A method for suppressing temperature fluctuation of an electric cooker

Abstract

This invention relates to a method for suppressing temperature fluctuations in an electric cooker, aiming to solve the problems of temperature overshoot and insufficient stability caused by thermal expansion and contraction during heating. The core technical solution includes: acquiring and processing strain signals using multiple sets of flexible piezoresistive strain sensors arranged on the bottom of the pot to extract micron-level deformation characteristics of the pot body; constructing a deformation characteristic mapping table related to overshoot risk levels by combining measured data from pots of various materials, determining the risk in real time and triggering feedforward power compensation control to dynamically adjust the heating power to suppress temperature oscillations and overshoot; achieving power recovery compensation during the rebound phase, and adaptively optimizing risk assessment and compensation parameters through periodic data and satisfaction scores. This solution improves the temperature control accuracy and stability during the heating process of the electric cooker, reduces overshoot, and enhances heat preservation performance.

Description

Technical Field

[0001] This invention relates to the field of temperature control technology for electric cooking appliances, and in particular to a method for suppressing temperature fluctuations in an electric cooker. Background Technology

[0002] Currently, electric cooking appliances, especially electric cookers, widely employ closed-loop proportional-integral-derivative (PID) regulation or its improved algorithms based on temperature sensors to dynamically adjust heating power for temperature control. With the development of intelligent home appliances, this field is gradually shifting from fixed-value heating to refined temperature control. Some new products have also introduced ambient temperature compensation, zoned power decoupling, and predictive modeling strategies to improve heat preservation accuracy and user experience. In practical engineering applications, traditional temperature control methods commonly use sensors such as Pt100 and NTC thermistors to detect the temperature of the pot body or local areas inside the pot, reflecting the system's thermal state, and then adjusting the heater output power accordingly to achieve target temperature stability. However, this type of technology has several problems that are difficult to fundamentally avoid.

[0003] First, the electric heating element and the metal pot itself have significant thermal inertia. When the temperature sensor detects a value higher than the target, it issues a power reduction command. However, the heat stored in the pot has not yet been released, resulting in a non-negligible time lag between the power adjustment and the actual temperature response. When the heater is turned off or the power is reduced, the residual heat of the pot still acts on the medium, causing an overall temperature overshoot. This manifests as the temperature rapidly exceeding the target threshold and then falling back, resulting in temperature fluctuations. Repeated PID feedback may further cause secondary or even multiple temperature oscillations, which not only makes it difficult to ensure the heat preservation effect but also affects the flavor and food safety.

[0004] Secondly, existing compensation algorithms based on thermoelectric coupling and data-driven prediction typically require complex heat transfer process models, material thermal parameter matrices, or neural network prediction structures to compensate for the stability deficiencies of simple PID control. While these methods can reduce overshoot under laboratory conditions, in actual mass production and home use, they suffer from weak model robustness due to limitations in manufacturing tolerances, the diversity of pot structures, and changes in the usage environment. Furthermore, they place high demands on the computing power and storage resources of embedded control hardware, leading to increased costs and difficulties in widespread adoption.

[0005] Furthermore, some recent technologies in the industry have attempted to improve local temperature response speed by introducing methods such as zoned heating, fuzzy rule bases, or multi-sensor coordination. However, none of these methods have substantially eliminated the fundamental problem of temperature overshoot caused by the thermal inertia of the pot itself. Their versatility and debugging complexity have also become bottlenecks for widespread adoption. In addition, existing technologies always use temperature as the only feedback variable, which cannot achieve a direct correlation between the actual mechanical response and thermal state of the pot, easily overlooking the hidden overshoot risk under seemingly stable temperatures. Summary of the Invention

[0006] This application provides a method for suppressing temperature fluctuations in an electric cooker, which aims to solve one of the problems or issues of the prior art mentioned in the background section.

[0007] This application provides a method for suppressing temperature fluctuations in an electric cooker, specifically including: S1: Acquire the original voltage signals of four sets of flexible piezoresistive strain sensing units evenly distributed along the circumference of the bottom of the metal pot body of the electric cooker, and perform noise reduction and normalization processing on the original voltage signals to generate standardized real-time strain time series data.

[0008] S2: Based on the standardized real-time strain time series data, the current strain change rate is calculated using the sliding window difference method, and the synchronization of signals from adjacent sensing units is judged by combining spatial consistency verification logic, so as to generate an effective deformation feature vector characterizing the thermal expansion start-up state.

[0009] S3: Based on the measured data of typical thermal expansion, cooling and contraction cycles of pots of different materials under the target insulation temperature in the pot body thermal deformation feature library, construct a mapping table between deformation features and overshoot risk level, including rise slope, peak hysteresis period and rebound decay time, to generate a dynamic threshold benchmark for risk assessment.

[0010] S4: Input the effective deformation feature vector into the deformation feature and overshoot risk level mapping table for matching and comparison. If the strain change rate of multiple adjacent sensing units exceeds the dynamic threshold benchmark in multiple consecutive sampling periods and the duration of each exceeding the threshold meets the minimum judgment window requirement, it is determined to be a high-risk deformation mode, and a control command signal to trigger feedforward compensation action is generated.

[0011] S5: In response to the control command signal, reduce the current heating power according to the preset gradient coefficient, and simultaneously start the power freeze window to prevent conventional proportional-integral-derivative feedback regulation from intervening, so as to generate a smooth power output curve in a state of suppressed oscillation.

[0012] S6: At the end of the power freeze window, monitor in real time whether the standardized real-time strain timing data has exceeded the peak value and entered the rebound stage. If it is confirmed that the rebound stage has been entered, generate recovery control parameters that include a slight power rebound amount.

[0013] S7: Based on the positive correlation between the slight power recovery amount and the current rebound rate, the conventional proportional-integral-derivative closed-loop control that is reintroduced after the power freeze is lifted is superimposed and corrected to generate a compensated heating power setpoint that offsets the temperature lag caused by the release of residual heat from the pot.

[0014] S8: Archive the deformation response curve, actual temperature fluctuation trajectory and user satisfaction score data during this insulation cycle, and update the risk threshold and compensation parameters in the deformation characteristics and overshoot risk level mapping table to generate an optimized dynamic threshold benchmark for the next control cycle.

[0015] The method for suppressing temperature fluctuations in an electric cooker provided in this application has the following beneficial effects: (1) By taking the mechanical deformation of the pot body as a directly observable proxy variable of the thermal inertia process, a mapping table of “deformation characteristics → overshoot risk level” based on the measured strain time sequence characteristics is constructed. This effectively overcomes the response lag problem caused by the reliance on temperature sensor feedback in traditional heat preservation control and significantly advances the timing of identifying the heating overshoot trend. Combined with the sliding window differential calculation and spatial consistency verification mechanism, a high sensitivity and high reliability of thermal expansion start-up is achieved at a sampling frequency of 20ms, avoiding false triggering caused by local noise or single-point failure. Thus, without increasing the complexity of thermal coupling modeling, the system’s adaptability to different materials of pots (such as 304 stainless steel, aluminum alloy, and composite multilayer structure) in dynamic thermal cycling is greatly improved. This effectively solves the defects of control instability and large temperature fluctuation caused by poor model generalization and complicated parameter tuning in the existing technology.

[0016] (2) A feedforward-driven power intervention strategy is introduced. After a high-risk deformation mode is detected, the heating power is immediately reduced and a "power freeze window" is started. During this period, the superimposed intervention of PID feedback regulation is actively shielded to effectively prevent conflict and oscillation between feedforward and feedback control actions, and to ensure the clear boundary and dynamic stability of the control action. When the freeze window ends, the intelligent decision-making recovery strategy is based on whether the strain signal has entered the rebound stage. If it has rebounded, a micro-amplitude power compensation positively correlated with the residual heat decay rate is injected simultaneously to accurately offset the temperature downward deviation caused by thermal inertia lag. This realizes a paradigm upgrade from "passive correction" to "active pre-control", significantly improves the temperature stability and trajectory smoothness during the heat preservation stage, significantly reduces the temperature overshoot amplitude and oscillation frequency, and improves the consistency and comfort of the user experience.

[0017] (3) An integrated adaptive learning module is used to continuously optimize the risk threshold classification and compensation parameter configuration by utilizing the complete deformation response curve, actual temperature trajectory and user satisfaction score archived for each heat preservation cycle. This enables the system to have a closed-loop evolution capability that continuously improves with the increase in usage frequency. This design does not rely on high-precision temperature prediction models, fuzzy rule base reconstruction or partition decoupling, which greatly reduces the computing power requirements of the controller and the engineering deployment threshold. At the same time, it retains good compatibility with diverse pot types and cooking scenarios, showing excellent robustness and scalability. The overall solution takes physically interpretable mechanical response as the core of perception, gets rid of the excessive reliance on indirect derivation of variables in traditional methods, and builds a lightweight, highly reliable and easily generalizable new intelligent heat preservation control architecture. It is particularly suitable for home appliance scenarios with high requirements for safety and user experience, and has outstanding technological advancement and industrial application prospects.

[0018] The combined effects of these technologies have formed a new type of heat preservation control system that uses the thermal deformation of the pot itself as a sensing benchmark, feedforward intervention as the core means, and adaptive optimization as a continuous evolution mechanism. This system achieves forward control response, simplified regulation logic, and robust system performance, fundamentally improving the operating quality and user experience of electric cookers under complex real-world conditions. Attached Figure Description

[0019] Figure 1 This is the main flowchart of a method for suppressing temperature fluctuations in an electric cooker.

[0020] Figure 2 This is a sub-flowchart of a method for suppressing temperature fluctuations in an electric cooker.

[0021] Figure 3 This is another sub-flowchart of a method for suppressing temperature fluctuations in an electric cooker. Detailed Implementation

[0022] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0023] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0024] like Figure 1 As shown, this application provides a method for suppressing temperature fluctuations in an electric cooker, specifically including: S1: Acquire the original voltage signals of four sets of flexible piezoresistive strain sensing units evenly distributed along the circumference of the bottom of the metal pot body of the electric cooker, and perform noise reduction and normalization processing on the original voltage signals to generate standardized real-time strain time series data.

[0025] S2: Based on the standardized real-time strain time series data, the current strain change rate is calculated using the sliding window difference method, and the synchronization of signals from adjacent sensing units is judged by combining spatial consistency verification logic, so as to generate an effective deformation feature vector characterizing the thermal expansion start-up state.

[0026] S3: Based on the measured data of typical thermal expansion, cooling and contraction cycles of pots of different materials under the target insulation temperature in the pot body thermal deformation feature library, construct a mapping table between deformation features and overshoot risk level, including rise slope, peak hysteresis period and rebound decay time, to generate a dynamic threshold benchmark for risk assessment.

[0027] S4: Input the effective deformation feature vector into the deformation feature and overshoot risk level mapping table for matching and comparison. If the strain change rate of multiple adjacent sensing units exceeds the dynamic threshold benchmark in multiple consecutive sampling periods and the duration of each exceeding the threshold meets the minimum judgment window requirement, it is determined to be a high-risk deformation mode, and a control command signal to trigger feedforward compensation action is generated.

