Semiconductor ceramic sintering temperature control method and system

Abstract

This invention relates to the field of temperature control technology, and discloses a method and system for controlling the sintering temperature of semiconductor ceramics. The method acquires the furnace temperature, target temperature curve, heating output, heating rate, holding time, and furnace zone temperature difference to form the current furnace process temperature sequence. Combined with historical thermal response sequences of similar furnaces, the method estimates the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state. Based on this, a virtual billet temperature sequence is derived, and the virtual billet temperature is used as feedback to predict the billet temperature change trend in subsequent control cycles. A heating output correction is generated based on furnace temperature constraints, heating rate constraints, and overshoot risk constraints. This scheme transforms temperature control from furnace temperature measurement and following to billet equivalent thermal state control, reducing the risk of insufficient or overcompensated hysteresis compensation caused by inconsistencies between furnace temperature and billet thermal history.

Description

Technical Field

[0001] This invention relates to the field of temperature control technology, and discloses a method and system for controlling the sintering temperature of semiconductor ceramics. Background Technology

[0002] Sintering of semiconductor ceramics typically requires controlling heating, holding, and cooling according to a preset temperature profile. The temperature control results directly affect the densification process of the green body, grain growth, and the consistency of subsequent electrical properties. Existing sintering furnaces mostly use furnace temperature measurements as feedback. The temperature controller compares the deviation between the actual furnace temperature and the target temperature, then adjusts the heating output to ensure the furnace temperature follows the preset profile. In specific control, the control system usually records the furnace temperature, target temperature, heating output, heating rate, and holding time, and executes control according to a fixed heating slope, set holding temperature, and holding time. This method uses whether the furnace temperature reaches the set profile as the criterion, clearly defines the controlled object, has a mature execution process, and is suitable for conventional sintering processes with stable thermal response and minimal changes in loading conditions.

[0003] In semiconductor ceramic sintering scenarios, the green body has a certain thickness, heat capacity, and internal heat transfer path. The temperature measurement point in the furnace reflects the ambient temperature inside the furnace, which is not equivalent to the actual heating state inside the green body. Existing control methods typically do not perform online estimation of the green body's equivalent temperature, thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity, nor do they use historical thermal response sequences of similar furnace runs for temperature inversion in the current furnace run. Even if the furnace temperature operates according to the set curve, the interior of the green body may still be below the target thermal state due to heat absorption hysteresis, or continue to heat up after the control output is reduced due to residual heat release. Especially in the heating stage, the target temperature approach stage, the critical densification stage, and the holding stage, there are deviations between the furnace temperature curve and the green body's thermal history. Relying solely on furnace temperature measurement feedback is insufficient to characterize the true sintering process of the green body.

[0004] Current semiconductor ceramic sintering temperature control uses furnace temperature as the direct feedback object, failing to establish a correspondence between furnace temperature changes, heating output changes, and the equivalent thermal state of the green body. This results in the control system adjusting the furnace ambient temperature rather than the actual thermal state of the green body during sintering. When the furnace temperature approaches or reaches the target temperature, the green body may still be in a state of heat absorption hysteresis; when the heating output decreases, the green body may continue to absorb residual heat from the furnace. This causes a discrepancy between the target temperature curve and the green body's thermal history, making temperature compensation and overshoot suppression in the critical sintering stage lack reliable basis, and consequently causing fluctuations in the green body's densification degree, grain growth state, and electrical performance stability. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for controlling the sintering temperature of semiconductor ceramics, which can solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for controlling the sintering temperature of semiconductor ceramics includes: acquiring the furnace temperature, target temperature curve, heating output, heating rate, holding time and furnace zone temperature difference during the sintering process of semiconductor ceramics, and forming a process temperature sequence corresponding to the current furnace batch; Based on the process temperature sequence and the thermal response sequence of similar historical furnaces, estimate the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state of the current furnace. Based on the thermal inertia coefficient, the heat transfer hysteresis, and the equivalent heat capacity state, the virtual blank temperature sequence of the semiconductor ceramic blank is inverted; Using the virtual billet temperature sequence as feedback, the billet temperature change trend in subsequent control cycles is predicted in conjunction with the target temperature curve. Based on the green body temperature change trend, furnace temperature constraint, heating rate constraint, and overshoot risk constraint, a heating output correction amount is generated, and the heating output of the sintering process is adjusted according to the heating output correction amount.

[0007] Preferably, estimating the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity of the current furnace cycle includes: dividing the process temperature sequence into time windows according to the temperature rise segment, the target temperature approach segment, and the heat preservation segment, and extracting the response slope, response delay, and temperature residual between the furnace temperature change and the heating output change in each time window. The thermal inertia coefficient is obtained by matching the deviation of the response slope with the benchmark response slope in the historical thermal response sequence of similar furnaces. The response delay is aligned and compared with the set change node of the target temperature curve to obtain the heat transfer hysteresis. The temperature residual and the cumulative heating output are combined and fitted to obtain the equivalent heat capacity state.

[0008] Preferably, the virtual blank temperature sequence of the semiconductor ceramic blank is inverted, including: using the furnace temperature of the current control cycle as the external thermal boundary, using the thermal inertia coefficient to limit the response amplitude of heat entering the blank, using the heat transfer hysteresis to limit the time offset of the furnace temperature change to the blank equivalent temperature, and using the equivalent heat capacity state to limit the blank temperature change amplitude under the same heating output. The equivalent heat absorption and temperature increment of the billet are calculated recursively according to the control cycle sequence. The recursively obtained equivalent billet temperature is then continuously corrected with the virtual billet temperature of the previous control cycle to generate a virtual billet temperature sequence.

[0009] Preferably, the generation of heating output correction includes: dividing the target temperature curve into a heating segment, a critical densification segment, a heat preservation segment, and a cooling segment; Within each control cycle, the trend of billet temperature change in subsequent control cycles is predicted based on the deviation between the virtual billet temperature sequence and the target billet temperature. Based on the billet temperature change trend, calculate the furnace temperature constraint, heating rate constraint, overshoot risk constraint, and thermal history deviation constraint respectively; By jointly solving for each constraint term, the heating output correction amount that satisfies the current sintering stage limitation conditions is obtained.

