A pressure intelligent compensation adjusting method for a multi-stage adjustable atomizing nozzle

Abstract

This invention relates to the field of industrial process control technology, specifically disclosing a method for intelligent pressure compensation and adjustment of a multi-stage adjustable atomizing nozzle. First, nozzle inlet pressure and medium flow rate data are collected synchronously. Then, a state vector characterizing the fluid's dynamic properties is constructed and updated in real time based on sliding window covariance analysis and dynamic fusion. Next, an optimal valve control sequence is generated using a combination of pattern matching and rolling optimization. Smooth output of control commands is achieved by introducing a smoothing factor based on control trends through dynamic calculation. Finally, closed-loop correction of the state vector is completed using prediction deviation and adaptive correction strategies. This invention effectively solves the parameter mismatch problem of traditional fixed-parameter PID control when dealing with nonlinear and time-varying fluid characteristics, improving system stability and equipment lifespan while ensuring control accuracy.

Description

Technical Field

[0001] This invention relates to the field of industrial process control technology, specifically to a method for intelligent pressure compensation and adjustment of a multi-stage adjustable atomizing nozzle. Background Technology

[0002] In existing control technologies for multi-stage adjustable atomizing nozzles, PID control algorithms based on fixed parameters are commonly used to regulate the nozzle's working pressure. This method adjusts the valve opening by detecting pressure deviations and performing proportional, integral, and derivative calculations. However, fluid media in real-world industrial applications often exhibit complex rheological properties, especially when dealing with non-Newtonian fluids, whose apparent viscosity changes significantly with shear rate. This results in strong nonlinearity and time-varying dynamic characteristics of the controlled object. This inherent contradiction between the uncertainty of the controlled object model and the fixedness of the controller parameters leads to a severe deterioration in control quality when traditional PID control operates under varying conditions.

[0003] Existing technologies lack the ability to overcome the parameter mismatch problem that occurs when fixed-parameter PID controllers handle changes in the dynamic characteristics of fluids. When fluid characteristics change (such as shear thinning of non-Newtonian fluids) and the controlled object model changes, the original control parameters cannot adapt to the new dynamic characteristics, resulting in system response lag, increased overshoot, or even continuous oscillation. This makes the controller itself an unstable excitation source for the system, seriously affecting the stability of atomization effect and the consistency of product quality. Summary of the Invention

[0004] The purpose of this invention is to provide a method for intelligent pressure compensation adjustment of a multi-stage adjustable atomizing nozzle to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A method for intelligent pressure compensation adjustment of a multi-stage adjustable atomizing nozzle includes the following steps:

[0007] S1: Obtain the real-time operating parameter set of the nozzle system; the real-time operating parameter set includes the real-time pressure value at the nozzle inlet and the real-time flow rate of the medium flowing into the nozzle;

[0008] S2: Based on the real-time operating parameter set, construct and update a state vector describing the current fluid dynamic characteristics;

[0009] S3: Using the current state vector and valve opening as initial conditions, combined with the preset pressure target value, perform rolling optimization calculation within the finite prediction time domain; the goal of the rolling optimization calculation is to solve for an optimal valve control sequence that minimizes the cumulative deviation between the predicted system pressure trajectory and the pressure target value within the prediction time domain, while constraining the valve's action amplitude.

[0010] S4: Extract the first control quantity from the optimal valve control sequence as an immediate control command and output it to the nozzle control valve actuator to adjust the valve opening;

[0011] S5: In the next sampling period, collect a new set of real-time operating parameters, calculate the prediction deviation between the predicted pressure value and the actual pressure value at the previous moment, and use the prediction deviation to perform feedback correction on the state vector.

[0012] As a further aspect of the present invention: S2 specifically includes:

[0013] S210: Obtain the real-time pressure value and real-time medium flow rate value at the current time and multiple previous sampling times to form a pressure parameter sequence and a flow rate parameter sequence;

[0014] S220: By calculating the covariance relationship between the pressure parameter sequence and the flow parameter sequence over time, an instantaneous change vector reflecting the sensitivity of pressure to instantaneous flow rate changes is generated;

[0015] S230: The instantaneous change vector is fused with the state vector of the previous moment, and the derived parameters are recursively updated to construct the state vector of the current moment.

