Flywheel post-machining thermal deformation residual stress release and return control system

Abstract

The application discloses a flywheel post-processing thermal deformation residual stress release and return control system and method, relates to the advanced manufacturing and industrial artificial intelligence control technical field, and the steps of the method comprise the following steps: obtaining flywheel initial compressive stress and asynchronous space-time sensing data in sequence, and extracting a thermal hysteresis dislocation index, a warping misjudgment separation degree and a targeted compressive stress retention rate in parallel; the above-mentioned multi-source features are injected into a waveform optimization model to calculate a coupling degradation flow form degree, and a variable-frequency pulse heating waveform is output by combining a anti-whip bias matrix inversion; finally, the reciprocal of the flow form degree is monotonically integrated and mapped to generate a return position confidence index, so as to trigger the system to stop accurately. The application effectively avoids the structural resonance caused by high-frequency thermal pulses, avoids the indiscriminate dissipation of the preset compressive stress resisting centrifugal force, and realizes the safe release of residual stress and the size return.

Description

Technical Field

[0001] This invention relates to the field of advanced manufacturing and industrial artificial intelligence control technology, specifically to a control system for the release and return of residual stress from thermal deformation after flywheel processing. Background Technology

[0002] During the machining and forming process of large flywheels, residual processing stresses with extremely uneven distribution inevitably arise within them. These stresses must be relieved and dimensionally corrected through subsequent heat treatment processes. However, flywheels typically exhibit abrupt changes in cross-sectional geometry, with a thick outer rim and a thin central spoke, making them highly susceptible to anisotropic thermal responses during heat treatment. This inherent difference in heat capacity transforms the traditional stress relief process into a strongly coupled, multidimensional, nonlinear process involving heat transfer, stress dissipation, and macroscopic deformation. This macroscopically, it can easily lead to uncontrollable spatiotemporal misalignment and structural degradation risks.

[0003] To address the aforementioned residual stress elimination and deformation return control, existing technologies typically employ isothermal heat treatment schemes based on static process curves. Specifically, these methods often rely on surface thermocouples to collect single-dimensional temperature data and combine this with conventional PID algorithms to adjust the heating power, attempting to achieve slow creep release of internal stress through long-term global heat preservation. Simultaneously, some processes utilize rigid pressure plates and other tooling to apply multi-dimensional solid constraints to the flywheel, aiming to forcibly suppress its axial warping and force the outer diameter dimension to return to its original position during the cooling phase.

[0004] However, the solutions of the aforementioned existing technologies still suffer from the following extremely serious technical defects when dealing with complex multi-scale and multi-dimensional entity evolution: First, there is a risk of stress reorganization caused by thermal response misalignment. Due to the lack of dynamic characteristic inversion of the deep transient heat transfer differences between the rim and the spokes, single surface temperature measurement often leads the control system to misjudge the internal heat penetration state, thereby causing severe temperature difference stretching in the temperature rise and fall transition zone, resulting in secondary thermal stress tearing; Second, there is a problem of indiscriminately eliminating beneficial pre-stress. Traditional methods pursue the ultimate stress zeroing or blindly... The insulation creep was ignored, and the real-time monitoring of the high-value pre-set compressive stress on the flywheel surface used to resist high-speed centrifugal force was neglected. This caused the beneficial stock to be excessively dissipated under the conventional algorithm, which greatly reduced the fatigue life of the flywheel. Third, there is a loophole in the coupling of high-frequency resonance and multi-dimensional deformation. Conventional one-dimensional radial displacement monitoring cannot remove the three-dimensional saddle-shaped warping artifacts generated during the heating and recovery process. Moreover, the fixed frequency heating pulse is very easy to excite the low-order modal resonance of the thin spokes. This makes the final dimensional return and shutdown decision based on the severely contaminated distorted data.