[0010] Preferably, when estimating the thermal inertia coefficient, heat transfer hysteresis and equivalent heat capacity of the current furnace, the method further includes: constructing waste heat response characteristics based on the continuous rise of furnace temperature after the heating output changes from a changing state to a stable state within the time window. Based on the diffusion change of the temperature difference in the furnace area within the same time window, a heat distribution uniformity characteristic is constructed. The waste heat response characteristics, the heat distribution equilibrium characteristics, the response slope, the response delay, and the temperature residual are combined to form a thermal inertia state vector, and the thermal inertia state vector is used to synchronously correct the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state.

[0011] Preferably, when generating the virtual billet temperature sequence, the method further includes: shifting the furnace temperature sequence over time according to the heat transfer hysteresis to obtain a hysteresis thermal boundary sequence; The amplitude attenuation process of the hysteresis thermal boundary sequence is performed based on the thermal inertia coefficient to obtain the equivalent input temperature sequence of the billet. Based on the equivalent heat capacity state, the equivalent input temperature sequence of the billet is subjected to cumulative heat absorption constraint to obtain the initial virtual billet temperature; The initial virtual billet temperature is compared with the virtual billet temperature of the historical control cycle to check for monotonicity and continuity, and temperature points that do not meet the check conditions are corrected by recursive write-back.

[0012] Preferably, when solving the various constraint terms jointly, the method further includes: using virtual billet temperature tracking deviation and heating rate constraint as the main constraints in the heating stage; using furnace zone temperature difference constraint, thermal history deviation constraint and overshoot risk constraint as the main constraints in the critical densification stage; using virtual billet temperature fluctuation amplitude and cumulative heat input deviation as the main constraints in the heat preservation stage; and using virtual billet temperature decrease rate and residual heat release trend as the main constraints in the cooling stage. The solution weights for reconstructing the heating output correction amount are based on the main constraints of the current sintering stage.

[0013] Preferably, when synchronously correcting the thermal inertia state vector, the method further includes: performing similarity matching between the thermal inertia state vector of the current furnace and the thermal inertia state vector corresponding to the qualified furnace in the same historical furnace, and determining the deviation direction of the current furnace relative to the reference thermal response. When the deviation direction indicates an enhanced heat transfer hysteresis in the current furnace, the heat transfer hysteresis is increased and the billet temperature increment corresponding to the same heating output is reduced. When the deviation direction indicates that the heat absorption response of the current furnace is accelerated, the heat transfer hysteresis is reduced and the overshoot risk constraint is strengthened in the solution of the heating output correction.

[0014] Preferably, when reconstructing the solution weights for the heating output correction, the method further includes: calculating the instantaneous deviation, prediction deviation, and cumulative thermal history deviation between the virtual billet temperature and the target billet temperature within the current control cycle; When the instantaneous deviation is in the same direction as the predicted deviation, the heating output correction amount is limited to a continuous compensation type correction; When the instantaneous deviation is opposite to the predicted deviation and the cumulative thermal history deviation exceeds the preset allowable range, the heating output correction amount is limited to a correction that suppresses overcompensation. When the instantaneous deviation, prediction deviation, and cumulative thermal history deviation are all within the allowable range, the heating output change trend of the previous control cycle is maintained.

[0015] A semiconductor ceramic sintering temperature control system includes: a temperature process data acquisition module, used to acquire furnace temperature, target temperature curve, heating output, heating rate, holding time and furnace zone temperature difference during the semiconductor ceramic sintering process, and form a process temperature sequence corresponding to the current furnace. The thermal inertia state estimation module is used to estimate the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state of the current furnace based on the process temperature sequence and the thermal response sequence of similar furnaces in the past. The virtual blank temperature inversion module is used to invert the virtual blank temperature sequence of the semiconductor ceramic blank based on the thermal inertia coefficient, heat transfer hysteresis and equivalent heat capacity state. The predictive control module is used to predict the trend of billet temperature change in subsequent control cycles by using the virtual billet temperature sequence as feedback object and combining it with the target temperature curve. The heating output adjustment module is used to generate a heating output correction amount based on the billet temperature change trend, furnace temperature constraint, heating rate constraint and overshoot risk constraint, and adjust the heating output of the sintering process according to the heating output correction amount.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By acquiring furnace temperature, target temperature curves, heating output, heating rate, holding time, and furnace zone temperature difference, and combining this with historical thermal response sequences of similar furnace cycles, the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state of the current furnace cycle are estimated. Based on this, a virtual billet temperature sequence is derived, expanding the feedback object of temperature control from furnace temperature measurements to the equivalent thermal state of the billet. The virtual billet temperature sequence is used to predict the billet temperature change trend in subsequent control cycles, generating heating output corrections constrained by furnace temperature, heating rate, and overshoot risk. This ensures that the heating output no longer simply follows furnace temperature deviations but is corrected based on billet thermal hysteresis and heat capacity state. This reduces the deviation between furnace temperature and the billet's thermal history, making temperature compensation during heating, holding, and approaching the target temperature stages more closely match the actual heating process of the billet.

[0017] 2. In further implementation, by dividing the time window, extracting the response slope, comparing response delays, fitting temperature residuals, constructing residual heat response characteristics, and constructing heat distribution equilibrium characteristics, the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity can be synchronously corrected, making the virtual billet temperature inversion independent of a single temperature deviation. By dividing the target temperature curve into heating, critical densification, holding, and cooling segments, and solving for the weights of the reconstructed constraint terms for different stages, the heating output correction can match the thermal response characteristics of the sintering stage. Through the joint judgment of instantaneous deviation, prediction deviation, and cumulative thermal history deviation, control states such as continuous compensation, suppression of overcompensation, and maintenance of output trends can be distinguished, reducing temperature fluctuations caused by excessive hysteresis compensation, and improving the continuity of thermal history within a furnace and the consistency of thermal response between different furnaces. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the overall process for controlling the sintering temperature of semiconductor ceramics; Figure 2 Flowchart for estimating the thermal inertia state of the current furnace; Figure 3 Flowchart for virtual billet temperature inversion; Figure 4 This is a flowchart of predictive control and heating output correction based on virtual billet temperature. Detailed Implementation

[0019] In one embodiment, please refer to Figure 1A method for controlling the sintering temperature of semiconductor ceramics includes acquiring the furnace temperature, target temperature curve, heating output, heating rate, holding time, and furnace zone temperature difference during the semiconductor ceramic sintering process to form a process temperature sequence corresponding to the current furnace. Based on the process temperature sequence and the thermal response sequence of similar historical furnaces, the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state of the current furnace are estimated. Based on the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state, a virtual blank temperature sequence of the semiconductor ceramic blank is inverted. Using the virtual blank temperature sequence as a feedback object, the blank temperature change trend in subsequent control cycles is predicted in combination with the target temperature curve. Based on the blank temperature change trend, furnace temperature constraints, heating rate constraints, and overshoot risk constraints, a heating output correction is generated, and the heating output of the sintering process is adjusted according to the heating output correction.