[0016] As a further aspect of the present invention: S220 specifically includes:

[0017] S221: Divide the pressure parameter sequence and the flow parameter sequence into multiple overlapping sliding data windows respectively;

[0018] S222: For each corresponding sliding data window, calculate the covariance between pressure data points and flow data points within the sliding data window, and arrange the covariance calculation results of all windows in chronological order to form a covariance sequence;

[0019] S223: Differentiate the covariance sequence to obtain the covariance rate of change sequence, and use the covariance rate of change sequence as the instantaneous change vector.

[0020] As a further aspect of the present invention: S3 specifically includes:

[0021] S310: Based on the derived parameters in the state vector and the current valve opening, generate multiple pressure prediction sequences corresponding to multiple consecutive moments in the future;

[0022] S320: Calculate the cumulative pressure deviation between the pressure prediction sequence and the pressure target value, and calculate the valve opening change required to generate the pressure prediction sequence.

[0023] S330: Multiply the cumulative pressure deviation and the corresponding valve opening change by the dynamically adjusted weighting coefficient respectively, and then add them together to obtain the comprehensive evaluation value of each pressure prediction sequence;

[0024] S340: Select the valve control sequence corresponding to the pressure prediction sequence with the smallest comprehensive evaluation value as the optimal valve control sequence.

[0025] As a further aspect of the present invention: S310 specifically includes:

[0026] S311: Pre-set library of pressure change patterns under various typical fluid dynamic characteristics, where each pressure change pattern is associated with a historical state vector segment;

[0027] S312: Perform dynamic time warping matching between the current state vector and each historical state vector segment in the pressure change pattern library, and calculate the matching degree with each historical state vector segment.

[0028] S313: Select the pressure change pattern corresponding to the historical state vector segment with the highest matching degree as the basic prediction pattern, and adjust the amplitude of the basic prediction pattern according to the ratio between the current valve opening and the historical valve opening.

[0029] S314: Based on the scaled and adjusted pressure change pattern, and combined with the current real-time pressure value, generate multiple pressure prediction sequences.

[0030] As a further aspect of the present invention: S4 specifically includes:

[0031] S410: Obtain the optimal valve control sequence, and extract the first two control variables in the optimal valve control sequence as the current control variable and the preparation control variable for the next moment;

[0032] S420: Calculate the difference between the current control quantity and the prepared control quantity at the next moment to obtain the trend of control quantity change;

[0033] S430: Calculate the corresponding smoothing factor based on the absolute value of the trend of change of the control quantity;

[0034] S440: The current control quantity is weighted and averaged with the actual control command at the previous moment, where the weight is determined by the smoothing factor, to generate the instantaneous control command at the current moment and output it to the regulating valve actuator.

[0035] As a further aspect of the present invention: S430 specifically includes:

[0036] S431: Divide the absolute value of the trend of change of the control quantity into multiple consecutive numerical intervals, and associate each numerical interval with a basic smoothing coefficient.

[0037] S432: Select the corresponding basic smoothing coefficient based on the numerical range of the absolute value of the current control variable change trend;

[0038] S433: Obtain the smoothing factor sequence actually used in the historical control cycle and calculate the statistical characteristic value of the smoothing factor sequence;

[0039] S434: Calculate the ratio of the statistical characteristic value to the preset fixed benchmark value to obtain the historical smoothing correction amount, and multiply the historical smoothing correction amount with the basic smoothing coefficient to obtain the smoothing factor.

[0040] As a further aspect of the present invention: S5 specifically includes:

[0041] S510: Calculate the ratio of the prediction deviation to the derived parameters in the current state vector to obtain the unit parameter deviation.

[0042] S520: Based on the absolute value of the unit parameter deviation, the corresponding correction influence factor is determined through a preset mapping relationship;

[0043] S530: Multiply the prediction bias, the correction impact factor and a fixed learning step size to obtain the state vector correction amount;

[0044] S540: Add the state vector correction amount to the state vector at the previous moment to complete the feedback correction of the current state vector.