[0005] Therefore, the present invention provides a control system for the release and return of residual stress caused by thermal deformation after flywheel processing. Summary of the Invention

[0006] The purpose of this invention is to provide a control system for the release and return of residual stress from thermal deformation after flywheel processing, so as to solve the existing problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a control system for the release and return of residual stress from thermal deformation after flywheel machining, comprising: The multi-source asynchronous spatiotemporal data acquisition module is configured to acquire the initial compressive stress before the flywheel processing and real-time asynchronous thermal field and deformation sensing data. The multi-scale feature decoupling and extraction module is configured to run multiple computational branches in parallel based on the sensor data to extract the anisotropic thermal response hysteresis index, warpage interference distortion degree and target compressive stress retention rate that characterize the internal state of the flywheel. The waveform optimization model construction module is configured to inject the sub-parameters extracted by the multi-scale parameter decoupling and extraction module into the information-constrained leakage echo state network to construct a waveform optimization model, calculate the evolutionary manifold reflecting the thermo-mechanical coupling state, and generate a frequency conversion pulse heating waveform with the goal of minimizing the evolutionary manifold in the next moment. The state evolution monotonic mapping and homing decision module is configured to perform time-axis integral mapping on the evolutionary manifold, generate a monotonically increasing homing confidence index, and output a system shutdown command based on the index.

[0008] A further improvement of the present invention is that the multi-source asynchronous spatiotemporal data acquisition module further includes: The static constitutive parameter analysis unit is configured to obtain the pre-set initial compressive stress on the outer surface of the flywheel via the industrial control interface. The thermal inertia time constant was extracted based on the constitutive equation of the flywheel material. and rated circumferential tensile stress ; An asynchronous concurrent ring buffer unit is configured to receive signals with a first sampling frequency. Real-time rim-mapped temperature data, and data with a second sampling frequency. Real-time measurement of total deformation and rotation phase angle data, among which The first and second sampling frequencies are used to mark the corresponding data streams with independent hardware timestamps and perform concurrent caching.

[0009] A further improvement of the present invention is that the multi-scale feature decoupling and extraction module further includes: The asynchronous stress prevention unit is configured to calculate the inversion temperature of the deep layer of the spokes based on historical rim temperature data and an inverse heat conduction algorithm, and perform integral calculations using the thermal inertia time constant to output a thermal hysteresis misalignment index. Quantitatively characterize and prevent the risk of structural tearing caused by misalignment of heat capacity between the rim and the spokes; The warp interference decoupling unit is configured to perform modal cosine projection calculations based on the real-time measured total deformation and rotation phase angle, separating the axial warp component in three-dimensional space from the one-dimensional measurement data and outputting the warp interference distortion degree. With pure radial shrinkage rate ; The compressive stress targeted retention unit is configured to calculate the tensile stress release equivalent based on the cumulative heat input history, and, in conjunction with the initial compressive stress and the rated circumferential tensile stress, calculate the targeted compressive stress retention rate. .

[0010] A further improvement of the present invention is that the waveform optimization model construction module further includes: The multi-scale feature mapping injection unit is configured to map and inject a ternary state feature vector containing the thermal hysteresis misalignment index, warpage interference distortion degree and target compressive stress retention rate into the neuron node region with different leakage rate parameters in the reservoir of the waveform optimization model of the base leakage echo state network. The evolutionary manifold solver is configured to extract the dynamic state matrix of the reservoir and calculate the triaxial field coupled stress evolutionary manifold. ; The dynamic penalty optimization and waveform inversion unit is configured to calculate the predicted manifold based on the ridge regression algorithm. The network weights are minimized, and a three-dimensional non-standard heating waveform containing pulse duty cycle, carrier frequency, and dead time is output to the actuator based on the updated network weights.

[0011] A further improvement of this invention is that, in the multi-scale feature mapping injection unit, the topology of the reservoir in the leakage echo state network is configured as follows: It includes a first neuron group whose leakage rate is less than a first preset threshold, and the data input of the first neuron group is related to the retention rate of the target compressive stress that characterizes the slow evolution. Corresponding and connected; The second neuron group contains a leakage rate less than or equal to a second preset threshold and greater than or equal to a first preset threshold. The data input terminal of the second neuron group is connected to the anisotropic thermal response hysteresis index. Corresponding and connected; It includes a third neuron group whose leakage rate is greater than a second preset threshold, and the data input terminal of the third neuron group is connected to the warping interference distortion degree characterizing high-frequency evolution. Corresponding and connected; Wherein, the second preset threshold is greater than the first preset threshold.

[0012] A further improvement of this invention is that, in the evolutionary manifold solution unit, the evolutionary manifold... The computational model is expressed as follows: in, , , For the preset regularization weights, The evolution gradient of the thermal hysteresis dislocation index. The evolution gradient of the target compressive stress retention rate; This is a cross-coupled multiplication penalty operator.