[0020] In this embodiment, the semiconductor ceramic sintering temperature control system includes a temperature process data acquisition module, a thermal inertia state estimation module, a virtual billet temperature inversion module, a predictive control module, and a heating output adjustment module. The temperature process data acquisition module is used to obtain the furnace temperature, target temperature curve, heating output, heating rate, holding time, and furnace zone temperature difference from existing sintering control data and form the process temperature sequence of the current furnace. The thermal inertia state estimation module is used to estimate the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state based on the process temperature sequence and the thermal response sequence of similar historical furnaces. The virtual billet temperature inversion module is used to calculate the virtual billet temperature sequence. The predictive control module is used to predict the billet temperature change trend in subsequent control cycles. The heating output adjustment module is used to generate heating output correction and adjust the heating output of the sintering process.

[0021] In this embodiment, the process temperature sequence is recorded according to the control cycle. Each control cycle corresponds to a process data group. The process data group includes the furnace temperature sampling value, the target value of the target temperature curve in the same control cycle, the heating output, the heating rate obtained from the furnace temperature difference between adjacent control cycles, the accumulated holding time, and the furnace temperature difference formed between multiple furnace zones. The thermal response sequence of the same type of historical furnace is selected from historical process data that matches the current semiconductor ceramic material, sintering curve type, and loading state. The thermal response sequence is not directly used as the control curve to replace the current curve, but is used as the benchmark sequence for the thermal inertia estimation of the current furnace, so that the current furnace can still identify the heat absorption hysteresis and residual heat release state of the billet when the furnace temperature measurement value normally follows the target temperature.

[0022] In this embodiment, the furnace temperature is used as external thermal environment data, the heating output is used as control execution data, the heating rate is used as furnace thermal response change data, the holding time is used as thermal history data, and the furnace zone temperature difference is used as furnace internal heat distribution deviation data. All types of data are unified to the same time base within the same control cycle. If there is an instantaneous sampling disturbance within a certain control cycle, the median or local mean of the adjacent control cycle is used for replacement. The replacement process retains the original sampling mark and participates in subsequent confidence judgment to avoid abnormal sampling points directly entering the virtual billet temperature inversion and causing temperature state jumps.

[0023]

[0024] ; Where w represents the time window and i represents the control cycle number. This represents the heating output during the i-th control cycle. This represents the average heating output within the time window w. This represents the furnace temperature during the i-th control cycle. This represents the average temperature difference in the furnace within the time window w. This represents a stable term to prevent the denominator from being zero. This represents the furnace temperature response slope corresponding to a unit heating output disturbance. This represents the target value of the target temperature curve in the i-th control cycle. Let d represent the set of allowed delays for comparison, and let d represent the number of candidate delay cycles. This represents the response delay that minimizes the sum of squared deviations between the furnace temperature sequence and the target temperature sequence. For positive and If it is still close to zero, then A low value indicates that the heat has not yet formed a synchronized temperature rise. If the sum of squares decreases after increasing the candidate d, then... A larger value indicates that there is a strong heat transfer lag in the current furnace batch.

[0025] In this embodiment, the thermal inertia state of the current furnace can be initially estimated based on the response slope and response delay. Specifically, the process temperature sequence of the current furnace is compared in phase with the thermal response sequence of similar furnaces in the past. The comparison objects include the temperature residual when the target temperature is close, the amount of furnace temperature continuing to rise after the heating output changes, the output required to maintain the target temperature during the heat preservation stage, and the convergence trend of the furnace temperature difference. When the current furnace requires a higher heating output to obtain the same furnace temperature rise response in the similar target temperature range, the thermal inertia coefficient is increased. When the current furnace still maintains an upward trend after reducing the heating output, the heat transfer hysteresis is increased. When there is a continuous deviation between the temperature residual and the cumulative heating output, the equivalent heat capacity state is increased.

[0026] In this embodiment, the virtual billet temperature sequence is used to characterize the equivalent heating state of the semiconductor ceramic billet under the current control cycle. During the calculation, the furnace temperature is not directly identified as the billet temperature. Instead, it is recursively calculated based on the response amplitude of the furnace temperature change to the billet, the time offset, and the heat capacity constraint. The starting point of the recursion can be the initial temperature at which the billet and the furnace reach equilibrium at the start of sintering. In each subsequent control cycle, a new equivalent billet temperature is calculated based on the virtual billet temperature of the previous control cycle, the furnace temperature after hysteresis processing, the current heating output, and the equivalent heat capacity state. If the recursion result forms a non-physical jump with the adjacent control cycle, the correction amplitude is jointly limited by the thermal inertia coefficient and the heating rate constraint.

[0027]

[0028] ; Where k represents the current control period, This represents the thermal inertia state quantity of the current furnace in the kth control cycle. and These represent the lower and upper limits of the thermal inertia state quantity, respectively. This represents the baseline response slope corresponding to the qualified furnaces among similar furnace batches in history. This indicates the residual heat response characteristic formed by the continued rise in furnace temperature after the heating output transitions from a changing state to a steady state. This represents the uniform heat distribution characteristics formed by the diffusion of temperature differences in the furnace area, a, b, c, This represents the weight coefficient of the corresponding feature. Indicates the baseline equivalent heat capacity. This indicates the equivalent heat capacity state during the current control cycle. Less than and , If both increase, then Move towards the upper limit and make As the temperature increases, the virtual billet temperature increment corresponding to the same heating output in the control calculation decreases accordingly.

