An air-cooled system adaptive control method based on residual heat attenuation model

Abstract

This invention relates to the field of industrial process control technology, specifically to an adaptive control method for an air-cooled system based on a residual heat attenuation model. The method includes: filtering real-time temperature measurements of the components to output an estimated temperature and instantaneous cooling rate; constructing a phase-change-free reference model to calculate the theoretical temperature; calculating the latent heat release intensity index based on the deviation between the estimated and theoretical temperatures and the instantaneous cooling rate; calculating the cross-sectional geometric thermal inertia factor based on the component's geometric parameters; calculating a proportional gain term based on the deviation between the estimated temperature and the target cooling curve temperature; constructing a negative phase-change damping term using the latent heat release intensity index and the cross-sectional geometric thermal inertia factor; and generating a fan frequency command. This invention solves the technical problems of existing technologies failing to identify cooling misjudgments caused by latent heat of phase change and failing to achieve differentiated cooling control for workpieces with different geometric characteristics.

Description

Technical Field

[0001] This invention relates to the field of industrial process control technology, and in particular to an adaptive control method for an air-cooled system based on a waste heat attenuation model. Background Technology

[0002] Large wind turbine ring forgings, especially deep-step L-shaped and T-shaped flanges and gear rings, are key load-bearing components in wind power generation equipment and are typically made of medium-carbon alloy steel. These workpieces accumulate extremely high levels of heat after rolling and forming, requiring precise air-cooling processes to control their cooling rate and obtain a uniform pearlitic and ferrite microstructure. This ensures that the workpiece meets the required low-temperature impact strength and fracture toughness standards.

[0003] In existing industrial settings, air-cooled control systems primarily employ PID closed-loop feedback control to regulate fan speed. The controller directly reads the surface temperature of the components using an infrared thermometer as the controlled variable feedback signal, comparing the measured temperature with a set curve. When the temperature exceeds the set value, the controller outputs a command to increase the fan speed; conversely, it decreases the fan speed when the temperature falls below the set value. This control strategy is effective for handling conventional uniform cooling processes, but it has significant limitations when dealing with wind turbine components exhibiting complex phase transformation characteristics. When wind turbine components cool to a specific temperature range, a solid-state phase transformation from austenite to pearlite or ferrite occurs, releasing a large amount of latent heat. This causes the surface temperature to decrease more slowly or even rise. Traditional PID controllers cannot recognize this physical process and often misinterpret it as insufficient heat dissipation, leading to an increase in fan speed. Forced rapid cooling can result in the formation of abnormal bainitic or martensitic structures on the surface, triggering thermal stress cracks.

[0004] Furthermore, the corners of irregularly shaped workpieces have thick walls and slow heat dissipation, while the edges have thin walls and fast heat dissipation. Existing control methods lack a systematic evaluation of the cross-sectional geometry. Using the same control parameters for workpieces of different shapes can easily lead to excessive cooling for compact, thick-walled irregularly shaped parts, resulting in thermal stress cracking, which is difficult to meet the production requirements of high-quality wind power components. Summary of the Invention

[0005] To address the technical problems of existing technologies failing to identify cooling misjudgments due to the latent heat of phase change and failing to achieve differentiated cooling control for workpieces with different geometric features, this invention provides an adaptive control method for air-cooled systems based on a residual heat attenuation model, comprising: The real-time temperature measurement value of the ring is obtained, the real-time temperature measurement value is filtered, and the estimated temperature and instantaneous cooling rate at the current moment are output. A no-phase-change baseline model is constructed, and the theoretical temperature at the current moment if no phase change occurs is calculated. Based on the deviation between the estimated temperature and the theoretical temperature and the instantaneous cooling rate, the latent heat release intensity index of phase change is calculated. The latent heat release intensity index of phase change is positively correlated with both the deviation and the instantaneous cooling rate. Obtain the geometric parameters of the ring component, and calculate the cross-sectional geometric thermal inertia factor based on the geometric parameters. The cross-sectional geometric thermal inertia factor is positively correlated with the heat dissipation difficulty and cross-sectional distortion degree of the ring component. The proportional gain term is calculated based on the deviation between the estimated temperature and the target cooling curve temperature. A negative phase change damping term is constructed using the latent heat release intensity index of phase change and the cross-sectional geometric thermal inertia factor. The basic sustaining frequency, the proportional gain term, and the phase change damping term are combined to generate the fan frequency command.

[0006] This invention eliminates the interference of ambient noise by filtering real-time temperature measurements, providing accurate state estimation of the controlled variable for the closed-loop control system. By introducing the latent heat release intensity index of phase change and the cross-sectional geometric thermal inertia factor as feedforward compensation signals into the control loop, the control system actively adapts to the microstructure transformation and macroscopic geometric characteristics of the material, avoiding the controller from outputting destructive strong cooling commands due to misjudgment during the intense phase change period.

[0007] Preferably, the filtering process is implemented using a discrete Kalman filter state observer, including: defining a system state vector containing the estimated temperature and instantaneous cooling rate; based on the physical laws of heat conduction, using the product of the estimated temperature at the previous moment and the instantaneous cooling rate at the previous moment and the sampling period as the prior estimate at the current moment; using the real-time temperature measurement value at the current moment to correct the prior estimate, with the goal of minimizing the estimation error covariance, and outputting the estimated temperature and instantaneous cooling rate at the current moment.

