basic specific heat capacity. The mapping relationship is expressed by the following formula: ; in, Indicates the first The baseline heating time required for each timing step; This represents the preset thermodynamic compensation coefficient in the equipment's underlying control system; This represents the heating response time constant of the corresponding equipment; This indicates the equivalent specific heat capacity of the material in the pot at the current time step; This indicates the total loading mass of the material in the pot at the current time step; Indicates the maximum rated output power; This represents the target temperature rise span required by the relative state constraint variables in the extracted thermodynamic control factors; This represents the duty cycle parameter of the baseline heating power calculated in this step.

[0072] In step S230, the adaptive compilation module combines the mechanical motion parameter vector in the device capability vector set to perform speed mapping on the stir-frying and mixing action factors. The adaptive compilation module parses the relative state constraint variables in the state vector, extracts the relative stir-frying intensity factor, and places it into the motor speed limit range of the target bottom mechanism for stretching calculation.

[0073] This process transforms the unified relative intensity description in the cloud into an absolute control pulse frequency for a 360° drum or planetary mixer motor. The adaptive compilation module performs interpolation and scaling calculations based on the upper and lower limits of motor speed constraints calibrated in the local device capability profile, and outputs the final target absolute operating speed sent to the drive motor.

[0074] In step S240, the adaptive compilation module maps the solenoid valve opening duration to the basic feeding action factor by combining the material delivery parameter vector in the equipment capability vector set. The adaptive compilation module locates the independent material box channel where the target seasoning is located based on the operation object feature value in the state vector, and calculates the solenoid valve opening duration of the corresponding spraying channel by proportional conversion based on the basic feeding quality required in the recipe semantics and the unit time reference flow rate parameter of the spraying port recorded in the local equipment capability file.

[0075] In step S250, the adaptive compilation module concatenates and encapsulates the calculated baseline heating time, baseline heating power duty cycle, target absolute operating speed, and solenoid valve opening time into an array according to the timeline of the cooking process. After encapsulation, the adaptive compilation module outputs a basic hardware execution sequence, which enables multi-source heterogeneous devices to obtain a unified underlying parameter basis for executing the semantics of the current recipe.

[0076] See Figure 7 The adaptive compilation module is configured to perform secondary mapping calculations by introducing variable conditions from the set of food nutrition constraint vectors on the basis of the basic hardware execution sequence. This process aims to make the execution results conform to the set dietary health goals by adjusting the opening time of the underlying solenoid valve, heating power and mechanical speed in a coordinated manner.

[0077] In step S260, the adaptive compilation module calls the allergen logic masking vector from the food nutrition constraint vector set to filter the material delivery action factors in the basic hardware execution sequence based on channel connectivity. The adaptive compilation module extracts the Boolean release variable from the masking vector and performs logical matching and numerical combination with the initial solenoid valve opening duration of the corresponding spraying channel.

[0078] When an allergen conflict is detected, the opening duration command of the solenoid valve in the corresponding channel is directly reset to zero, and the system forcibly cuts off the spraying action of the seasoning containing the allergen from the underlying driving mechanism.

[0079] In step S270, the adaptive compilation module performs quantitative reduction adjustment on the basic hardware execution sequence based on the quantitative limit vector in the set of food nutrition constraint vectors. The adaptive compilation module classifies and summarizes the total demand mass of a single nutrient category throughout the entire cooking process cycle, compares it with the upper limit threshold in the quantitative limit vector, and calculates the material attenuation coefficient of each nutrient category.

[0080] Based on the obtained material attenuation coefficient, the adaptive compilation module proportionally reduces the opening time of the solenoid valve in the corresponding nutrient channel, generating the final safe energization time after quantitative constraint recalculation. The adaptive compilation module synchronously completes the parameter reduction and adjustment for the oil, sodium, and sugar channels according to the same logic.

[0081] In step S280, the adaptive compilation module triggers a correlation compensation mechanism to address the thermodynamic medium changes caused by the reduction adjustment. During the cooking process, oil and water act as heat transfer media, and the reduction in their mass will cause a sudden increase in the local heat absorption rate of the food. The adaptive compilation module extracts the material attenuation coefficient calculated in step S270 and uses it to perform reverse compensation correction on the reference heating power duty cycle and motor stirring speed in the same time sequence step.

[0082] The correlation compensation calculation is expressed by the following formula: ; ; in, This indicates the corrected target electromagnetic heating power duty cycle; This indicates the duty cycle of the baseline heating power calculated in the original basic hardware execution sequence; This indicates the system's built-in power thermal compensation coefficient; This indicates the target drive motor's operating speed after compensation; This indicates the originally set absolute operating speed; Indicates the kinematic speed compensation coefficient; This item represents the relative reduction in the amount of oil added.