[0029] In this embodiment, please refer to Figure 3 The virtual billet temperature sequence is generated in a recursive manner. The heat transfer lag given by the thermal inertia state estimation module determines the reading position of the furnace temperature sequence. The equivalent heat capacity state determines the magnitude of the change in billet temperature from the furnace thermal boundary. The thermal inertia coefficient determines the response intensity of the external thermal boundary entering the billet. The heating output is used as the heat input compensation quantity to participate in the same cycle calculation. The recursive result is then continuously compared with the virtual billet temperature of the previous control cycle to ensure that the virtual billet temperature will not change inconsistently with the sintering heat process due to single-point disturbance of furnace temperature.

[0030] ; in, This represents the virtual billet temperature in the k-th control cycle. This indicates the virtual billet temperature of the previous control cycle. This represents the heat exchange intensity calculated from the thermal inertia coefficient. Indicates the time interval between adjacent control cycles. Indicates the equivalent heat capacity state. Indicated by heat transfer hysteresis The furnace temperature after relocation This represents the input coefficient that influences the heating output on the equivalent heat absorption of the billet. This represents the heating output in the k-th control cycle. Greater than Then the parenthesized item is positive and the result is positive. Rise, if near and If the temperature is lowered, the recursive result is mainly maintained by the existing thermal state of the billet. Increasing the temperature will compress the temperature increment generated by the same temperature difference and the same heating output.

[0031] Furthermore, after generating the virtual billet temperature sequence, the virtual billet temperature inversion module reconstructs the correspondence between the virtual billet temperature sequence and the target temperature curve. The target temperature curve is not directly used as the furnace temperature follow-up target, but is converted into a target billet temperature sequence. In the heating stage, the target billet temperature is allowed to have a controlled lag relative to the furnace target temperature. In the target temperature approach stage, the difference between the target billet temperature and the furnace target temperature gradually converges. In the heat preservation stage, the target billet temperature is judged based on the stability of thermal history. In the cooling stage, the target billet temperature is judged based on the residual heat release trend.

[0032] In this embodiment, the predictive control module uses the virtual billet temperature sequence as the feedback object, and predicts the billet temperature change trend in subsequent control cycles based on the virtual billet temperature, target billet temperature, thermal inertia state quantity, furnace zone temperature difference, heating rate, and heating output change trend in the current control cycle. The prediction process adopts a rolling time domain method. In the current control cycle, only the heating output correction amount corresponding to the most recent cycle is executed. In subsequent control cycles, the calculation is recalculated after the next acquisition of the process temperature sequence, so as to avoid the prediction result being fixed for a long time and unable to respond to changes in thermal inertia state.

[0033] ; in, Let represent the rolling control objective function for the k-th control cycle, N represent the number of prediction cycles, j represent the relative sequence number within the prediction cycle, and s represent the current sintering stage identifier. This indicates the weight of the virtual billet temperature following deviation. This indicates the constraint weight for changes in heating output. Indicates the overshoot risk constraint weight. This indicates the cumulative thermal history deviation constraint weight. This represents the overshoot risk during the (k+j)th prediction period. This represents the cumulative thermal history deviation within the (k+j)th prediction period. If the current period is in the critical densification stage, then... and Taking a higher weight allows overshoot risk and thermal history deviation to become the dominant constraints, if the current phase is in the heating phase. The compensation direction is jointly limited by constraints related to the heating rate.

[0034] In this embodiment, please refer to Figure 4 The heating output adjustment module generates a heating output correction amount based on the solution of the objective function. The solution process simultaneously satisfies the furnace temperature constraint, the heating rate constraint, and the overshoot risk constraint. The furnace temperature constraint is used to limit the actual furnace temperature from deviating from the temperature control range allowed by the sintering equipment. The heating rate constraint is used to limit the temperature rise change in adjacent control cycles from exceeding the direction of change allowed by the process curve. The overshoot risk constraint is used to determine whether the predicted virtual billet temperature will exceed the target billet temperature and cause excessive heat history. Before applying the heating output correction amount, it is also compared with the heating output change trend of the previous control cycle to avoid unfounded reverse oscillations in the output between adjacent control cycles.

[0035] Furthermore, in the time window division, the process temperature sequence is divided into a temperature rise segment, a target temperature approach segment, and a heat preservation segment. The temperature rise segment is used to extract the response slope of the furnace temperature change caused by the change in heating output. The target temperature approach segment is used to extract the response delay of the furnace temperature relative to the target temperature curve. The heat preservation segment is used to extract the relationship between the output required to maintain the furnace temperature and the temperature residual. Some overlap between the time windows is allowed so that the thermal inertia state estimation does not break down because the time window boundary is exactly at the temperature inflection point.

[0036] In a preferred embodiment, the extraction process of response slope, response delay, and temperature residual is matched with the thermal response sequence of historical similar furnaces. The benchmark response slope in historical similar furnaces comes from the process temperature sequence of qualified furnaces. The deviation between the response slope and the benchmark response slope is used to calculate the thermal inertia coefficient. The response delay is aligned and compared with the set change node of the target temperature curve to calculate the heat transfer hysteresis. The temperature residual and the cumulative heating output are jointly fitted to calculate the equivalent heat capacity state. The positive and negative directions of the residual are retained during the joint fitting. A positive residual indicates that the furnace temperature is higher than the target but the thermal state of the billet may still be lagging. A negative residual indicates that the furnace temperature is lower than the target and it is necessary to distinguish between insufficient heating and increased heat capacity.

[0037] In this embodiment, please refer to Figure 2 When jointly fitting the temperature residual and the cumulative heating output, correlation judgment within a sliding window is used instead of single-point residual judgment. When the cumulative heating output increases but the furnace temperature residual does not converge for a long time, the equivalent heat capacity state is increased. When the cumulative heating output decreases but the furnace temperature continues to rise, the waste heat response characteristic is increased. When the furnace temperature difference expands within the same window, the heat distribution equilibrium characteristic is increased. The above characteristics are entered into the thermal inertia state vector, so that the thermal inertia coefficient, heat transfer hysteresis and equivalent heat capacity state can be synchronously corrected within the same control cycle.