[0028] S5: In response to the control command signal, reduce the current heating power according to the preset gradient coefficient, and simultaneously start the power freeze window to prevent conventional proportional-integral-derivative feedback regulation from intervening, so as to generate a smooth power output curve in a state of suppressed oscillation.

[0029] S6: At the end of the power freeze window, monitor in real time whether the standardized real-time strain timing data has exceeded the peak value and entered the rebound stage. If it is confirmed that the rebound stage has been entered, generate recovery control parameters that include a slight power rebound amount.

[0030] S7: Based on the positive correlation between the slight power recovery amount and the current rebound rate, the conventional proportional-integral-derivative closed-loop control that is reintroduced after the power freeze is lifted is superimposed and corrected to generate a compensated heating power setpoint that offsets the temperature lag caused by the release of residual heat from the pot.

[0031] S8: Archive the deformation response curve, actual temperature fluctuation trajectory and user satisfaction score data during this insulation cycle, and update the risk threshold and compensation parameters in the deformation characteristics and overshoot risk level mapping table to generate an optimized dynamic threshold benchmark for the next control cycle.

[0032] Step S1: Acquire the original voltage signals of four sets of flexible piezoresistive strain sensing units evenly distributed circumferentially on the bottom of the metal pot of the electric cooker, and perform noise reduction and normalization processing on the original voltage signals to generate standardized real-time strain time series data. Specifically, this includes: S1.1: Obtain the original voltage signal output by four sets of flexible piezoresistive strain sensing units evenly distributed along the circumferential direction on the bottom of the metal pot of the electric cooker. Discretize the original voltage signal using an analog-to-digital conversion interface at a high frequency sampling rate to generate an original digital sequence containing mixed features of thermal deformation information and high-frequency electromagnetic interference.

[0033] The input conditions are analog voltage signals output by four sets of flexible piezoresistive strain sensing units evenly distributed circumferentially along the bottom of the metal pot of the electric cooker. Each sensing unit is in a high-temperature working environment and has millivolt-level sensitivity. The signals contain mixed characteristics of thermal deformation information and high-frequency electromagnetic interference.

[0034] The analog voltage signals of each sensing unit are connected to the multi-channel buffer input of the analog-to-digital conversion interface. The sampling frequency is set based on the thermal inertia characteristics of the pot body and the results of interference spectrum analysis to ensure that the sampling frequency is greater than twice the frequency of the highest expected deformation signal to satisfy the Nyquist sampling theorem.

[0035] During the analog-to-digital conversion process, a full-channel synchronous triggering mechanism is invoked to ensure that the signals of the four sensing units are quantized at the same sampling time, so as to form initial digital data with spatial synchronization characteristics.

[0036] For each acquired analog signal, quantization mapping is performed, and the voltage amplitude is converted into a digital coded value through linear segmentation within a preset input range. The number of bits for encoding is determined according to the required strain resolution.

[0037] During the digital output stage, the outputs of each channel are spliced ​​together into a multi-vector record according to the sampling time sequence and a channel identifier and timestamp are added to form the original digital sequence, which retains the complete mixed characteristics of thermal deformation information and high-frequency electromagnetic interference.

[0038] By using high-frequency synchronous acquisition and precise quantization mapping processing, the analog voltage signal from the previous step is converted into a raw digital sequence that can be used for subsequent noise removal and normalization operations, thereby achieving high-fidelity thermal deformation information capture.

[0039] For example, in one embodiment, the static output range of the four sets of flexible piezoresistive strain sensing units is 0 mV to 500 mV, the analog-to-digital conversion interface is configured with 16-bit precision, the full-scale input corresponds to a numerical range of 0 to 65535, and the sampling frequency is set to 5 kHz to cover the maximum frequency component of the pot deformation signal, 2 kHz. During the acquisition process, the synchronous trigger delay of the four channels is controlled within 1 μs. The four-channel voltage data of each sampling point are quantized and mapped to generate digital values, for example, 100 mV corresponds to a mapped value of 13107, and 200 mV corresponds to a mapped value of 26214. The multi-vector records are arranged in chronological order and accompanied by channel numbers and millisecond-level timestamps to form the original digital sequence. Verification shows that this sequence can significantly improve the resolution of the thermal deformation signal after subsequent wavelet denoising processing and effectively preserve the strain data characteristics corresponding to the micron-level curvature change of the pot bottom.

[0040] S1.2: Based on the original digital sequence, a moving average filter is used to smooth small-amplitude random noise, and wavelet threshold denoising technology is combined to remove transient pulse interference caused by the switching action of the heating element, so as to generate a clean strain voltage sequence after filtering out high-frequency noise.

[0041] S1.3: For the clean strain voltage sequence, call the zero-point drift compensation coefficients of each sensing unit pre-stored in the controller, and perform baseline correction calculation to eliminate the sensor static offset error caused by long-term high-temperature operation, so as to generate a calibrated strain voltage sequence with corrected zero-point deviation.

[0042] For the clean strain voltage sequence after filtering out high-frequency noise, the zero-point drift compensation coefficients pre-stored in the controller's non-volatile memory and uniquely bound to each flexible piezoresistive strain sensing unit are retrieved, and a mapping table of sensing unit index and compensation coefficient is established as the index basis for subsequent baseline correction.

[0043] The clean strain voltage sequence is grouped according to the sensing unit channel, the zero drift compensation coefficient in the corresponding mapping table is read, and a static offset elimination operation is performed on each sampling point.

[0044] The corrected channel voltage values ​​are reconstructed into a time series. The drift-compensated data segments are then stitched together in time stamp order to form a continuous calibration voltage sequence, ensuring that the output sequence is consistent with the input sequence in the time domain and has no sampling delay.

[0045] For the reconstructed calibration voltage sequence, a global baseline alignment operation is performed to calculate the average voltage reference value of all channels in the initial temperature range, and this value is used as a reference for uniform translation adjustment to eliminate residual baseline inconsistencies between channels.

[0046] Through the baseline correction and zero-point drift compensation described above, the clean strain voltage sequence output in the previous step is transformed into a calibrated strain voltage sequence with corrected zero-point deviation and consistent baseline across channels, thus achieving accurate input conditions for the subsequent physical strain calculation stage.

[0047] For example, in a 304 stainless steel electric cooker prototype, the zero-point drift compensation coefficients of the four sets of sensing units are 0.018V, 0.022V, 0.019V, and 0.021V, respectively, with a sampling frequency of 500Hz. The original amplitude of a sampling point in a certain channel of the clean strain voltage sequence is 0.245V, and the corrected value of 0.227V is calculated by calling the compensation coefficients. The same correction is applied to all sampling points in this channel sequentially to form the drift compensation voltage sequence for that channel. Subsequently, during the timestamp alignment process, the compensation sequences of the four channels are reconstructed according to the sampling time sequence to form a complete calibration strain voltage sequence. The average reference voltage of 0.225V under the initial temperature range is calculated during global baseline alignment, and the sequences of each channel are shifted to ensure that the baselines of each channel are consistent. It has been verified that the fluctuation amplitude of the corrected calibration sequence is significantly reduced under static no-thermal deformation conditions, with the maximum amplitude stabilizing within the range of ±0.002V, providing a stable input for accurate mapping to physical strain in the subsequent S1.4 step.

[0048] S1.4: Based on the calibration strain voltage sequence and the sensitivity calibration curves of each flexible piezoresistive strain sensing unit in the standard temperature range, perform linear mapping transformation calculation to convert the voltage amplitude into a physical strain variable characterizing the micron-level curvature change of the bottom of the pot, so as to generate a real-time strain measurement value sequence with clear physical meaning.

[0049] For the calibrated strain voltage sequence after zero-point drift compensation, the sensitivity calibration curve data of each flexible piezoresistive strain sensing unit in the standard temperature range is called in the controller as a mapping reference to determine the strain proportional coefficient corresponding to each voltage amplitude.

[0050] Each discrete data point in the calibration strain voltage sequence is interpolated according to the sensitivity calibration curve of its respective sensing unit to eliminate the step mapping error caused by the sampling gap of the calibration curve and form a high-resolution voltage-strain correspondence matrix.

[0051] Based on the aforementioned correspondence matrix, a linear mapping transformation method is used to convert the unit voltage value into a physical strain variable that characterizes the micron-level curvature change of the bottom of the pot.

[0052] Perform time series integrity checks on the obtained dependent variable sequence, remove isolated outliers caused by sampling jitter, and rearrange the time index to ensure that subsequent processing modules have strict sampling consistency when calling time series data.

[0053] The strain variable sequence after time integrity verification is input into the physical meaning identification module, and metadata including sensor unit number, spatial location index and sampling timestamp is added as the final output to generate a real-time strain measurement value sequence with clear physical meaning.

[0054] By using linear mapping calculations driven by sensitivity calibration curves and outlier removal, we ensure that the strain measurement values ​​converted from voltage signals have accurate physical units and spatial positioning information, thus achieving precise quantification from voltage signals to physical deformation.

[0055] For example, under a certain heat preservation mode, the sensitivity calibration curves of the four sets of flexible piezoresistive strain sensing units on the bottom of the 304 stainless steel pot body show proportionality coefficients of 2.15, 2.12, 2.10, and 2.14 μm / m·V, respectively. -1 The calibration voltage sequence after zero-point drift compensation is [1.024V, 1.018V, 1.021V, 1.019V]. After interpolation to eliminate curve sampling gaps, the strain values ​​are [2.202 μm / m, 2.158 μm / m, 2.144 μm / m, 2.179 μm / m]. No outliers were found after time sequence integrity verification. After adding the sensor unit number and sampling timestamp, a real-time strain measurement value sequence was formed. This sequence was mapped to the zero-to-one interval in the subsequent S1.5 normalization process, eliminating differences in pot material, ensuring the comparability of different pot types in the S2 deformation rate calculation, and the accuracy of physical units significantly improved the accuracy of high-risk pattern recognition.

[0056] S1.5: Perform maximum and minimum value normalization processing on the real-time strain measurement value sequence to uniformly map the dimensions of different material pots and different sensing units to the dimensionless range of zero to one, so as to generate standardized real-time strain time series data that eliminates individual differences and dimensional influences.

[0057] Step S2: Based on the standardized real-time strain time-series data, the current strain change rate is calculated using the sliding window difference method, and the synchronization of signals from adjacent sensing units is determined by spatial consistency verification logic to generate an effective deformation feature vector characterizing the thermal expansion initiation state. Specifically, this includes: S2.1: Perform a sliding window truncation operation on the standardized real-time strain time series data to generate a local strain data sequence containing continuous sampling points, which serves as the basic input object for differential operations.

[0058] S2.2: Based on the local strain data sequence, the central difference method is used to perform point-by-point slope calculation processing to generate a set of original strain change rates characterizing the instantaneous thermal expansion rate of each sensing unit.

[0059] The input object is a local strain data sequence containing continuous sampling points generated by step S2.1. Each element of the data sequence corresponds to the standardized strain measurement value of a specific flexible piezoresistive strain sensing unit at a certain moment.