[0045] As a further aspect of the present invention: S520 specifically includes:

[0046] S521: Divide the absolute value of the unit parameter deviation into multiple continuous numerical intervals, and pre-set a basic influence factor for each interval;

[0047] S522: Calculate the difference between the current unit parameter deviation and the unit parameter deviation at the previous moment, and determine the direction of the deviation change trend based on the sign of the difference;

[0048] S523: Based on the direction of the deviation change trend, the selected basic influencing factors are directionally adjusted to obtain preliminary influencing factors;

[0049] S524: Obtain the actual impact factor sequence at multiple historical moments, select the value with the highest frequency as the frequency correction value, and merge the preliminary impact factor with the frequency correction value to generate the final corrected impact factor.

[0050] The beneficial effects of this invention are:

[0051] (1) By constructing a state vector based on real-time pressure-flow covariance analysis, dynamic identification and tracking of the viscous characteristics of complex media such as non-Newtonian fluids are realized; combined with the predictive control based on pattern matching and dynamic weight adjustment mechanism, the control system can adapt to the changes in fluid characteristics, effectively overcome the parameter memory and instability problems of traditional PID control under nonlinear conditions such as shear thinning, improve the pressure control accuracy, and eliminate overshoot and oscillation phenomena.

[0052] (2) By introducing a smoothing factor dynamic calculation mechanism based on the trend of control quantity change, combined with a feedback correction strategy with directional correction, the frequent action and impact of valve mechanism are significantly suppressed while maintaining rapid response to pressure fluctuations. Experimental results show that this method reduces the action frequency of regulating valve under the condition of maintaining the same control accuracy, effectively prolongs the life of actuator, and avoids pressure pulsation caused by sudden changes in control commands. Attached Figure Description

[0053] The invention will now be further described with reference to the accompanying drawings.

[0054] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Please see Figure 1 As shown, this invention provides a method for intelligent pressure compensation adjustment of a multi-stage adjustable atomizing nozzle, comprising the following steps:

[0057] S1: Obtain the real-time operating parameter set of the nozzle system; the real-time operating parameter set includes the real-time pressure value at the nozzle inlet and the real-time flow rate of the medium flowing into the nozzle;

[0058] S2: Based on the real-time operating parameter set, construct and update a state vector describing the current fluid dynamic characteristics;

[0059] S3: Using the current state vector and valve opening as initial conditions, combined with the preset pressure target value, perform rolling optimization calculation within the finite prediction time domain; the goal of the rolling optimization calculation is to solve for an optimal valve control sequence that minimizes the cumulative deviation between the predicted system pressure trajectory and the pressure target value within the prediction time domain, while constraining the valve's action amplitude.

[0060] S4: Extract the first control quantity from the optimal valve control sequence as an immediate control command and output it to the nozzle control valve actuator to adjust the valve opening;

[0061] S5: In the next sampling period, collect a new set of real-time operating parameters, calculate the prediction deviation between the predicted pressure value and the actual pressure value at the previous moment, and use the prediction deviation to perform feedback correction on the state vector.

[0062] In S1, the data acquisition process of the real-time operating parameter set is as follows: by using a pressure sensor tightly installed on the nozzle inlet pipe, the fluid pulsating pressure signal at the nozzle inlet is continuously acquired at a millisecond sampling period to obtain the unprocessed raw pressure data sequence; an electromagnetic flow sensor is used to synchronously detect the medium velocity flowing through the pipe cross section, and the acquired velocity signal is multiplied by the pre-calibrated pipe cross-sectional area parameter to calculate the accurate instantaneous volumetric flow rate value.

[0063] In S2, firstly, the historical and current data sequences that form the basis of the analysis are obtained. Real-time pressure values ​​and real-time medium flow rates from the current moment and multiple previous sampling moments are read from the storage unit. The pressure values ​​arranged in chronological order are then combined into a pressure parameter sequence, denoted as: ;in Indicates the current time Pressure value, The sequence length is preset to 50 sampling points. Similarly, the synchronously collected medium flow rates are grouped into a flow parameter sequence, denoted as: ;in Indicates the current time The flow rate value. Sequence length. The settings are based on control requirements, for example, when the sampling frequency is 100Hz. A value of 50 can be used to cover 0.5 seconds of data history. These two sequences form the basis of the input data for subsequent calculations.