[0038] Furthermore, when generating the virtual billet temperature sequence, the furnace temperature sequence is time-shifted based on the heat transfer hysteresis to obtain the hysteresis thermal boundary sequence. The amplitude of the hysteresis thermal boundary sequence is attenuated based on the thermal inertia coefficient to obtain the equivalent input temperature sequence of the billet. The equivalent input temperature sequence of the billet is constrained by cumulative heat absorption based on the equivalent heat capacity state to obtain the initial virtual billet temperature. The initial virtual billet temperature is compared with the virtual billet temperature of the historical control cycle for monotonicity and continuity. Temperature points that do not meet the test conditions are recursively written back for correction.

[0039] In this embodiment, the time shift processing determines the delay period for the furnace temperature to enter the equivalent calculation of the billet by using the heat transfer hysteresis. When the heat transfer hysteresis is zero, the furnace temperature of the current control cycle is directly used as the thermal boundary input. When the heat transfer hysteresis is positive, the furnace temperature of the previous few control cycles is read as the thermal boundary input. The amplitude attenuation processing scales the difference between the hysteresis thermal boundary and the virtual billet temperature of the previous control cycle according to the thermal inertia coefficient. The cumulative heat absorption constraint limits the change amplitude of the billet temperature in each control cycle according to the equivalent heat capacity state. The monotonicity test is used to determine whether the virtual billet temperature shows an abnormal decrease opposite to the direction of heat input during the heating stage. The continuity test is used to determine whether the change of virtual billet temperature in adjacent control cycles exceeds the allowable range of the heating rate constraint.

[0040] In a preferred embodiment, the recursive writeback correction searches for the nearest reliable temperature point backward from the abnormal temperature point, and re-executes the hysteresis thermal boundary reading, amplitude decay, and cumulative heat absorption constraint calculation from the reliable temperature point until the abnormal temperature point meets the monotonicity and continuity test. If there is a sudden change in furnace temperature or heating output within the writeback interval, the overshoot risk flag of the corresponding control cycle is increased, and the overshoot risk flag is converted into a constraint on the heating output correction amount in subsequent predictive control.

[0041] Furthermore, the target temperature curve is divided into a heating segment, a critical densification segment, a holding segment, and a cooling segment. Different combinations of constraints are used for each sintering stage. In the heating segment, the virtual billet temperature tracking deviation and heating rate constraint are the main constraints. In the critical densification segment, the furnace zone temperature difference constraint, thermal history deviation constraint, and overshoot risk constraint are the main constraints. In the holding segment, the virtual billet temperature fluctuation amplitude and cumulative heat input deviation are the main constraints. In the cooling segment, the virtual billet temperature decrease rate and residual heat release trend are the main constraints. The predictive control module reconstructs the solution weights of the heating output correction based on the current sintering stage.

[0042] In this embodiment, the control objective of the heating stage is not to make the furnace temperature quickly approach the target temperature, but to make the virtual billet temperature stably approach the target billet temperature along the direction of the target billet temperature. If the virtual billet temperature lags behind but the heating rate constraint still allows for compensation, the heating output correction amount increases. If the furnace temperature is close to the target temperature but the virtual billet temperature is still lagging behind, the heating output correction amount maintains the heat input through limited compensation. If the heating rate is close to the constraint boundary, the predictive control module reduces the compensation magnitude and distributes the compensation in subsequent control cycles.

[0043] In this embodiment, the control objective of the critical densification stage is to jointly limit the thermal history deviation, furnace temperature difference, and overshoot risk. If the virtual billet temperature is lower than the target billet temperature but the furnace temperature difference increases, the heating output correction amount will not increase directly according to the temperature deviation, but will compress the compensation range according to the furnace temperature difference constraint. If the predicted virtual billet temperature will exceed the target billet temperature in subsequent control cycles, the overshoot risk constraint will take the lead. If the cumulative thermal history deviation is still insufficient, the heating output correction amount will be maintained slightly within the overshoot risk constraint range to avoid premature weakening of heat input in the critical densification stage due to instantaneous furnace temperature reaching the target.

[0044] In this embodiment, the control objective of the heat preservation section revolves around the fluctuation range of the virtual billet temperature and the cumulative heat input deviation. If the furnace temperature is near the target temperature but the virtual billet temperature still lags behind, the control calculation continues to retain the compensation amount. If the virtual billet temperature has met the target billet temperature and the cumulative heat input deviation is close to zero, the heating output correction amount tends to be flat. If the furnace temperature continues to rise after the heating output decreases and the predicted value of the virtual billet temperature continues to increase, the residual heat release trend enters the constraint of the heat reduction input direction. The cooling section then limits the thermal stress induction conditions during the cooling process based on the rate of decrease of the virtual billet temperature.

[0045]

[0046] ; in, This represents the cumulative thermal history deviation during the k-th control cycle. This indicates the lower limit temperature for participation in thermal history calculations. This indicates taking the larger value between the number inside the parentheses and zero. This indicates the heating output correction amount. Indicates instantaneous deviation, Indicates prediction bias. Indicates the amount of overshoot risk. , , , The solution weights represent the changes in the sintering stage s. Below Then the corresponding period is not included in the thermal history deviation, if negative and If negative Tend to increase heating output, if For positive and Increase It tends to reduce heating output.

[0047] Furthermore, when synchronously correcting the thermal inertia state vector, the thermal inertia state vector of the current furnace is matched with the thermal inertia state vector of the qualified furnace in the same historical furnace. The matching content includes response slope, response delay, temperature residual, residual heat response characteristics and heat distribution equilibrium characteristics. The similarity matching does not select a historical curve to directly replace the current control curve, but determines the deviation direction of the current furnace relative to the reference thermal response. The deviation direction is used to determine the correction direction of heat transfer hysteresis, billet temperature increment and overshoot risk constraint.

[0048] In this embodiment, if the current furnace heat response relative to the reference thermal response is slower than the furnace temperature response under the same heating output, the furnace temperature continues to rise more after the heating output stabilizes, and the furnace temperature difference converges more slowly, then it is determined that the heat absorption lag of the current furnace heat is enhanced. The thermal inertia state estimation module increases the heat transfer lag and reduces the billet temperature increment corresponding to the same heating output. The virtual billet temperature inversion module reads the furnace temperature of an earlier control cycle as the lag thermal boundary in the recursive calculation. The predictive control module allows limited compensation in the heating stage and checks the overshoot risk in advance in the stage when the target temperature is close.