[0060] The central difference operation module is called on the local strain data sequence to sequentially select the adjacent sample values ​​of each data point to form a difference operation pair, and the slope is calculated using the sampling interval time parameter as the divisor.

[0061] Set the sampling interval time parameter to the reciprocal of the controller's sampling frequency to ensure that the slope unit remains the strain amplitude per second.

[0062] Using the central difference formula in, Let be the instantaneous thermal expansion rate at the i-th sampling point. and These represent the strain amplitudes at adjacent sampling points. This is the sampling interval.

[0063] The rate values ​​output by the formula are categorized and stored according to the sensor unit number, forming a set of original strain change rates covering all sensor units.

[0064] During the transformation process, bidirectional differential degradation is applied to the first and last sampling points, and extrapolated values ​​are used to fill in missing neighboring points to avoid slope distortion at data boundaries.

[0065] A one-time integration is performed on the generated set of original strain change rates to ensure that the data structure is consistent with the input format of the subsequent S2.3 synchronization comparison logic.

[0066] By using the central difference method and boundary correction processing, the local strain data sequence from the previous step is transformed into a set of original strain change rates with clear physical meaning, thereby achieving high-precision quantification of the instantaneous thermal expansion rate of each sensing unit.

[0067] For example, for a local strain data sequence with a sampling frequency of 50 Hz, the sampling interval time is... With a time limit of 0.02 seconds, the adjacent amplitudes at the 10th sampling point of a flexible piezoresistive strain sensor were 0.452 and 0.447, respectively, yielding a calculated rate of 0.125 strain units per second. This value, after being aggregated and combined with the rate of change of the other three groups of sensors at the same sampling point, constituted the original strain rate set, verifying that it can significantly improve the rate resolution capability of the thermal expansion start-up state under real-world operating conditions.

[0068] S2.3: Perform neighborhood synchronization comparison logic on the values ​​of the four sets of flexible piezoresistive strain sensing units in the original strain change rate set to generate a synchronization state flag bit array that marks the consistency of spatial distribution.

[0069] S2.4: Based on the synchronization status flag array, filter the effective strain change rate subset that meets the threshold of the number of adjacent units to generate spatial consistency verification result data that excludes single-point abnormal interference.

[0070] The synchronization status flag array is read as the initial input data for spatial consistency verification. A group-by-group neighborhood matching operation is performed on this flag array, performing a logical AND operation on the synchronization flag status of each sensing unit and its circumferentially adjacent units to form a matching result matrix that satisfies the threshold for the number of adjacent units. Records in the matching result matrix that only have single-point triggering and do not meet the threshold condition are removed to eliminate the influence of single-point abnormal interference signals. During the removal process, a binary mask generation method based on the target threshold is used, with the mask value corresponding to high synchronization records in the flag array and low synchronization records set to zero. The strain rate change set after masking is reorganized, combining continuous rate change data that meet the threshold condition into an effective rate change subset according to spatial distribution rules. Through this reorganization, spatial consistency verification result data specifically used for subsequent encapsulation of structured information is generated, realizing the conversion of rate change data from the full set to an effective subset.

[0071] By performing neighbor-to-neighbor matching operations, mask removal, and data recombination, the original strain rate set from the previous step is transformed into spatial consistency verification result data that excludes single-point anomaly interference and meets the spatial synchronization judgment conditions, thus realizing the reliable data screening before judging high-risk deformation modes.

[0072] For example, the synchronization status flag array of the four sets of flexible piezoresistive strain sensing units is set to [1,1,0,1], where "1" indicates synchronous triggering and "0" indicates no synchronous triggering. The neighborhood matching threshold is set to at least two adjacent units that are synchronously triggered. In the matching operation, the first and second sets meet the condition, the fourth set meets the condition, and the third set is eliminated because it is not triggered. When generating the mask, the mask value corresponding to the flag bits of the first, second, and fourth sets is 1, and the mask value corresponding to the third set is 0. The strain rate set is [0.0021,0.0019,0.0005,0.0023], which is obtained after masking as [0.0021,0.0019,0,0.0023], and then reorganized into an effective rate subset [0.0021,0.0019,0.0023]. In this scenario, all three groups of adjacent units meet the threshold conditions, and the spatial consistency verification results fully retain the high-reliability change rate information. The effective deformation feature vector of the subsequent encapsulation can significantly improve the stability and accuracy of the thermal expansion start-up state determination.

[0073] S2.5: Based on the spatial consistency verification result data encapsulation, structured information including timestamp, average rate of change and synchronization confidence is generated to generate an effective deformation feature vector characterizing the thermal expansion initiation state.

[0074] Based on the spatial consistency verification results, the timestamp generation module is invoked to append system clock values ​​to all strain rate of change data points that meet the threshold conditions, in order to form traceable time stamp information.

[0075] An arithmetic mean operation is performed on the effective strain rate of change subset with attached timestamps. The strain rate of change values ​​of each adjacent sensing unit are weighted and fused according to the spatial consistency weighting coefficient to obtain the average rate of change index.

[0076] Using the synchronization confidence calculation module, the current synchronization confidence value is calculated based on the number of stable durations of the synchronization status flag bits and the flag bit distribution density in the spatial consistency verification results, according to the preset confidence mapping function.

[0077] A structured encapsulation logic is adopted to combine timestamp information, average rate of change index, and synchronization confidence value into a single data structure in a fixed three-field format, ensuring that the field sequence matches the analytical model of the subsequent risk assessment module.

[0078] The combined structured information is written into the effective deformation feature vector cache for subsequent matching and comparison between deformation features and risk mapping relationship table. Full field encapsulation ensures data integrity and parsing accuracy, enabling a reliable quantitative description of the thermal expansion initiation state.

[0079] like Figure 2 As shown, step S3 involves: based on measured data of typical thermal expansion, cooling, and contraction cycles of pots made of different materials at the target insulation temperature, constructed a mapping table between deformation characteristics and overshoot risk levels, including the rise slope, peak hysteresis period, and rebound decay time, to generate a dynamic threshold benchmark for risk assessment. Specifically, this includes: S3.1: Obtain the original measured dataset of thermal expansion, cooling and contraction cycle processes of pots of different materials recorded in the pre-stored pot body thermal deformation feature library at a specific heat preservation target temperature. Use data cleaning to remove abnormal noise and missing segments in the original measured dataset and perform interpolation repair to generate standardized multi-material pot body thermal deformation time series benchmark data.

[0080] The pot body thermal deformation feature library is a database in the non-volatile memory of a pre-installed electric cooker controller. It stores measured strain time-series data of typical thermal expansion-cooling contraction cycles of pot bodies made of different materials (such as 304 stainless steel, aluminum alloy, ceramic coating, etc.) at a specific heat preservation target temperature. This feature library provides the raw data foundation for step S3, used to extract deformation features such as rise slope, peak hysteresis period, and rebound decay time, and further constructs a mapping table between deformation features and overshoot risk levels to achieve risk assessment based on thermal deformation features.

[0081] The boiler body thermal deformation feature library adopts a hierarchical index structure and consists of the following elements: Material Category Index: Classified by pot body material, such as 304 stainless steel, 430 stainless steel, aluminum alloy, cast iron, etc.

[0082] Thermal insulation target temperature index: For each material, data is stored for different thermal insulation setting temperatures (such as 60℃, 70℃, 80℃, 90℃).

[0083] Thermal cycling process data recording: Each record contains a strain-time curve for a complete thermal expansion-cooling-contraction cycle, with sampling time in the order of seconds, and the strain is micro-strain or normalized value.

[0084] Feature annotation fields: Each record is accompanied by manually or automatically annotated rise slope (micro-strain / second), peak hysteresis period (second), rebound decay time constant (second), and corresponding temperature overshoot amplitude (°C), which are used for subsequent mapping relationship table construction.

[0085] Construction method: (a) Data Acquisition: In a standard laboratory environment, using the same electric cooker prototype as the product, a strain sensing unit (same as S1) was installed. The pot was heated to the set insulation temperature and then allowed to cool naturally. The strain change curve of the complete thermal expansion, cooling, and contraction cycle was recorded, and the temperature of the pot bottom was recorded using a thermocouple. The experiment was repeated multiple times for different materials and different insulation target temperatures. After removing outlier data, the average value was taken.

[0086] (II) Feature Extraction and Labeling: For the strain curve of each cycle, sliding window extreme value detection is used to determine the rising segment, peak plateau segment, and rebound descent segment. Linear regression is performed on the rising segment to calculate the slope; the duration of the peak plateau segment is calculated; and exponential fitting is performed on the rebound descent segment to obtain the time constant. The corresponding temperature overshoot amplitude (i.e., the maximum deviation of the temperature from the set value) is recorded. The feature parameters of each cycle are stored together with the original curve data in the feature library.

[0087] (iii) Data storage: Utilizing a structured database (such as SQLite) or binary files, the data is embedded in the controller firmware. A query interface is provided to support rapid retrieval of corresponding strain time-series reference data based on material and target insulation temperature.

[0088] Through the aforementioned thermal deformation feature library of the pot body, the thermal deformation patterns of pot bodies of different materials under different heat preservation target temperatures are digitally stored, providing a standardized and reusable data foundation for subsequent risk level determination and dynamic threshold generation.

[0089] For the pre-stored database of pot body thermal deformation features, which contains raw measured datasets of the thermal expansion, cooling, and contraction cycles of pots made of different materials at a specific target insulation temperature, the input object is determined to be the raw sampling sequence of multi-material pots containing strain-time curves. Based on the raw sequences acquired by sensor sampling, the data cleaning module is invoked to perform abnormal noise removal. Sampling points that exceed the physically reasonable range or exhibit abrupt amplitude changes are identified as noise and deleted, while the noise location index is recorded for subsequent interpolation repair. For time-series data with missing segments after cleaning, a piecewise interpolation method is used to construct an interpolation model for each missing interval. Cubic spline interpolation is selected to ensure the continuity of the curve slope, and the missing values ​​are filled in through interpolation prediction. The complete time-series data after filling is input into the benchmarking unit, and amplitude adjustment is performed according to the standard reference curve of material type and target insulation temperature to eliminate strain amplitude deviations of different materials in the same temperature range. A uniform time axis resolution is selected, and resampling is used to standardize the time axis of pot body data with different sampling frequencies, aligning all data at the same sampling interval. Through the above processing method, the original measured dataset from the previous step is transformed into standardized multi-material pot thermal deformation time series benchmark data with good consistency, so as to achieve stable input conditions for subsequent feature extraction and risk mapping.