[0064] Secondly, based on the above data sequence, the instantaneous change vector reflecting the dynamic relationship between pressure and flow is calculated. The pressure parameter sequence... With flow parameter sequence The data is divided into multiple overlapping sliding windows. The length of each window is... The default is 10 sampling points, and the overlap between adjacent windows is... One sampling point, meaning the window step size is 1 sampling point. For example, the first window covers indices 1 to 10. The second window covers index 2 to And so on. Regarding the first... A sliding data window, whose internal pressure data subsequence is: The traffic data subsequence is: The covariance between pressure and flow data points within this window is calculated using the following formula: ;in, Indicates the first The covariance value of each window. yes The arithmetic mean, yes The arithmetic mean, It is the window length. and These represent the first and second elements within the window, respectively. Each pressure and flow rate value. Calculate the covariance of all windows. Arranged in chronological order, they form a covariance sequence: Differentiate the covariance sequence to obtain the covariance rate of change sequence.

[0065] The rate of change at each point is calculated using the first-order backward difference method. For the nth point in the sequence... elements ( ≥2), its differential value The calculation formula is: ;in, Indicates the first Rate of change of covariance at points and These are the th elements in the covariance sequence. and the One element, It is the sampling time interval, for example, at a sampling frequency of 100Hz. =0.01 seconds. For the first element of the sequence. Set it as The value is used to maintain a consistent sequence length. Finally, the sequence obtained by differentiation is: ;

[0066] Finally, the fusion and update of the state vectors are completed. The resulting instantaneous change vector is then processed. The state vector stored in the previous control cycle Perform fusion. State vector It is a multidimensional vector whose dimensions are the same as those of the instantaneously changing vector. Same, for example, dimension is The fusion process uses a weighted average method, for the th element in the state vector... Each component, the value after fusion The calculation formula is: ;in, Represents the first vector after fusion. One portion, It is an instantaneously changing vector The One portion, It is the th moment of the state vector at the previous moment. One portion, This is a preset fusion coefficient, ranging from 0 to 1; for example, a value of 0.3 is used to balance the influence of old and new data. This is a derived parameter characterizing the viscosity properties of fluids. Perform recursive updates and derive parameters It is a scalar component of the state vector, and its update formula is: ;in, Indicates the derived parameters at the current time. This represents the derived parameters from the previous time step. This is a preset forgetting factor, with a value between 0 and 1. For example, a value of 0.9 can be used to emphasize the continuity of historical data. From the instantaneously changing vector The feature values ​​extracted are calculated. The arithmetic mean is obtained. Finally, the fused vector is... With the updated derived parameters Combine them to construct the complete state vector at the current moment. This provides accurate fluid state information for subsequent optimized control.

[0067] In S3, firstly, based on the derived parameters in the state vector and the current valve opening, a pressure prediction sequence corresponding to multiple consecutive future time points is generated. A pre-set pressure change pattern library contains pressure change patterns under 20 typical fluid dynamic characteristics. Each pressure change pattern is associated with a historical state vector segment of 50 sampling points. The current state vector is dynamically time-warped and matched with each historical state vector segment in the pressure change pattern library. The matching degree is calculated as follows: a Euclidean distance matrix is ​​constructed between each point in the current state vector and the historical state vector segment. A dynamic programming algorithm is used to find the warped path with the minimum cumulative distance, and the reciprocal of this cumulative distance is taken as the matching degree. The pressure change pattern corresponding to the historical state vector segment with the highest matching degree is selected as the basic prediction pattern. The amplitude of the basic prediction pattern is scaled and adjusted according to the ratio between the current valve opening and the historical valve opening. The scaling factor is equal to the ratio of the current valve opening to the historical valve opening. Based on the scaled and adjusted pressure change pattern, and combined with the current real-time pressure value, multiple pressure prediction sequences are generated by adding the difference between the current real-time pressure value and the starting point of the basic prediction pattern to each point of the basic prediction pattern. Each sequence contains the predicted pressure values ​​for the next 30 sampling times.