[0013] A further improvement of the present invention is that the dynamic penalty optimization and waveform inversion unit further includes: Calculate the warping interference distortion degree The first derivative is used as the resonance coefficient, and the output of the cross-coupling multiplication penalty operator is monitored and denoted as the mutation coefficient. When the resonance coefficient is greater than a set resonance critical value, or the mutation coefficient is greater than a set mutation threshold, a coefficient related to the first derivative is introduced into the objective function of the ridge regression algorithm. The dynamic penalty bias matrix with negative correlation of evolution gradient is used; the network weights are updated based on the objective function after applying the dynamic penalty bias matrix, and the carrier frequency contained in the non-standard heating waveform generated by constraint inversion is configured to avoid the inherent frequency range of the flywheel structure.

[0014] A further improvement of this invention is that the specific configuration of the state evolution monotonic mapping and homing decision module includes: obtaining the pure radial contraction rate. And based on the pure radial shrinkage rate The fluctuation variance generates a deformation stability penalty function ; for the evolutionary manifold The reciprocal of the integral is integrated along the time axis, and the integral result is multiplied by the deformation stability penalty function. Obtain the dynamic safety return location confidence index ; Monitor the dynamic safety return location confidence index When its calculation cycle is greater than or equal to the preset safety threshold for N consecutive cycles, a system shutdown command is triggered.

[0015] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described control system for the release and return of residual stress from thermal deformation after flywheel machining.

[0016] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned control system for the release and return of residual stress from thermal deformation after flywheel machining.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention first introduces a multi-scale, multi-dimensional entity feature decoupling and extraction module to extract thermal hysteresis misalignment index, warpage misjudgment separation degree, and targeted compressive stress retention rate in parallel. This solves three major problems in traditional processes: secondary stress tearing caused by asynchronous thermal response between the rim and spokes, misjudging beneficial pre-compression stress on the surface by indiscriminate insulation, and misjudging three-dimensional warpage artifacts as one-dimensional radial dimensional shrinkage. This invention achieves multi-dimensional accurate quantification of the complex thermodynamic state of the flywheel and perfect separation of spatial distortion interference, ensuring that the system can be controlled based on pure, real, objective multi-dimensional entity parameters, and significantly improving the fatigue life and dimensional return accuracy of the finished product.

[0018] 2. By constructing a waveform optimization model and combining it with an anti-excitation constraint bias matrix model, the divergence of the optimization algorithm caused by strong coupling across time scales and the control problem of low-order modal resonance of the flywheel structure easily excited by high-frequency thermal pulses were solved. Forced carrier frequency cutting for vibration avoidance and damping convergence were achieved when generating the inverted heating waveform. At the same time, by performing a definite integral monotonic mapping on the coupling deterioration manifold, the deterioration tension of transient fluctuations was transformed into a monotonic safety confidence, which solved the shutdown misjudgment caused by local high-frequency disturbances and ensured the safety of the release and recovery process. Attached Figure Description

[0019] Figure 1 This is a framework diagram of a control system for the release and return of residual stress from thermal deformation after flywheel processing according to the present invention; Figure 2 This is a dynamic response curve of the multi-scale decoupling and anti-vibration control of the present invention. Detailed Implementation

[0020] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0021] The term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone.

[0022] Example 1 Figure 1This embodiment illustrates a framework diagram of a control system for the release and return of residual stress from thermal deformation of a flywheel after machining. The system includes: The module includes a multi-source asynchronous spatiotemporal data acquisition module, a multi-scale feature decoupling and extraction module, a waveform optimization model construction module, and a state evolution monotonic mapping and retracement decision module. In conventional flywheel heat treatment processes, the control system often relies solely on surface temperature sensors and statically set heating and cooling curves for open-loop or simple proportional-integral-derivative (PI-DE) adjustments. This conventional approach ignores the initial stress state left over from the flywheel processing, and the sensor's data acquisition dimensions are extremely limited. This results in a severe disconnect between the control commands and the actual internal stress state of the objective structure, easily leading to localized material yielding or structural fracture during the stress release process.

[0023] To overcome the blind spots caused by static curves and single sensors in conventional methods, this solution introduces a multi-source asynchronous spatiotemporal data acquisition module. This module is configured to acquire the initial compressive stress before flywheel machining, as well as real-time asynchronous thermal field and deformation sensing data. Specifically, this module includes a static constitutive parameter analysis unit, which reads the preceding turning feed rate and machining pattern from the manufacturing execution system through the industrial control interface, thereby obtaining the preset initial compressive stress on the outer surface of the flywheel. Simultaneously, the thermal inertia time constant is extracted based on the material constitutive equation of the flywheel. and rated circumferential tensile stress Preferably, the rated circumferential tensile stress Based on the flywheel's ultimate burst speed and centrifugal load, the values ​​can be calculated and set by those skilled in the art, and used as a benchmark reference for subsequent safety boundaries.