[0090] For example, when the target insulation temperature of a 304 stainless steel pot is 80℃, the original strain data contains 20 data points with a sampling frequency of 0.5 seconds within 10 seconds. Among them, the 6th and 7th points show a sudden increase in curvature change of 0.12 micrometers, exceeding the physically reasonable threshold of 0.08 micrometers. Therefore, these two noise points are marked and removed during the data cleaning stage. The missing interval formed after removal is filled using cubic spline interpolation, combined with three valid points before and after the missing points to construct an interpolation model, resulting in a smooth and continuous strain curve. The completed data is then standardized and adjusted with an amplitude scaling factor of 0.95 to make its strain amplitude consistent with the reference amplitude of the aluminum alloy pot when it is kept at 80℃. For the data with the original sampling interval of 0.5 seconds, it is adjusted to a uniform sampling interval of 0.2 seconds through resampling interpolation, generating a standardized time series of 50 data points. The rebound decay time constant calculated on this series can be used as the input feature for the subsequent risk mapping table construction. The application effect shows that the strain curve has no discontinuous abrupt changes, all material data are comparable year-on-year, and the feature extraction error is significantly reduced.

[0091] S3.2: Based on the standardized multi-material pot body thermal deformation time series benchmark data, the sliding window extreme value detection and linear regression fitting techniques are used to perform feature extraction operations on the strain signal rising segment, peak plateau segment and rebound falling segment of each complete thermal cycle, so as to generate a three-dimensional deformation feature vector set containing the rising slope value, peak lag time duration and rebound decay time constant.

[0092] Based on the standardized multi-material pot body thermal deformation time series reference data, the processing object is set as a strain signal sequence covering the complete thermal expansion-cooling contraction cycle, with pot body material type and target insulation temperature as index conditions. A sliding window extreme value detection is applied to the input time series data, with a fixed window width and number of sampling points. Local maxima and minima of the strain signal are located by scanning window by window to ensure accurate capture of the starting point of the rising segment, the peak segment, and the end point of the rebound segment. Linear regression fitting is performed on the detected rising segment data segments, using strain value as the dependent variable and time point as the independent variable to obtain the slope parameter of the fitted line, which serves as the rising slope value of the pot body's thermal expansion stage. The peak duration is calculated for the peak plateau segment data segments. Time difference calculation is used to subtract the peak start time from the peak end time to obtain the peak lag period, characterizing the continuous effect of the pot body deformation peak on temperature maintenance. Exponential regression fitting is performed on the rebound descent segment data segments, fitting the strain value change to an exponential decay function. The time constant parameter is extracted as the rebound decay time constant, used to characterize the cooling contraction rate. The aforementioned slope, hysteresis period, and time constant are stored as a set of three-dimensional deformation feature vectors categorized by material, serving as the input benchmark for subsequent risk mapping analysis. This processing method transforms the standardized thermal deformation time-series benchmark data from the previous step into structured indicators with dynamic characteristics of rise rate, peak persistence, and shrinkage decay, enabling the quantitative extraction of the dynamic characteristics of thermal inertia in multi-material pot bodies.

[0093] For example, at a target insulation temperature of 90℃ for a 304 stainless steel pot body, strain time-series data for the complete thermal expansion-cooling contraction cycle were collected. The sampling frequency was 50Hz, and the window width was set to 25 sampling points. The sliding window extreme value detection located the starting time of the rising segment at 2.4 seconds, the peak time at 18.6 seconds, and the ending time of the springback segment at 35.2 seconds. The length of the rising segment data fragment was 16.2 seconds, and linear regression fitting yielded a slope of 0.0025 microstrain / second; the peak lag time was calculated as 6.8 seconds using the time difference; the springback segment data fragment length was 16.6 seconds, and exponential regression fitting yielded a time constant of 4.2 seconds. In the aluminum alloy pot body sample, the same method yielded a rising slope of 0.0031 microstrain / second, a peak lag time of 5.4 seconds, and a springback time constant of 3.7 seconds. Both sets of feature data were stored as three-dimensional vectors for risk mapping analysis. During the verification process, it was found that pot types with larger upward slope values ​​and shorter peak lag periods have a significantly increased risk of temperature overshoot at high temperatures. Pots with smaller rebound time constants show a more pronounced temperature lag during the cooling process. This feature extraction method significantly improves the sensitivity and stability of risk assessment on actual prototypes.

[0094] S3.3: Call the temperature fluctuation trajectory record data archived in the historical experiment, and perform spatiotemporal alignment and correlation between each set of deformation feature data in the three-dimensional deformation feature vector set and the corresponding actual temperature overshoot amplitude data. Use statistical correlation analysis method to calculate the contribution weight of each deformation feature component to the temperature overshoot amplitude, so as to generate a weighted correlation matrix that characterizes the correlation strength between deformation features and overshoot risk.

[0095] S3.4: Based on the contribution weight distribution pattern in the weighted correlation matrix, a multi-level risk classification threshold range is set, and the data points in the three-dimensional deformation feature vector set are mapped to three overshoot risk level categories of low risk, medium risk and high risk according to their comprehensive risk scores, so as to generate a preliminary deformation feature and overshoot risk level classification mapping dataset.

[0096] Given the input conditions of the three-dimensional deformation feature vector set and the weighted correlation matrix, the contribution weights corresponding to each feature component in the matrix are called to perform a comprehensive risk score calculation. The rising slope, peak lag period, and rebound decay time of a single vector are superimposed into a single risk indicator according to their weight ratios.

[0097] The entire set of risk indicators is statistically sorted, and the quantile analysis method is used to determine the low, medium and high risk division points. The threshold range corresponding to the quantile is then adjusted in combination with the material characteristic parameters.

[0098] Based on the threshold range, the risk indicator values ​​are mapped to three risk categories, and a unique risk level identifier is assigned to each category to form a classification code.

[0099] Cross-validation is performed on the mapping results. If an imbalance in the class distribution is found, the threshold range is dynamically adjusted to ensure that the sample distribution is stable under each risk level.

[0100] The final category identifier is appended to the original 3D deformation feature vector to generate a preliminary mapping dataset containing feature values ​​and risk levels.

[0101] By using the quantile analysis and weighted superposition processing described above, the three-dimensional deformation feature vector from the previous step is transformed into mapping data with risk level information, thereby achieving a quantitative expression of risk assessment.

[0102] S3.5: Perform boundary smoothing and logical consistency verification on the preliminary deformation feature and overshoot risk level classification mapping dataset, and construct a structured lookup table containing the rising slope threshold range, peak lag period judgment interval and rebound decay time critical value to generate the final dynamic threshold benchmark for real-time risk judgment, namely the deformation feature and overshoot risk level mapping relationship table.

[0103] Numerical boundary smoothing is performed on the preliminary deformation feature and overshoot risk level classification mapping dataset. A bidirectional iterative mean filter is used to smooth and correct continuous data points at the risk level division boundaries, preventing sharp abrupt changes in the data distribution curve near the boundaries. Logical consistency is checked on the smoothed risk threshold set. An interval inclusion verification method is used to compare the upper and lower limits of each risk level threshold, ensuring that the partitions of the rising slope threshold, peak lag threshold, and rebound decay time threshold do not overlap and conform to a monotonically increasing pattern. For the verified risk threshold set, the parameter encapsulation module is invoked to map the rising slope threshold range to the corresponding fields of a structured lookup table in interval form. The peak lag judgment interval is converted to a standardized time unit and filled into the peak field of the lookup table. The rebound decay time critical value is quantized with fixed precision and stored in the rebound field of the lookup table, forming a ternary feature threshold mapping table available for real-time querying. During the lookup table generation process, an index optimization-based fast retrieval index using a three-dimensional composite key is established to achieve millisecond-level response when the controller performs risk determination. Through the above processing and assembly, the preliminary mapping data output from the previous step is transformed into a dynamic threshold benchmark with stable numerical characteristics, rigorous logical partitioning, and fast retrieval capabilities, thereby achieving real-time and accurate determination of high-risk deformation patterns.

[0104] For example, in a heat preservation control experiment of a 304 stainless steel pot, the initial mapping dataset has rising slope thresholds of 0.008 and 0.012, with strain units in micrometers per second; peak lag thresholds of 1.8 seconds and 2.4 seconds; and a rebound decay time threshold of 3.2 seconds. Mean filtering is performed on the rising slope threshold set, replacing the 0.008 near the boundary with the calculated result of 0.5 × 0.008 + 0.012, i.e., 0.010, to reduce the abrupt change. Consistency checks are performed on the peak lag thresholds, confirming that the upper limit of the low-risk range (1.8 seconds) is less than the lower limit of the medium-risk range (2.0 seconds), conforming to interval monotonicity. The smoothed rising slope threshold range [0.010, 0.012] is mapped to the "slope" field of a lookup table, and the peak lag judgment range [2.0, 2.4] seconds is converted into a standardized number of sampling points (e.g., 100 to 120 points if the sampling frequency is 20ms), and mapped to the "peak" field. The rebound decay time threshold of 3.2 seconds was quantized to 3.200 seconds and filled into the "Bounce" field. An index was built using a three-dimensional key combination (slope, peak, bounce), and the retrieval time was measured to be 0.85 milliseconds in controller testing. This structured lookup table significantly improves matching efficiency in subsequent risk assessment calls and maintains the numerical stability of high-risk assessment outputs.

[0105] like Figure 3As shown, step S4 involves inputting the effective deformation feature vector into the deformation feature and overshoot risk level mapping table for matching and comparison. If the strain change rate of multiple adjacent sensing units exceeds the dynamic threshold benchmark in multiple consecutive sampling periods and the duration of each exceeding the threshold meets the minimum judgment window requirement, it is determined to be a high-risk deformation mode, and a control command signal to trigger feedforward compensation is generated. Specifically, this includes: S4.1: Based on the pre-stored mapping table between deformation features and overshoot risk level, multi-dimensional feature extraction is performed on the rising slope value, peak lag duration and rebound decay time parameters in the currently input effective deformation feature vector to generate a standard deformation feature dataset to be compared.

[0106] S4.2: Perform a first-order difference operation on the rising slope values ​​in the standard deformation feature dataset using a sliding time window to generate an instantaneous strain rate sequence characterizing the accelerating trend of thermal expansion.

[0107] S4.3: Logically compare the extreme points in the instantaneous strain rate of change sequence with the preset dynamic threshold benchmark. If the instantaneous strain rate of change sequence exceeds the dynamic threshold benchmark and the duration meets the minimum judgment window requirement, a preliminary judgment flag for high-risk deformation mode is generated.

[0108] The extreme points of the instantaneous strain rate of change sequence are located, and the maximum and minimum values ​​appearing in the current sliding decision window are extracted as the reference benchmark for the fluctuation amplitude.

[0109] The extreme point value is compared with the critical value of the rising slope in the dynamic threshold benchmark by difference, and the instantaneous rate of change exceeding the threshold is calculated to quantify the risk intensity.

[0110] The duration determination method compares the number of sampling points that continuously meet the condition of exceeding the threshold amplitude with the minimum determination window length to filter risk event segments that meet the time dimension constraints.

[0111] Construct a logical judgment matrix, cross-validate the risk intensity and duration judgment results, and retain only the risk event flag bits that simultaneously meet the magnitude and time conditions.

[0112] Call the deformation feature index corresponding to the time when the flag occurs to generate a preliminary judgment flag for high-risk deformation mode, which is used for subsequent spatial consistency logic verification and triggering feedforward compensation actions.

