Intelligent control method and system for wind power bearing machining

Abstract

The present application relates to the technical field of manufacturing and industrial process control, and particularly relates to an intelligent control method and system for wind power bearing machining. The method comprises the following steps: obtaining monitoring data in a laser cladding process; determining a predicted temperature at a next time based on a temperature dynamic prediction model and in combination with the monitoring data; generating an adjustment amount of laser power and wire feeding speed through model predictive control strategy optimization according to a deviation between the predicted temperature and a target temperature; the method further calculates a prediction residual between the predicted temperature and an actual temperature at the next time, and online evaluates and adjusts model parameters of the temperature dynamic prediction model based on the prediction residual to eliminate model mismatch; and finally, the adjustment amount is applied to closed-loop control of a laser and a wire feeding system. Through adaptive adjustment of the prediction model, the present application can overcome the influence of sudden changes in working conditions and improve the accuracy of temperature prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of manufacturing and industrial process control, and particularly relates to an intelligent control method and system for wind power bearing processing. BACKGROUND

[0002] Copper-steel composite thrust bearing is applied to the fields of wind power and metallurgy due to its excellent wear resistance and mechanical strength. Laser cladding of the copper layer is a core process for realizing a high-performance composite structure, and the quality of the laser cladding directly determines the service life and operation reliability of the bearing.

[0003] At present, the traditional laser cladding method generally adopts fixed process parameters of laser power and wire feeding speed. However, in the actual production process, complex factors such as random fluctuations of the initial temperature of the steel substrate, nonlinear changes of the ambient temperature, and batch differences of material thermal properties can easily cause severe fluctuations of the temperature in the cladding area. Such temperature instability can directly lead to significant fluctuations of the metallurgical bonding strength of the copper layer, increase the probability of defects such as cracks, and finally cause a significant decline in the cladding quality, which is difficult to meet the stringent requirements of high-end equipment manufacturing industry on product high consistency and high reliability.

[0004] In related technologies, the idea of model-based predictive control is introduced, that is, a temperature prediction model is established, and the control quantity is adjusted in advance according to the deviation between the predicted future temperature and the target temperature. However, the temperature prediction model in these methods is usually established based on offline data calibration or theoretical derivation, and the model parameters are fixed. In the dynamic and complex laser cladding process, once the actual working condition or material property changes, the fixed prediction model will be mismatched with the actual physical process. Such model mismatch will lead to inaccurate temperature prediction, and then the control system makes wrong adjustments, which not only cannot realize accurate temperature control, but also may exacerbate temperature fluctuations and damage the stability of the control system. SUMMARY

[0005] To solve the technical problem of how to construct a temperature prediction model that can adapt to dynamic changes of working conditions and adjust model parameters when mismatched, and based on which to solve the adjustment amount of laser power and wire feeding speed to ensure the temperature control accuracy of laser cladding of copper-steel composite thrust bearing, the present application provides solutions in the following aspects.

[0006] In a first aspect, the present application provides an intelligent control method for wind power bearing processing, which comprises the following steps:

[0007] The process involves acquiring monitoring data during laser cladding, including actual temperature, laser power, and wire feeding speed. Based on a pre-defined dynamic temperature prediction model for the cladding zone and the monitoring data, the predicted temperature for the next moment is determined. Based on the deviation between the predicted temperature and the pre-defined target temperature, and the laser power and wire feeding speed adjustments to be solved, a model predictive control strategy is used for optimization to generate the laser power and wire feeding speed adjustments. The prediction residual between the predicted temperature and the actual temperature acquired at the next moment is calculated. Based on the prediction residual, the degree of mismatch in the dynamic temperature prediction model is evaluated. According to the degree of mismatch, the model parameters of the dynamic temperature prediction model are adjusted. The laser power and wire feeding speed adjustments are then applied to update the laser and wire feeding system, thereby achieving control over the laser cladding process for wind turbine bearings.

[0008] This invention solves the core problem of model mismatch in existing technologies—namely, the static nature of the predictive model and its inability to cope with changing operating conditions—by constructing a complete adaptive closed-loop control framework encompassing acquisition, prediction, control, and feedback correction. Compared to simple open-loop control or pure feedback control, this method achieves forward-looking prediction and regulation of temperature. Compared to traditional model predictive control, this method introduces a mechanism for online adjustment of the model itself based on the prediction residuals. This enables the control system to continuously learn and adapt to dynamic changes during the actual cladding process, thereby maintaining high-precision temperature control even under complex operating conditions, significantly improving the robustness, adaptability, and final cladding quality stability of the control system.