[0008] By utilizing the prediction and update mechanism of Kalman filtering, the actual temperature change trajectory and instantaneous cooling rate of the workpiece can be accurately reconstructed in an industrial environment filled with water vapor and oxide scale interference, providing a clean and reliable data foundation for subsequent phase transition identification and control decisions.

[0009] Preferably, the real-time temperature measurement of the ring component is obtained by multiple sets of infrared thermal imagers arranged in the air-cooling station. The infrared thermal imagers are evenly arranged along the circumference of the ring component, and the monitoring area covers the inner corner of the ring component's cross-section.

[0010] By evenly distributing multiple points and focusing on covering the inner corner hot spot area, it is possible to capture abnormal temperature rise signals at the point of maximum wall thickness in the first instance, avoiding the failure to detect local thermal stress concentration points due to single-point temperature measurement.

[0011] Preferably, the phase-change-free reference model is constructed based on an exponential decay law, with the ambient temperature as the asymptotic lower limit and the difference between the initial temperature and the ambient temperature as the base. The theoretical temperature at the current moment is calculated according to the law of exponential decay over time, and the rate of decay is determined by the comprehensive heat transfer coefficient.

[0012] By establishing a calculable theoretical baseline based on the exponential decay law, a physically meaningful reference standard is provided for subsequent identification of latent heat of phase transition through residuals, avoiding confusion between temperature deviations during the phase transition period and normal heat dissipation fluctuations.

[0013] Preferably, the calculation of the latent heat release intensity index of the phase change includes: A fixed-length historical sliding window is extracted from the current time. The residual between the estimated temperature and the theoretical temperature at each time point within the historical sliding window is calculated. The residual is normalized and processed by root mean square to obtain statistical characteristic values ​​that characterize the degree of temperature deviation. A nonlinear gain term is constructed for the instantaneous cooling rate. The statistical characteristic value is multiplied by the nonlinear gain term to obtain the latent heat release intensity index of the phase change.

[0014] By characterizing the statistical intensity of temperature deviation by the root mean square value of the normalized residual within the sliding window, and combining it with the nonlinear weighting of the cooling rate, the internal phase transformation process of the material, which cannot be directly observed, is transformed into a calculable evaluation index, enabling the control system to sense the transformation of metallographic structure.

[0015] Preferably, the nonlinear gain term adopts a logarithmic function form, such that the growth rate of the nonlinear gain term gradually decreases as the instantaneous cooling rate increases.

[0016] Preferably, the geometric parameters include the outer surface area, volume, maximum wall thickness, and minimum wall thickness of the ring.

[0017] Preferably, the calculation of the cross-sectional geometric thermal inertia factor includes: constructing a thermal inertia term reflecting the overall heat dissipation difficulty using the ratio of the outer surface area to the volume of the ring component; constructing a shape correction term reflecting the degree of cross-sectional distortion using the ratio of the cross-sectional wall thickness range to the average wall thickness; multiplying the thermal inertia term with the shape correction term in a positive exponential form using a negative exponent to obtain the cross-sectional geometric thermal inertia factor, such that the smaller the specific surface area or the greater the difference in cross-sectional wall thickness, the larger the value of the cross-sectional geometric thermal inertia factor.

[0018] Preferably, the generation of the fan frequency command includes: calculating a positive frequency increment based on the difference between the estimated temperature and the target cooling curve temperature using proportional control logic; multiplying the phase change latent heat release intensity index, the cross-sectional geometric thermal inertia factor, the wind field utilization coefficient, and the phase change compensation gain coefficient used to match the latent heat release intensity and cross-sectional complexity of the ring material to obtain a negative frequency adjustment amount; adding the positive frequency increment to the base maintenance frequency and then subtracting the negative frequency adjustment amount, and performing amplitude limiting processing on the calculation result to obtain the fan frequency command.

[0019] Through the counterbalancing mechanism of positive frequency increment and negative frequency adjustment, the increase in wind speed is automatically suppressed when the phase transition is violent or the workpiece geometry is complex. This prevents the high temperature caused by the release of latent heat from being misjudged as insufficient heat dissipation and causing overcooling, so that the cooling process can maintain the target cooling rate and suppress thermal stress cracks.

[0020] Preferably, the wind field utilization coefficient is determined based on the relative position of the fan and the ring component, specifically by calculating the effective cooling area ratio corresponding to the coverage area of ​​the airflow on the surface of the ring component.