[0049] In this embodiment, if the current furnace heat response relative to the reference thermal response is faster, has a smaller response delay, and shows a smaller increase in furnace temperature after the heating output decreases, it is determined that the current furnace heat absorption response is faster. The thermal inertia state estimation module reduces the heat transfer lag and increases the overshoot risk constraint in the solution of the heating output correction. The virtual billet temperature inversion module shortens the furnace temperature time shift distance in the recursive calculation. The prediction control module applies stronger suppression to the prediction trend of exceeding the target billet temperature in the critical densification section and the heat preservation section.

[0050] Furthermore, when reconstructing the solution weights for the heating output correction, the instantaneous deviation, predicted deviation, and cumulative thermal history deviation between the virtual billet temperature and the target billet temperature within the current control cycle are calculated. The instantaneous deviation is used to characterize the direction of deviation of the billet thermal state in the current control cycle, the predicted deviation is used to characterize the direction of change of the billet thermal state in subsequent control cycles, and the cumulative thermal history deviation is used to characterize the degree of deviation of the thermal process that has already occurred. The three deviations are not simply added together, but the sign direction and allowable range are first determined, and then it is determined whether the heating output correction belongs to continuous compensation type correction, suppression of overcompensation type correction, or maintenance of output trend.

[0051] In this embodiment, when the instantaneous deviation and the predicted deviation are in the same direction, it indicates that the current virtual billet temperature deviation will continue to exist in subsequent control cycles, and the heating output correction is limited to continuous compensation type correction. If both the instantaneous deviation and the predicted deviation are negative and the cumulative thermal history deviation is negative, the heating output correction is generated in the increasing direction. If both the instantaneous deviation and the predicted deviation are positive and the cumulative thermal history deviation is positive, the heating output correction is generated in the decreasing direction. When the instantaneous deviation and the predicted deviation are in opposite directions and the cumulative thermal history deviation exceeds the allowable range, the heating output correction is limited to suppress overcompensation type correction, and the reverse compensation magnitude is limited according to the overshoot risk amount.

[0052] In this embodiment, when the instantaneous deviation, prediction deviation, and cumulative thermal history deviation are all within the allowable range, the heating output adjustment module maintains the heating output change trend of the previous control cycle and only cuts off the output change that exceeds the furnace temperature constraint or heating rate constraint. Maintaining the output trend is not the same as fixing the output amount, but rather retaining the change part that meets the constraint in the original output change direction. If the virtual billet temperature deviates from the target billet temperature again in the next control cycle, the prediction control module re-enters continuous compensation or suppresses overcompensation calculation.

[0053] In a preferred embodiment, the thermal response sequences of similar historical furnaces are screened for compliance before entering the thermal inertia state estimation. The screening criteria include the quality inspection conclusions corresponding to the historical furnaces, the integrity of the temperature curves, and the continuity of process data. The quality inspection conclusions are used to exclude historical furnaces that have shown sintering deviations. The integrity of the temperature curves is used to exclude historical furnaces that lack key data for the heating, holding, or cooling stages. The continuity of process data is used to exclude historical furnaces with too many sampling interruptions. The screened historical furnaces form a benchmark thermal response library.

[0054] In this embodiment, the reference thermal response library is indexed according to the type of sintering material, the type of target temperature curve, and the thermal response characteristics. After the current furnace starts running, the temperature process data acquisition module selects candidate historical furnaces based on the current target temperature curve. The thermal inertia state estimation module then selects a reference sequence from the candidate historical furnaces based on the response slope of the early heating segment and the response delay of the target temperature approach segment. If the early process data of the current furnace is not sufficient to complete the matching, the robust mean in the reference thermal response library is used as the initial reference, and the matching is re-matched as the data increases in subsequent control cycles.

[0055] Furthermore, the temperature process data acquisition module processes the furnace zone temperature difference by taking the difference between multiple furnace zone temperatures within the same control cycle, the direction of the difference change, and the duration of the difference as input. It does not require the addition of new hardware acquisition objects. After the furnace zone temperature difference enters the thermal distribution equilibrium characteristic, it is used to reflect whether the furnace internal thermal environment has become equilibrium when the furnace temperature measurement point meets the standard. When the furnace zone temperature difference continues to expand, the virtual billet temperature inversion module reduces the degree of acceptance of a single furnace temperature measurement value, and the predictive control module increases the solution weight of the furnace zone temperature difference constraint in the critical densification section.

[0056] In this embodiment, the overshoot risk is determined by the virtual billet temperature prediction, the target billet temperature, the heating rate, and the residual heat response characteristics. If the virtual billet temperature prediction is lower than the target billet temperature but the residual heat response characteristics are high, the overshoot risk is still increased. If the virtual billet temperature prediction is higher than the target billet temperature and the heating output is still increasing, the overshoot risk becomes the dominant constraint. If the virtual billet temperature prediction is lower than the target billet temperature and the residual heat response characteristics are low, the overshoot risk remains low and compensation is allowed according to the virtual billet temperature deviation.

[0057] In a preferred embodiment, thermal history consistency verification is performed after each control cycle. The verification objects include cumulative thermal input deviation, critical temperature zone dwell time, and virtual billet temperature continuity. The cumulative thermal input deviation is obtained based on the integral difference between the target billet temperature sequence and the virtual billet temperature sequence above the lower limit temperature of thermal history calculation. The critical temperature zone dwell time is obtained based on the cumulative number of cycles after the virtual billet temperature enters the preset temperature zone. The virtual billet temperature continuity is obtained based on the temperature change amplitude of adjacent control cycles. When any verification object exceeds the allowable range, the verification result enters the objective function weight reconstruction of the next control cycle.