[0009] Preferably, the predicted temperature at the next moment satisfies the following relationship:

[0010] ;

[0011] in, It is the cladding zone Predicted temperature at any given time; It is the cladding zone The actual temperature at that moment; yes The effect coefficient of laser power on temperature at any given time; It is the cladding zone Laser power at any given moment; yes Time's up The amount of laser power adjustment at any given time; yes The coefficient of influence of constant wire feed speed on temperature; It is the cladding zone The wire feeding speed at any given moment; yes Time's up the wire feed speed adjustment amount at the time instant t; is the predicted temperature at the time instant t is the wire feed speed adjustment amount at the time instant t is the heat loss coefficient at the time instant t is the preset prediction time step.

[0012] The temperature dynamic prediction model is embodied as a linear mathematical expression with clear structure and explicit physical meaning. The model not only reasonably includes three key physical factors affecting temperature change, i.e. the heating effect of laser power, the cooling effect of wire feed speed and the heat dissipation effect to the environment, but also makes the subsequent optimization solution and parameter identification efficient and easy to implement in calculation due to its linear form. This provides a solid foundation for the engineering application of the entire complex control algorithm, while ensuring the model description ability, effectively reducing the requirement for processor computing capacity, and enhancing the practicability and real-time performance of the method.

[0013] Preferably, the model parameters of the temperature dynamic prediction model include the influence coefficient of laser power on temperature, the influence coefficient of wire feed speed on temperature and the heat loss coefficient.

[0014] Preferably, the optimization solution by the model predictive control strategy includes: constructing an objective function, and solving the laser power adjustment amount and the wire feed speed adjustment amount by minimizing the objective function; the objective function satisfies the relationship: ; wherein, is the predicted temperature at the time instant t is the preset target temperature of the cladding zone; is the predicted temperature at the time instant t is is the laser power adjustment amount at the time instant t is the wire feed speed adjustment amount at the time instant t is the predicted temperature at the time instant t is the wire feed speed adjustment amount at the time instant t is the predicted temperature at the time instant t , are weight factors of the laser power adjustment amount and the wire feed speed adjustment amount, respectively.

[0015] The application constructs and optimizes a specific objective function. The design of the objective function is extremely ingenious. It not only pursues the minimization of the deviation between the predicted temperature and the target temperature, but also suppresses the drastic changes of the laser power adjustment amount and the wire feed speed adjustment amount by introducing weight factors. This multi-objective optimization strategy effectively avoids the frequent oscillation of the control system due to the pursuit of rapid response, makes the control process more smooth and stable, and is also conducive to prolonging the service life of the laser and other actuators.

[0016] Preferably, the process of obtaining the laser power adjustment and the wire feed speed adjustment includes: applying preset control constraints to the laser power adjustment and the wire feed speed adjustment; and obtaining the laser power adjustment and the wire feed speed adjustment that satisfy their corresponding control constraints.

[0017] Preferably, the step of evaluating the mismatch degree of the temperature dynamic prediction model based on the prediction residual includes: using the previous time step... The current time is defined as a time window. The average absolute value of the prediction residuals within this time window is calculated to obtain the prediction bias. The mismatch degree of the temperature dynamic prediction model is the normalized prediction bias. The prediction bias at the current time is... Satisfying the relation: ;in, It is the first time within the current time window. The actual temperature at that moment, It is the first time within the current time window. Predicted temperature at any time It is the size of the time window at the current moment. It is the absolute value symbol.

[0018] This invention effectively smooths out random interference caused by single-point measurement noise by calculating the average absolute value of the prediction residuals within a preset time window. Compared to using instantaneous residuals, the obtained prediction bias more realistically and stably reflects the overall performance deviation of the prediction model over a recent period. This provides a more reliable and robust decision-making basis for subsequent model parameter adjustments, avoiding the oversensitivity of adaptive mechanisms to noise and frequent misjudgments.