[0021] The technical solution of the present invention has the following beneficial technical effects: This invention introduces a noise filtering mechanism based on a state observer into the control system. It extracts accurate estimated temperature and instantaneous cooling rate from noisy measurement data as the controlled variable feedback signal of the control system. Then, using the theoretical temperature of the phase change reference model without phase change as the reference baseline, it transforms the phase change latent heat release process, which cannot be directly observed inside the material, into a calculable phase change latent heat release intensity index. Combined with the cross-sectional geometric thermal inertia factor, which reflects the geometric heat dissipation characteristics of the workpiece, a negative phase change damping feedforward compensation term is constructed in the proportional feedback control loop. This forms a composite control structure combining feedback and feedforward, which automatically suppresses the frequency conversion command of the fan when the phase change is severe or the cross-sectional distortion of the workpiece is serious. This avoids the surface structure abnormalities and thermal stress cracks caused by the traditional PID control system misjudging the latent heat release as insufficient heat dissipation and blindly applying strong cooling. It achieves precise adaptive control and adjustment of large wind turbine ring forgings, which takes into account both the formation of the target metallographic structure and the prevention of thermal stress cracking. Attached Figure Description

[0022] Figure 1 This is a flowchart of an adaptive control method for an air-cooled system based on a waste heat attenuation model, according to the present invention. Figure 2 This is a schematic diagram of the 3D model of the spatiotemporal evolution of latent heat of phase transition based on Kalman filtering in this invention; Figure 3 This is a schematic diagram of the phase plane analysis of the stability of the control system based on the residual heat decay model in this invention. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0024] This invention discloses an adaptive control method for an air-cooled system based on a waste heat attenuation model, referring to... Figure 1 This includes steps S101 to S104: S101. Obtain the real-time temperature measurement value of the ring component, filter the real-time temperature measurement value, and output the estimated temperature and instantaneous cooling rate at the current moment.

[0025] It should be noted that the surface temperature of wind turbine components after rolling and forming can reach hundreds of degrees Celsius. Water vapor and oxide scale present in industrial environments can obstruct and interfere with infrared thermography, resulting in significant random fluctuations in the directly acquired temperature data. Using unprocessed measurements directly for control decisions can easily cause severe fluctuations in the wind turbine frequency. Therefore, this embodiment introduces a discrete Kalman filter state observer to extract smooth and reliable temperature state information from the noisy measurement data.

[0026] Specifically, for example, wind turbine ring components with deep-stepped L-shaped cross-section flanges, after being rolled and formed, enter the air-cooling station. The system evenly distributes four sets of infrared thermal imagers along the circumference of the ring component at this station, focusing on monitoring the inner corners of the L-shaped cross-section flange and the gear ring, i.e., the hot spot area. Because the inner corner has the greatest wall thickness and the slowest heat dissipation, it is the most prone to thermal stress concentration. Focusing on the inner corner as the key temperature monitoring point allows for the immediate capture of abnormal temperature rise signals. The system is configured with a sampling period... Temperature measurements are collected in real time, with a time interval of 0.5 seconds. In other embodiments, the sampling period can be set according to the actual application scenario and requirements. And sampling period The value range is [0.1, 2]. A sampling period that is too short will increase the data processing load and the noise ratio, while a sampling period that is too long will delay the response to temperature changes.

[0027] The Discrete Kalman Filter (DKF) state observer is a recursive optimal estimation algorithm based on a state-space model. It is a well-known technique and will not be elaborated upon here. The definition includes the estimated temperature. With instantaneous cooling rate The system status, where For the first The estimated temperature at that moment, in °C; For the first The instantaneous cooling rate at any given time is expressed in °C / s. Based on the physical laws of heat conduction, the state transition equation is constructed as follows:

[0028] In other words, the estimated temperature at the current moment is equal to the estimated temperature at the previous moment plus the temperature change caused by the instantaneous cooling rate at the previous moment within one sampling period. When the value is negative, it indicates that the temperature is trending downwards, and the estimated temperature after superposition is lower than the previous time step; when When the temperature approaches zero, it indicates that the temperature change is becoming more gradual.

[0029] The initial state setting of the Kalman filter state observer is: initial estimated temperature. Take the first real-time temperature measurement value acquired by the infrared thermal imager. Initial instantaneous cooling rate Set as The algorithm gradually converges to the true cooling rate through subsequent iterations. A large diagonal value is used in the initial estimation error covariance matrix to assign a high confidence weight to the measured values ​​in the initial stage. The prediction and update iterative mechanism of the Kalman filter is utilized to process the real-time temperature measurements. By integrating the prediction model and correcting prediction bias through the gain matrix, the goal is to minimize the estimation error covariance, and output the optimal temperature estimate for the current time after noise removal. and smooth instantaneous cooling rate .

[0030] S102. Construct a baseline model without phase change, calculate the theoretical temperature at the current moment if no phase change occurs, and calculate the latent heat release intensity index of phase change based on the degree of deviation between the estimated temperature and the theoretical temperature and the instantaneous cooling rate. The latent heat release intensity index of phase change is positively correlated with the degree of deviation and the instantaneous cooling rate.

[0031] When wind turbine components are cooled to a specific temperature range, a solid-state phase transformation from austenite to pearlite or ferrite occurs. This solid-state phase transformation process releases a large amount of latent heat, causing the surface temperature of the workpiece to decrease more slowly or even rise. Traditional PID controllers cannot distinguish between this temperature rise caused by the internal phase transformation and the actual insufficient heat dissipation, and are prone to making incorrect accelerated cooling decisions. Therefore, this embodiment constructs a theoretical decay model that does not consider the influence of phase transformation as a reference baseline. By comparing the deviation between the actual observed temperature and the theoretical temperature, the intensity of the phase transformation inside the material is inferred.