[0068] Secondly, the cumulative pressure deviation between each pressure prediction sequence and the target pressure value is calculated, along with the valve opening change required to generate each pressure prediction sequence. The cumulative pressure deviation is calculated by subtracting the target pressure value from the predicted pressure value at each moment in the pressure prediction sequence and taking the absolute value, then summing the absolute values ​​across all moments. The valve opening change is calculated by determining the valve control sequence required to generate the pressure prediction sequence and calculating the sum of the absolute values ​​of the valve opening differences between adjacent moments in that valve control sequence. These two calculation processes quantify the pressure tracking accuracy and actuator operation cost, respectively.

[0069] Then, the cumulative pressure deviation and the corresponding valve opening change amplitude are multiplied by dynamically adjusted weighting coefficients and summed to obtain the comprehensive evaluation value of each pressure prediction sequence. The dynamic adjustment of the weighting coefficients is based on the rate of change of derived parameters in the state vector. When the rate of change of derived parameters is greater than the preset threshold of 0.05, the weighting coefficient for valve opening change amplitude is increased, and the weighting coefficient for cumulative pressure deviation is decreased. When the rate of change of derived parameters is less than or equal to the preset threshold of 0.05, the default weighting coefficient configuration is used. The default value for the weighting coefficient of cumulative pressure deviation is 0.7, and the default value for the weighting coefficient of valve opening change amplitude is 0.3. The adjustment range of the weighting coefficients is directly proportional to the magnitude of the rate of change of derived parameters, and the adjustment coefficient ranges from 0.1 to 0.5.

[0070] Finally, the valve control sequence corresponding to the pressure prediction sequence with the smallest comprehensive evaluation value is selected as the optimal valve control sequence from all pressure prediction sequences. This selection process involves comparing the comprehensive evaluation values ​​of each pressure prediction sequence, and the valve control sequence corresponding to the pressure prediction sequence with the smallest comprehensive evaluation value is taken as the output. This valve control sequence contains the valve opening setpoints for the next 10 control cycles, providing a clear instruction sequence for subsequent control execution.

[0071] In S4, firstly, the optimal valve control sequence is obtained and key control variables are extracted. The optimal valve control sequence is derived from the aforementioned rolling optimization calculation, and this sequence contains valve opening setpoints for the next 10 control cycles. The first two control variables are extracted from this sequence, with the first being the current control variable and the second being the pre-emptive control variable for the next moment. The current control variable is the valve opening command to be executed immediately, and the pre-emptive control variable for the next moment is a reference value used to predict the control trend for the next cycle. Both control variables are dimensionless standardized values, ranging from 0 to 1, where 0 represents a fully closed valve and 1 represents a fully open valve.

[0072] Secondly, the trend of control quantity change is calculated. The trend of control quantity change is obtained by calculating the difference between the current control quantity and the planned control quantity for the next time step. The specific calculation process is as follows: subtract the current control quantity from the planned control quantity for the next time step; the difference is the trend of control quantity change. This difference is a positive or negative value; a positive sign indicates an increasing trend, and a negative sign indicates a decreasing trend. The absolute value indicates the strength of the trend. For example, if the current control quantity is 0.6 and the planned control quantity for the next time step is 0.7, then the trend of control quantity change is 0.1, indicating that the control quantity is showing an increasing trend.

[0073] Then, based on the absolute value of the control quantity's change trend, the corresponding smoothing factor is calculated. The absolute value of the control quantity's change trend is divided into three consecutive numerical intervals: the first interval is 0 to 0.05, with an associated base smoothing coefficient of 0.1; the second interval is 0.05 to 0.1, with an associated base smoothing coefficient of 0.3; and the third interval is above 0.1, with an associated base smoothing coefficient of 0.5. The corresponding base smoothing coefficient is selected based on the numerical interval in which the current control quantity's change trend falls. A sequence of smoothing factors actually used in historical control cycles is obtained, containing the values ​​of smoothing factors actually used in the most recent 20 control cycles. The statistical characteristic value of this sequence is calculated; the statistical characteristic value is the arithmetic mean of all values ​​in the sequence. The ratio of the statistical characteristic value to a preset fixed benchmark value (0.5) is calculated and used as the historical smoothing correction. The base smoothing coefficient is multiplied by the historical smoothing correction, and the product is used as the final smoothing factor.

