Abrasion state monitoring and early warning system for rotary nozzle of unmanned aerial vehicle
By combining multi-source data acquisition and spray fluid dynamics characteristic inversion module with CFD simulation, the wear status of nozzles can be directly monitored, which solves the problem of inaccurate wear monitoring in existing technologies, realizes accurate assessment and timely early warning of nozzle wear, and improves the reliability and economy of UAV operations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-12
- Publication Date
- 2026-03-27
AI Technical Summary
Existing drone rotating nozzle wear monitoring technology is susceptible to environmental interference, making it impossible to accurately assess wear levels and remaining lifespan, resulting in inaccurate timing of maintenance.
The system employs a multi-source data acquisition module, a spray fluid dynamics characteristic inversion module, a multi-modal fusion wear assessment module, and an environmental adaptive early warning decision module. It combines CFD simulation to invert the wear parameters of the nozzle, directly monitors the geometric changes of the flow channel by acquiring the spray flow field through high-speed images, and accurately judges the wear level by combining a dual-channel deep learning network.
It enables physical monitoring of nozzle wear, improves the comprehensiveness and accuracy of monitoring, ensures timely and reliable early warning, reduces misjudgments and resource waste, and enhances the continuity and economy of UAV operations.
Smart Images
Figure CN121740141A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of unmanned aerial vehicles, and particularly relates to an unmanned aerial vehicle rotating nozzle wear state monitoring and early warning system. BACKGROUND
[0002] Unmanned aerial vehicle rotating nozzle wear monitoring refers to a systematic technical solution for dynamically monitoring and evaluating the real-time state of a rotating nozzle in an unmanned aerial vehicle agricultural plant protection, fire extinguishing and other operation scenarios through integrated sensor technology, data analysis algorithm and Internet of Things platform. As a core executive component, the wear degree of the rotating nozzle directly affects the spray uniformity, operation efficiency and equipment life. The traditional manual inspection method has problems such as low efficiency, large error and delayed response. The monitoring system usually installs acceleration sensors, pressure sensors or high-precision cameras inside or around the nozzle, collects key parameters such as nozzle speed, vibration frequency, flow change and atomized particle distribution in real time, combines edge computing and cloud big data analysis, constructs a wear characteristic model, realizes early warning and accurate quantitative evaluation of abnormal states such as nozzle bearing sealing, blade deformation and flow passage blockage, and pushes wear trend reports and maintenance suggestions through a visual interface, which can effectively reduce the probability of sudden failure downtime, prolong the service life of the nozzle by 20% to 30%, and at the same time ensure the operation accuracy and safety, and provide data support for intelligent equipment operation and maintenance management.
[0003] However, the existing technology mainly relies on indirect signal monitoring such as vibration current to monitor the wear of the nozzle, which is easily disturbed by environmental factors such as unmanned aerial vehicle flight noise, liquid impurities and the like, cannot directly associate the geometric change of the flow passage caused by wear, and lacks direct inversion of wear physical nature parameters and multi-modal feature fusion, which makes it difficult to accurately evaluate the wear grade and remaining life, and leads to inaccurate maintenance timing. SUMMARY
[0004] The purpose of the present application is to solve the above-mentioned problems, and to provide an unmanned aerial vehicle rotating nozzle wear state monitoring and early warning system.
[0005] The technical scheme adopted by the present application is as follows: an unmanned aerial vehicle rotating nozzle wear state monitoring and early warning system, the system comprising: a multi-source data acquisition module, a data preprocessing module, a spray fluid dynamics characteristic inversion module, a multi-modal fusion wear evaluation module, an environment adaptive early warning decision module and a man-machine interaction module.
[0006] The spray fluid dynamics characteristic inversion module is internally provided with a flow field modeling and simulation submodule, a flow field measurement and feature extraction submodule, a target function construction and optimization submodule and a wear parameter output and quantization submodule.
[0007] The spray fluid dynamics characteristic inversion module is internally provided with a flow field modeling and simulation submodule, a flow field measurement and characteristic extraction submodule, and a wear parameter output and quantization submodule.