[0024] In addition, the multi-source asynchronous spatiotemporal data acquisition module also includes an asynchronous concurrent ring buffer unit, which receives data with a first sampling frequency. Real-time rim-mapped temperature data, and data with a second sampling frequency. The system measures the total deformation and rotation phase angle data in real time. In actual operation, the response to thermal field changes is slow; therefore, the first sampling frequency... The preferred setting is 5 Hz to 10 Hz; and the response of deformation flutter is extremely fast, with the second sampling frequency... The preferred sampling frequency is between 2 kHz and 5 kHz. This unit marks the corresponding data stream with independent hardware timestamps and performs concurrent caching according to the different sampling frequencies mentioned above. This asynchronous concurrent design is different from the forced resampling alignment method in conventional data fusion, thus completely preserving the original frequency domain properties of high-frequency warp modes and low-frequency heat conduction, providing a distortion-free underlying data foundation for subsequent multi-scale decoupling.

[0025] After acquiring the aforementioned multi-source asynchronous data, the system enters the state extraction phase. Conventional surface temperature measurement schemes face the challenge of stress reorganization caused by the asynchronous thermal responses of the rim and spokes. Flywheels are typically designed with a thick outer rim to store kinetic energy and thin, lightweight spokes in the middle to reduce weight. During the heat release and cooling process, the thin spokes have a small heat capacity and contract extremely quickly, while the thick rim has a large heat capacity and expands continuously. If a conventional system only monitors the average surface temperature and assumes that heat release is complete, accelerating cooling will introduce extremely dangerous radial tensile stress at the transition radius between the two, potentially leading to rotor breakage.

[0026] To address the aforementioned issues, this solution introduces a multi-scale parameter decoupling and extraction module. This module includes an asynchronous stress prevention unit, configured to calculate the inversion temperature of the deep layers of the spokes based on historical rim temperature data and an inverse heat conduction algorithm. It then performs integral calculations using the aforementioned thermal inertia time constant to output the thermal hysteresis misalignment index. Its calculation formula is expressed as: in, Let t be the integration time variable, and t be the current total running time. The temperature is the surface temperature of the rim. The deep inversion temperature of the spokes is obtained through the inverse algorithm. The target temperature is set for the current stage. This index normalizes the absolute value of the transient temperature difference with the target temperature and integrates it over time with the reciprocal of the thermal inertia time constant as the decay weight. It can quantitatively characterize and prevent the risk of structural tearing caused by the misalignment of the heat capacity between the rim and the spokes.

[0027] At the same time, the flywheel is also prone to three-dimensional saddle-shaped spatial warping during the heat release process. Conventional systems often misinterpret this axial warping projection with rotational phase as radial contraction with a smaller diameter, and then apply incorrect local heating feedback to forcibly flatten it, resulting in extremely high bending potential energy being locked inside the flywheel.

[0028] To eliminate the aforementioned types of coupling misjudgments, the warp interference decoupling unit is triggered. Based on the real-time measured total deformation and rotation phase angle, it performs modal cosine projection calculation, separating the axial warp component in three-dimensional space from the one-dimensional measurement data and outputting the warp interference distortion degree. Its calculation formula is expressed as: ,in, The axial warping spatial amplitude is separated by the modal matrix. The rotation phase angle is obtained through a rotary encoder. The total deformation measured in real time by the sensor.

[0029] Using the above method, the system successfully separated the distortion interference term and simultaneously output the pure radial shrinkage rate after filtering out the interference term. This prevents the system from misjudging the size return due to dimensional degradation.

[0030] Furthermore, traditional stress relief algorithms pursue an extreme zero-stress state. This indiscriminate elimination can inadvertently eliminate beneficial compressive stresses pre-installed in the flywheel during the initial manufacturing process to resist centrifugal force, leading to a significant decrease in the flywheel's fatigue life. To avoid this drawback, the compressive stress targeted retention unit calculates the equivalent tensile stress release based on the accumulated heat input history. and combined with initial compressive stress and rated circumferential tensile stress Calculate the target compressive stress retention rate Its calculation formula is expressed as This formula monitors the ratio of remaining beneficial stress to ultimate burst load in real time, preventing the surface prestress resisting centrifugal force from being excessively dissipated under conventional elimination algorithms.