[0083] Through the above calculations, when the amount of oil added is limited due to nutritional constraints, the adaptive compilation module will simultaneously reduce the output power of the electromagnetic heating and increase the stirring speed of the underlying mechanical mechanism to compensate for the uneven heating of the food caused by the reduction of the heat transfer medium and prevent local scorching.

[0084] In step S290, the adaptive compilation module integrates the various control parameters after channel filtering, quantitative reduction and correlation compensation calculation. The adaptive compilation module encapsulates the recombined time control parameters, power control parameters, motor speed parameters and on / off duration parameters of each solenoid valve in an array, and outputs an adjustment state execution parameter set carrying nutritional constraints as the underlying running instructions for subsequent execution and safety boundary detection.

[0085] See Figure 8 The adaptive compilation module is configured to intercept out-of-limit operations and check for timing mutual exclusion on the adjusted execution parameter set after nutrient constraint recalculation before issuing instructions to the underlying layer. This process relies on the security protection boundary parameters in the local device capability file to ensure that the generated parameter set will not induce device failure or operational safety accidents.

[0086] In step S310, the adaptive compilation module performs thermodynamic extreme value verification and amplitude limiting processing on the adjusted state execution parameter set. The adaptive compilation module scans the target electromagnetic heating power duty cycle and expected continuous heating duration in each timing step frame by frame. The adaptive compilation module extracts the maximum rated output power, the highest safe temperature limit value, and the anti-dry burning protection trigger limit duration from the local equipment capability file as the judgment criteria.

[0087] The adaptive compiler module uses a built-in thermodynamic integral function to predict the cumulative temperature rise of the pot under the current instruction combination. This thermodynamic integral function is expressed as: ; in, To estimate the temperature; This is the initial temperature detected by the current system; The overall thermal efficiency of electromagnetic heating; This is the maximum rated output power; To adjust the duty cycle of the target electromagnetic heating power in the set of parameters for the execution state; The specific heat capacity of the material; Total material mass; The expected continuous heating duration is determined by the adaptive compilation module. When the estimated temperature in a certain timing step exceeds the maximum safe temperature limit, the heating power of that timing step is forcibly attenuated and limited to ensure that the final output power does not exceed the maximum rated output power specified in the equipment file.

[0088] For the safety protection limit logic of heating time, the adaptive compilation module detects the total cumulative mass variable of the material inside the pot under the current state. When the total mass variable of the material is zero and the continuous heating time required by the instruction is greater than the trigger limit time of the anti-dry burning protection, the adaptive compilation module forcibly cuts off the excess instruction delay, thereby establishing the safe final continuous heating time.

[0089] In step S320, the adaptive compilation module performs a time-series mutual exclusion check on the mechanical structure and material delivery actions of the adjusted execution parameter set. Since the underlying hardware includes a stepless electric multi-angle tilting mechanism and a multi-channel solenoid valve spray port, if the sequence of actions of the mechanism overlaps at the same time, it is very easy to cause the seasoning liquid to spray to the outside of the pot or cause a short circuit in the machine body. The adaptive compilation module establishes a multi-concurrent action interference matrix to check the tilting angle parameters of the tilting pot and the on / off state parameters of the solenoid valve within the same execution frame.

[0090] The adaptive compilation module determines in real time whether the tilt angle of the tilting mechanism required by the current instruction sequence exceeds the calibrated safe working tilt angle threshold. When it is determined that the current tilt angle is not within the limit of a large tilting posture, the original solenoid valve opening instruction is retained. When it is determined that the tilt angle has exceeded the limit, the adaptive compilation module forcibly overwrites and clears the content in the concurrent matrix and cuts off the material injection instruction for that period.

[0091] In step S330, the adaptive compilation module encapsulates the various instruction parameters (specifically including: the duty cycle of the target electromagnetic heating power after the amplitude limiting process, the final continuous heating time, the target absolute operating speed, the on / off time parameters of each solenoid valve, and the tilt angle of the tilting mechanism) after the amplitude limiting process and mutual exclusion check. The adaptive compilation module converts the control parameters into a hexadecimal machine code queue that can be directly read by the microcomputer numerical control unit according to the data frame format of the underlying bus communication protocol.

[0092] For the specific underlying queue scheduling mechanism of protocol data encapsulation, those skilled in the art can configure it based on conventional embedded control systems. Its memory queue scheduling is a well-known technology in this field and will not be elaborated here.