[0074] Finally, the instantaneous control command for the current moment is generated. The current control quantity is averaged with the actual control command from the previous moment using a weighted average, with the weight determined by a smoothing factor. Specifically, the instantaneous control command for the current moment equals the smoothing factor multiplied by the current control quantity, plus one minus the smoothing factor multiplied by the actual control command from the previous moment. For example, if the smoothing factor is 0.3, the current control quantity is 0.7, and the actual control command from the previous moment is 0.65, then the instantaneous control command for the current moment is 0.3 multiplied by 0.7 plus 0.7 multiplied by 0.65, resulting in 0.665. The calculated instantaneous control command is then output to the control valve actuator, which adjusts the valve opening based on this command value.

[0075] In S5, at the start of the next sampling period, a new set of real-time operating parameters is acquired. This step is completed synchronously by pressure and flow sensors. The pressure sensor acquires the real-time pressure value at the nozzle inlet, and the flow sensor acquires the real-time flow rate of the medium flowing into the nozzle. These acquired data, after filtering, constitute the new set of real-time operating parameters. Based on these new parameter values, the prediction deviation between the predicted pressure value and the actual pressure value at the previous moment is calculated. The prediction deviation is calculated by subtracting the currently acquired actual pressure value from the predicted pressure value stored at the previous moment; the difference is the prediction deviation. This prediction deviation reflects the accuracy of the prediction at the previous moment and is the basis for feedback correction.

[0076] The ratio of the prediction deviation to the derived parameter in the current state vector is calculated to obtain the unit parameter deviation. The derived parameter in the current state vector is a parameter characterizing the fluid's viscosity properties, obtained through the previous state vector update process. The unit parameter deviation is calculated by dividing the absolute value of the prediction deviation by the value of the derived parameter in the current state vector. For example, if the prediction deviation is 0.2 MPa and the derived parameter is 0.5, then the unit parameter deviation is 0.4. This calculation process normalizes the prediction deviation, eliminating the influence of different numerical ranges of derived parameters under different fluid states, thus making the subsequent correction process more adaptable.

[0077] Based on the absolute value of the unit parameter deviation, the corresponding correction influence factor is determined through a preset mapping relationship. The absolute value of the unit parameter deviation is divided into three continuous numerical intervals: the first interval is 0 to 0.2, with a preset basic influence factor of 0.1; the second interval is 0.2 to 0.5, with a preset basic influence factor of 0.3; and the third interval is above 0.5, with a preset basic influence factor of 0.5. The corresponding basic influence factor is selected according to the numerical interval of the current unit parameter deviation. The difference between the current unit parameter deviation and the previous unit parameter deviation is calculated. The sign of the difference determines the direction of the deviation change trend: a positive difference indicates an increasing deviation, and a negative difference indicates a decreasing deviation. Based on the direction of the deviation change trend, the selected basic influence factor is directionally adjusted. When the deviation change trend is increasing, the basic influence factor is multiplied by 1.2; when the deviation change trend is decreasing, the basic influence factor is multiplied by 0.8 to obtain the preliminary influence factor. The actual influence factor sequence for 20 historical moments is obtained, and the value with the highest frequency is selected as the frequency correction value. The preliminary impact factor and the frequency correction value are combined by multiplying the preliminary impact factor by 0.7 and adding the frequency correction value by 0.3 to generate the final corrected impact factor.

[0078] The state vector correction is obtained by multiplying the prediction bias, the correction influence factor, and the fixed learning step size. The fixed learning step size is 0.01, a value determined experimentally to control the correction magnitude. The calculation process for the state vector correction is as follows: multiply the prediction bias by the correction influence factor, and then multiply by the fixed learning step size. For example, if the prediction bias is 0.2 MPa, the correction influence factor is 0.3, and the fixed learning step size is 0.01, then the state vector correction is 0.0006 MPa. This calculation process ensures that the correction can respond to the prediction bias in a timely manner without causing system instability due to excessive correction.