[0008] In a preferred embodiment, the multi-source data acquisition module includes a vibration sensing unit, a motor current monitoring unit, a fluid parameter sensing unit, a high-speed image acquisition unit, and an environmental sensing unit. The vibration sensing unit is configured with a piezoelectric acceleration sensor installed on the spray head motor housing, supporting three-axis measurement and a sampling rate of 1 kHz. The motor current monitoring unit is connected in series with a Hall effect current sensor in the spray head motor power supply loop, with an accuracy controlled at ±1%. The fluid parameter sensing unit integrates a miniature pressure transmitter and an ultrasonic flowmeter, with the pressure transmitter installed on the spray head inlet pipeline, a range of 0-1 MPa, and the ultrasonic flowmeter installed on the spray head upstream pipeline, a range of 0.1-10 L / min. The high-speed image acquisition unit consists of a 1000 fps miniature high-speed camera and an LED fill light, with a camera resolution of 640x480, installed below the side of the spray head. The environmental sensing unit includes a temperature and humidity sensor, a liquid conductivity sensor, and a pH sensor, integrated and installed at the outlet of the chemical tank.
[0009] In a preferred embodiment, the data preprocessing module includes a vibration signal preprocessing unit, a current signal preprocessing unit, a fluid parameter preprocessing unit, and an image data preprocessing unit. The vibration signal preprocessing unit uses a wavelet threshold denoising method to suppress noise in the original vibration signal, and then extracts frequency domain and time domain features through Fourier transform and wavelet packet decomposition. The current signal preprocessing unit is configured with a sliding window filtering module, with a window size of 50 ms, and calculates time domain features such as mean, variance, and pulse factor of the filtered signal. The fluid parameter preprocessing unit removes the trend items of the pressure and flow signals through polynomial fitting, and further calculates the fluctuation coefficient and peak deviation. The image data preprocessing unit includes an edge detection module and a particle image velocimetry algorithm module, the edge detection module performs Canny algorithm on the spray head appearance image to extract contour features, and the particle image velocimetry algorithm module calculates two-dimensional velocity distribution and turbulence intensity after deblurring processing of high-speed spray images.
[0010] In a preferred embodiment, the flow field modeling and simulation submodule includes a geometric parameter input unit, a CFD simulation engine, and an ideal characteristic output unit.
[0011] The geometric parameter input unit receives the original outlet diameter, cone angle, surface roughness, and physical properties such as the density and viscosity of the liquid medicine of the un-worn nozzle. The CFD simulation engine has a three-dimensional grid generator and a fluid solver inside. The three-dimensional grid generator uses tetrahedral unstructured grid and encrypts the flow passage wall. The fluid solver integrates the k-ε turbulence model and the VOF two-phase flow model to simulate the flow field distribution under different inlet pressure and flow rate. The ideal characteristic output unit converts the simulation obtained flow field data into a standardized data matrix, and outputs the spatially gridded theoretical velocity field, spray direction angle and turbulence intensity, with the grid size being 0.5mm x 0.5mm.
[0012] In a preferred embodiment, the flow field measurement and feature extraction submodule includes a PIV image processing unit, a flow field gridding unit, and a multi-feature extraction unit. The PIV image processing unit receives the spray image sequence taken by the high-speed camera, and calculates the particle displacement by the particle matching algorithm based on the cross-correlation method to generate the actual velocity vector field. The flow field gridding unit spatially aligns the actual velocity vector field according to the same grid size as the ideal flow field, ensuring that the comparison basis of the theoretical and actual flow fields is consistent. The multi-feature extraction unit extracts three types of core features from the gridded actual flow field, namely the velocity magnitude, the direction angle, and the turbulence intensity. The direction angle is defined as the included angle between the spray velocity vector and the nozzle axis, and the turbulence intensity is calculated by the ratio of the root mean square of velocity fluctuation to the average velocity.
[0013] In a preferred embodiment, the target function construction and optimization submodule includes three core functional units: a target function construction unit, an optimization algorithm execution unit, and a convergence judgment and parameter output unit.
[0014] The target function construction unit constructs a comprehensive error function based on the multi-feature differences between the ideal flow field and the actual flow field. First, the theoretical features output by the ideal flow field modeling submodule are called: the velocity field U ideal , the direction angle a ideal , and the turbulence intensity I ideal , and the experimental features output by the actual flow field measurement submodule are called: U actual , a actual , and I actual . The relative deviations of the three types of features are fused by the weighted squared error formula. The velocity field error reflects the change of the flow passage cross-sectional area, the direction angle error reflects the inclination of the cone surface wear, and the turbulence intensity error reflects the change of the inner wall roughness. The weight coefficients are determined according to the sensitivity experiment of the features to wear.
[0015] Optimization algorithm execution unit: take the wear parameters (outlet diameter D, cone angle θ, surface roughness Ra) as optimization variables, and minimize the objective function. Genetic algorithm is used to realize iterative optimization: a set of wear parameter combinations are randomly generated in the initialization stage; the selection operation retains the parameter combinations with smaller target function values; the crossover operation generates new combinations through parameter recombination (such as the crossover probability of D is set to 0.7); the mutation operation randomly disturbs the parameters (such as the mutation step of Ra is 5% of the original value), and the optimal solution is approached through multiple generations of evolution.