[0031] However, after obtaining the above three key state parameters, in order to reduce the thermal hysteresis dislocation index, the system usually needs to apply high-frequency pulse heating to wait for internal heat conduction; however, the thermal shock wave brought by the high-frequency pulse will excite the low-order mode resonance of the thin plate, resulting in a sharp deterioration of the warping interference distortion; at the same time, the transient thermal shock peak will cause local yielding of the material, resulting in a cliff drop in the target compressive stress retention rate.

[0032] Conventional optimization algorithms are prone to numerical solution explosion when faced with the strong coupling contradictions across multiple time scales, including millisecond-level deformation, second-level heat transfer, and hour-level stress loss. Therefore, this invention introduces a waveform optimization model construction module, which includes a multi-scale parameter mapping injection unit configured to inject parameters including thermal hysteresis misalignment index. Warp interference distortion and the retention rate of targeted compressive stress The ternary state parameter vectors are respectively mapped to the neuron node regions with different leakage rate parameters injected into the information-constrained leakage echo state network (PI-LIESN) reservoir.

[0033] The specific configuration is as follows: the storage pool includes: The first neuron group has a leakage rate that is extremely small and less than a first preset threshold, preferably set between 0.01 and 0.05. The data input of this neuron group is related to the retention rate of the target compressive stress, which characterizes the slow evolution over several hours. Corresponding connections endow it with long-term memory capabilities; The second neuron group has a leakage rate less than or equal to a second preset threshold and greater than or equal to a first preset threshold, preferably set between 0.05 and 0.8. This input is related to the anisotropic thermal response hysteresis index that evolves on a second-by-second basis. Corresponding and connected; The third neuron group has a leakage rate greater than the second preset threshold, preferably set between 0.8 and 0.99, approaching a state of transient forgetting. This input is related to the warping interference distortion degree, which characterizes millisecond-level high-frequency evolution. Corresponding and connected.

[0034] This topology configuration enables the time constants of nodes within the network to form an isomorphic mapping with the different thermal evolution cycles of objective entities, which can resolve the computational crash problem caused by multi-scale asynchronous data.

[0035] Subsequently, the evolutionary manifold solving unit extracts the dynamic state matrix of the reservoir and calculates the triaxial field coupled stress evolutionary manifold. Its model is represented as: .in, , , The preset regularization weights are preferably set to 0.3, 0.5, and 0.2, respectively. These values ​​tend to favor the protection of compressive stress retention rate and structural integrity. This represents the evolution gradient of the thermal hysteresis dislocation index at the current time step. The evolution gradient of the target compressive stress retention rate; For cross-coupled multiplication penalty operators, To penalize the amplification factor, it is preferably set to twice the base of the natural constant. The design principle is that when the control action attempts to forcibly reduce the temperature difference with extreme thermal shock, but causes the beneficial compressive stress to undergo a gradient jump yielding, the gradient inner product of the two increases rapidly. Through exponential calculation, the manifold exhibits an explosive growth, thereby negating this kind of collapse control at the numerical calculation level.

[0036] In some possible embodiments, such as in the specific application scenario of manufacturing heavy-duty composite energy storage flywheels, the spokes, made of a carbon fiber and metal composite structure, are extremely sensitive to high-frequency excitation. In this case, the intervention of the dynamic penalty optimization and waveform inversion unit is particularly crucial. This unit calculates the warpage interference distortion degree. The first time derivative as the resonance coefficient ,Right now Simultaneously, monitor the output of the cross-coupling multiplication penalty operator and record it as the mutation coefficient. ,Right now When the resonance coefficient Greater than the resonance critical value or abrupt change coefficient pre-set according to modal analysis. When the abrupt change threshold is greater than the threshold set according to the yield strength, it indicates that the current heat input is stimulating structural resonance or causing material softening.