[0019] Preferably, adjusting the model parameters of the temperature dynamic prediction model according to the degree of mismatch includes: determining an adaptive forgetting factor based on the degree of mismatch, wherein the adaptive forgetting factor is negatively correlated with the degree of mismatch; and updating the model parameters of the temperature dynamic prediction model using a recursive least squares algorithm with the adaptive forgetting factor and based on the prediction residual.

[0020] This invention establishes a negative correlation between the adaptive forgetting factor and the degree of model mismatch. This design perfectly resolves the inherent contradiction between response speed and noise resistance faced by the traditional recursive least squares algorithm with a fixed forgetting factor. When the system undergoes sudden changes or the mismatch is high, the forgetting factor automatically decreases, enabling the algorithm to quickly track changes; when the system is stable and the mismatch is low, the forgetting factor automatically increases, enhancing the ability to suppress noise. This intelligent dynamic balancing mechanism makes the model parameter update process both agile and stable, which is key to improving the overall system performance.

[0021] Preferably, the updating of the model parameters of the temperature dynamic prediction model comprises: determining a correction term based on the product of the Kalman gain and the prediction residual; and adding the model parameters at the previous time to the correction term to obtain the updated model parameters at the current time.

[0022] Preferably, the initial value of the model parameters of the temperature dynamic prediction model is obtained by collecting a plurality of sets of sample data in a historical laser cladding process and solving the plurality of sets of sample data by using a least square method.

[0023] In a second aspect, the present application provides an intelligent control system for wind power bearing machining, which comprises a memory and a processor, and the memory stores computer program instructions which, when executed by the processor, implement the intelligent control method for wind power bearing machining according to the first aspect of the present application.

[0024] By using the above technical solution, the intelligent control method for wind power bearing machining according to the first aspect of the present application is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and the use is facilitated.

[0025] The present application has the following beneficial effects: The present application constructs a complete model adaptive closed loop: through the steps of calculating the prediction residual, evaluating the mismatch degree and adjusting the model parameters, the prediction model can be online self-corrected according to the real-time prediction error. This design makes the control system no longer rely on a perfect and unchangeable model, but can continuously learn and adapt to the dynamic changes and uncertainties in the process, thereby showing strong robustness and environmental adaptability in complex and variable industrial sites. The present application realizes intelligent adjustment by making the adaptive forgetting factor negatively correlated with the model mismatch degree: when the mismatch degree is high, the forgetting factor is automatically reduced, so that the algorithm focuses more on the current data, thereby quickly updating the model parameters and realizing agile tracking of dynamic changes; when the mismatch degree is low, the forgetting factor is automatically increased, so that the algorithm focuses more on the historical data, effectively suppressing the interference of measurement noise on the model and ensuring the stability of the control. The objective function defined in the present application not only minimizes the temperature deviation, but also strives to minimize the change amplitude of the control quantity. This can effectively avoid the violent oscillation of the control output, making the control process smoother. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A flowchart of an intelligent control method for wind power bearing machining provided for an embodiment of the present application;

[0027] Figure 2 A structural block diagram of an intelligent control system for wind power bearing machining provided for an embodiment of the present application. DETAILED DESCRIPTION

[0028] The first aspect of the embodiment of the present application provides an intelligent control method for wind power bearing machining, as shown in the figure, the method comprises steps S100-S600: Figure 1

[0029] In step S100, monitoring data in the laser cladding process is acquired, and the monitoring data comprises actual temperature, laser power and wire feeding speed.

[0030] It should be noted that in the copper-steel composite thrust bearing machining process, the copper layer laser cladding is a key process to ensure the mechanical strength of the product. The core of this process is to melt the copper wire by laser and form a metallurgical bond with the steel matrix. This step is the basis for closed-loop control. Real-time and accurate acquisition of core state variables such as temperature of the cladding area and current output of the control system such as laser power and wire feeding speed is a prerequisite for subsequent state prediction, optimization solution and model correction. Only feedback control based on the true state of the current process can effectively suppress the impact of various disturbances on the cladding quality.

[0031] Specifically, a temperature measuring device such as an infrared thermometer or a two-color pyrometer is deployed near the laser cladding head to measure the center temperature of the cladding molten pool in real time, which is denoted as the actual temperature. At the same time, the laser power and wire feeding speed are acquired from the laser controller and wire feeding system. The laser cladding head is referred to as the cladding area below.