[0016] Convergence judgment and parameter output unit: monitor the change of the objective function in the optimization process, and output the optimal wear parameters after convergence. Double convergence conditions are set: one is that the target function value JJ is less than the preset error threshold (5%, which corresponds to the theoretical and actual deviation of the flow field characteristics within the engineering allowable range); the other is that the JJ value change of continuous 5 generations of optimization is less than 10 −4 (avoid local optimum). After convergence, the wear parameters (D*, θ*, Ra*) obtained by the current optimization are passed to the wear parameter quantization submodule to complete the inverse closed loop.
[0017] The multi-feature fusion objective function formula is:
[0018] ;
[0019] In the formula:
[0020] J(D, θ, Ra) represents the target function value (dimensionless), which represents the comprehensive error between the ideal and actual flow fields, and the smaller the value, the more matched the wear parameters and the actual state are;
[0021] D, θ, Ra represent the wear parameters to be optimized (outlet diameter of the nozzle, cone angle, and surface roughness, respectively);
[0022] N represents the number of flow field grid sampling points (dimensionless), which takes the effective grid in the range of 10-50 mm range of the spray cone (usually N=1000\sim2000N=1000∼2000, which ensures that the flow field characteristics cover the core area of the spray);
[0023] U actual,i ,U ideal,i represents the actual / ideal speed of the i-th grid point (m / s), and the ideal speed Uideal,i is calculated by the current wear parameters (D, θ, Ra) through CFD simulation;
[0024] α actual,i ,α ideal,i represents the actual / ideal spray direction angle of the i-th grid point, and the direction angle is the included angle between the spray velocity vector and the nozzle axis;
[0025] I actual,i ,Iideal,i Ri represents the actual / ideal turbulence intensity (dimensionless) of the i-th grid point, defined as the ratio of the root mean square of velocity fluctuation to the mean velocity;
[0026] ωU, ωα, ωI represent the characteristic weight coefficients (dimensionless, satisfying ωU+ωα+ωI=1), ωU=0.5, ωα=0.3, ωI=0.2, because the velocity field is most sensitive to wear.
[0027] In a preferred embodiment, the wear parameter output and quantification module includes a parameter comparison unit, a wear index generation unit and a data standardization unit. The parameter comparison unit calls the original parameters of the un-worn nozzle and calculates the difference with the optimized wear parameters. The wear index generation unit outputs three core wear indexes based on the difference calculation results, which are the outlet diameter increase value, the cone angle change value and the surface roughness increment. The outlet diameter increase value is the difference between the worn diameter and the original diameter, the cone angle change value is the difference between the worn cone angle and the original cone angle, and the surface roughness increment is the difference between the worn roughness and the original roughness. The data standardization unit converts the above wear indexes into dimensionless parameters, and realizes standardization processing through the ratio of wear indexes to original parameters, which is convenient for subsequent wear grade evaluation module to call directly.
[0028] In a preferred embodiment, the multi-modal fusion wear evaluation module includes a feature vector construction unit, a dual-channel deep learning network unit and a wear grade and life prediction unit. The feature vector construction unit receives multi-source features from the data preprocessing module, including vibration spectrum entropy, current pulse factor, pressure fluctuation coefficient, spray velocity gradient, and inverted diameter increase value and surface roughness increment, and integrates them into a high-dimensional feature vector. The dual-channel deep learning network unit includes a sensor feature extraction channel and a wear profile feature extraction channel. The sensor feature extraction channel adopts a three-layer fully connected network and uses a ReLU activation function, and the wear profile feature extraction channel is configured with a one-dimensional CNN structure and a convolution kernel size of 5x1. The features of the two channels are spliced and input into the subsequent processing layer. The wear grade and life prediction unit integrates a softmax classifier and an LSTM network. The softmax classifier outputs a 0-4 grade wear grade, and the LSTM network predicts the remaining service life of the nozzle based on historical feature sequences.