[0113] Through the above processing method, the instantaneous strain rate of change sequence from the previous step is transformed into structured risk indicator data with high risk preliminary judgment capability, realizing accurate conversion from numerical detection to pattern recognition.

[0114] S4.4: Combine the preliminary judgment flag of the high-risk deformation mode with the spatial consistency verification result in the effective deformation feature vector for joint logic verification. If it is confirmed that the adjacent sensing units are synchronously triggered and meet the judgment conditions of the high-risk deformation mode, then generate the final high-risk deformation mode confirmation signal.

[0115] The execution objects for joint logic verification, combining the preliminary judgment flag of high-risk deformation mode with the spatial consistency verification results in the valid deformation feature vector, are: the high-risk deformation mode preliminary judgment flag data structure derived from the previous steps, and the synchronization status flag array output by the spatial consistency comparison logic. The high-risk deformation mode preliminary judgment flag and the synchronization status flag array are loaded into the joint judgment register area in the data bus interface, and a bitwise AND operation is performed to generate a preliminary joint judgment mask matrix. The risk judgment logic unit is called, and based on the preset spatial consistency threshold condition, the cell count result of the joint judgment mask matrix is ​​compared with the threshold. If the count value is greater than or equal to the set threshold, a spatial consistency satisfaction flag is generated. A logical AND matching verification is performed between the spatial consistency satisfaction flag and the high-risk deformation mode preliminary judgment flag. If both are valid, a final high-risk deformation mode confirmation signal trigger bit is set in the judgment buffer. Timing stability is checked on the trigger bit to ensure that it is valid for three consecutive sampling periods before a confirmation signal is officially output to the next control instruction generation module. Through the above-mentioned joint logic verification process, the high-risk mode judgment result of the previous step is transformed into a final confirmation signal after spatial consistency verification, thereby eliminating single-point anomaly triggers and ensuring the robustness and technical effectiveness of the judgment.

[0116] For example, in an electric cooker with a 304 stainless steel pot body and a target insulation temperature of 85℃, the high-risk deformation mode preliminary judgment flag obtained from the pre-detection is in a valid state. The synchronization status flag array output by the four sets of flexible piezoresistive strain sensing units is [1,1,1,0], and the spatial consistency threshold is set to 3. After loading the two types of flags into the joint judgment register, a bitwise AND operation is performed to obtain the mask matrix [1,1,1,0], whose valid cell count is 3, satisfying the threshold condition. The spatial consistency satisfaction flag and the high-risk preliminary judgment flag are logically ANDed to obtain the final trigger bit as valid. The timing stability detector is used to verify that the trigger bit remains valid for three consecutive sampling periods (20ms per period), confirming the high-risk deformation mode. The final confirmation signal is output to the feedforward compensation strategy calling module, which drives the power execution unit to reduce the heating power value by 20% according to the gradient coefficient in subsequent steps and starts the power freeze window for 1.8 seconds. In this embodiment, joint logic verification ensures the reliability of temperature overshoot risk assessment, avoids erroneous power adjustment caused by momentary interference from a single sensor, significantly reduces the amplitude of system temperature fluctuation, and significantly improves thermal stability.

[0117] S4.5: In response to the final high-risk deformation mode confirmation signal, call the corresponding gradient coefficient configuration parameters in the feedforward compensation strategy library to generate a control command signal that triggers the feedforward compensation action, including a power reduction magnitude command and a power freeze window start command.

[0118] The feedforward compensation strategy library is a parameter configuration library pre-installed in the non-volatile memory of the electric cooker controller. It stores feedforward compensation action parameters corresponding to different risk levels (low, medium, and high) or different deformation modes (such as the range of rising slope and expected peak hysteresis). When S4 determines a high-risk deformation mode, the controller calls this strategy library to obtain the power reduction gradient coefficient and the power freeze window duration to generate a control command signal that triggers the feedforward compensation action, thereby achieving the effect of anticipating and suppressing temperature overshoot.

[0119] The feedforward compensation strategy library uses a table structure, with each record containing the following fields: Risk level identifier: low risk, medium risk, high risk (or directly use a combination of deformation feature thresholds as the index key).

[0120] Rise slope range: the corresponding strain rate range (unit: microstrain / second).

[0121] Typical value of peak lag period: used to help determine the timing of compensation (optional).

[0122] Power reduction gradient coefficient: ranging from 0 to 1, indicating the percentage by which the current heating power should be reduced (e.g., 0.2 indicates a 20% reduction).

[0123] Power freeze window duration: in seconds, representing the length of time that feedback adjustment needs to be shielded after feedforward compensation.

[0124] Applicable material types: 304 stainless steel, aluminum alloy, etc. (optional, can be omitted if material is not specified).

[0125] Construction method: (1) Offline calibration experiment: For typical pot materials (such as 304 stainless steel), different degrees of thermal expansion processes were artificially simulated during the heat preservation stage. The effect of suppressing temperature overshoot was recorded by adjusting the power reduction and freezing window duration.

[0126] With the goal of minimizing temperature fluctuations and shortening recovery time, the optimal gradient coefficient and freeze window duration were optimized for different risk levels.

[0127] (2) Parameter import: The calibrated parameters are written into the controller memory according to the risk level and then fixed in the firmware.

[0128] (3) Query interface: Based on the risk level or deformation characteristic value confirmed by S4, the controller retrieves the matching strategy record and returns the corresponding power reduction gradient coefficient and freeze window duration.

[0129] The input consists of the final high-risk deformation mode confirmation signal generated in the preceding steps, and the gradient coefficient configuration parameters preset for different risk levels in the feedforward compensation strategy library.

[0130] The parameter retrieval logic inside the controller is invoked to locate the gradient coefficient record that perfectly matches the current risk level from the feedforward compensation strategy library, and the gradient coefficient value in the record is read as the core input for power attenuation calculation.

[0131] The power reduction magnitude is obtained by multiplying the gradient coefficient with the current heating power setting.

[0132] The difference between the power reduction amount and the current heating power setting is calculated to form the target power value after reduction.

[0133] Based on the risk level-based freeze strategy table, the corresponding freeze window duration parameter is retrieved, and this parameter, together with the power reduction magnitude instruction, is encapsulated into a structured control instruction message.

[0134] Add a power freeze window start flag to the control command message to ensure that the receiver immediately enters the freeze lock state after performing power adjustment.

[0135] Through the above processing method, the high-risk mode determination result of the previous step is transformed into a control command with power reduction range and freeze window activation information, so as to achieve the expected technical effect of advanced trigger power adjustment.

[0136] Step S5: In response to the control command signal, reduce the current heating power according to a preset gradient coefficient, and simultaneously activate a power freeze window to prevent conventional proportional-integral-derivative feedback regulation from intervening, thereby generating a smooth power output curve in a state of suppressed oscillation. Specifically, this includes: S5.1: Obtain the control command signal for triggering feedforward compensation action generated in the previous steps and the current real-time heating power setting value. Based on the preset table of thermal inertia gradient coefficients of pot body material, perform linear attenuation calculation on the current real-time heating power setting value to generate a target reduced heating power value with overshoot suppression capability.

[0137] The thermal inertia gradient coefficient table for the pot body material is a parameter lookup table pre-installed in the electric cooker controller. It is used to linearly map the instantaneous rebound rate vector (characterizing the acceleration of the pot body's cooling and contraction) to a reference value for a small power recovery amount. This table is calibrated based on the heat capacity, heat dissipation characteristics, and historical experimental data of pot bodies of different materials to ensure that the power recovery amount is in an appropriate proportion to the current rebound rate, thereby offsetting the temperature lag caused by residual heat release.

[0138] This coefficient table represents a two-dimensional mapping relationship and includes: Material category index: such as 304 stainless steel, 430 stainless steel, aluminum alloy, etc.; Springback rate range: The range of springback rate (micro-strain / second) divided by step size, such as 0~0.001, 0.001~0.002, etc.; Linear mapping coefficient k: Unit (power / rate), that is, the amount of power recovery corresponding to each unit rebound rate, usually a positive value; Limiting boundary: The upper limit of the slight power recovery (to prevent excessive recovery).

[0139] Construction method: (1) Heat dissipation test calibration: For pots of different materials, the relationship between the rebound rate and the actual power recovery amount that needs to be compensated is measured during the heat preservation stage. Linear regression fitting is used to obtain the mapping coefficient k; if the nonlinear relationship is obvious, it can be piecewise linearized, that is, different k values ​​are used for different rate ranges.

[0140] (2) Coefficient table storage: The material, rate range, coefficient k and limit boundary are written into the controller’s non-volatile memory and stored in read-only form.

[0141] (3) Query and calculation: The controller searches for the corresponding mapping coefficient k in the coefficient table based on the current pot material and the instantaneous rebound rate vector (or weighted average rate), and then calculates the baseline value of the micro-power recovery amount by formula P=k×Δε.

[0142] S5.2: Receive the target reduced heating power value and drive the power execution unit to make a step adjustment. At the same time, start the power freeze window timer based on the risk level identifier in the control command signal to generate a power regulation lock flag that is in an active state.

[0143] The heating power value after the target is reduced is received as input and passed to the drive interface circuit of the power execution unit for step adjustment command loading and execution, ensuring that the output power of the heating element completes a smooth transition from the current set value to the reduced value within the preset time constant range.

[0144] The digital power adjustment module inside the power execution unit is invoked to map the target reduced heating power value to the corresponding pulse width modulation duty cycle parameter, and the duty cycle is applied in the hardware timer period update to drive the heating element to be in a power reduction state.

[0145] Read the risk level identifier field in the control command signal, load the duration configuration from the power freeze window timer module according to the freeze window duration parameter corresponding to the risk level, and reset the timer state to zero to start precise timing.

[0146] At the moment the timer starts, the power regulation lock flag register is set to the active state. The activation of the flag will be used to block the correction signal path of the conventional proportional-integral-derivative feedback regulation module during the freeze period.

[0147] The interrupt service routine of the frozen window timer continuously monitors the status of the lock flag, keeping it active until the window countdown is complete, ensuring that all closed-loop correction operations are forcibly excluded by hardware logic during the freeze period.

[0148] Through the above-mentioned drive control and state locking processing method, the target heating power value after the previous step is reduced is transformed into the power reduction output state of the power execution unit, and a power regulation lock flag bit under the freeze control condition is generated, so as to achieve the expected technical effect that the heating power is only dominated by feedforward compensation during the freeze period.

[0149] For example, in the heat preservation scenario of an electric cooker with a rated power of 800W and a pot body made of 304 stainless steel, the preliminary steps calculated a target reduced heating power of 640W. The power execution unit uses a PWM drive module with a resolution of 0.1% to map 640W to an 80% duty cycle and updates the duty cycle parameter in a timer with a period of 20ms. The risk level in the control command signal is marked as "high risk," corresponding to a freeze window duration of 2.0 seconds. After loading this parameter, the timer is started immediately and the power adjustment lock flag is set. The PID correction signal path during the freeze window is hardware shielded, allowing the heating element to maintain a stable output of 640W during these 2.0 seconds. During the verification process, the temperature fluctuation trajectory was monitored, and it was found that the temperature curve within the freeze window decreased slowly and steadily, with a significant reduction in oscillation amplitude. The lock status indicator light remained on throughout the process, and after the freeze ended, the rebound stage judgment logic was entered, achieving the dual effects of smooth power transition and oscillation suppression.