[0058] In this embodiment, the temperature process data acquisition module, thermal inertia state estimation module, virtual billet temperature inversion module, predictive control module, and heating output adjustment module form a closed-loop processing flow during operation. The temperature process data acquisition module outputs the current furnace process temperature sequence, the thermal inertia state estimation module outputs the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state, the virtual billet temperature inversion module outputs the virtual billet temperature sequence, the predictive control module outputs the billet temperature change trend and constraint term solution results, and the heating output adjustment module outputs the heating output correction amount. In the next control cycle, the actual furnace temperature and heating output are collected again, and the prediction results of the previous cycle are corrected.

[0059] In this embodiment, the closed-loop processing flow is established around the difference between the furnace temperature and the equivalent thermal state of the billet. The furnace temperature feedback is responsible for providing the external thermal boundary, the historical thermal response sequence of similar furnaces is responsible for providing the reference thermal response, the thermal inertia state estimation is responsible for converting the deviation of the current furnace from the reference into a calculable state quantity, the virtual billet temperature inversion is responsible for converting the furnace temperature and heating output into the equivalent thermal state of the billet, the predictive control is responsible for judging the risk of thermal hysteresis and overshoot in subsequent control cycles, and the heating output adjustment is responsible for converting the judgment result into the heating output correction amount.

[0060] In a preferred embodiment, if the process temperature sequence of the current furnace lacks furnace area temperature difference data within a certain control cycle, the thermal inertia state estimation module uses the thermal distribution equilibrium feature within the previous confidence window as a temporary input and reduces the weight of the thermal distribution equilibrium feature in the thermal inertia state vector. If furnace area temperature difference data is missing for several consecutive control cycles, the predictive control module no longer uses the furnace area temperature difference constraint as the main constraint, but increases the constraint weight of the virtual billet temperature fluctuation amplitude and overshoot risk, avoiding missing data from participating in the solution of the main constraints.

[0061] In this embodiment, if there are multiple similar reference sequences in the historical matching results of the current furnace batch, the thermal inertia state estimation module performs weighted fusion of the response slope, response delay and residual heat response characteristics of the multiple reference sequences. The weight is determined by the similarity between the current furnace batch and the reference sequence in the early heating stage and the stage where the target temperature is close. The reference sequence with higher similarity gets higher weight, while the reference sequence with lower similarity only provides boundary constraints. The fused reference thermal response is used to calculate the thermal inertia state quantity without changing the target temperature curve of the current furnace batch.

[0062] Furthermore, if the current control cycle is in the target temperature approach segment and the furnace temperature is already close to the target temperature, but the virtual billet temperature is still lower than the target billet temperature, the predictive control module judges the subsequent billet temperature trend that may occur after continued heating based on the heat transfer hysteresis. When the prediction result shows that the overshoot risk constraint will not be exceeded after compensation, the heating output adjustment module generates a correction amount in the restricted increase direction. When the prediction result shows that the residual heat release will cause the virtual billet temperature to exceed the target billet temperature in the subsequent control cycle, the heating output adjustment module generates a correction amount in the decrease direction or the maintenance direction.

[0063] In a preferred embodiment, if the current control cycle is in the cooling phase, the virtual billet temperature inversion module continues to perform recursive calculations without stopping the inversion due to the decrease in heating output. The thermal boundary of the cooling phase is jointly constituted by the furnace temperature decreasing trend and the waste heat response characteristics. The predictive control module judges whether the cooling process is too fast or the release is delayed based on the virtual billet temperature decreasing rate and the waste heat release trend. When the virtual billet temperature decreasing rate exceeds the constraint range, the output correction amount is limited to the direction of slowing down the change in heat input. When the waste heat release is still strong, the output correction amount is limited to the direction of suppressing the continued heating.

[0064] In this embodiment, during specific operation, the temperature process data acquisition module generates the latest process data set at the end of each control cycle. The thermal inertia state estimation module incorporates the latest process data set into the sliding time window and updates the response slope, response delay, temperature residual, residual heat response characteristics, and heat distribution equilibrium characteristics. The virtual billet temperature inversion module recursively derives the virtual billet temperature based on the updated thermal inertia state variables. The predictive control module solves the objective function based on the recursive result and the target billet temperature sequence. The heating output adjustment module combines the solved heating output correction amount with the current heating output amount and sends it to the sintering temperature control execution terminal.

[0065] In this embodiment, all processing steps are based on the existing temperature control data and heating output data of the semiconductor ceramic sintering process. The furnace temperature, target temperature curve, heating output, heating rate, holding time, and furnace zone temperature difference together constitute the data basis for inversion and predictive control. The virtual billet temperature sequence, thermal inertia state quantity, cumulative thermal history deviation, and overshoot risk quantity together constitute the control basis for heating output correction, so that the temperature control object is transformed from a single furnace temperature measurement value into a control object constrained by both the furnace thermal boundary and the equivalent thermal state of the billet.

Claims

1. A method for controlling the sintering temperature of a semiconductor ceramic, characterized by, include: The furnace temperature, target temperature curve, heating output, heating rate, holding time and furnace zone temperature difference during the semiconductor ceramic sintering process are obtained to form a process temperature sequence corresponding to the current furnace batch. Based on the process temperature sequence and the thermal response sequence of similar historical furnaces, estimate the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state of the current furnace. Based on the thermal inertia coefficient, the heat transfer hysteresis, and the equivalent heat capacity state, the virtual blank temperature sequence of the semiconductor ceramic blank is inverted; Using the virtual billet temperature sequence as feedback, the billet temperature change trend in subsequent control cycles is predicted in conjunction with the target temperature curve. Based on the green body temperature change trend, furnace temperature constraint, heating rate constraint, and overshoot risk constraint, a heating output correction amount is generated, and the heating output of the sintering process is adjusted according to the heating output correction amount.

2. The method of claim 1, wherein the sintering temperature is controlled to be in the range of 1,000 to 1,300°C. Estimating the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity of the current furnace cycle includes: dividing the process temperature sequence into time windows according to the temperature rise segment, the target temperature approach segment, and the heat preservation segment, and extracting the response slope, response delay, and temperature residual between the furnace temperature change and the heating output change in each time window. The thermal inertia coefficient is obtained by matching the deviation of the response slope with the benchmark response slope in the historical thermal response sequence of similar furnaces. The response delay is aligned and compared with the set change node of the target temperature curve to obtain the heat transfer hysteresis. The temperature residual and the cumulative heating output are combined and fitted to obtain the equivalent heat capacity state.