[0032] At this point, the actual temperature, laser power and wire feeding speed in the laser cladding process are acquired.

[0033] In step S200, a preset temperature dynamic prediction model of the cladding area is used to determine the predicted temperature at the next time based on the monitoring data.

[0034] It should be noted that the heat transfer in the laser cladding process is a complex dynamic process, and traditional control methods cannot cope with disturbances such as initial temperature of the steel matrix, environmental temperature and material batch differences, resulting in loss of temperature control. Therefore, building a dynamic model that can accurately predict the temperature of the cladding area is a prerequisite and foundation for closed-loop intelligent control.

[0035] Specifically, based on the principle of energy conservation, a discrete-time dynamic recursive model describing the temperature change of the cladding area is established. The model relates the predicted temperature at the next time to the actual temperature at the current time, laser power, wire feeding speed and environmental heat dissipation.

[0036] According to the above construction logic, the predicted temperature at the next time predicted by the temperature prediction model satisfies the following relationship:

[0037] ;​

[0038] wherein, is the predicted temperature at time ; is the actual temperature at time ; is the influence coefficient of laser power on temperature at time ; is the laser power at time ; is the laser power adjustment amount from time to time ; is the influence coefficient of wire feeding speed on temperature at time ; is the wire feeding speed at time ; is the wire feeding speed adjustment amount from time to time ; is the heat loss coefficient from time to time ; is the preset predicted time step, in units of .

[0039] In this relationship, represents the change in temperature caused by the energy input contributed by the total laser power in the time , the greater the laser power, the greater the value, and the higher the temperature prediction value. represents the temperature drop caused by the heat absorbed by the melted wire fed in the time , the faster the wire feeding speed, the more heat absorbed, and the lower the temperature prediction value. follows the Newtonian cooling law, which describes the temperature decrease caused by natural heat dissipation driven by the temperature difference between the molten pool and the environment in the time , the higher the actual temperature at the current time, the greater the temperature difference, the faster the heat dissipation, and the lower the temperature prediction value.

[0040] It should be noted that, , are parameters that need to be controlled in the laser cladding process, and how to make the values of these two parameters more reasonable is described in detail in the subsequent step S300. ​​​

[0041] It should also be noted that the model parameters , , are dynamically changing over time. Their initial values , , can be obtained by collecting multiple sets of input laser power, wire feed speed and output temperature data during the historical cladding process, and fitting them using offline least squares method. To cope with more complex changes in online operation, the present application introduces a recursive least squares method with a forgetting factor in subsequent step S400 to perform online adaptive updating of these parameters.

[0042] At this point, the next time prediction temperature is obtained based on the preset temperature dynamic prediction model of the cladding zone.

[0043] Step S300, based on the deviation of the predicted temperature and the preset target temperature, and the laser power adjustment amount and the wire feed speed adjustment amount to be solved, the model predictive control strategy is optimized and solved to generate the laser power adjustment amount and the wire feed speed adjustment amount.

[0044] It should be noted that after obtaining the temperature dynamic prediction model, in order to realize accurate and smooth control of the cladding temperature, the present embodiment further constructs a temperature control optimizer based on the model predictive control strategy. MPC is an advanced control strategy that predicts future system behavior using a model and determines the optimal control input by solving an optimization problem, so as to prospectively handle time delay, constraints and multi-objective optimization problems.

[0045] Specifically, within each control period to , the optimizer takes minimizing the deviation of the predicted temperature and the target temperature as the primary goal, while taking into account the adjustment amplitude of the control input to avoid system fluctuations.

[0046] According to the above logic, the objective function satisfies the relationship:

[0047] ;

[0048] Where, is the predicted temperature of the cladding zone at time ; is the preset target temperature of the cladding zone; is the laser power adjustment amount from time to time ; is the wire feed speed adjustment amount from time to time ; , are weight factors of laser power adjustment and wire feed speed adjustment respectively.

[0049] In the relationship, is the square of temperature prediction error, whose goal is to make the predicted temperature as close as possible to the preset target temperature. and are square penalty terms for controlling laser power adjustment and wire feed speed adjustment respectively, which are used to suppress the sharp changes of laser power and wire feed speed, so as to ensure the stability of the control process and prolong the life of the actuator.