[0029] In a preferred embodiment, the environment adaptive early warning decision module includes an environment parameter mapping unit, a pre-warning threshold dynamic adjustment unit and a hierarchical pre-warning instruction generation unit. The environment parameter mapping unit receives the temperature and humidity, liquid conductivity and pH value data output by the environment sensing module, and calculates the influence coefficient of environmental factors on the wear rate through a mapping model trained by historical data sets. The pre-warning threshold dynamic adjustment unit dynamically corrects the residual life threshold corresponding to the wear grade according to the influence coefficient output by the environment parameter mapping unit, such as lowering the residual life threshold of moderate wear in a high-impurity liquid environment. The hierarchical pre-warning instruction generation unit generates a three-level pre-warning instruction according to the real-time wear grade output by the multi-modal fusion module and the corrected threshold, and the 1st level pre-warning triggers the ground station yellow prompt, the 2nd level pre-warning starts the buzzer alarm and displays the replacement countdown, and the 3rd level pre-warning sends the unmanned aerial vehicle automatic return instruction.
[0030] In a preferred embodiment, the human-computer interaction module includes a ground station data visualization unit, a data storage and management unit and a remote interaction interface unit. The ground station data visualization unit configures a wear state dashboard, a sensor curve display window and a spray image display area, the dashboard displays the 0-4 wear grade in real time, the curve window draws the vibration spectrum and current waveform diagram, and the image display area displays the PIV processed spray speed field thermal map. The data storage and management unit includes an unmanned aerial vehicle end SD card cache module and a cloud database module, the SD card cache stores the original data and feature data of the last 3 operations, and the cloud database uses MySQL to store the historical wear trend curve and pre-warning record. The remote interaction interface unit provides a mobile phone APP interface and a PC management platform interface, the mobile phone APP interface supports viewing real-time pre-warning information and historical wear data, and the PC management platform interface allows exporting wear analysis reports and generating maintenance plan suggestions.
[0031] In summary, due to the adoption of the above technical solutions, the present application has the following advantages:
[0032] 1. In the present application, the physical nature of the spray head wear state is monitored through the innovative design of the spray fluid dynamics characteristic inversion module. Traditional wear monitoring relies on indirect signals such as vibration and current, which are easily disturbed by unmanned aerial vehicle flight noise, liquid impurities, etc. However, this module captures the spray flow field through high-speed image acquisition, combines CFD simulation to invert core wear parameters such as outlet diameter cone angle and surface roughness, directly relates to the geometric changes of the flow channel caused by wear, avoids the technical difficulty of installing sensors on rotating parts, and makes the monitoring results closer to the actual wear state of the spray head. The multi-source data acquisition module synchronously acquires auxiliary signals such as vibration, current, pressure and flow, and the physical parameters obtained by inversion form a complement, further improving the comprehensiveness of monitoring and reducing false positives caused by single signal anomalies.
[0033] 2、In the present application, the system's environment adaptive early warning decision and multi-modal fusion evaluation mechanism ensure the timeliness and reliability of the early warning. The multi-modal fusion module fuses the inversion wear parameters and vibration spectrum entropy current pulse factor and other characteristics, accurately judges the wear grade through the double-channel deep learning network, and the environment adaptive module adjusts the early warning threshold in real time according to the conductivity, temperature and humidity of the liquid and other real-time conditions, avoids the lag of early warning caused by accelerated wear in harsh environments such as high-impurity liquid, and the like. The man-machine interaction module displays the wear state, remaining life and spray flow field image in real time through the ground station, combines the graded early warning instructions, so that the operator can intuitively master the nozzle condition, arrange maintenance and replacement in advance, reduce the interruption of operation caused by sudden failure of the nozzle, avoid resource waste caused by excessive maintenance, and improve the continuity and economy of unmanned aerial vehicle plant protection operation. BRIEF DESCRIPTION OF DRAWINGS
[0034] Fig. 1 is the overall system block diagram of the present application;
[0035] Fig. 2 is the system block diagram of the spray fluid dynamics characteristic inversion module in the present application. DETAILED DESCRIPTION
[0036] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application, and are not used to limit the present application.
[0037] Example:
[0038] Reference Figs. 1-2 The unmanned aerial vehicle rotating nozzle wear state monitoring and early warning system comprises a multi-source data acquisition module, a data preprocessing module, a spray fluid dynamics characteristic inversion module, a multi-modal fusion wear evaluation module, an environment adaptive early warning decision module and a man-machine interaction module.
[0039] The spray fluid dynamics characteristic inversion module is internally provided with a flow field modeling and simulation submodule, a flow field measurement and feature extraction submodule, a target function construction and optimization submodule, and a wear parameter output and quantization submodule.
[0040] The spray fluid dynamics characteristic inversion module is internally provided with a flow field modeling and simulation submodule, a flow field measurement and feature extraction submodule, and a wear parameter output and quantization submodule.