[0093] The adaptive compilation module outputs the encapsulated execution sequence, which is the set of absolute action instructions issued by the system to the cooking execution and perception module. At this point, the device-independent recipe semantics in the cloud have been completely converted into a local control code string that takes into account the target hardware's unique parameter capabilities, user nutritional and health constraints, and the absolute boundary safety of device operation.

[0094] See Figure 9 The cooking execution and perception module is configured to synchronously collect multi-dimensional environmental parameters through the sensing unit during the process of controlling the underlying hardware devices to run according to the execution sequence, and build a real-time status dataset to provide absolute data support for the closed-loop dynamic correction of the system.

[0095] In step S410, the microcomputer numerical control unit parses the execution sequence issued by the adaptive compilation module and sends action level signals to the execution unit according to the preset machine clock cycle. While activating the electromagnetic heating component, the multi-channel solenoid valve nozzle and the stirring motor, the microcomputer numerical control unit sends synchronous wake-up commands to each data acquisition channel of the sensing unit 32 to establish a timestamp mapping between the sensor data stream and the execution action.

[0096] In step S420, the sensing unit uses an infrared temperature sensor and a K-type temperature sensor to monitor the thermodynamic state of the pot and the food inside. The infrared temperature sensor collects the surface radiation temperature of the food in a non-contact, high-frequency manner, while the K-type temperature sensor is attached to the outer wall of the pot or embedded in the bottom heating area to collect the base conduction temperature of the pot. The microcomputer numerical control unit performs weighted fusion and Kalman filtering on the two temperature signals and outputs the actual comprehensive temperature for the current clock cycle.

[0097] The formula for calculating the actual combined temperature is as follows: ; in, Indicates in The actual overall temperature read by the underlying system at any given time; express The surface temperature value is constantly collected and converted by an infrared temperature sensor; express The substrate temperature value is constantly collected by a K-type temperature sensor; This represents the temperature measurement weighting coefficient; the system dynamically assigns values ​​to μμ based on the stirring speed and pot loading capacity indicated in the current execution sequence, and the dynamic calculation rules are given by the following formula: ; in, The built-in baseline static weights; This represents the forward gain coefficient for rotational speed. The stirring speed is the speed assigned to the currently executing sequence; This is the load reverse attenuation coefficient; This represents the current loading capacity of the pot as reported by the sensor. A dynamic value is calculated using the above rules to accurately correct for surface temperature measurement errors caused by changes in exposed area under different material tumbling conditions.

[0098] In step S430, the sensing unit monitors the operating resistance and material state of the underlying mechanical mechanism through the motor current sensor. When the drum motor or planetary stirring motor executes the stir-frying action command in the execution sequence, the change in the state of the ingredients in the pot will be directly reflected in the fluctuation of the load torque at the motor shaft end. For example, in the sauce reduction stage, the increase in material viscosity caused by water evaporation will cause the operating resistance to rise.

[0099] The motor current sensor collects the instantaneous measured operating current of the stator winding. The microcomputer numerical control unit combines the no-load reference current at the current corresponding speed and the rated operating current limit value extracted from the local equipment capability file, and converts the collected data into mechanical load parameters that characterize the actual resistance of the food in the pot by normalizing the difference ratio.

[0100] In step S440, the sensing unit uses the flow meter device to monitor the material flow of the independent material box feeding mechanism. When the execution sequence triggers the opening of the specific multi-way solenoid valve nozzle, the liquid flow meter configured at the end of the pipeline or the front position of the pump body synchronously starts time integration calculation. The microcomputer numerical control unit continuously reads the pulse count value fed back by the flow meter device to obtain the current actual cumulative mass of feeding.

[0101] The microcomputer numerical control unit compares the target feed quality originally required for the specific channel to be output in the execution sequence with the actual outflow quality obtained by integrating the current instantaneous flow rate over time, and extracts the actual feed deviation variable that characterizes the current process execution status.

[0102] For the data sampling mechanism of high-precision analog-to-digital converters and the pulse capture principle of flow sensors, those skilled in the art can perform circuit board-level design based on conventional industrial control hardware circuits. The underlying analog-to-digital signal acquisition process is a well-known technology in this field and will not be elaborated here.

[0103] In step S450, the cooking execution and sensing module extracts various collected parameters within the same clock cycle (specifically including the actual comprehensive temperature, mechanical load parameters, and actual feeding deviation variables obtained from real-time monitoring), and combines them with the currently recorded actual heating output power parameters to generate a multi-dimensional real-time status data matrix.