[0032] Specifically, a phase-change-free benchmark model is constructed based on Newton's law of cooling, with a theoretical temperature... The calculation formula is as follows:

[0033] in, The ambient temperature is collected in real time by an ambient temperature sensor located in the air-cooled workstation. The initial temperature of the ring component when it enters the air-cooling station is obtained by an infrared thermal imager when the ring component first arrives at the station. The overall heat transfer coefficient was obtained by static calibration of samples of the same material under the same wind speed conditions before the cooling test, and the unit is s. - ¹, The value range is [0.0001, 0.001]. When the deviation is small, the theoretical temperature decays too slowly, and the sensitivity of the residual signal to the phase transition decreases. If the value is too large, the theoretical temperature drops too quickly, and false residuals will also be generated during the normal heat dissipation phase, interfering with phase transition identification. This refers to the cumulative cooling time from the start of cooling to the current moment. During actual control, the fan frequency is maintained around the base frequency. The actual variation in the overall heat transfer coefficient is limited due to ambient temperature fluctuations. Furthermore, the temperature deviation caused by the release of latent heat from phase change is far greater than the theoretical temperature drift caused by wind speed fluctuations. Therefore, a phase-change-free benchmark model constructed using a statically calibrated overall heat transfer coefficient can effectively distinguish between phase change signals and normal heat dissipation fluctuations. In one embodiment, the overall heat transfer coefficient can be corrected online based on the current fan frequency command to further improve the tracking accuracy of the theoretical temperature. This relationship is based on ambient temperature. To approach the lower limit, with excess temperature As the attenuation base, when The larger the value, the faster the exponential decay and the faster the theoretical temperature approaches the ambient temperature, when the cumulative cooling time... As it approaches infinity, the theoretical temperature approaches... .

[0034] Furthermore, during a phase transition, the release of latent heat hinders the temperature decrease, leading to a change in the actual estimated temperature. Higher than theoretical temperature This embodiment constructs a latent heat release intensity index for phase change. To assess the severity of a phase transition, the calculation process involves the following steps: First, calculate the normalized residual of the estimated temperature relative to the theoretical temperature at each moment within the historical sliding window:

[0035] in, For the time index within the sliding window, ; The sliding window length is defined in engineering practice as three levels: short window [30, 40], medium window [40, 50], and long window (50, 60]. The short window is sensitive to transient response but lacks statistical robustness, while the long window is statistically robust but responds slowly to the onset of phase change. In this embodiment, a medium value of 50 is selected for the phase change time span of the L-shaped cross-section flange. The value range is [30, 60]. When the number of sampling points is insufficient at the beginning of the cooling process... That is, The sliding window automatically collapses to The upper bound of the summation in the root mean square calculation formula is determined by... Replace with The corresponding number of terms to be summed is determined by Replace with The denominator is synchronously replaced with To avoid referencing historical data that has not yet been generated; as sampling progresses to Afterwards, the window is restored to its full length. ; This is a safety constant, expressed in °C, and is related to the excess temperature in the denominator. The dimensions are consistent, which is used to prevent the denominator from approaching zero. If the value is too small, the normalized residual value may still be abnormally amplified; if it is too large, it will compress the dynamic range. The value range is [1, 10]. Normalized residual The numerator is the deviation between the estimated temperature and the theoretical temperature, and the denominator is the excess temperature of the theoretical temperature relative to the ambient temperature plus a safety margin. The physical meaning of normalization is to transform the absolute temperature deviation into a proportional deviation relative to the current remaining cooling space, so that the deviations at different stages of the cooling process are comparable.

[0036] Then, the root mean square of the normalized residuals is processed to obtain statistical characteristic values ​​representing the degree of temperature deviation:

[0037] Statistical eigenvalues This is the root mean square value of the normalized residual within the sliding window, calculated in the same way as the normalized root mean square error commonly used in signal processing. The root mean square of the residual square is used for temperature deviation assessment because the square of the residual corresponds to the unsteady temperature field distortion caused by the release of latent heat during phase transition at the energy level, characterizing the intensity of temperature deviation caused by the latent heat released during the transformation from internal austenite to pearlite or ferrite. This approach of identifying phase transitions by detecting deviations of the actual temperature curve from the theoretical baseline has mature practices in the field of thermal analysis; for example, differential scanning calorimetry uses a similar baseline deviation detection principle to identify the phase transition temperature range of materials. The greater the deviation between the estimated temperature and the theoretical temperature, the more pronounced the temperature deviation. The larger.

[0038] Next, a nonlinear gain term with respect to the instantaneous cooling rate is constructed:

[0039] in, The value is a natural constant, ensuring that the output value of the logarithmic term is greater than or equal to 1. For reference cooling rate, it is determined based on the typical natural cooling rate of the ring material in the no-phase-change stage. When the value is taken within the range of [0.5, 5], the output of the logarithmic gain term is within a reasonable dynamic range; in this embodiment, it is taken as 1℃ / s. It should be noted that the absolute value of the instantaneous cooling rate... Substituting the logarithmic function and the natural constant Before addition, the values ​​have been divided by the reference cooling rate. Dimensionless processing was performed, and the nonlinear mapping was performed in pure numerical form. The nonlinear gain term adopts the form of a logarithmic function, which conforms to the continuous cooling transformation law in materials science that the greater the cooling rate, the more intense the phase change driving force and latent heat release. This makes the growth rate of the nonlinear gain term gradually decrease as the instantaneous cooling rate increases, reflecting the huge heat difference represented by a small temperature difference under high wind speed conditions.