[0037] At this point, this unit will introduce a relation to the objective function of the ridge regression algorithm. The dynamic penalty bias matrix is ​​used to negatively correlate evolutionary gradients. This matrix acts as an algebraic repulsion term, updating network weights based on the objective function after applying the dynamic penalty bias matrix. This forces the optimization direction away from the current dangerous solution space. The loss formula for the objective function after incorporating this matrix can be expressed as: in, This represents the objective function loss value that the waveform optimization model aims to minimize within this iteration period; The label matrix represents the expected evolution state of the system, which is used to guide the safe convergence target of various indicators to stabilize. This represents the dynamic state matrix output from the aforementioned multi-scale parametric mapping injection unit to the leakage echo state network reservoir. This represents the weight matrix of the network layer to be solved and updated, which ultimately determines the variable frequency pulse heating waveform parameters output to the actuator; This is a standard prediction fitting error term, used to constrain how closely the network output approximates the desired state; The preset base regularization coefficient, and These are combined to form the basic penalty term, used to prevent the network weights from overfitting; This represents the resonance coefficient currently monitored by the system, and its value is equal to the warping interference distortion degree. The time-order first derivative is used to quantify the swiftness of the current modal divergence; Indicates the retention rate of targeted compressive stress. Temporal evolution gradient; For the preset negative correlation mapping operator, configured as when When the trend is downward, a proportionally amplified positive penalty base is output to characterize the objective state that the beneficial preload stress is facing excessive dissipation. Together, they constitute the dynamic penalty bias matrix terms used to prevent structural resonance and material softening.

[0038] In conventional open-loop or constant-parameter control systems, high-frequency thermal pulses often output blindly, easily causing the actual control action to overlap with the inherent response frequency of the physical entity. Based on the objective function constructed above, when the industrial control system detects an intensifying trend of spatial resonance (i.e., (significantly increased) and the beneficial preload stress faces loss (i.e., mapping operator) When the output value increases dramatically, this dynamic penalty bias matrix term applies to the entire objective function. The algebraic weighting in the calculation exhibits a non-linear surge. This algebraic exclusion mechanism forces the waveform optimization model to actively abandon the current weight update path that causes distortion at the underlying architecture of the optimization calculation, thereby enabling the carrier frequency generated in the final inversion to actively avoid the corresponding risk blind zone. This process transforms the structural resonance disaster into a quantifiable and operable algebraic penalty term, achieving coordinated control of anti-vibration and anti-yielding, and effectively ensuring the safety and stability of the residual stress release process.

[0039] Ultimately, this unit enables the prediction of manifold. With minimization as the objective, the updated network weight mapping outputs a three-dimensional non-standard heating waveform containing pulse duty cycle, carrier frequency, and dead time to the actuator. This restricts the generated carrier frequency to actively avoid the flywheel's inherent frequency range, achieving coordinated control of anti-vibration and anti-yield.

[0040] Figure 2 The dynamic response curves of multi-scale decoupling and anti-vibration control are shown, representing the changes in radial displacement sensor readings during high-frequency pulse heating. Figure 2 As shown, in the conventional approach without the introduction of the anti-vibration constraint bias matrix (i.e., the thin gray solid line in the figure), due to the aliasing of high-frequency thermal pulses with the low-order modes of the structure, the mixed deformation amount, including warping distortion, read by the radial displacement sensor exhibits obvious divergent oscillation amplification. However, in the solution provided in this embodiment, when the system reaches the preset trigger point (i.e., the 5th second in the figure), the system detects the excitation tendency and instantly introduces the anti-vibration constraint bias matrix into the objective function.

[0041] As can be seen from the thick black line in the figure, after the bias matrix is ​​triggered, the system quickly performs carrier frequency switching to avoid the resonant frequency band, transforming the previously uncontrollable divergent oscillations into damped rapid convergence. More importantly, after an extremely short decay period, the control curve smoothly matches the pure radial contraction rate stripped away by the warped interference decoupling unit. (i.e., the thick black dashed line in the figure). The above comparison intuitively demonstrates the control stability of this system in filtering out spatial distortion interference and curbing structural resonance divergence at its source.

[0042] During waveform optimization and control execution, the final shutdown decision becomes an issue. Local thermal disturbances during heating or high-frequency noise from sensors can affect the evolving manifold. The system exhibits severe non-monotonic fluctuations in the transient state. If the completion of the recovery is judged directly based on the manifold or size at a single moment, it can easily lead to premature system shutdown, preventing the residual stress from being fully released. Therefore, the fluctuating transient parameters must be transformed into globally accumulated evaluation parameters.

[0043] The state evolution monotonic mapping and homing decision module is triggered, which obtains the pure radial contraction rate separated by the deformation interference decoupling unit. And based on the pure radial shrinkage rate The variance of the fluctuation within a set sliding time window generates a deformation stability penalty function. The range of this function is limited to 0 and 1, and its value approaches zero when the size fluctuates drastically.