[0104] This matrix fully maps the objective data of the equipment at the current execution site. The system sends it to the memory comparison register of the microcomputer numerical control unit as the input basis for triggering dynamic correction and anti-dry burning protection.

[0105] The cooking execution and perception module is configured to use a closed-loop control algorithm to intervene in the operating parameters of the underlying actuators in real time after acquiring a multi-dimensional real-time status data matrix. This process ensures that the actual cooking state closely follows the theoretical target set in the execution sequence, and performs hardware-level safety cutoff when an out-of-limit trend is detected.

[0106] In step S460, the microcomputer numerical control unit performs subdivided firepower compensation based on the actual comprehensive temperature. The microcomputer numerical control unit extracts the target temperature setpoint of the current time step in the execution sequence, performs a difference operation between it and the actual comprehensive temperature in the real-time status data matrix, and generates a temperature deviation variable.

[0107] When materials are added in a concentrated manner or when the ambient temperature changes suddenly, causing a temperature jump inside the pot, the microcomputer numerical control unit calls the internal proportional-integral-derivative control algorithm and uses the extracted dynamic temperature deviation variable to perform closed-loop mathematical gain calculation to obtain the detailed firepower compensation amount.

[0108] Subsequently, the microcomputer numerical control unit superimposes the dynamically calculated subdivided firepower compensation amount with the original output power control parameters in the execution sequence to generate the final dynamic output power control duty cycle that is dynamically sent to the electromagnetic heating component.

[0109] For the coefficient tuning and anti-integral saturation limiting logic of the proportional-integral-derivative control algorithm, those skilled in the art can perform conventional calibration based on the specific electrical characteristics of the heating coil. The parameter tuning of its closed-loop controller is a well-known technology in this field and will not be elaborated here.

[0110] In step S470, the microcomputer numerical control unit (NC unit) adaptively adjusts the stirring speed based on the mechanical load parameters. The NC unit reads the mechanical load parameters from the real-time status data matrix. During process stages such as reducing the sauce, which cause an increase in material viscosity, the mechanical load parameters rise accordingly. The NC unit sets a load reference threshold. When the actual mechanical load parameters deviate from this threshold, the NC unit extracts the instantaneous load difference and, combined with the built-in speed-load feedback coupling strength, performs reverse intervention compensation on the preset target operating speed in the execution sequence.

[0111] The microcomputer numerical control unit directly sends the dynamically fine-tuned actual operating speed to the underlying motor. This mechanism overcomes local sticking resistance by increasing the speed, keeping the material turning evenly.

[0112] In step S480, the microcomputer numerical control unit performs dynamic correction of the on / off time of the multi-way solenoid valve based on the actual feeding deviation variable. The microcomputer numerical control unit detects the opening status of each spraying channel in the current timing step. When the system clock reaches the original solenoid valve closing time of a certain channel's execution sequence, the microcomputer numerical control unit compares the actual feeding deviation variable of that channel at this time.

[0113] When the actual feed deviation exceeds the set allowable quality error range, it indicates that the actual outflow rate has not met the standard due to fluctuations in pipeline air pressure or liquid viscosity. The microcomputer numerical control unit forcibly blocks the solenoid valve closing command at the bottom layer and directly calculates the additional extension of the solenoid valve opening time by combining the instantaneous measured mass flow rate at the original disconnection time point and the remaining feed deviation variable. After continuously maintaining the solenoid valve on for the calculated extension time, the microcomputer numerical control unit sends the solenoid valve cut-off level again.

[0114] In step S490, the microcomputer numerical control unit performs safety boundary inspection and intervention of hardware operation status, constructs the highest priority hardware interrupt program, compares the actual comprehensive temperature with the highest safe temperature limit value in the local equipment capability file in real time, and simultaneously monitors the output power duty cycle of the electromagnetic heating component and the mechanical load parameters fed back by the sensing unit.

[0115] When the system determines that the actual overall temperature has reached the maximum safe temperature limit, or when the electromagnetic heating component is in a high-power output state and the mechanical load parameter is lower than the set empty pot resistance lower limit threshold and continues to exceed the anti-dry-burning protection trigger limit duration, the microcomputer numerical control unit immediately generates the highest level protection interrupt signal. This interrupt signal bypasses the software scheduling queue and directly pulls down the gate drive level of the insulated gate bipolar transistor of the electromagnetic heating component, cutting off the heating circuit. The microcomputer numerical control unit locks the position of the mechanical actuator and outputs an alarm fault code to the touch screen and communication module until a manual reset command is received. This mechanism constructs the last line of defense for the underlying control system.