[0040] Finally, multiplying the statistical eigenvalues ​​by the nonlinear gain term yields the latent heat release intensity index of the phase transition:

[0041] The greater the deviation between the estimated temperature and the theoretical temperature, the more statistical characteristic values... The larger the value, the greater the latent heat release intensity index of the phase change; when the absolute value of the instantaneous cooling rate is larger, the nonlinear gain term... The larger the value, the greater the latent heat release intensity index of phase change. Therefore, the latent heat release intensity index of phase change is positively correlated with the degree of deviation and the instantaneous cooling rate, which is in line with the laws of physics: even if the temperature difference is not large at high wind speeds, the actual heat contained is very different, requiring a higher assessment of phase change intensity.

[0042] From the mathematical properties of logarithmic functions, even when the instantaneous cooling rate is extremely high, the growth of the nonlinear gain term tends to be gradual, preventing abrupt changes in control commands and ensuring the stability of the control system.

[0043] Under another operating condition, if no phase change occurs, the estimated temperature is basically consistent with the theoretical temperature, the normalized residual approaches zero, and the latent heat release intensity index of the phase change approaches zero, indicating that no phase change is currently occurring.

[0044] Thus, by comparing the actual estimated temperature with the theoretical temperature of the reference model without phase change, and by introducing nonlinear gain weighting based on the instantaneous cooling rate, the internal phase change process of the material that cannot be directly observed can be transformed into a calculable latent heat release intensity index, enabling the control system to sense the metallographic structure transformation and solving the technical problem that traditional control methods cannot identify the release of internal latent heat.

[0045] S103. Obtain the geometric parameters of the ring component, and calculate the cross-sectional geometric thermal inertia factor based on the geometric parameters. The cross-sectional geometric thermal inertia factor is positively correlated with the heat dissipation difficulty and cross-sectional distortion degree of the ring component.

[0046] It should be noted that the heat dissipation characteristics of ring components of different shapes vary greatly. Ring components with thin-walled, uniform cross-sections have a large specific surface area and small wall thickness differences, resulting in rapid and uniform heat dissipation. In contrast, ring components with deep-stepped L-shaped cross-sections have a large volume and corner wall thicknesses that far exceed edge wall thicknesses, leading to slow heat dissipation and a tendency for excessive local temperature differences to cause thermal stress cracking. Existing control methods typically apply the same control parameters to all workpieces, neglecting the influence of geometric structure on thermal conduction inertia. Therefore, this embodiment introduces a cross-sectional geometric thermal inertia factor, transforming complex geometric features into a dimensionless comprehensive correction coefficient, providing a basis for subsequent differentiated control.

[0047] Specifically, the system reads geometric parameters from the process document: outer surface area of ​​the ring component. ,volume Maximum wall thickness of L-shaped cross-section Minimum wall thickness Cross-sectional geometric thermal inertia factor The calculation formula is as follows:

[0048] in, The standard reference specific surface area constant is determined statistically based on the typical geometric dimensions of similar ring components. The range of values ​​for is [3, 8]; If the value is too small, the thermal inertia term will be too small, and the geometric factor will not reflect the heat dissipation difficulty of compact workpieces. If the value is too large, it will cause the thermal inertia term to be too high, resulting in the control system applying excessively strong wind speed suppression to all workpieces. The average wall thickness of the cross section is , and All are shape sensitivity indices. The value range is [0.5, 1.5]. In this embodiment... Setting it to 1, the influence of specific surface area on heat dissipation difficulty has a moderate weight. The larger the surface area, the greater the weight of the influence of the heat dissipation difficulty; The value range of is [1, 3]. Choosing a value of 2 amplifies the effect of wall thickness non-uniformity on shape correction to a quadratic degree, which aligns with engineering experience regarding stress concentration at the corners of L-shaped cross-sections. The larger the value, the higher the weight of the influence of wall thickness non-uniformity on shape correction.

[0049] In other words, the cross-sectional geometric thermal inertia factor In the calculation formula, the ratio of the outer surface area of ​​the ring component to its volume, i.e., the specific surface area, is used to construct a thermal inertia term that reflects the overall heat dissipation difficulty, based on the ratio of the outer surface area to the volume of the ring component to the standard reference value. Since the exponent is... When the specific surface area is smaller relative to the standard reference value, the value of the thermal inertia term is larger, indicating that the workpiece is more compact and more difficult to dissipate heat; consequently, the difference in cross-sectional wall thickness... With average wall thickness The ratio of these values ​​constructs a shape correction term reflecting the degree of cross-sectional distortion. The greater the difference in wall thickness, the larger the shape correction term, indicating a more non-uniform cross-section and a greater likelihood of localized temperature differences. Multiplying the thermal inertia term with the shape correction term using a negative exponent results in greater difficulty in overall heat dissipation or a more non-uniform cross-section, thus increasing the cross-sectional geometric thermal inertia factor. The larger the value, the better.