[0150] S5.3: Monitor the activation status of the power regulation lock flag and force the correction amount output by the conventional proportional-integral-derivative feedback regulation module within the freeze time window to zero, so as to generate a shielded invalid feedback regulation signal.

[0151] The active power regulation lock flag is received as the trigger condition for feedback channel shielding within the freeze window.

[0152] Invoke the real-time output correction data stream of the proportional-integral-derivative feedback adjustment module and establish an event listener binding with the power adjustment lock flag status.

[0153] While the power regulation lock flag is detected to remain active, the correction value is zeroed out, resetting the heating power correction value currently calculated by the feedback regulation module to the zero vector.

[0154] By using forced assignment, the correction value after being set to zero replaces the feedback component in the original closed-loop control command, forming an intermediate command stream without PID feedback signal.

[0155] The validity flag overwrite operation is performed on the intermediate instruction stream, marking it as a shielded state and outputting it to the subsequent heating power superposition operation unit as a shielded invalid feedback adjustment signal.

[0156] By using the shielding method described above, the heating power value after the target reduction in the previous step is combined with the control heating power reference data under undisturbed conditions, so as to realize a control link dominated only by feedforward compensation during the freeze window.

[0157] For example, during the heat preservation stage of a 304 stainless steel electric cooker, the freezing window duration is set to 2 seconds, and the power adjustment lock flag is generated by the timer triggered by sub-step S5.2. The correction value output by the PID feedback adjustment module is 5W. The zeroing process directly changes the correction value to 0 through assignment operation, replacing the feedback component in the original instruction stream, so that the power control value consists only of the feedforward compensation component. At this time, the heating power value after the target is reduced is 180W. After being superimposed with the zeroed correction value, it is maintained at a constant output of 180W. This is converted into a pulse width modulation waveform by the drive module. During the freezing window, the temperature fluctuation of the pot body is significantly reduced, the risk of temperature overshoot is effectively suppressed, and the conventional closed-loop control smoothly recovers after the window ends.

[0158] S5.4: The heating power value after the target reduction is superimposed with the shielded invalid feedback adjustment signal to eliminate the disturbance variable introduced by the feedback loop, so as to generate an intermediate smooth power control quantity dominated only by feedforward compensation.

[0159] The heating power value after the target reduction is received is used as the main input signal for intermediate power calculation, while a shielded invalid feedback adjustment signal is simultaneously introduced as an auxiliary input for disturbance elimination. Signal amplitude analysis is performed on both sets of input signals to confirm that the amplitude characteristics of the invalid feedback signal within the freeze window fully conform to the output specifications of the zero-setting process, ensuring that no additional dynamic components are introduced during the superposition process. A numerical superposition operation is used to synchronize the heating power value after the target reduction and the invalid feedback signal on the time axis, ensuring that each sampling point in the superposition corresponds to the same physical moment, avoiding calculation errors caused by timing misalignment. The heating power synthesis formula is called when performing power value calculation: Where P is the intermediate smoothing power control quantity. The value of the heating power after the target reduction is the value at the nth sampling time. This represents the value of the invalid feedback adjustment signal at the corresponding time. Based on the freeze window status flag, the invalid feedback signal is confirmed to be zero throughout the entire process, simplifying the formula to: The result of the synthesis operation is stored in the intermediate power control register, and the power control output buffer is updated synchronously for use in subsequent pulse width modulation mapping conversion. Through the above processing method, the result of the previous step is transformed into a disturbance-free intermediate smooth power control quantity, realizing stable power regulation output dominated by feedforward compensation.

[0160] For example, when performing this sub-step during the heat preservation stage of the 304 stainless steel pot, the previously calculated target reduced heating power value is 820W. The amplitude of the invalid feedback signal within the freezing window is 0W after being zeroed out. The two sets of signals are time-aligned at a sampling frequency of 50Hz, with each sampling point corresponding to a 20ms interval. The intermediate smooth power control value is obtained by superimposing the formulas. This control value is directly used as the input to the pulse width modulation module, mapped in the PWM converter according to the duty cycle parameters corresponding to the 820W power, ultimately driving the heating element to maintain a stable output. Testing shows that under these conditions, the pot temperature fluctuation amplitude is significantly reduced, the power output curve remains smooth and oscillating within the freezing window, overshoot caused by thermal inertia is effectively suppressed, and user experience performance is significantly improved.

[0161] S5.5: Perform pulse width modulation signal mapping conversion on the intermediate smooth power control quantity and output it to the heating element drive circuit to generate a smooth power output curve that is finally in a state of suppressed oscillation.

[0162] The target of the pulse width modulation mapping conversion of the intermediate smooth power control value is the intermediate power control value after feedforward compensation attenuation and shielding of feedback disturbances, and the PWM control parameter configuration of the current heating element drive circuit. The intermediate smooth power control value is input to the PWM modulation module, and a linear mapping function from power value to duty cycle is established based on a preset switching frequency and duty cycle range, ensuring that the mapping curve is monotonic and free of saturation distortion across the entire power range. This mapping function is called to convert the intermediate smooth power control value into the corresponding duty cycle value, and a limiting process is performed to ensure that the duty cycle does not exceed the safety boundary. The limited duty cycle value is converted into the initial count value and comparison value of the PWM timer, and the synchronization register is updated to ensure that the PWM waveform is accurately output in the next modulation cycle. Edge synchronization and phase correction are performed on the PWM output waveform to eliminate phase drift caused by asynchronous loading of control commands and improve the stability of power regulation. The synchronously corrected PWM waveform is directly sent to the gate input of the power transistor in the heating element drive circuit, driving the heating element to maintain a controlled smooth power state within the freeze window.

[0163] By converting the intermediate smooth power control value output from the previous step into a duty cycle command executable by the heating element drive circuit through PWM modulation mapping and safety limiting processing, the final smooth power curve is output in a state of suppressed oscillation within the power freeze window.

[0164] Step S6: At the end of the power freeze window, monitor in real time whether the standardized real-time strain time series data has exceeded the peak value and entered the rebound stage. If it is confirmed that the rebound stage has been entered, generate recovery control parameters containing a slight power rebound amount. Specifically, this includes: S6.1: Obtain a standardized real-time strain time series data segment at the moment the power freeze window closes, and use sliding extreme value search to perform local maximum point location processing on the data segment to generate a deformation peak timestamp and the corresponding peak strain amplitude characterizing the thermal expansion of the pot body reaching the limit state.

[0165] S6.2: Based on the deformation peak timestamp, extract the subsequent standardized real-time strain time series data subsequence for a preset duration, and use a first-order backward difference operator to perform point-by-point slope calculation processing on the data subsequence to generate an instantaneous strain rate vector reflecting the cooling and contraction rate of the pot body.

[0166] S6.3: Based on the sign characteristics of the instantaneous strain rate of change vector, execute the springback stage entry judgment logic. If the instantaneous strain rate of change of multiple consecutive sampling points is detected to change from positive to negative and the absolute value exceeds the preset noise threshold, it is confirmed that the pot body has entered the thermal shrinkage springback stage, so as to generate an effective springback stage trigger flag.

[0167] The sign characteristics of the instantaneous strain rate of change vector are determined, and the instantaneous strain rate of change vector generated in the previous sub-step is input into the sign analysis module and positive and negative flag values ​​are extracted point by point to form a sign sequence.

[0168] A continuity detection operation is performed on the symbol sequence, and a preset threshold for the number of sampling points is called to determine whether the symbols of multiple consecutive sampling points change from positive to negative, thus forming a potential rebound stage candidate sequence.

[0169] Amplitude gating logic is applied to the absolute value of the instantaneous strain rate of change corresponding to the candidate sequence. The absolute value is determined by using a noise threshold to determine whether it exceeds a preset noise threshold, thus eliminating sign flipping interference caused by measurement error.

[0170] After the sign transition mode and amplitude determination conditions are met, a confirmation flag is generated to indicate that the pot body has entered the thermal shrinkage and springback stage, and it is encapsulated as a valid springback stage trigger flag.

[0171] By using the above-mentioned sign flipping and amplitude gating joint judgment processing method, the instantaneous strain rate of change vector of the previous step is transformed into a confirmation flag of the rebound stage, realizing the accurate identification of the pot body rebound stage and the triggering condition for subsequent power recovery calculation.

[0172] For example, when the power freezing window of a certain heat preservation cycle is about to end, the instantaneous strain rate of change vector generated by the first-order backward difference of the standardized real-time strain time series data shows a sign change from positive to negative at 5 consecutive sampling points. The sampling frequency is 20ms / point, corresponding to a time of 100ms, and the absolute value of each sampling point exceeds the preset noise threshold of 0.002μm / μm. The rate of change vector is input into the sign analysis module to obtain the sign sequence [+,+,+,-,-,-,-,-,-,-]. The continuity detection module confirms that the sign change mode meets the positive to negative judgment condition. After amplitude gating operation, all negative values ​​meet the threshold requirements, and finally a valid rebound stage trigger flag is generated. In this scenario, the matching of the sign flip mode and the amplitude condition ensures that the recognition result is free from measurement noise interference, providing a reliable trigger condition for the linear mapping calculation of the micro-amplitude power recovery reference value in S6.4. The application effect is that the recognition delay of the rebound stage in the temperature recovery response chain is significantly shortened, and the control accuracy is significantly improved.

[0173] S6.4: In response to the rebound stage trigger flag, extract the instantaneous strain rate of change value at the current moment, input it into a preset linear mapping function for proportional conversion processing, so as to generate a reference value for the slight power recovery amount to offset the temperature lag caused by residual heat release.

[0174] Under the condition that the power freeze window has ended and the effective rebound stage trigger flag has been generated, the numerical component of the instantaneous strain rate of change vector output by the previous sub-step S6.3 at the current time is called as the input parameter.

[0175] The instantaneous strain rate of change is input into the linear mapping function module pre-installed in the controller firmware. This module contains a set of proportional coefficient parameters based on the thermal properties of the pot material and the historical rebound rate-power recovery calibration curve.

[0176] Before performing proportional conversion calculations, sign determination and amplitude limitation processing are performed on the instantaneous strain rate of change values. Only data points that are confirmed as negative contraction rates and whose amplitudes are higher than the noise threshold are retained to ensure the validity and safety of the input.

[0177] Based on the above effective instantaneous strain rate of change, the calculation is performed using the proportional conversion formula defined in MathML, as follows: Where P is the baseline value of the micro-power recovery, k is the proportional coefficient obtained from the material and strain characteristics, and Δε is the current instantaneous strain change rate.

[0178] The P-value obtained by the proportional conversion calculation is subjected to a safety limit check to prevent excessive power rebound under extreme rebound rates.