[0116] The learning iteration module is configured to start a data collection and fusion process after the end command of a single cooking process is triggered. This process aims to combine objective operational deviations at the execution level with subjective sensory evaluations of the final dish to construct a comprehensive feedback data sample for optimizing underlying parameters.

[0117] In step S510, the learning iteration module extracts the underlying execution record file of the entire cooking cycle from the storage register of the microcomputer numerical control unit. The learning iteration module performs dimensionality reduction and feature extraction on the time series data in the underlying execution record file to construct the process execution feature vector.

[0118] The learning iteration module reads the actual comprehensive temperature, actual execution power duty cycle, actual feeding deviation variables, and various compensation parameters generated during the dynamic correction phase.

[0119] To quantify the degree to which the machine's execution process deviates from the ideal state, the learning iteration module uses an integral model to calculate the thermodynamic cumulative error factor and the material delivery cumulative error factor. The learning iteration module obtains the relative deviation ratio between the instantaneous actual comprehensive temperature within the actual total time of a single cooking session and the preset theoretical target temperature in the execution sequence, and integrates this deviation ratio over the entire time domain to accurately calculate the thermodynamic cumulative error factor.

[0120] The learning iteration module uses the same time integration operation mechanism to continuously accumulate the overall material delivery deviation to obtain the cumulative error factor of material delivery. Finally, the learning iteration module packages and outputs the process execution feature vector containing multiple error factors.

[0121] Step S520: The learning iteration module acquires the raw sensory evaluation data for the finished cooking product and converts it into a quantitative feature vector. The raw sensory evaluation data is obtained from form scoring inputs collected by the touch screen connected to the microcomputer numerical control unit, as well as feedback messages from user mobile terminals received through the communication interface. The evaluation dimensions are broken down into doneness and texture, saltiness deviation, color deviation, and overall satisfaction.

[0122] The learning iteration module converts descriptive semantics or star ratings into numerical vectors with positive and negative polarities based on a pre-set mapping table. For example, when the evaluation system receives feedback that the salt content is too high, the corresponding saltiness deviation variable is assigned a positive deviation value.

[0123] When feedback indicating that the food is undercooked is received, the corresponding doneness / texture variable is assigned a negative deviation value. The user interface form structure design and the message parsing process for external terminal communication can be developed by those skilled in the art based on conventional software application layer protocols. The terminal human-computer interaction and data reception mechanisms are well-known technologies in the field and will not be elaborated upon here.

[0124] In step S530, the learning iteration module performs validity checks and outlier cleaning on the extracted process execution feature vector and evaluation quantization feature vector, and reads the underlying hardware fault register flag bits during the cooking cycle.

[0125] When a serious hardware fault code indicating a sensor disconnection, motor stall, or heating coil overcurrent is detected, the learning iteration module marks the corresponding process execution feature vector as an invalid sample and discards it to avoid including equipment anomalies caused by non-parameter configuration in subsequent learning calculations.

[0126] In step S540, the learning iteration module aligns and concatenates the process execution feature vector and the evaluation quantification feature vector through effectiveness detection to construct a comprehensive feedback data matrix. Within the process execution feature vector merged into the matrix, the learning iteration module compiles the calculated thermodynamic cumulative error factor, material delivery cumulative error factor, the total number of times the closed-loop dynamic correction procedure is triggered in a single cooking cycle, and the absolute time difference between the actual total time and the theoretical time.

[0127] Simultaneously, the learning iteration module gathers the ripeness and texture deviation values, saltiness deviation values, color deviation values, and the final overall satisfaction weight score from the evaluation quantitative feature vector side.

[0128] The learning iteration module writes the assembled comprehensive feedback data matrix into a non-volatile storage area, completing a single archiving. This comprehensive feedback data matrix establishes a mathematical mapping relationship between the underlying control deviation and the upper-level finished product defects, forming a complete basic data structure for triggering error reverse settlement and parameter updates.

[0129] The learning iteration module is configured to, after constructing the comprehensive feedback data matrix, call the internal error back-calculation model to perform closed-loop correction calculations on the underlying execution parameters. This process converts multi-dimensional deviation data into executable control code increments, thereby generating independent control parameter versions tailored to the specific characteristics of the equipment.

[0130] In step S550, the learning iteration module reads the process execution feature vector and evaluation quantification feature vector from the comprehensive feedback data matrix. The learning iteration module uses these vectors as input node data for the error back-calculation model. This error back-calculation model adopts a multivariate linear matrix mapping structure to establish a gradient correction relationship between the upper-level sensory evaluation deviation, the middle-level execution error feedback, and the lower-level basic control parameters.