[0050] In the lumped parameter method of heat transfer, the geometric properties of heat dissipation from an object are usually expressed by characteristic lengths. To represent, compared to Olympiad math The characteristic length is the core parameter. In this embodiment, the first term of the cross-sectional geometric thermal inertia factor is the ratio of the specific surface area to the standard reference value and a negative exponent, so that compact workpieces with smaller specific surface areas obtain larger thermal inertia values, which is suitable for the quantitative comparison of heat dissipation difficulty between workpieces of different specifications. The second term introduces the ratio of the wall thickness range to the average wall thickness, reflecting the influence of cross-sectional wall thickness differences on the internal temperature gradient and thermal stress distribution during the cooling process. That is, the greater the wall thickness difference, the steeper the internal temperature gradient of the cross-section, and the greater the resulting thermal stress.

[0051] Under another operating condition, if the ring is a uniformly thin-walled component with a small difference in wall thickness and a large specific surface area, then the cross-sectional geometric thermal inertia factor... It will approach or even be less than 1, the cross-sectional geometric thermal inertia factor The larger the value, the thicker and heavier the workpiece, and the more severe the cross-sectional distortion, making heat dissipation more difficult.

[0052] The cross-sectional geometric thermal inertia factor compresses the geometric heat dissipation characteristics of different workpieces into a dimensionless value, enabling subsequent control links to distinguish between thin-walled parts that are easy to cool and thick-walled irregular parts that are difficult to cool, thus providing a basis for differentiated fan frequency adjustment.

[0053] S104. Calculate the proportional gain term based on the deviation between the estimated temperature and the target cooling curve temperature. Construct a negative phase change damping term using the latent heat release intensity index of phase change and the cross-sectional geometric thermal inertia factor. Combine the base sustaining frequency, the proportional gain term, and the phase change damping term to generate the fan frequency command.

[0054] It should be noted that traditional PID control adjusts the fan speed solely based on temperature deviation. When the temperature is too high due to latent heat release from the phase change, it may mistakenly and drastically increase the fan speed for forced cooling, causing abnormal surface structures and thermal stress cracks. Therefore, this embodiment introduces a negative phase change damping term, composed of the latent heat release intensity index and the cross-sectional geometric thermal inertia factor, on top of proportional control. This term automatically suppresses the increase in fan speed when the phase change is intense or the workpiece geometry is complex, achieving an intelligent cooling suppression mechanism.

[0055] From the perspective of control theory, the above phase transition damping term This constitutes the feedforward compensation loop of the fan frequency conversion control system. It transforms the internal disturbance of latent heat release due to phase change, which cannot be directly measured, and the workpiece's geometric heat dissipation characteristics into a feedforward compensation signal, which is then combined with the feedback proportional control term based on temperature deviation. A feedback-feedforward composite control structure is formed. Unlike conventional feedforward compensation, which is usually designed for directly measurable external disturbances (such as ambient temperature fluctuations), this embodiment extends the source of the feedforward signal to the internal phase transition state of the material indirectly inferred through residual analysis and the heat dissipation characteristics of the workpiece calculated through geometric parameters. This embeds knowledge of metal phase transition thermodynamics into the bottom layer of the control loop, enabling the feedback-feedforward composite structure to simultaneously cope with the dual coupling of geometric features and phase transition heat release.

[0056] Specifically, the wind turbine frequency command The calculation process is as follows: first calculate the frequency adjustment amount. :

[0057] Based on this, a base sustaining frequency is superimposed and amplitude limiting is applied:

[0058] in, This is the frequency adjustment amount. The base sustaining frequency is determined based on the rated operating parameters of the air-cooled unit fan and the baseline cooling requirements corresponding to the ring material. Considering that the base sustaining frequency needs to balance the minimum cooling driving force and the adjustment margin of the phase change damping term, The value range is [20, 40] Hz, and in this embodiment, 30 Hz is used. The proportional gain characterizes the frequency adjustment magnitude corresponding to a unit temperature deviation. In this embodiment... The set value is 1.5 Hz / ℃. In other embodiments, the value can be adjusted within the range of [0.5, 3] Hz / ℃ depending on the ring material and cooling rate requirements. The target cooling curve temperature is the temperature-time relationship pre-set in the cooling process specification at the current moment. The corresponding temperature value, This is the phase transition compensation gain coefficient. There is a positive correlation between the suppression strength of the phase transition damping term, i.e. If the value is too small, the suppression effect will be insufficient to counteract the high temperature signal during the phase transition period. If the value is too high, the airflow speed will be excessively reduced, extending the cooling cycle. The value range is [50, 200]. The specific value is determined by comprehensively considering the latent heat release intensity of the ring material and the cross-sectional complexity. The latent heat release intensity index of phase change. The cross-sectional geometric thermal inertia factor; The wind field utilization coefficient reflects the actual coverage of the airflow on the surface of the ring component. It is used for situations where the airflow only covers a local area of ​​the workpiece. Take the smaller value, when the airflow completely covers the workpiece surface. It is close to 1, and its value range is [0.5, 1]. This indicates frequency limiting, which restricts the calculated wind turbine frequency command to the minimum frequency allowed by the wind turbine hardware. and maximum frequency Within the specified range, to prevent inverter overload or unexpected fan stoppage. The value range of is [0, 5]. The value range is [45, 50], and the specific value is determined according to the rated parameters of the wind turbine hardware.