[0041] The multi-source data acquisition module includes a vibration sensing unit, a motor current monitoring unit, a fluid parameter sensing unit, a high-speed image acquisition unit, and an environmental sensing unit. The vibration sensing unit is configured with a piezoelectric acceleration sensor, which is installed on the nozzle motor housing, supports three-axis measurement, and has a sampling rate of 1 kHz. The motor current monitoring unit is connected with a Hall effect current sensor in series to the nozzle motor power supply loop, and the accuracy is controlled within ±1%. The fluid parameter sensing unit integrates a miniature pressure transmitter and an ultrasonic flowmeter. The pressure transmitter is installed on the nozzle inlet pipeline, with a range of 0-1 MPa, and the ultrasonic flowmeter is installed on the nozzle upstream pipeline, with a range of 0.1-10 L / min. The high-speed image acquisition unit is composed of a 1000 fps miniature high-speed camera and an LED fill light. The camera has a resolution of 640x480 and is installed below the side of the nozzle. The environmental sensing unit includes a temperature and humidity sensor, a liquid conductivity sensor, and a pH sensor, which are integrated and installed at the outlet of the chemical tank.
[0042] The data preprocessing module includes a vibration signal preprocessing unit, a current signal preprocessing unit, a fluid parameter preprocessing unit, and an image data preprocessing unit. The vibration signal preprocessing unit uses wavelet threshold denoising method to suppress noise of the original vibration signal, and extracts frequency domain and time domain features through Fourier transform and wavelet packet decomposition. The current signal preprocessing unit is configured with a sliding window filtering module, with a window size of 50 ms. The mean, variance, and pulse factor of the filtered signal are calculated as time domain features. The fluid parameter preprocessing unit removes the trend items of the pressure and flow signals by polynomial fitting, and further calculates the fluctuation coefficient and peak deviation. The image data preprocessing unit includes an edge detection module and a particle image velocimetry algorithm module. The edge detection module extracts the contour features of the nozzle appearance image by performing Canny algorithm. The particle image velocimetry algorithm module calculates the two-dimensional velocity distribution and turbulence intensity after deblurring the high-speed spray image.
[0043] The flow field modeling and simulation submodule includes a geometric parameter input unit, a CFD simulation engine, and an ideal feature output unit.
[0044] The geometric parameter input unit receives the original outlet diameter, cone angle, surface roughness of the un-worn nozzle, and the physical properties of the liquid such as density and viscosity. The CFD simulation engine has a three-dimensional grid generator and a fluid solver inside. The three-dimensional grid generator uses tetrahedral unstructured grid and encrypts the flow passage wall. The fluid solver integrates k-ε turbulence model and VOF two-phase flow model to simulate the flow field distribution under different inlet pressure and flow rate. The ideal feature output unit converts the simulated flow field data into a standardized data matrix, outputs the spatially gridded theoretical velocity field, spray direction angle, and turbulence intensity, with a grid size of 0.5 mm x 0.5 mm.
[0045] The flow field measurement and feature extraction submodule includes a PIV image processing unit, a flow field gridding unit, and a multi-feature extraction unit. The PIV image processing unit receives the spray image sequence captured by the high-speed camera, calculates the particle displacement to generate the actual velocity vector field through the particle matching algorithm based on the cross-correlation method. The flow field gridding unit aligns the actual velocity vector field in space according to the same grid size as the ideal flow field, ensuring that the comparison basis of the theoretical and actual flow fields is consistent. The multi-feature extraction unit extracts three core features from the gridded actual flow field, namely the velocity magnitude, the direction angle, and the turbulence intensity. The direction angle is defined as the included angle between the spray velocity vector and the nozzle axis, and the turbulence intensity is calculated by the ratio of the root mean square of the velocity fluctuation to the average velocity.
[0046] The objective function construction and optimization submodule includes three core functional units: an objective function construction unit, an optimization algorithm execution unit, and a convergence determination and parameter output unit.
[0047] The objective function construction unit constructs a comprehensive error function based on the multi-feature differences between the ideal flow field and the actual flow field. First, the theoretical features output by the ideal flow field modeling submodule are called: the velocity field U ideal , the direction angle α ideal , and the turbulence intensity I ideal , and the experimental features output by the actual flow field measurement submodule are called: U actual , α actual , and I actual . The relative deviations of the three features are fused through the weighted squared error formula: the velocity field error reflects the change in the flow passage cross-sectional area, the direction angle error reflects the inclination of the cone surface wear, and the turbulence intensity error reflects the change in the inner wall roughness. The weight coefficients are determined according to the sensitivity experiments of the features to wear.