[0044] Next, the state evolution monotonic mapping and homing decision module performs monotonic mapping calculations on the evolutionary manifold degree. The reciprocal of the integral is integrated along the time axis, and the result is multiplied by the deformation stability penalty function. Obtain the dynamic safety return location confidence index Its calculation formula is expressed as Where K is the system sensitivity gain constant, preferably 0.05; the integral part accumulates the safe state duration throughout the entire processing history.

[0045] Because integral operations inherently possess smoothing and cumulative effects, coupled with the exponential inversion design, the confidence index is necessarily a monotonically increasing parameter between 0 and 100%. Only when the deformation depth converges... When the confidence level is close to 1, it approaches the true cumulative value. The system continuously monitors this dynamic safe return location confidence index. When the number of consecutive calculation cycles (preferably 100 cycles) is greater than or equal to the preset safety threshold (preferably 99.5%), a system shutdown command is triggered, thereby ensuring the reliable completion of the flywheel residual stress release and dimensional return process from a low-bias global perspective.

[0046] Example 2 This embodiment provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the aforementioned control system for the release and return of residual stress from thermal deformation after flywheel machining by calling the computer program stored in the memory.

[0047] The electronic device can vary considerably depending on its configuration or performance. It may include one or more processors (Central Processing Units, CPUs) and one or more memories, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement the flywheel machining thermal deformation residual stress release and return control system provided in the above-described method embodiment. The electronic device may also include other components for implementing the device's functions; for example, it may have wired or wireless network interfaces and input / output interfaces for data input and output. Details will not be elaborated upon in this embodiment.

[0048] Example 3 This embodiment proposes a computer-readable storage medium on which an erasable and rewritable computer program is stored. When the computer program runs on the computer device, it causes the computer device to execute the aforementioned control system for the release and return of residual stress from thermal deformation after flywheel processing.

[0049] For example, computer-readable storage media can be read-only memory (ROM), random access memory (RAM), compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage devices.

[0050] The threshold and weight settings involved in this embodiment can be set by default according to the present invention, or can be set by those skilled in the art.

[0051] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0052] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0053] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0054] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0055] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A control system for the release and return of residual stress from thermal deformation after flywheel machining, characterized in that: The system includes: The multi-source asynchronous spatiotemporal data acquisition module is configured to acquire the initial compressive stress before the flywheel processing and real-time asynchronous thermal field and deformation sensing data. The multi-scale feature decoupling and extraction module is configured to run multiple computational branches in parallel based on the sensor data to extract the anisotropic thermal response hysteresis index, warpage interference distortion degree and target compressive stress retention rate that characterize the internal state of the flywheel. The waveform optimization model construction module is configured to inject the sub-parameters extracted by the multi-scale parameter decoupling and extraction module into the information-constrained leakage echo state network to construct a waveform optimization model, calculate the evolutionary manifold reflecting the thermo-mechanical coupling state, and generate a frequency conversion pulse heating waveform with the goal of minimizing the evolutionary manifold in the next moment. The state evolution monotonic mapping and homing decision module is configured to perform time-axis integral mapping on the evolutionary manifold, generate a monotonically increasing homing confidence index, and output a system shutdown command based on the index.

2. The control system for residual stress release and return after thermal deformation of a flywheel after machining, as described in claim 1, is characterized in that: The multi-source asynchronous spatiotemporal data acquisition module further includes: The static constitutive parameter analysis unit is configured to obtain the pre-set initial compressive stress on the outer surface of the flywheel via the industrial control interface. The thermal inertia time constant was extracted based on the constitutive equation of the flywheel material. and rated circumferential tensile stress ; An asynchronous concurrent ring buffer unit is configured to receive signals with a first sampling frequency. Real-time rim-mapped temperature data, and data with a second sampling frequency. Real-time measurement of total deformation and rotation phase angle data, among which The first and second sampling frequencies are used to mark the corresponding data streams with independent hardware timestamps and perform concurrent caching.