[0059] When the estimated temperature is higher than the target cooling curve temperature, the proportional gain term is positive, tending to increase the fan frequency to accelerate cooling. Simultaneously, if the latent heat release intensity index of the phase change is high, the phase change damping term is also large. After the proportional gain term and the phase change damping term offset each other, the actual increase in fan frequency is suppressed. When the estimated temperature is basically the same as the target cooling curve temperature and no phase change occurs, both the proportional gain term and the phase change damping term approach zero, and the fan frequency remains near the base sustaining frequency.

[0060] It should be further explained that the wind farm utilization rate coefficient Based on the relative positions of the fan and the ring component, the effective cooling area ratio corresponding to the coverage area of ​​the airflow on the surface of the ring component is calculated. When the coverage area increases, the wind field utilization coefficient increases, thereby increasing the weight of the negative phase change damping term when calculating the fan frequency command. This is to match the requirements for thermal stress protection strength under large-area cooling conditions and prevent the overall thermal shock during phase change from being aggravated by excessive cooling area.

[0061] This explanation uses a typical control cycle of a deep-step L-section flange within the phase transformation temperature range as an example. The ambient temperature is set. The initial temperature was 25℃. The temperature is 880℃, and the overall heat transfer coefficient is... It is 0.0003s -1 Sampling period The sliding window length is 0.5s. The safety constant is 50. The value is 5, and the reference cooling rate is... The base sustaining frequency is 1℃ / s. 30Hz, proportional gain The phase transition compensation gain coefficient is 1.5 Hz / ℃. 120Hz, wind farm utilization coefficient The minimum frequency is 0.8. 0Hz, maximum frequency The frequency is 50Hz. The workpiece geometry parameters are: outer surface area... 3.2m 2 ,volume It is 0.65m 3 Maximum wall thickness Minimum wall thickness is 180mm. The specific surface area is 60 mm, and the standard reference specific surface area constant is... 5m -1 Shape sensitivity index =1, The value is 2.

[0062] First, calculate the cross-sectional geometric thermal inertia factor. :

[0063] A value greater than 1 indicates that the L-shaped flange is a type of workpiece that is difficult to dissipate heat and has severe cross-sectional distortion.

[0064] Assuming the current time is 2000s into the cooling process, within the phase transformation temperature range of austenite to pearlite. The theoretical temperature calculated by the no-phase-transformation benchmark model: ℃. Due to the release of latent heat from the phase change, the estimated temperature output by the Kalman filter is... The instantaneous cooling rate is 530℃, which is about 35.8℃ higher than the theoretical temperature. The target cooling curve temperature is -0.15℃ / s. It is 520℃.

[0065] Calculate the latent heat release intensity index of phase change To simplify the example, we assume that the root mean square value of the normalized residuals at each time point within the sliding window is the statistical characteristic value. The statistical calculation yielded a value of 0.072, indicating that the estimated temperature within the window was generally about 7.2% higher than the theoretical temperature. (Nonlinear gain term) Phase change latent heat release intensity index .

[0066] Substitute into the formula for calculating frequency adjustment: proportional gain term Hz, tending to increase the fan frequency to accelerate cooling.

[0067] Phase change damping term Hz, which suppresses the increase in wind speed.

[0068] Frequency adjustment amount Hz.

[0069] Fan frequency command Hz.

[0070] The above results indicate that although the estimated temperature was 10°C higher than the target cooling curve temperature when the phase transition occurred, the proportional control should have increased the fan frequency to 45Hz. However, the phase transition damping term offset most of the frequency increment, resulting in a fan frequency of only 15.37Hz, which effectively avoided applying excessive cooling to the workpiece during the phase transition period.

[0071] In contrast, without introducing a phase change damping term, the wind turbine frequency command would be... At frequencies close to the maximum, forced rapid cooling may cause abnormal bainitic or martensitic structures to form on the surface.

[0072] Thus, by introducing a negative phase transformation damping term that includes the latent heat release intensity index and the cross-sectional geometric thermal inertia factor into the proportional control loop, the increase in wind speed is automatically suppressed when the phase transformation is intense or the workpiece shape is complex. This prevents the overcooling caused by the misjudgment of the high temperature due to latent heat release as insufficient heat dissipation. As a result, large wind turbine ring forgings can maintain a reasonable cooling rate to obtain the target metallographic structure during the cooling process, while avoiding thermal stress cracks caused by excessive temperature difference between the surface and the interior. This ensures the smooth convergence of the overall cooling process and the consistency of product quality.

[0073] Reference Figure 2The horizontal axis represents the cooling process time axis, in minutes; the vertical axis represents the L-shaped cross-section position coordinates, in mm; and the vertical axis represents the latent heat release intensity index of the phase change. During the approximately 30 to 50 minute cooling process, a significant bulge appears on the surface near the thick-walled hot spot region, indicating a violent solid-state phase change and concentrated latent heat release in this spatiotemporal region. However, in the thin-walled edge region and the early and late stages of cooling, the surface approaches a zero plane, indicating that the phase change has not yet occurred or has been largely completed. The above analysis verifies that the latent heat release intensity index of the phase change in this embodiment can accurately locate anomalies in the phase change thermal state across the spatiotemporal dimensions.