[0048] The optimization algorithm execution unit takes the wear parameters (the outlet diameter D, the cone angle θ, and the surface roughness Ra) as the optimization variables and minimizes the objective function. The genetic algorithm is used to realize iterative optimization: a set of wear parameter combinations is randomly generated in the initialization stage; the selection operation retains the parameter combinations with smaller objective function values; the crossover operation generates new combinations through parameter recombination (e.g., the crossover probability of D is set to 0.7); the mutation operation randomly perturbs the parameters (e.g., the mutation step of Ra is 5% of the original value), and the optimal solution is approached through multiple generations of evolution.
[0049] The convergence determination and parameter output unit monitors the change of the objective function during the optimization process and outputs the optimal wear parameters after convergence. Double convergence conditions are set: one is that the objective function value JJ is less than the preset error threshold (5%, corresponding to the theoretical and actual deviations of the flow field features within the engineering allowable range); the other is that the JJ value change of continuous 5 generations of optimization is less than 10 −4(Avoid local optimum). After convergence, the current optimization of the wear parameter (D*, θ*, Ra*) is passed to the wear parameter quantification submodule to complete the inverse closed loop.
[0050] The multi-feature fusion objective function formula is:
[0051] ;
[0052] In the formula:
[0053] J(D, θ, Ra) represents the objective function value (dimensionless), which represents the comprehensive error between the ideal and actual flow field, and the smaller the value, the more matched the wear parameters and the actual state;
[0054] D, θ, Ra represent the wear parameters to be optimized (respectively, the nozzle outlet diameter, the cone angle, and the surface roughness);
[0055] N represents the number of flow field grid sampling points (dimensionless), and the effective grid in the range of 10-50 mm range of the spray cone (usually N=1000\sim2000N=1000\sim2000, to ensure that the flow field characteristics cover the spray core area) is taken;
[0056] U actual,i ,U ideal,i represents the actual / ideal speed of the i-th grid point (m / s), and the ideal speed Uideal,i is calculated by the current wear parameter (D, θ, Ra) through CFD simulation;
[0057] α actual,i ,α ideal,i represents the actual / ideal spray direction angle of the i-th grid point, and the direction angle is the included angle between the spray velocity vector and the nozzle axis;
[0058] I actual,i ,I ideal,i represents the actual / ideal turbulence intensity of the i-th grid point (dimensionless), which is defined as the ratio of the root mean square of velocity fluctuation to the average velocity;
[0059] ωU, ωα, ωI represent the feature weight coefficients (dimensionless, ωU+ωα+ωI=1), ωU=0.5, ωα=0.3, ωI=0.2, because the velocity field is most sensitive to wear.
[0060] The wear parameter output and quantization submodule includes a parameter comparison unit, a wear index generation unit, and a data standardization unit. The parameter comparison unit calls the original parameters of the un-worn nozzle and calculates the difference with the optimized wear parameters. The wear index generation unit outputs three core wear indexes based on the difference calculation results, which are the outlet diameter increase value, the cone angle change value, and the surface roughness increment. The outlet diameter increase value is the difference between the worn diameter and the original diameter, the cone angle change value is the difference between the worn cone angle and the original cone angle, and the surface roughness increment is the difference between the worn roughness and the original roughness. The data standardization unit converts the above wear indexes into dimensionless parameters by standardizing the ratio of wear indexes to original parameters, which facilitates the subsequent wear grade evaluation module to directly call.
[0061] The multi-modal fusion wear evaluation module includes a feature vector construction unit, a dual-channel deep learning network unit, and a wear grade and life prediction unit. The feature vector construction unit receives multi-source features from the data preprocessing module, including vibration spectrum entropy, current pulse factor, pressure fluctuation coefficient, spray velocity gradient, and inverted diameter increase value and surface roughness increment, and integrates them into a high-dimensional feature vector. The dual-channel deep learning network unit includes a sensor feature extraction channel and a wear profile feature extraction channel. The sensor feature extraction channel uses a three-layer fully connected network with a ReLU activation function, and the wear profile feature extraction channel is configured with a one-dimensional CNN structure with a convolution kernel size of 5x1. The features of the two channels are concatenated and input into the subsequent processing layer. The wear grade and life prediction unit integrates a softmax classifier and an LSTM network. The softmax classifier outputs a 0-4 wear grade, and the LSTM network predicts the remaining service life of the nozzle based on historical feature sequences.