[0079] The state vector correction is added to the state vector from the previous moment to complete the feedback correction of the current state vector. The state vector from the previous moment is read from memory and contains multiple components, each representing a different fluid dynamic characteristic. When adding the state vector correction to the state vector from the previous moment, each corresponding component is added separately. For example, if the state vector from the previous moment is [0.5, 0.3, 0.2] and the state vector correction is [0.001, 0.0005, 0.0003], then the corrected state vector will be [0.501, 0.3005, 0.2003]. By adding components one by one, precise adjustment of the state vector is achieved, allowing the state vector to better reflect the actual fluid dynamic characteristics.

[0080] The working principle of this invention is as follows: First, real-time pressure and medium flow values ​​at the nozzle inlet are synchronously collected by pressure and flow sensors to form a set of real-time operating parameters. Second, based on the current and historical operating parameter sequences, instantaneous change vectors are obtained through sliding window covariance calculation and differentiation processing, and then merged and updated with the state vector of the previous moment to construct a state vector characterizing the current fluid dynamic characteristics. Then, using the current state vector and valve opening as initial conditions, multiple pressure prediction sequences are generated by matching the pressure change pattern library through dynamic time warping. Combined with rolling optimization calculations based on dynamic weight adjustment, the valve control sequence that optimizes the pressure tracking deviation and valve action amplitude is solved. Next, the control quantity is extracted from the optimal sequence, and combined with the calculation of a smoothing factor based on the change trend, a smooth, instantaneous control command is generated to drive the regulating valve. Finally, in the next sampling period, the unit parameter deviation is calculated using the ratio of the prediction deviation to the state vector derived parameters. The state vector correction amount is determined through a correction influence factor with directional adjustment and frequency correction, completing the feedback correction of the state vector and forming closed-loop control.

[0081] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method of pressure smart compensation adjustment of a multi-stage adjustable atomizing nozzle, characterized in that, Includes the following steps: S1: Obtain the real-time operating parameter set of the nozzle system; The set of real-time operating parameters includes the real-time pressure value at the nozzle inlet and the real-time flow rate of the medium flowing into the nozzle. S2: Based on the real-time operating parameter set, construct and update a state vector describing the current fluid dynamic characteristics; S3: Using the current state vector and valve opening as initial conditions, combined with the preset pressure target value, perform rolling optimization calculation within the finite prediction time domain; the goal of the rolling optimization calculation is to solve for an optimal valve control sequence that minimizes the cumulative deviation between the predicted system pressure trajectory and the pressure target value within the prediction time domain, while constraining the valve's action amplitude. S4: Extract the first control quantity from the optimal valve control sequence as an immediate control command and output it to the nozzle control valve actuator to adjust the valve opening; S5: In the next sampling period, collect a new set of real-time operating parameters, calculate the prediction deviation between the predicted pressure value and the actual pressure value at the previous moment, and use the prediction deviation to perform feedback correction on the state vector.

2. The method of claim 1, wherein, S2 specifically includes: S210: Obtain the real-time pressure value and real-time medium flow rate value at the current time and multiple previous sampling times to form a pressure parameter sequence and a flow rate parameter sequence; S220: By calculating the covariance relationship between the pressure parameter sequence and the flow parameter sequence over time, an instantaneous change vector reflecting the sensitivity of pressure to instantaneous flow rate changes is generated; S230: The instantaneous change vector is fused with the state vector of the previous moment, and the derived parameters are recursively updated to construct the state vector of the current moment.

3. The method of claim 2, wherein the pressure compensation adjustment is made by adjusting the pressure of the fluid to the nozzle. S220 specifically includes: S221: Divide the pressure parameter sequence and the flow parameter sequence into multiple overlapping sliding data windows respectively; S222: For each corresponding sliding data window, calculate the covariance between pressure data points and flow data points within the sliding data window, and arrange the covariance calculation results of all windows in chronological order to form a covariance sequence; S223: Differentiate the covariance sequence to obtain the covariance rate of change sequence, and use the covariance rate of change sequence as the instantaneous change vector.