[0050] The values of weight factors and can be adjusted according to process requirements. For example, when laser cladding repair of high-precision workpieces is performed, if the temperature error is within 80 , the values of and may be appropriately reduced, such as and , so that the controller quickly increases the laser power to pull the temperature back to the vicinity of the target value within 1-2 time steps, avoiding defects caused by excessively low temperature. However, when laser surface quenching of large components is performed, it is necessary to ensure that the temperature rises and falls slowly to avoid deformation or cracking of the components due to excessive thermal stress. If the temperature error is within 20 , the weights are set as: and , so that the controller focuses more on limiting the sharp changes of power and smoothly reaches the target temperature within multiple time steps, avoiding thermal shock caused by sudden changes in power.

[0051] It should be noted that in order to ensure that the control action is within the allowable range of the physical device and meets the process safety specifications, constraints must be imposed on the control adjustment.

[0052] Specifically, the control constraints are as follows:

[0053] ;

[0054] wherein, is the laser power adjustment from time to time ; is the wire feed speed adjustment from time to time .

[0055] Based on the objective function and the constraints, the minimum and The optimization problem is a typical quadratic programming problem, which can be solved efficiently in each control period by mature numerical optimization algorithms. How to solve it is not described here.

[0056] So far, through the optimization solution of model predictive control, the laser power adjustment amount and the wire feeding speed adjustment amount for the next control period are generated.

[0057] Step S400, calculate the prediction residual between the predicted temperature and the actual acquired actual temperature at the next time; based on the prediction residual, evaluate the mismatch degree of the temperature dynamic prediction model; and adjust the model parameters of the temperature dynamic prediction model according to the mismatch degree.

[0058] It should be noted that this step is the key to realizing adaptive control and improving robustness of the present application. The laser cladding process will be affected by unmodeled dynamics and disturbances such as material batch differences, local heat dissipation condition changes of the workpiece, nozzle wear, etc., which will cause the preset temperature dynamic prediction model to gradually mismatch, the prediction accuracy to decrease, and then affect the control effect. By comparing the predicted value with the true value online and using the deviation to continuously correct the model parameters, the prediction model can always maintain the best approximation to the real process, thereby ensuring that the control system can still maintain high performance in a changing environment.

[0059] Specifically, the present embodiment defines a short-term prediction deviation degree , which takes the time window of the current time as the time window of the previous time, and measures the recent performance of the model by calculating the average of the absolute error between the model predicted value and the actual measured value in the latest time window. The prediction deviation degree satisfies the relationship:

[0060] ;

[0061] Wherein, is the actual temperature at the time in the time window of the current time, is the predicted temperature at the time in the time window of the current time, is the size of the time window of the current time, is the absolute value symbol.

[0062] As a preferred embodiment, the value of the window size is a trade-off. A smaller , such as 10, can respond faster to changes, but is more sensitive to measurement noise; a larger , such as 20, is more robust to noise, but responds more slowly. In the present embodiment, the window size To balance response speed and stability.

[0063] It should be noted that, The numerical range of this value is not fixed, making it inconvenient to use directly for subsequent logical judgments. Therefore, this embodiment uses the calculated value... Perform normalization to make its values ​​within The interval. Normalization can be achieved using the Sigmoid function. The Sigmoid function is a current technique and will not be discussed further here.

[0064] Thus, the degree of mismatch in the temperature dynamic prediction model has been obtained.

[0065] Step S500: Adjust the model parameters of the temperature dynamic prediction model according to the degree of mismatch.

[0066] It should be noted that if the model parameters in step S200 , , Maintaining a constant temperature prediction method will struggle to adapt to dynamic fluctuations caused by changes in raw material batches, surface treatment methods, and environmental conditions. This will prevent the system from consistently and accurately reflecting its actual temperature response characteristics, leading to decreased temperature prediction accuracy and impacting control performance. This invention employs a recursive least squares algorithm with a variable forgetting factor (VFF-RLS). By dynamically adjusting the weights of historical data, it assigns greater influence to recent data, enabling rapid tracking of time-varying system characteristics and online dynamic updating of model parameters. This effectively improves the model's adaptability to complex operating conditions.