[0131] The embedded multiple linear regression algorithm structure and the setting of its initial weight bias parameters can be developed and built by those skilled in the art based on conventional machine learning architectures. Its basic matrix regression calculation and gradient optimization mechanism are well-known technologies in the field and will not be described in detail here.

[0132] In step S560, the learning iteration module calculates the correction increment of each core underlying control parameter through the error reverse settlement model. The learning iteration module extracts the ripeness and texture deviation and saltiness deviation in the evaluation quantification feature vector, and combines the thermodynamic cumulative error factor and material delivery cumulative error factor in the process execution feature vector to perform reverse compensation settlement on the benchmark heating power duty cycle and the basic opening time of the solenoid valve for the next execution of the same recipe.

[0133] The specific logic for calculating the incremental correction of core parameters is based on a negative feedback equation with dimension transformation, and is represented by the following mathematical model: ; ; in, This indicates the calculated result for the first... The correction increment of the duty cycle of the reference heating power in each timing step; This represents the dimensional weighting coefficient for maturity and quality (unit: duty cycle increment / unit subjective score). This represents the cooked texture deviation value extracted from the quantitative feature vector. When the food is undercooked, it is a negative deviation value. Through equation negative feedback calculation, a positive increment can be obtained to increase the heating power duty cycle. Represents the dimensionless thermodynamic error weighting coefficient; The thermodynamic cumulative error factor extracted from the process execution feature vector is represented by a negative cumulative value when the actual temperature is less than the preset theoretical target temperature, which in turn increases the duty cycle. Indicates that for the first Controlling the first timing step The correction increment for the opening time of the solenoid valve in a specific spray channel; This represents the dimensional weighting coefficient for the deviation of saltiness from normality (unit: seconds / unit subjective rating). This represents the saltiness deviation value extracted from the quantitative feature vector. When the taste is too bland, it is a negative value. The reverse calculation obtains a positive time increment to extend the valve opening time. This represents the weighting coefficient for material delivery error with dimensions. This represents the cumulative error factor of material input extracted from the process execution feature vector. It is a negative value when the actual input amount is low.

[0134] The dimensional definition of the above weight coefficients completely eliminates the problem of inconsistent dimensions on the left and right sides of the equation, and the inverse operator built into the correction formula ensures the correctness of the underlying parameter correction logic.

[0135] In step S570, the learning iteration module adds the calculated correction increments to the original local execution parameter sequence of the corresponding recipe. To prevent the parameters from diverging without limit and causing the underlying hardware to run out of bounds, the learning iteration module calls the local device capability profile preset by the data acquisition module to perform a forced safety limit on the new parameter sequence after the addition. The learning iteration module assigns the obtained correction increments to the original baseline heating power duty cycle before correction, and performs bidirectional threshold truncation clamping on the calculated sum based on the lower and upper limits of the heating output power calibrated in the capability profile, thereby generating a reliable target baseline heating power duty cycle after the update limit.

[0136] Simultaneously, the learning iteration module adds the material feeding correction increment to the original solenoid valve opening duration and performs zeroing protection to prevent the value from being negative, so as to finally lock the target solenoid valve opening duration.

[0137] In step S580, the learning iteration module restructures and encapsulates the various control parameters that have passed the safety limit according to the underlying action sequence. The learning iteration module generates a target device-specific cooking parameter set and assigns a unique version verification code and timestamp identifier to the parameter set. The learning iteration module writes the target device-specific cooking parameter set into the non-volatile memory of the local microcomputer numerical control unit for offline caching.

[0138] When the system receives a data request instruction with the same dish identifier and nutritional constraints again, the adaptive compilation module will prioritize calling the target device's dedicated cooking parameter set as the output benchmark for the underlying execution sequence, thus completing the closed-loop iterative optimization of the control parameters of the commercial cooking equipment.

[0139] The learning iteration module is configured to generate a set of cooking parameters specific to the target device locally, and then perform local storage scheduling and cloud synchronization operations. This mechanism maintains the availability of parameters for the local device when it is offline and establishes a cooking parameter sharing channel between multiple homogeneous hardware devices.

[0140] In step S591, the learning iteration module assigns an incrementing local version number to the newly generated target device-specific cooking parameter set, and attaches the local device's hardware identification code and generation timestamp. The microcomputer numerical control unit allocates a fixed-capacity parameter cache area in the non-volatile memory to store different versions of the parameter set.