[0179] The baseline value of the power recovery after the limit is output to the recovery control parameter generation mechanism in this step, serving as the core numerical basis for offsetting the temperature lag caused by residual heat release.

[0180] By using a proportional conversion process, the instantaneous strain rate of change value from the previous step is transformed into a baseline value for a small power recovery that has a clear physical meaning and is adapted to the thermal shrinkage characteristics of the pot body, thereby achieving the ability to actively compensate for the lag in temperature decrease.

[0181] For example, during the heat preservation cycle of a 304 stainless steel electric cooker, the instantaneous strain change rate recorded at the moment the power freeze window ended was -0.0025 mm / mm·s. After verification of the sign and amplitude, this was confirmed as a valid input. The controller invoked the preset proportional coefficient k = 1800 W·mm / mm·s, and calculated P = -0.0025 × 1800 = -4.5W according to the formula. Taking the absolute value and limiting it to within 5 W, the final output power recovery baseline value was 4.5 W. In the subsequent recovery control stage, this value was superimposed on the base power setting, which significantly improved temperature stability during the rebound stage, effectively suppressing the rate of temperature drop. During the verification process, the actual temperature fluctuation amplitude was reduced to within 0.15 ℃, and the user-perceived heat preservation taste became significantly more stable.

[0182] S6.5: Integrate the reference value of the micro-power recovery amount with the current system base power setting value, and perform parameter superposition and fusion operation to generate recovery control parameters containing dynamic compensation information, which serve as the initial input command for conventional proportional-integral-derivative closed-loop control after the power freeze is lifted.

[0183] The input consists of the baseline value of the slight power recovery generated through step S6.4 and the current system base power setting value in the controller, both of which have unified units and dimensions.

[0184] The baseline value of the slight power recovery is input to the parameter fusion calculation module. This module uses vectorized superposition to add the two power parameters component by component, so as to inject dynamic compensation components while maintaining the basic power level.

[0185] During the superposition calculation, the baseline value of the micro-power recovery is scaled proportionally according to the thermal inertia coefficient corresponding to the pot material to ensure that the compensation amount matches the base power in terms of amplitude and response speed.

[0186] The proportionally scaled baseline value of the slight power recovery and the base power setting value are fused and summed at the same timestamp. The fusion summation formula is as follows: Where P is the recovery control parameter, P base Base power setting, P comp This is a scaled-corrected baseline value for the slight power recovery.

[0187] The fused recovery control parameters are bound to the power freeze release signal to generate a recovery control parameter data package containing dynamic compensation information, and marked as the initial input command for conventional proportional-integral-derivative closed-loop control.

[0188] Through the above-mentioned fusion processing method, the result of the previous step is transformed into power setting data with real-time compensation function, so as to achieve a smooth transition of closed-loop control after unfreezing.

[0189] For example, during the heat preservation cycle of an electric cooker, the baseline value of the slight power recovery generated in step S6.4 is 0.35W, the base power setting value is 15.00W, and the corresponding material thermal inertia coefficient is 0.9. After scaling, P comp The value is 0.315W. The superposition and fusion formula is as follows: P=15.00+0.315, resulting in a recovery control parameter of 15.315W. This parameter, along with the unfreezing signal, is synchronously input into a conventional PID control module. In the early stage of the rebound phase, the rate of temperature decrease is significantly slowed down, the fluctuation range of the insulation temperature is greatly reduced, the output power curve shows a smooth transition, and ultimately, user feedback shows a significant reduction in perceived temperature difference and improved stability of the insulation experience.

[0190] Step S7: Based on the positive correlation between the slight power recovery and the current rebound rate, the conventional proportional-integral-derivative closed-loop control, which is reintroduced after the power freeze is lifted, is superimposed and corrected to generate a compensated heating power setpoint that offsets the temperature lag caused by the release of residual heat from the pot. Specifically, this includes: S7.1: Obtain standardized real-time strain time series data at the end of the power freeze window, and perform second-order difference operation on the standardized real-time strain time series data to extract the instantaneous rebound rate vector characterizing the thermal contraction acceleration of the pot body.

[0191] S7.2: Based on the instantaneous rebound rate vector, call the pre-stored linear mapping coefficient table to perform weighted scaling processing to generate a reference value for the slight power recovery amount used to offset the residual heat release effect.

[0192] Based on the input condition of the instantaneous rebound rate vector, the linear mapping coefficient table pre-stored in the controller's non-volatile memory is used as the parameter data source. A search and matching process is performed according to the material type, thermal inertia level, and historical compensation effect tags to determine the mapping coefficient set under the current operating condition. Point-by-point weighting is performed on each component of the instantaneous rebound rate vector. The mapping coefficient set is multiplied by the rate component according to the material type index to form a weighted rate vector. The weighted rate vector is then summed and divided by the number of components to generate a weighted average rate value. This process can be represented as: in, This is the weighted average rate value. For mapping coefficients, For the i-th component of the velocity vector, The total number of components is used. The weighted average rate value is input into a proportional conversion function optimized based on the heat capacity and heat dissipation curve of the pot material to calculate the baseline value of the slight power recovery. A limiting judgment is performed on the calculation result. If it exceeds the preset safety boundary, it is adjusted to the boundary value to form the final baseline value output data of the slight power recovery. Through weighted scaling and proportional conversion, a stable conversion from the instantaneous rebound rate vector to the baseline value of power recovery is achieved, ensuring that the compensation signal is targeted and controllable for the temperature drop caused by residual heat release.

[0193] S7.3: Receive the actual temperature deviation signal fed back by the temperature sensor at the current moment, and use the conventional proportional-integral-derivative control algorithm to perform closed-loop adjustment calculation on the actual temperature deviation signal to generate a basic heating power adjustment command.

[0194] It receives the current temperature signal output by the temperature sensor and converts it into a digital temperature sample value through an analog-to-digital converter interface for use in subsequent control calculations.

[0195] The difference between the digital temperature sample value and the current insulation target temperature setting is calculated to generate an actual temperature deviation signal, which serves as the core input parameter for closed-loop regulation.

[0196] The temperature deviation signal is processed by calling the preset proportional, integral, and derivative coefficients, and each operation is executed sequentially according to the proportional-integral-derivative control algorithm. The proportional term is directly multiplied by the deviation signal, the integral term accumulates the sum of historical deviations and multiplies it by the integral coefficient, and the derivative term calculates the current deviation change rate and multiplies it by the derivative coefficient.

[0197] The results of the proportional, integral, and differential terms are algebraically superimposed to generate an unlimited initial heating power adjustment command, which is used to characterize the basic power correction amount based on temperature deviation.

[0198] The initial heating power adjustment command is subjected to upper and lower limit safety threshold verification. Command values ​​that exceed the safety range are forcibly truncated to within the safety boundary to ensure that the output command will not cause overload or failure under any abnormal conditions.

[0199] Through the above processing method, the temperature deviation signal from the previous step is converted into a basic heating power adjustment command based on conventional proportional-integral-derivative closed-loop calculation with safety constraints, thereby achieving precise conventional correction control of the pot body temperature.

[0200] For example, during the heat preservation stage of a certain model of electric cooker, the temperature sensor detects a current temperature of 82.4℃, and the target heat preservation temperature is 85.0℃. The difference is calculated to yield a temperature deviation of -2.6℃. The proportional coefficient is set to 0.5, the integral coefficient to 0.1, and the derivative coefficient to 0.05, with a sampling period of 0.02 seconds. The proportional term calculates to -1.3W, the integral term accumulates a historical deviation of -15.0℃·second, corresponding to a power correction of -1.5W, and the derivative term shows a deviation change rate of -0.2℃ / second, corresponding to a power correction of -0.01W. The sum of these three terms yields an initial heating power adjustment command of -2.81W. After upper and lower limit safety threshold checks, with a safety range set to -5W to +5W, the command value does not exceed this range and is therefore directly output as the basic heating power adjustment command. In the test, after the instruction was superimposed on the current power setting, the temperature response speed of the pot was significantly improved. The temperature rose back to the target value within 10 seconds without overshoot, which verified the control effect of this step in terms of closed-loop stability and safety.

[0201] S7.4: Perform algebraic superposition and fusion processing on the reference value of the micro-power recovery amount and the basic heating power adjustment command to generate a composite heating power control signal that includes thermal inertia feedforward compensation components.

[0202] This sub-step receives the micro-amplitude power recovery reference value and the basic heating power adjustment command generated from the previous sub-step as input signal sources. For the micro-amplitude power recovery reference value, the controller's built-in parameter scaling module is invoked to perform amplitude correction calculations, ensuring its units are consistent with the basic heating power adjustment command for algebraic operations. A one-to-one data mapping relationship is established between the corrected micro-amplitude power recovery reference value and the basic heating power adjustment command, ensuring consistent sampling timestamps for each component during fusion and avoiding timing misalignments. These two values ​​are input to the composite power control signal generation module, which invokes the algebraic superposition processing unit to perform the fusion calculation. Numerical accuracy optimization processing is performed on the fused composite heating power control signal, employing floating-point truncation and error correction strategies to eliminate accumulated computational errors during the fusion process. The optimized composite heating power control signal is encapsulated into a command data packet, and the logic segment containing the thermal inertia feedforward compensation component is marked to maintain compensation effectiveness during subsequent amplitude limiting and safety constraint processing. Through algebraic superposition and fusion processing, the reference value of the micro-power recovery amount and the basic heating power adjustment command of the previous step are transformed into a composite heating power control signal containing thermal inertia feedforward compensation components, thereby realizing real-time compensation of the residual heat release effect of the pot body and improving temperature stability.

[0203] For example, in the heat preservation control scenario of an electric cooker with a 304 stainless steel pot, the baseline value for the slight power recovery is 0.35 W, which remains 0.35 W after scaling correction. This value, aligned with the basic heating power adjustment command of 25.40 W via timestamp, is input to the algebraic superposition processing unit. The fusion calculation formula is: P = 25.40 + 0.35, resulting in a composite heating power control signal of 25.75 W. This control signal is truncated to three decimal places via floating-point processing and encapsulated into an instruction data packet before being transmitted to the execution module. This setting significantly improves the temperature drop damping in the latter half of the heat preservation cycle, reducing the temperature lag caused by the release of residual heat from the pot, thereby achieving a significant reduction in temperature fluctuation amplitude in actual testing.

[0204] S7.5: Perform amplitude limiting verification and safety boundary constraint processing on the composite heating power control signal to generate the final output compensated heating power setpoint and send it to the heating execution unit.

[0205] A limiting verification operation is performed on the composite heating power control signal, which includes a thermal inertia feedforward compensation component. The controller's preset safe power upper and lower limit values ​​are called, and the input signal is compared with these limits. If the limit is exceeded, the signal is truncated to the corresponding critical value range to prevent overload operation. Based on the power setpoint after limiting, the safety boundary constraint module is invoked to compare the power change rate with the allowable thermal shock threshold of the pot material in real time. The slope of the power change curve is smoothed using a constraint function to avoid structural fatigue caused by sudden large power adjustments. After limiting and safety boundary constraints are completed, the corrected power control signal is input to the pulse width modulation mapping unit. The mapping function converts the power setpoint into a PWM duty cycle, ensuring that the drive is within the acceptable electrical parameter range of the heating element. Synchronization detection is performed on the generated PWM duty cycle signal, and phase locking is performed with the current system clock signal to avoid control interference caused by conflicts between the drive signal and the sampling timing. The power control signal, after limiting verification, boundary constraints, PWM mapping, and timing locking, is output to the drive circuit of the heating execution unit to achieve safe and stable compensated heating power output.