4. The intelligent pressure compensation and adjustment method for a multi-stage adjustable atomizing nozzle according to claim 1, characterized in that, S3 specifically includes: S310: Based on the derived parameters in the state vector and the current valve opening, generate multiple pressure prediction sequences corresponding to multiple consecutive moments in the future; S320: Calculate the cumulative pressure deviation between the pressure prediction sequence and the pressure target value, and calculate the valve opening change required to generate the pressure prediction sequence. S330: Multiply the cumulative pressure deviation and the corresponding valve opening change by the dynamically adjusted weighting coefficient respectively, and then add them together to obtain the comprehensive evaluation value of each pressure prediction sequence; S340: Select the valve control sequence corresponding to the pressure prediction sequence with the smallest comprehensive evaluation value as the optimal valve control sequence.

5. The intelligent pressure compensation and adjustment method for a multi-stage adjustable atomizing nozzle according to claim 4, characterized in that, S310 specifically includes: S311: Pre-set library of pressure change patterns under various typical fluid dynamic characteristics, where each pressure change pattern is associated with a historical state vector segment; S312: Perform dynamic time warping matching between the current state vector and each historical state vector segment in the pressure change pattern library, and calculate the matching degree with each historical state vector segment. S313: Select the pressure change pattern corresponding to the historical state vector segment with the highest matching degree as the basic prediction pattern, and adjust the amplitude of the basic prediction pattern according to the ratio between the current valve opening and the historical valve opening. S314: Based on the scaled and adjusted pressure change pattern, and combined with the current real-time pressure value, generate multiple pressure prediction sequences.

6. The intelligent pressure compensation and adjustment method for a multi-stage adjustable atomizing nozzle according to claim 1, characterized in that, S4 specifically includes: S410: Obtain the optimal valve control sequence, and extract the first two control variables in the optimal valve control sequence as the current control variable and the preparation control variable for the next moment; S420: Calculate the difference between the current control quantity and the prepared control quantity at the next moment to obtain the trend of control quantity change; S430: Calculate the corresponding smoothing factor based on the absolute value of the trend of change of the control quantity; S440: The current control quantity is weighted and averaged with the actual control command at the previous moment, where the weight is determined by the smoothing factor, to generate the instantaneous control command at the current moment and output it to the regulating valve actuator.

7. The intelligent pressure compensation and adjustment method for a multi-stage adjustable atomizing nozzle according to claim 6, characterized in that, S430 specifically includes: S431: Divide the absolute value of the trend of change of the control quantity into multiple consecutive numerical intervals, and associate each numerical interval with a basic smoothing coefficient. S432: Select the corresponding basic smoothing coefficient based on the numerical range of the absolute value of the current control variable change trend; S433: Obtain the smoothing factor sequence actually used in the historical control cycle and calculate the statistical characteristic value of the smoothing factor sequence; S434: Calculate the ratio of the statistical characteristic value to the preset fixed benchmark value to obtain the historical smoothing correction amount, and multiply the historical smoothing correction amount with the basic smoothing coefficient to obtain the smoothing factor.

8. The intelligent pressure compensation and adjustment method for a multi-stage adjustable atomizing nozzle according to claim 1, characterized in that, S5 specifically includes: S510: Calculate the ratio of the prediction deviation to the derived parameters in the current state vector to obtain the unit parameter deviation. S520: Based on the absolute value of the unit parameter deviation, the corresponding correction influence factor is determined through a preset mapping relationship; S530: Multiply the prediction bias, the correction impact factor and a fixed learning step size to obtain the state vector correction amount; S540: Add the state vector correction amount to the state vector at the previous moment to complete the feedback correction of the current state vector.

9. The intelligent pressure compensation adjustment method for a multi-stage adjustable atomizing nozzle according to claim 8, characterized in that, The S520 specifically includes: S521: Divide the absolute value of the unit parameter deviation into multiple continuous numerical intervals, and pre-set a basic influence factor for each interval; S522: Calculate the difference between the current unit parameter deviation and the unit parameter deviation at the previous moment, and determine the direction of the deviation change trend based on the sign of the difference; S523: Based on the direction of the deviation change trend, the selected basic influencing factors are directionally adjusted to obtain preliminary influencing factors; S524: Obtain the actual impact factor sequence at multiple historical moments, select the value with the highest frequency as the frequency correction value, and merge the preliminary impact factor with the frequency correction value to generate the final corrected impact factor.

Images