[0067] Specifically, this includes steps S510-S520:

[0068] Step S510: Calculate the adaptive forgetting factor based on the degree of mismatch.

[0069] It should be noted that the degree of mismatch identified in step S400 needs to be translated into direct control of the model parameter update rate. The adaptive forgetting factor adjustment mechanism in this embodiment is precisely the bridge connecting the degree of mismatch and the model parameter update rate.

[0070] Based on the above logic, The adaptive forgetting factor at time satisfies the following relationship:

[0071] ;

[0072] in, Is The adaptive forgetting factor, calculated at each step, is used for the next VFF-RLS update. It is the normalized prediction bias. , These are the preset upper and lower limits of the forgetting factor.

[0073] In this relationship, the forgetting factor The value of is negatively correlated with the normalized prediction bias. When the system is stable, the normalized prediction bias... ,but In this case, the VFF-RLS algorithm assigns higher weights to historical data, resulting in slow and stable model parameter updates. However, when a sudden change occurs in the system... ,but At this point, the VFF-RLS algorithm will quickly forget the past data and give higher weights to the new data, thereby achieving rapid correction and reconvergence of the model parameters.

[0074] It should be added that, in order to ensure the performance of the algorithm, It should be close to 1 to ensure the stationarity of model parameter estimation when the system is stable, and 0.995 is preferred. This determines the maximum adaptation speed when the system undergoes a mutation. A smaller value results in faster adaptation but may introduce instability. A value of 0.95 is preferred.

[0075] Step S520: Apply the adaptive forgetting factor to execute the VFF-RLS algorithm to update the model parameters.

[0076] To apply the VFF-RLS algorithm, the temperature prediction model in step S200 is first rewritten into a standard linear regression form:

[0077] ;

[0078] The variables are defined as follows:

[0079] ;

[0080] ;

[0081] ;

[0082] in, Indicates in Available at any time arrive The actual temperature change at any given moment; express The regression vector constructed at each moment; It is the cladding zone The actual temperature at that moment; It is the cladding zone The actual temperature at that moment; It is the cladding zone Laser power at any given moment; It is the cladding zone The wire feeding speed at any given moment; It is the preset prediction time step.

[0083] This is the vector of model parameters to be estimated. It is a time-varying vector, and its structure at any given time is as follows: The goal of this algorithm is to calculate, at each time step, based on the new measurement data... The optimal estimate. Specifically, it refers to the algorithm used in the current iteration step of the VFF-RLS algorithm. The estimated values ​​of the model parameters at known times.

[0084] The VFF-RLS algorithm is used to process the model parameter vector. Perform real-time updates. The update process is as follows:

[0085] First, calculate the Kalman gain. , Satisfying the relation:

[0086] ;

[0087] in, yes Kalman gain vector at time step; yes The covariance matrix at time t; It is a column vector that contains... All known input and status information that affects the system output at any given time; It is a forgetting factor; It is a column vector The transpose of is a row vector used for matrix and vector multiplication operations to conform to the rules of linear algebra.

[0088] This relation is used for calculation. Kalman gain vector at time step In the VFF-RLS algorithm, It plays a crucial role in determining the extent to which the prediction error introduced by new measurements should be trusted when updating model parameters. A larger... This means that new data will have higher weights and the parameter adjustments will be more significant; conversely, it will be more inclined to maintain the original parameter estimates.

[0089] Then, update the model parameter estimates. Satisfying the relation:

[0090] ;

[0091] where, is the model parameter vector at time t; is the model parameter vector at time t; is the Kalman gain at time t; is the actual temperature change from time t to time t+1, which is also the output value at time t+1; is the transpose of column vector , which is a row vector, used for matrix and vector multiplication to comply with the rules of linear algebra.

[0092] In this relationship, is the prediction residual, which is the core driving force of the entire expression. The prediction made by the old parameters on the current input. The value inside the entire square bracket represents the difference between the actual measured value and the model predicted value. If the prediction residual is zero, it means that the current model is perfect, and the model parameters do not need to be updated. If the residual is not zero, the error signal will drive the parameters to adjust in the direction of reducing the error through the gain . This relationship is the core of the RLS algorithm, which performs the actual update of the model parameters. Its logic is new estimate = old estimate + gain x prediction error, which is a classic prediction-correction structure.