[0074] Reference Figure 3 The horizontal axis represents the deviation between the estimated temperature and the target cooling curve temperature, and the vertical axis represents the rate at which this deviation changes over time. Figure 3 The trajectory of existing PID technology exhibits large-amplitude spiral oscillations, indicating that the fan speed and temperature fluctuate repeatedly after the phase transition, making it difficult to converge to the target control steady-state region. The trajectory of the adaptive control algorithm of this invention, after being subjected to an initial disturbance, presents a smooth convergence curve with a small physical disturbance, and enters the target control steady-state region, proving that the control system has convergent stability after introducing the phase transition damping term.

[0075] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An adaptive control method for an air-cooled system based on a waste heat attenuation model, characterized in that, include: The real-time temperature measurement value of the ring is obtained, the real-time temperature measurement value is filtered, and the estimated temperature and instantaneous cooling rate at the current moment are output. A no-phase-change baseline model is constructed, and the theoretical temperature at the current moment if no phase change occurs is calculated. Based on the deviation between the estimated temperature and the theoretical temperature and the instantaneous cooling rate, the latent heat release intensity index of phase change is calculated. The latent heat release intensity index of phase change is positively correlated with both the deviation and the instantaneous cooling rate. Obtain the geometric parameters of the ring component, and calculate the cross-sectional geometric thermal inertia factor based on the geometric parameters. The cross-sectional geometric thermal inertia factor is positively correlated with the heat dissipation difficulty and cross-sectional distortion degree of the ring component. The proportional gain term is calculated based on the deviation between the estimated temperature and the target cooling curve temperature. A negative phase change damping term is constructed using the latent heat release intensity index of phase change and the cross-sectional geometric thermal inertia factor. The basic sustaining frequency, the proportional gain term, and the phase change damping term are combined to generate the fan frequency command.

2. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 1, characterized in that, The filtering process is implemented using a discrete Kalman filter state observer, including: defining a system state vector containing the estimated temperature and instantaneous cooling rate; based on the physical laws of heat conduction, using the product of the estimated temperature at the previous moment and the instantaneous cooling rate at the previous moment and the sampling period as the prior estimate for the current moment; using the real-time temperature measurement value at the current moment to correct the prior estimate, with the goal of minimizing the estimation error covariance, and outputting the estimated temperature and instantaneous cooling rate for the current moment.

3. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 1, characterized in that, The real-time temperature measurement of the ring component is obtained by multiple sets of infrared thermal imagers arranged in the air-cooling station. The infrared thermal imagers are evenly arranged along the circumference of the ring component, and the monitoring area covers the inner corner of the ring component's cross-section.

4. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 1, characterized in that, The phase-change-free benchmark model is constructed based on the exponential decay law. It uses the ambient temperature as the asymptotic lower limit and the difference between the initial temperature and the ambient temperature as the base. It calculates the theoretical temperature at the current moment according to the law of exponential decay over time. The rate of decay is determined by the comprehensive heat transfer coefficient.

5. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 1, characterized in that, The calculation of the latent heat release intensity index of the phase change includes: A fixed-length historical sliding window is extracted from the current time. The residual between the estimated temperature and the theoretical temperature at each time point within the historical sliding window is calculated. The residual is normalized and processed by root mean square to obtain statistical characteristic values ​​that characterize the degree of temperature deviation. A nonlinear gain term is constructed for the instantaneous cooling rate. The statistical characteristic value is multiplied by the nonlinear gain term to obtain the latent heat release intensity index of the phase change.

6. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 5, characterized in that, The nonlinear gain term adopts a logarithmic function form, which makes the growth rate of the nonlinear gain term gradually decrease as the instantaneous cooling rate increases.

7. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 1, characterized in that, The geometric parameters include the outer surface area, volume, maximum wall thickness, and minimum wall thickness of the ring.

8. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 1, characterized in that, The calculation of the cross-sectional geometric thermal inertia factor includes: constructing a thermal inertia term reflecting the overall heat dissipation difficulty using the ratio of the outer surface area to the volume of the ring component; constructing a shape correction term reflecting the degree of cross-sectional distortion using the ratio of the cross-sectional wall thickness range to the average wall thickness; multiplying the thermal inertia term with the shape correction term in a positive exponential form using a negative exponent to obtain the cross-sectional geometric thermal inertia factor, such that the smaller the specific surface area or the greater the difference in cross-sectional wall thickness, the larger the value of the cross-sectional geometric thermal inertia factor.

9. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 1, characterized in that, The generation of the fan frequency command includes: calculating a positive frequency increment based on the difference between the estimated temperature and the target cooling curve temperature using proportional control logic; multiplying the phase change latent heat release intensity index, the cross-sectional geometric thermal inertia factor, the wind field utilization coefficient, and the phase change compensation gain coefficient used to match the latent heat release intensity and cross-sectional complexity of the ring material to obtain a negative frequency adjustment amount; adding the positive frequency increment to the base maintenance frequency and then subtracting the negative frequency adjustment amount, and performing amplitude limiting processing on the calculation result to obtain the fan frequency command.

10. The adaptive control method for an air-cooled system based on a waste heat attenuation model according to claim 9, characterized in that, The wind field utilization coefficient is determined based on the relative position of the fan and the ring component, specifically by calculating the effective cooling area ratio corresponding to the coverage of the airflow on the surface of the ring component.

Images