[0206] By limiting the amplitude and constraining the safety boundary, the composite heating power control signal from the previous step is transformed into a final compensated heating power setpoint that meets the safe operating conditions and has smooth dynamic characteristics, thus achieving the dual technical effects of preventing overshoot and preventing power surges.

[0207] Step S8: Archive the deformation response curve, actual temperature fluctuation trajectory, and user satisfaction score data for this insulation cycle; update the risk threshold and compensation parameters in the deformation characteristic and overshoot risk level mapping table to generate an optimized dynamic threshold benchmark for the next control cycle. Specifically, this includes: S8.1: Obtain the archived deformation response curves, actual temperature fluctuation trajectories, and user satisfaction rating data during this insulation cycle, and perform time axis alignment and outlier removal processing on the multi-source heterogeneous data to generate a standardized periodic feature dataset.

[0208] S8.2: Based on the standardized periodic feature dataset, the deviation between the actual temperature overshoot amplitude and the pre-stored risk level judgment result is calculated using the error backpropagation mechanism, so as to generate a model correction gradient vector that characterizes the prediction accuracy of the current mapping table.

[0209] S8.3: Using the model to correct the gradient vector and combining it with the weight coefficients of the user satisfaction score data, perform adaptive weighted updates to iteratively correct the rising slope threshold, peak lag threshold, and rebound decay time threshold in the deformation feature and overshoot risk level mapping table, so as to generate an updated set of static risk thresholds.

[0210] S8.4: Based on the correlation between the updated static risk threshold set and the current thermal characteristic parameters of the pot body material, dynamically adjust the compensation parameter values ​​of the preset gradient coefficient and the power freeze window duration to generate a dynamic compensation parameter set adapted to the current working conditions.

[0211] S8.5: The updated static risk threshold set and the dynamic compensation parameter set are fused and encapsulated, and the old parameter configuration in the original deformation feature and overshoot risk level mapping table is replaced to generate an optimized dynamic threshold benchmark and store it in the controller's non-volatile memory for the next control cycle to call.

[0212] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0213] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the elements or objects preceding “comprising” or “including” encompass the elements or objects listed following “comprising” or “including” and their equivalents, and do not exclude other elements or objects. The “multiple” mentioned in the embodiments of this application refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0214] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for suppressing temperature fluctuations in an electric cooker, specifically comprising: S1: Acquire the original voltage signals of four sets of flexible piezoresistive strain sensing units evenly distributed along the circumference at the bottom of the metal pot body of the electric cooker, and denoise and normalize the original voltage signals to generate standardized real-time strain time series data. S2: Calculate the current strain change rate based on real-time strain time series data, determine the synchronization of signals from adjacent sensing units, and generate an effective deformation feature vector characterizing the thermal expansion start-up state. S3: Construct a mapping table between deformation features and overshoot risk levels based on the boiler body thermal deformation feature library and generate a dynamic threshold benchmark; S4: Input the effective deformation feature vector into the deformation feature and overshoot risk level mapping table for matching and comparison. If the strain change rate of multiple adjacent sensing units exceeds the dynamic threshold benchmark in multiple consecutive sampling periods and the duration of each exceeding the threshold meets the minimum judgment window requirement, it is judged as a high-risk deformation mode, and a control command signal that triggers feedforward compensation action is generated. S5: Responding to the control command signal, the current heating power is reduced according to the preset gradient coefficient, and the power freeze window is activated simultaneously to prevent the intervention of conventional proportional-integral-derivative feedback regulation, thereby generating a smooth power output curve in a state of suppressed oscillation. S6: At the end of the power freeze window, monitor whether the real-time strain timing data has exceeded the peak value and entered the rebound stage. If it is confirmed that the rebound stage has been entered, generate recovery control parameters that include a slight power rebound amount. S7: Generates a compensated heating power setting value based on the slight power recovery amount and the current rebound rate.

2. The method of claim 1, wherein the temperature fluctuation is suppressed by controlling the power supplied to the heating element. The boiler body thermal deformation feature library includes a material category index, a heat preservation target temperature index, thermal cycling process data records, and feature annotation fields.

3. The method of claim 1, wherein the temperature fluctuation is suppressed by controlling the power supplied to the heating element. The process of constructing a mapping table between deformation features and overshoot risk levels based on the boiler body thermal deformation feature library and generating a dynamic threshold benchmark involves, specifically, constructing a mapping table between deformation features and overshoot risk levels that includes the rise slope, peak hysteresis period, and rebound decay time of boiler bodies of different materials under the target insulation temperature, based on the measured data of the typical thermal expansion, cooling, and contraction cycle process of boiler bodies of different materials under the target insulation temperature, and generating a dynamic threshold benchmark.

4. The method of claim 1, wherein the temperature fluctuation is suppressed by controlling the power supplied to the heating element. The process of generating a compensated heating power setpoint based on the slight power recovery and the current rebound rate involves superimposing and correcting the conventional proportional-integral-derivative closed-loop control that is re-engaged after the power freeze is lifted, based on the positive correlation between the slight power recovery and the current rebound rate, to generate a compensated heating power setpoint that offsets the temperature lag caused by the release of residual heat from the pot.

5. The method of claim 1, wherein the temperature fluctuation is suppressed by controlling the power supplied to the heating element. Following S7, the following also includes: S8: Archive the deformation response curve, actual temperature fluctuation trajectory, and user satisfaction score data during this insulation cycle, and update the risk threshold and compensation parameters in the deformation characteristics and overshoot risk level mapping table to generate an optimized dynamic threshold benchmark for the next control cycle.

6. The method of claim 1, wherein the temperature fluctuation is suppressed by controlling the power supplied to the heating element. S3 specifically includes: Obtain the original measured dataset of thermal expansion, cooling and contraction cycle process of pots of different materials at a specific heat preservation target temperature from the pre-stored pot body thermal deformation feature library. Remove abnormal noise and missing segments in the original measured dataset and perform interpolation repair to generate standardized multi-material pot body thermal deformation time series benchmark data. A set of three-dimensional deformation feature vectors is generated based on the time-series benchmark data of thermal deformation of multi-material pot bodies; By calling up temperature fluctuation trajectory records from historical experiments, each set of deformation feature data in the three-dimensional deformation feature vector set is spatiotemporally aligned and correlated with the corresponding actual temperature overshoot amplitude data. The contribution weight of each deformation feature component to the temperature overshoot amplitude is calculated using statistical correlation analysis to generate a weighted correlation matrix characterizing the correlation strength between deformation features and overshoot risk. Based on the contribution weight distribution pattern in the weighted correlation matrix, a multi-level risk classification threshold range is set, and the data points in the three-dimensional deformation feature vector set are mapped to three overshoot risk level categories of low risk, medium risk and high risk according to their comprehensive risk scores, so as to generate a preliminary deformation feature and overshoot risk level classification mapping dataset. The preliminary deformation feature and overshoot risk level classification mapping dataset is subjected to boundary smoothing and logical consistency verification. A structured lookup table containing the rising slope threshold range, peak lag period judgment interval and rebound decay time critical value is constructed to generate the final dynamic threshold benchmark for real-time risk judgment, namely the deformation feature and overshoot risk level mapping relationship table.

7. The method of claim 6, wherein the temperature fluctuation is suppressed by controlling the power supplied to the heating element. The process of generating a three-dimensional deformation feature vector set based on multi-material pot body thermal deformation time series benchmark data specifically involves using standardized multi-material pot body thermal deformation time series benchmark data, employing sliding window extreme value detection and linear regression fitting techniques to perform feature extraction operations on the strain signal rising segment, peak plateau segment, and rebound falling segment of each complete thermal cycle, in order to generate a three-dimensional deformation feature vector set containing the rising slope value, peak lag time duration, and rebound decay time constant.

8. The method of claim 1, wherein the temperature fluctuation is suppressed to a range of 0.5 to 1.5°C. S4 specifically includes: Based on the mapping table between deformation features and overshoot risk level, multi-dimensional feature extraction is performed on the rising slope value, peak lag duration and rebound decay time parameters in the current input effective deformation feature vector to generate a standard deformation feature dataset to be compared. Generate instantaneous strain rate sequence based on standard deformation feature dataset; The extreme points in the instantaneous strain rate of change sequence are logically compared with the preset dynamic threshold benchmark. If the instantaneous strain rate of change sequence exceeds the dynamic threshold benchmark and the duration meets the minimum judgment window requirement, a preliminary judgment flag for high-risk deformation mode is generated. By combining the spatial consistency verification results of the preliminary judgment flag of high-risk deformation mode and the effective deformation feature vector, a joint logic verification is performed. If it is confirmed that the adjacent sensing units are synchronously triggered and meet the judgment conditions of high-risk deformation mode, the final high-risk deformation mode confirmation signal is generated. In response to the final high-risk deformation mode confirmation signal, the corresponding gradient coefficient configuration parameters in the feedforward compensation strategy library are invoked to generate a control command signal that triggers the feedforward compensation action, including a power reduction magnitude command and a power freeze window start command.

9. The method of claim 8, wherein the temperature fluctuation is suppressed by controlling the power supplied to the heating element. The generation of the instantaneous strain rate of change sequence based on the standard deformation feature dataset specifically involves performing a first-order difference operation on the rising slope values ​​in the standard deformation feature dataset to generate an instantaneous strain rate of change sequence that characterizes the accelerating trend of thermal expansion.

10. The method for suppressing temperature fluctuations in an electric cooker according to claim 1, characterized in that, S5 specifically includes: The system acquires the control command signal that triggers the feedforward compensation action and the current real-time heating power setting value. Based on the thermal inertia gradient coefficient table of the pot body material, it performs linear decay calculation on the current real-time heating power setting value to generate a target reduced heating power value with overshoot suppression capability. After receiving the target heating power value after reduction, the power execution unit is driven to make a step adjustment. At the same time, a power freeze window timer is started based on the risk level indicator in the control command signal to generate a power regulation lock flag that is in an active state. Monitor the activation status of the power regulation lock flag and force the correction amount output by the conventional proportional-integral-derivative feedback regulation module within the freeze time window to zero, so as to generate a shielded invalid feedback regulation signal. The heating power value after the target is reduced is superimposed with the shielded invalid feedback adjustment signal to eliminate the disturbance variables introduced by the feedback loop, so as to generate an intermediate smooth power control quantity dominated only by feedforward compensation. The intermediate smooth power control quantity is mapped and converted using pulse width modulation, and then output to the heating element drive circuit to generate a smooth power output curve that is finally in a state of suppressed oscillation.

Images