[0093] Finally, update the covariance matrix , the covariance matrix satisfies the relationship:

[0094] ;

[0095] where, is the covariance matrix at time t; is the forgetting factor; is the covariance matrix at time t; is the Kalman gain vector at time t; is the transpose of column vector , which is a row vector, used for matrix and vector multiplication to comply with the rules of linear algebra.

[0096] This relationship is used to update the covariance matrix, preparing for the calculation of the next control period ​​​​​​​​​​​The covariance matrix at time 1 is the updated covariance matrix, representing our estimates of the newly obtained model parameters. The degree of uncertainty. Due to the uncertainty in New measurement information is constantly being incorporated. We have a more certain understanding of the model parameters, therefore The value is usually higher than A smaller value indicates reduced uncertainty. It will be used to calculate the Kalman gain vector in the next cycle. Input time.

[0097] At this point, the model parameters of the temperature dynamic prediction model have been adjusted based on the degree of mismatch.

[0098] Step S600: Update the laser and wire feeding system using the laser power adjustment amount and wire feeding speed adjustment amount to achieve control over the laser cladding process of wind turbine bearings.

[0099] In each control cycle arrive At any given time, by solving the above constrained optimization problem, that is, finding a set of conditions that satisfy the objective function... Minimized and The system updates and outputs the control commands for the next moment:

[0100] ;

[0101] ;

[0102] in, It is the cladding zone The laser power is updated in real time. It is the cladding zone Laser power at any given moment; yes Time's up The amount of laser power adjustment at any given time; It is the cladding zone The wire feeding speed is updated in real time; It is the cladding zone The wire feeding speed at any given moment; yes Time's up The amount of adjustment for the wire feeding speed at any given moment.

[0103] Updated control quantity and The data is sent to the laser and wire feeding system for execution. This predict-optimize-execute process rolls over in each control cycle, forming a complete closed-loop feedback control system.

[0104] It is also necessary to point out that steps S400 to S500 constitute a self-adaptive closed loop. Suppose that at a certain moment, the laser cladding head moves from the thick-walled body area of the workpiece to a thinner edge position.

[0105] The heat dissipation condition of the local workpiece area has a physical mutation. Compared with the thick-walled body area with large heat capacity and good heat dissipation path, the heat of the thin-walled edge area is more likely to accumulate, and the heat dissipation speed is significantly slowed down. This means that the real parameters describing the thermal response law in the system have changed, especially the coefficient related to heat dissipation will suddenly decrease.

[0106] At this time, the model mismatch and the error increase: the original model parameters suitable for the thick-walled area are no longer accurate at this new position. The prediction error term of the model, that is, , will significantly increase in absolute value in the short term.

[0107] Self-adaptive response: according to the logic of step S400, the significant prediction error will immediately cause the normalized prediction deviation degree to quickly approach 1, thereby causing the self-adaptive forgetting factor to be quickly lowered to its lower limit value .

[0108] Parameter rapid correction: observing the relationship of the Kalman gain , the decrease of in the denominator will directly cause the absolute values of the components of the gain vector to increase. In the update relationship of the parameter model parameter estimate value , a larger gain will give the current prediction error term a higher weight. The final effect is that the model parameters will be adjusted by a large margin, so that the model can quickly abandon the thermal response characteristics learned in the thick-walled area before and quickly converge to the parameter value that can accurately describe the new characteristics of the current thin-walled edge.

[0109] So far, the present application realizes accurate and robust control of the temperature in the laser cladding area under complex working conditions by constructing a complete control closed loop of prediction-optimization-evaluation-adaptation, and significantly improves the consistency and processing stability of the wind power composite bearing product.

[0110] The second aspect of the embodiment provides an intelligent control system for wind power bearing processing, as shown in Figure 2 the intelligent control system for wind power bearing processing comprises a memory and a processor, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, the first aspect of the present application is realized. The intelligent control method for wind power bearing processing.

[0111] The intelligent control system for wind power bearing machining further comprises a communication bus and a communication interface and other components well known to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

[0112] In this application, the aforementioned memory can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device. For example, the computer-readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as resistive random access memory, dynamic random access memory, static random access memory, enhanced dynamic random access memory, high bandwidth memory, hybrid memory cube, or the like, or any other medium that can be used to store the desired information and that can be accessed by an application, module, or both. Any such computer storage media can be part of a device or accessible or connectable thereto.