[0141] When the data occupancy of the parameter cache reaches the preset space threshold, the learning iteration module starts the cache replacement program, which cleans up the old parameter files that have been called infrequently in the past according to the least recently used algorithm. For the specific code implementation of the read and write management and cache eviction strategy of the local storage file system, those skilled in the art can configure it according to the conventional embedded file system. Its memory block erasure and space reclamation mechanism is a well-known technology in the field and will not be described in detail here.

[0142] In step S592, the learning iteration module monitors the bandwidth utilization of the local network communication module. During idle network communication periods, the learning iteration module encapsulates the target device-specific cooking parameter set carrying the version number and hardware identification code, along with the comprehensive feedback data matrix that triggered this update, into an encrypted synchronization message. The learning iteration module then uploads this encrypted synchronization message to the device management pool on the cloud server via the wide area network.

[0143] In step S593, the cloud server receives and decrypts the synchronization message, extracts the hardware identification code, and searches the cloud database for a group of isomorphic commercial cooking equipment with the same thermodynamic boundary parameters and mechanical motion parameters. The cloud server extracts the exclusive cooking parameter sets uploaded by multiple isomorphic devices, performs parameter group aggregation calculation, and updates the basic preset control parameters corresponding to this specific model stored on the cloud side.

[0144] The cloud-based aggregation computing process can be represented by the following mathematical equation: ; in, Indicates after the first After the next iteration update, the cloud generates a set of benchmark parameters for this type of homogeneous device; This indicates the total number of homogeneous devices participating in this aggregation calculation; Indicates the first The first device uploaded Version of the target device-specific cooking parameter set; The system assigns the first The trust weight coefficient of the parameters uploaded by the device, and the system requirements. .

[0145] cloud server according to the The average overall satisfaction score in the historical comprehensive feedback data matrix of the equipment was used to... The dynamic allocation calculation shows that the more stable the user rating of a device, the higher the weight ratio of its uploaded parameters in the aggregation calculation.

[0146] Step S594, the cloud server will aggregate the generated The optimal parameters are distributed to other commercial cooking devices within the same group that have not yet generated local iterative data, enabling cross-device sharing of optimal parameters. The learning iteration module synchronously performs parameter anomaly circuit breaker monitoring.

[0147] When the cloud server detects that a certain parameter version has caused the overall satisfaction score to fall below the baseline on multiple homogeneous devices, or caused the devices to frequently trigger the thermodynamic extreme value verification limit, the cloud server will issue a version forced rollback command to the corresponding batch of devices.

[0148] Upon receiving a forced rollback command, the learning iteration module immediately suspends the currently executing task queue, erases the storage sector containing the target device's unique cooking parameter set that caused the exception, and the microcomputer numerical control unit resets the call pointer of the underlying execution sequence, rolling back to the storage address where the verification code of the previous verified stable version is located. This mechanism ensures that the cooking device maintains the underlying operational reliability and the minimum requirements for the finished product during the continuous optimization of parameters without human intervention.

Claims

1. A method for adaptive control of nutritional constraints in commercial cooking equipment, characterized in that, Includes the following steps: The data acquisition module is used to obtain cloud-based device-independent recipe semantics, ingredient nutritional constraint parameters, and local device capability profiles. The adaptive compilation module is used to convert the semantics of the cloud-based recipes that are irrelevant to the local device capability profile and the nutritional constraint parameters of the ingredients, thereby generating an execution sequence adapted to the current device. The cooking execution and perception module controls the operation of the underlying hardware device according to the execution sequence, collects real-time status data during the cooking execution process, and performs dynamic correction based on the real-time status data. The learning iteration module uses the real-time status data and finished product evaluation data to correct and calculate the equipment execution parameters for similar recipes, generating a set of cooking parameters specific to the target equipment.

2. The nutritional constraint adaptive control method for commercial cooking equipment according to claim 1, characterized in that, The step of using the data acquisition module to obtain cloud-device-independent recipe semantics includes: The data acquisition module is used to parse and process the initial recipe data package to extract recipe semantics that are irrelevant to the cloud device; The data acquisition module uses a tree-structured data structure to parse the cloud device-independent recipe semantics and read the action semantic instructions. The data acquisition module uses temporal logic to vectorize and reconstruct the action semantic instructions, outputting a state space vector sequence. The action semantic instruction consists of an action identifier, the characteristic value of the operation object, and the relative state constraint variable.

3. The nutrient constraint adaptive control method for commercial cooking equipment according to claim 2, characterized in that, The process of acquiring nutritional constraint parameters of food ingredients and local device capability profiles using the data acquisition module includes: The data acquisition module reads local underlying parameters, performs classification and structuring processing, generates the local device capability profile, and converts the local device capability profile into a set of device capability vectors. The data acquisition module is used to convert the raw data of the food nutrition constraint parameters into numerical constraint variables and output a set of food nutrition constraint vectors, which includes quantitative restriction variables and logical masking variables.