3. The method of claim 1, wherein the sintering temperature is controlled to be in the range of 1,000 to 1,300°C. The virtual blank temperature sequence of the semiconductor ceramic blank is inverted, including: using the furnace temperature of the current control cycle as the external thermal boundary, using the thermal inertia coefficient to limit the response amplitude of heat entering the blank, using the heat transfer hysteresis to limit the time offset of the furnace temperature change to the blank equivalent temperature, and using the equivalent heat capacity state to limit the blank temperature change amplitude under the same heating output. The equivalent heat absorption and temperature increment of the billet are calculated recursively according to the control cycle sequence. The recursively obtained equivalent billet temperature is then continuously corrected with the virtual billet temperature of the previous control cycle to generate a virtual billet temperature sequence.

4. The method of claim 1, wherein the sintering temperature is controlled to be in the range of 1,000 to 1,300°C. The generation of heating output correction includes: dividing the target temperature curve into a heating segment, a critical densification segment, a heat preservation segment, and a cooling segment; Within each control cycle, the trend of billet temperature change in subsequent control cycles is predicted based on the deviation between the virtual billet temperature sequence and the target billet temperature. Based on the billet temperature change trend, calculate the furnace temperature constraint, heating rate constraint, overshoot risk constraint, and thermal history deviation constraint respectively; By jointly solving for each constraint term, the heating output correction amount that satisfies the current sintering stage limitation conditions is obtained.

5. The method of claim 2, wherein the sintering temperature is controlled to be in the range of 1,000 to 1,300°C. When estimating the thermal inertia coefficient, heat transfer hysteresis and equivalent heat capacity of the current furnace, the method further includes: constructing waste heat response characteristics based on the continuous rise of furnace temperature after the heating output changes from a changing state to a stable state within the time window. Based on the diffusion change of the temperature difference in the furnace area within the same time window, a heat distribution uniformity characteristic is constructed. The waste heat response characteristics, the heat distribution equilibrium characteristics, the response slope, the response delay, and the temperature residual are combined to form a thermal inertia state vector, and the thermal inertia state vector is used to synchronously correct the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state.

6. The method of claim 3, wherein the sintering temperature is controlled to be in the range of 1,000 to 1,300°C. When generating the virtual billet temperature sequence, the method further includes: shifting the furnace temperature sequence over time according to the heat transfer hysteresis to obtain the hysteresis thermal boundary sequence; The amplitude attenuation process of the hysteresis thermal boundary sequence is performed based on the thermal inertia coefficient to obtain the equivalent input temperature sequence of the billet. Based on the equivalent heat capacity state, the equivalent input temperature sequence of the billet is subjected to cumulative heat absorption constraint to obtain the initial virtual billet temperature; The initial virtual billet temperature is compared with the virtual billet temperature of the historical control cycle to check for monotonicity and continuity, and temperature points that do not meet the check conditions are corrected by recursive write-back.

7. The method for controlling the sintering temperature of semiconductor ceramics according to claim 4, characterized in that, When solving the various constraints together, the method further includes: in the heating stage, the virtual billet temperature tracking deviation and heating rate constraint are the main constraints; in the critical densification stage, the furnace zone temperature difference constraint, thermal history deviation constraint and overshoot risk constraint are the main constraints; in the heat preservation stage, the virtual billet temperature fluctuation amplitude and cumulative heat input deviation are the main constraints; and in the cooling stage, the virtual billet temperature drop rate and residual heat release trend are the main constraints. The solution weights for reconstructing the heating output correction amount are based on the main constraints of the current sintering stage.

8. The method for controlling the sintering temperature of semiconductor ceramics according to claim 5, characterized in that, When synchronously correcting the thermal inertia state vector, the process also includes: performing similarity matching between the thermal inertia state vector of the current furnace and the thermal inertia state vector corresponding to the qualified furnace in the same historical furnace, and determining the deviation direction of the current furnace relative to the reference thermal response. When the deviation direction indicates an enhanced heat transfer hysteresis in the current furnace, the heat transfer hysteresis is increased and the billet temperature increment corresponding to the same heating output is reduced. When the deviation direction indicates that the heat absorption response of the current furnace is accelerated, the heat transfer hysteresis is reduced and the overshoot risk constraint is strengthened in the solution of the heating output correction.

9. The method for controlling the sintering temperature of semiconductor ceramics according to claim 7, characterized in that, When reconstructing the solution weights for the heating output correction, the method also includes: calculating the instantaneous deviation, prediction deviation, and cumulative thermal history deviation between the virtual billet temperature and the target billet temperature within the current control cycle. When the instantaneous deviation is in the same direction as the predicted deviation, the heating output correction amount is limited to a continuous compensation type correction; When the instantaneous deviation is opposite to the predicted deviation and the cumulative thermal history deviation exceeds the preset allowable range, the heating output correction amount is limited to a correction that suppresses overcompensation. When the instantaneous deviation, prediction deviation, and cumulative thermal history deviation are all within the allowable range, the heating output change trend of the previous control cycle is maintained.

10. A semiconductor ceramic sintering temperature control system, characterized in that, include: The temperature process data acquisition module is used to acquire the furnace temperature, target temperature curve, heating output, heating rate, holding time and furnace zone temperature difference during the semiconductor ceramic sintering process, forming a process temperature sequence corresponding to the current furnace. The thermal inertia state estimation module is used to estimate the thermal inertia coefficient, heat transfer hysteresis, and equivalent heat capacity state of the current furnace based on the process temperature sequence and the thermal response sequence of similar furnaces in the past. The virtual blank temperature inversion module is used to invert the virtual blank temperature sequence of the semiconductor ceramic blank based on the thermal inertia coefficient, heat transfer hysteresis and equivalent heat capacity state. The predictive control module is used to predict the trend of billet temperature change in subsequent control cycles by using the virtual billet temperature sequence as feedback object and combining it with the target temperature curve. The heating output adjustment module is used to generate a heating output correction amount based on the billet temperature change trend, furnace temperature constraint, heating rate constraint and overshoot risk constraint, and adjust the heating output of the sintering process according to the heating output correction amount.

Images