[0113] The above are preferred embodiments of the present application, which do not limit the protection scope of the present application, therefore: any equivalent changes made in the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. An intelligent control method for wind power bearing machining, characterized in that, The method comprises the steps of: acquiring monitoring data in a laser cladding process, the monitoring data comprising actual temperature, laser power and wire feeding speed; determining a predicted temperature at a next time based on a preset temperature dynamic prediction model of a cladding zone and in combination with the monitoring data, the predicted temperature satisfying a relationship: ； in, It is the cladding zone Predicted temperature at any given time; It is the cladding zone The actual temperature at that moment; yes The effect coefficient of laser power on temperature at any given time; It is the cladding zone Laser power at any given moment; yes Time's up The amount of laser power adjustment at any given time; yes The coefficient of influence of constant wire feed speed on temperature; It is the cladding zone The wire feeding speed at any given moment; yes Time's up The amount of adjustment for the wire feed speed at any given time; It is the cladding zone Time's up The heat loss coefficient at any given time; It is the preset prediction time step; Model parameters of the temperature dynamic prediction model comprise a laser power impact coefficient on temperature, a wire feeding speed impact coefficient on temperature and a heat loss coefficient; based on a deviation between the predicted temperature and a preset target temperature, and a laser power adjustment amount and a wire feeding speed adjustment amount to be solved, performing optimization and solving through a model prediction control strategy, comprising: constructing an objective function, and solving the laser power adjustment amount and the wire feeding speed adjustment amount by minimizing the objective function, comprising: further imposing a preset control constraint condition on the laser power adjustment amount and the wire feeding speed adjustment amount; the laser power adjustment amount and the wire feeding speed adjustment amount satisfy the corresponding control constraint condition after being solved; Objective function satisfies the relationship: ; wherein, is a preset target temperature of the cladding zone; , are weight factors of the laser power adjustment amount and the wire feed speed adjustment amount, respectively, to generate the laser power adjustment amount and the wire feed speed adjustment amount. calculating a prediction residual error between the predicted temperature and an actual temperature acquired at the next time; based on the prediction residual error, evaluating a mismatch degree of the temperature dynamic prediction model, comprising: Before the current moment The current time window is defined as a time window, and the average absolute value of the prediction residuals within the time window is calculated to obtain the prediction bias. the mismatch degree of the temperature dynamic prediction model is a normalized prediction deviation degree; The predicted deviation degree of the current moment is calculated according to the following formula: The relationship is satisfied: ; wherein, is the actual temperature at the current time instant, is the predicted temperature at the current time instant, is the actual temperature at the current time instant, is the predicted temperature at the current time instant, is the size of the time window at the current time instant, is the absolute value symbol; adjusting model parameters of the temperature dynamic prediction model according to the mismatch degree; applying the laser power adjustment amount and the wire feeding speed adjustment amount to update a laser and a wire feeding system, and realizing control over the wind power bearing laser cladding process.

2. The intelligent control method for wind power bearing machining according to claim 1, characterized in that, The adjusting of the model parameters of the temperature dynamic prediction model according to the mismatch degree comprises: determining an adaptive forgetting factor according to the mismatch degree, the adaptive forgetting factor being negatively correlated with the mismatch degree; updating the model parameters of the temperature dynamic prediction model by using a recursive least squares algorithm with the adaptive forgetting factor and according to the prediction residual error.

3. The intelligent control method for wind power bearing machining according to claim 2, characterized in that, The updating of the model parameters of the temperature dynamic prediction model comprises: determining a correction term based on a product of a Kalman gain and the prediction residual error; adding the model parameters at a previous time to the correction term to obtain updated model parameters at a current time.

4. The intelligent control method for wind power bearing machining according to claim 2, characterized in that, Initial values of the model parameters of the temperature dynamic prediction model are obtained by collecting a plurality of sample data in a historical laser cladding process and solving the plurality of sample data by using a least squares method.

5. An intelligent control system for wind power bearing machining, characterized in that, The intelligent control system for wind power bearing machining comprises a processor and a memory, and the memory stores computer program instructions, which realize the intelligent control method for wind power bearing machining according to any one of claims 1-4 when executed by the processor.

Images