3. The control system for residual stress release and return after thermal deformation of a flywheel after machining, as described in claim 1, is characterized in that: The multi-scale feature decoupling and extraction module further includes: The asynchronous stress prevention unit is configured to calculate the inversion temperature of the deep layer of the spokes based on historical rim temperature data and an inverse heat conduction algorithm, and perform integral calculations using the thermal inertia time constant to output a thermal hysteresis misalignment index. Quantitatively characterize and prevent the risk of structural tearing caused by misalignment of heat capacity between the rim and the spokes; The warp interference decoupling unit is configured to perform modal cosine projection calculations based on the real-time measured total deformation and rotation phase angle, separating the axial warp component in three-dimensional space from the one-dimensional measurement data and outputting the warp interference distortion degree. With pure radial shrinkage rate ; The compressive stress targeted retention unit is configured to calculate the tensile stress release equivalent based on the cumulative heat input history, and, in conjunction with the initial compressive stress and the rated circumferential tensile stress, calculate the targeted compressive stress retention rate. .

4. The control system for residual stress release and return after thermal deformation of a flywheel after machining, as described in claim 3, is characterized in that: The waveform optimization model construction module further includes: The multi-scale feature mapping injection unit is configured to map and inject a ternary state feature vector containing the thermal hysteresis misalignment index, warpage interference distortion degree and target compressive stress retention rate into the neuron node region with different leakage rate parameters in the reservoir of the waveform optimization model of the base leakage echo state network. The evolutionary manifold solver is configured to extract the dynamic state matrix of the reservoir and calculate the triaxial field coupled stress evolutionary manifold. ; The dynamic penalty optimization and waveform inversion unit is configured to calculate the predicted manifold based on the ridge regression algorithm. The network weights are minimized, and a three-dimensional non-standard heating waveform containing pulse duty cycle, carrier frequency, and dead time is output to the actuator based on the updated network weights.

5. The control system for residual stress release and return after thermal deformation of a flywheel after machining, as described in claim 4, is characterized in that: In the multi-scale feature mapping injection unit, the topology of the reservoir in the leakage echo state network is configured as follows: It includes a first neuron group whose leakage rate is less than a first preset threshold, and the data input of the first neuron group is related to the retention rate of the target compressive stress that characterizes the slow evolution. Corresponding and connected; The second neuron group contains a leakage rate less than or equal to a second preset threshold and greater than or equal to a first preset threshold. The data input terminal of the second neuron group is connected to the anisotropic thermal response hysteresis index. Corresponding and connected; It includes a third neuron group whose leakage rate is greater than a second preset threshold, and the data input terminal of the third neuron group is connected to the warping interference distortion degree characterizing high-frequency evolution. Corresponding and connected; Wherein, the second preset threshold is greater than the first preset threshold.

6. The control system for residual stress release and return after thermal deformation of a flywheel after machining, as described in claim 5, is characterized in that: In the evolutionary manifold solution unit, the evolutionary manifold... The computational model is expressed as follows: in, , , For the preset regularization weights, The evolution gradient of the thermal hysteresis dislocation index. The evolution gradient of the target compressive stress retention rate; This is a cross-coupled multiplication penalty operator.

7. The control system for residual stress release and return after thermal deformation of a flywheel after machining, as described in claim 4, is characterized in that: The dynamic penalty optimization and waveform inversion unit further includes: Calculate the warping interference distortion degree The first derivative is used as the resonance coefficient, and the output of the cross-coupling multiplication penalty operator is monitored and denoted as the mutation coefficient. When the resonance coefficient is greater than a set resonance critical value, or the mutation coefficient is greater than a set mutation threshold, a coefficient related to the first derivative is introduced into the objective function of the ridge regression algorithm. The dynamic penalty bias matrix with negative evolution gradient correlation is used; the network weights are updated based on the objective function after applying the dynamic penalty bias matrix, and the carrier frequency contained in the non-standard heating waveform generated by constraint inversion is configured to avoid the inherent frequency range of the flywheel structure.

8. The control system for residual stress release and return after thermal deformation of a flywheel after machining, as described in claim 1, is characterized in that: The specific configuration of the state evolution monotonic mapping and homing decision module includes: obtaining the pure radial contraction rate. And based on the pure radial shrinkage rate The fluctuation variance generates a deformation stability penalty function ; for the evolutionary manifold The reciprocal of the integral is integrated along the time axis, and the integral result is multiplied by the deformation stability penalty function. Obtain the dynamic safety return location confidence index ; Monitor the dynamic safety return location confidence index When its calculation cycle is greater than or equal to the preset safety threshold for N consecutive cycles, a system shutdown command is triggered.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements a control system for the release and return of residual stress from thermal deformation after flywheel processing, as claimed in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a control system for the release and return of residual stress from thermal deformation after flywheel processing, as claimed in any one of claims 1-8.

Images