4. The nutrient constraint adaptive control method for a commercial cooking equipment according to claim 3, characterized in that, The step of using an adaptive compilation module to convert the semantics of cloud-based recipes irrelevant to the local device based on the local device capability profile and the nutritional constraint parameters of the ingredients, and generating an execution sequence adapted to the current device, includes: The adaptive compilation module is used to decouple the state vectors in the state space vector sequence from the instructions, and to extract independent thermodynamic control factors, mechanical motion control factors, and material delivery control factors. The execution sequence is generated by using the adaptive compilation module to combine the thermodynamic parameter vectors in the equipment capability vector set with the mechanical motion control factor and the material delivery control factor for synchronous mapping calculation.

5. The nutrient constraint adaptive control method for a commercial cooking equipment according to claim 4, characterized in that, The cooking execution and sensing module includes a sensing unit and a microcomputer numerical control unit; The method of collecting real-time status data during the cooking process using the cooking execution and sensing module includes: The surface temperature value is collected using an infrared temperature sensor in the sensing unit, and the substrate temperature value is collected using a K-type temperature sensor in the sensing unit. The actual comprehensive temperature is extracted by using the microcomputer numerical control unit to perform a weighted calculation based on the surface temperature value, the substrate temperature value, and the temperature measurement weighting coefficient. The actual comprehensive temperature is used as the basic data for subsequent splicing to generate a multi-dimensional real-time status data matrix.

6. The nutrient constraint adaptive control method for a commercial cooking equipment according to claim 5, characterized in that, The method of collecting real-time status data during the cooking process using the cooking execution and sensing module also includes: The instantaneous measured operating current is monitored by the motor current sensor in the sensing unit, and the instantaneous measured operating current is converted into mechanical load parameters by the microcomputer numerical control unit. The actual cumulative mass of feed is obtained by using the flow meter device in the sensing unit, and the actual feed deviation variable is extracted by comparing the target feed mass with the actual outflow mass. The mechanical load parameters and the actual feeding deviation variables are concatenated into the real-time status data matrix.

7. The nutrient constraint adaptive control method for a commercial cooking equipment according to claim 6, characterized in that, The dynamic correction based on the real-time status data includes: The microcomputer numerical control unit is used to perform a difference operation between the target temperature setpoint in the execution sequence and the actual comprehensive temperature in the real-time status data matrix to generate a temperature deviation variable, and the temperature deviation variable is used to obtain the subdivided firepower compensation amount. The mechanical load parameters in the real-time status data matrix are read using the microcomputer numerical control unit. When the mechanical load parameters deviate from the load reference threshold, reverse intervention compensation is performed on the target operating speed in the execution sequence.

8. The nutrient constraint adaptive control method for a commercial cooking equipment according to claim 1, characterized in that, The step of using the learning iteration module to correct the equipment execution parameters of similar recipes based on the real-time status data and finished product evaluation data includes: The learning iteration module is used to extract the execution feature vector of the underlying execution record file construction process; The learning iteration module is used to obtain the raw data for finished product evaluation, and the raw data for finished product evaluation is converted into an evaluation quantification feature vector. The learning iteration module is used to align and concatenate the process execution feature vector and the evaluation quantification feature vector to construct a comprehensive feedback data matrix.

9. The nutrient constraint adaptive control method for a commercial cooking equipment according to claim 8, characterized in that, Also includes: The learning iteration module uses the process execution feature vector and evaluation quantization feature vector in the comprehensive feedback data matrix as input node data to calculate the correction increment of the underlying control parameters; The learning iteration module is used to superimpose the correction increment onto the original local execution parameter sequence of the corresponding recipe, and the local device capability file is called to perform a forced safety limit on the superimposed new parameter sequence to generate the target device-specific cooking parameter set.

10. The nutrient constraint adaptive control method for a commercial cooking equipment according to claim 9, characterized in that, After generating the target device's specific cooking parameter set, it also includes: The learning iteration module is used to encapsulate the target device-specific cooking parameter set and the comprehensive feedback data matrix into an encrypted synchronization message and upload it to the cloud server. The cloud server is used to retrieve a group of isomorphic commercial cooking equipment with the same thermodynamic boundary parameters and mechanical motion parameters from the cloud database, and the exclusive cooking parameter sets uploaded by multiple isomorphic devices are extracted to perform parameter group aggregation calculation.

Images