[0062] The environment adaptive early warning decision module includes an environmental parameter mapping unit, a warning threshold dynamic adjustment unit, and a hierarchical warning instruction generation unit. The environmental parameter mapping unit receives temperature and humidity, liquid conductivity, and pH value data from the environmental sensing module, and calculates the influence coefficient of environmental factors on wear rate through a mapping model trained by historical data sets. The warning threshold dynamic adjustment unit dynamically corrects the preset remaining life threshold corresponding to the wear grade based on the influence coefficient output by the environmental parameter mapping unit, such as lowering the moderate wear remaining life threshold in a high-impurity liquid environment. The hierarchical warning instruction generation unit generates a three-level warning instruction based on the real-time wear grade output by the multi-modal fusion module and the corrected threshold, i.e., level 1 warning triggers a yellow prompt on the ground station, level 2 warning starts a buzzer alarm and displays a replacement countdown, and level 3 warning sends an unmanned aerial vehicle automatic return instruction.
[0063] The human-computer interaction module includes a ground station data visualization unit, a data storage and management unit and a remote interaction interface unit. The ground station data visualization unit is configured with a wear state instrument panel, a sensor curve display window and a spray image display area. The instrument panel displays the 0-4 grade wear level in real time, the curve window draws the vibration spectrum and current waveform diagram, and the image display area displays the PIV processed spray speed field thermal map. The data storage and management unit includes an unmanned aerial vehicle end SD card cache module and a cloud database module. The SD card cache stores the original data and feature data of the last 3 operations, and the cloud database uses MySQL to store the historical wear trend curve and warning record. The remote interaction interface unit provides a mobile phone APP interface and a PC management platform interface. The mobile phone APP interface supports viewing real-time warning information and historical wear data, and the PC management platform interface allows export of wear analysis reports and generation of maintenance plan suggestions.
[0064] From the above, it can be seen that in the present application, through the innovative design of the spray fluid dynamics characteristic inversion module, the physical nature monitoring of the wear state of the spray head is realized. Traditional wear monitoring relies on indirect signals such as vibration and current, which is easily disturbed by unmanned aerial vehicle flight noise, liquid impurities, etc. However, the module collects the spray flow field through high-speed image acquisition, combines CFD simulation to invert core wear parameters such as outlet diameter cone angle surface roughness, directly relates to the geometric change of the flow passage caused by wear, avoids the technical difficulty of installing sensors on rotating parts, and makes the monitoring result closer to the actual wear state of the spray head. The multi-source data acquisition module synchronously acquires auxiliary signals such as vibration, current, pressure and flow, and the physical parameters obtained by inversion form a complement, further improving the comprehensiveness of the monitoring and reducing the misjudgment caused by single signal anomaly.
[0065] In the present application, the environmental adaptive early warning decision and multi-modal fusion evaluation mechanism of the system ensure the timeliness and reliability of the early warning. The multi-modal fusion module fuses the inversion wear parameters with vibration spectrum entropy, current pulse factor and other features, accurately judges the wear grade through a double-channel deep learning network, and the environmental adaptive module adjusts the early warning threshold in real time according to the liquid conductivity, temperature and humidity, etc., avoiding the lag of early warning caused by accelerated wear in harsh environments such as high-impurity liquid. The human-computer interaction module displays the wear state remaining life and spray flow field image in real time through the ground station, combined with the graded early warning instruction, so that the operator can intuitively master the spray head condition, arrange maintenance and replacement in advance, reduce the interruption of work caused by sudden failure of the spray head, avoid resource waste caused by excessive maintenance, and improve the continuity and economy of unmanned aerial vehicle plant protection operation.
[0066] It is to be noted that, in the present text, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", "has", "having", "includes", "including", or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or even inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a", "has... a", "includes... a", or "has... a" does not, without more constraints, preclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.
[0067] The above examples are merely intended to illustrate the technical solutions of the present application, but not to limit the same; even though the present application has been described in detail with reference to the foregoing examples, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing examples can be modified, or some technical features thereof can be replaced by equivalents; and such modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A UAV rotating nozzle wear condition monitoring and early warning system, characterized in that: The system includes: a multi-source data acquisition module, a data preprocessing module, a spray fluid dynamics characteristic inversion module, a multi-modal fusion wear assessment module, an environmental adaptive early warning decision module, and a human-computer interaction module; The spray fluid dynamics feature inversion module is internally configured with a flow field modeling and simulation submodule, a flow field measurement and feature extraction submodule, an objective function construction and optimization submodule, and a wear parameter output and quantification submodule; The spray fluid dynamics feature inversion module includes sub-modules for flow field modeling and simulation, flow field measurement and feature extraction, and wear parameter output and quantification.
2. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The multi-source data acquisition module includes a vibration sensing unit, a motor current monitoring unit, a fluid parameter sensing unit, a high-speed image acquisition unit, and an environmental sensing unit. The vibration sensing unit is equipped with a piezoelectric acceleration sensor, which is installed on the nozzle motor housing. The motor current monitoring unit connects a Hall effect current sensor in series with the nozzle motor power supply circuit, a pressure transmitter is installed in the nozzle inlet pipe, and an ultrasonic flow meter is installed in the nozzle upstream pipe. The high-speed image acquisition unit consists of a 1000fps miniature high-speed camera and an LED supplementary light, with the camera installed on the lower side of the nozzle. The environmental sensing unit includes a temperature and humidity sensor, a liquid conductivity sensor, and a pH sensor, which are integrated and installed at the medicine tank outlet.
3. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The data preprocessing module includes a vibration signal preprocessing unit, a current signal preprocessing unit, a fluid parameter preprocessing unit, and an image data preprocessing unit. The vibration signal preprocessing unit uses wavelet threshold denoising to suppress noise in the original vibration signal; The current signal preprocessing unit is equipped with a sliding window filtering module with a window size of 50ms. It calculates the mean, variance, and impulse factor of the filtered signal in the time domain. The fluid parameter preprocessing unit removes the trend terms of pressure and flow signals through polynomial fitting and further calculates the fluctuation coefficient and peak deviation. The image data preprocessing unit includes an edge detection module and a particle image velocimetry algorithm module. The edge detection module performs the Canny algorithm to extract contour features from the nozzle appearance image, and the particle image velocimetry algorithm module performs deblurring on the high-speed spray image and then calculates the two-dimensional velocity distribution and turbulence intensity.
4. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The flow field modeling and simulation submodule includes a geometric parameter input unit, a CFD simulation engine, and an ideal feature output unit. The geometric parameter input unit receives the original outlet diameter, cone angle, surface roughness, and physical properties of the liquid, such as density and viscosity, of the unworn nozzle. The CFD simulation engine is equipped with a 3D mesh generator and a fluid solver. The 3D mesh generator uses tetrahedral unstructured meshes and refines the flow channel walls. The fluid solver integrates the k-ε turbulence model and the VOF two-phase flow model to simulate the flow field distribution under different inlet pressures and flow rates. The ideal feature output unit converts the simulated flow field data into a standardized data matrix.
5. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The flow field measurement and feature extraction submodule includes a PIV image processing unit, a flow field meshing unit, and a multi-feature extraction unit.
6. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The objective function construction and optimization submodule includes: an objective function construction unit, an optimization algorithm execution unit, and a convergence determination and parameter output unit; The objective function construction unit constructs a comprehensive error function based on the multi-feature differences between the ideal flow field and the actual flow field; The optimization algorithm execution unit uses wear parameters as optimization variables to minimize the objective function; it employs a genetic algorithm to achieve iterative optimization. Convergence determination and parameter output unit: Monitors the change of the objective function during the optimization process, and outputs the optimal wear parameters after determining convergence; The multi-feature fusion objective function formula of the objective function construction and optimization submodule is as follows: ； In the formula: J(D,θ,Ra) represents the objective function value, which characterizes the combined error between the ideal and actual flow fields. The smaller the value, the better the wear parameters match the actual state. D, θ, Ra represent the wear parameters to be optimized; N represents the number of sampling points for the flow field grid, taking the effective grid within a range of 10–50 mm inside the spray cone; U actual,i U ideal,i This represents the actual / ideal velocity magnitude of the i-th grid point. The ideal velocity Uideal,i is calculated by CFD simulation using the current wear parameters (D,θ,Ra). α actual,i ,α ideal,i This represents the actual / ideal spray direction angle of the i-th grid point, where the direction angle is the angle between the spray velocity vector and the nozzle axis. I actual,i ,I ideal,i The actual / ideal turbulence intensity at the i-th grid point is defined as the ratio of the root mean square velocity fluctuation to the average velocity. ωU, ωα, ωI represent the characteristic weight coefficients, ωU=0.5, ωα=0.3, ωI=0.2, because the velocity field is most sensitive to wear.
7. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The wear parameter output and quantification submodule includes a parameter comparison unit, a wear index generation unit, and a data standardization unit. The parameter comparison unit calls the original parameters of the unworn nozzle and calculates the difference between them and the optimized wear parameters. The wear index generation unit outputs three core wear indices based on the difference calculation results: the increase in outlet diameter, the change in cone angle, and the increase in surface roughness. The data standardization unit converts the above wear indices into dimensionless parameters and achieves standardization by comparing the wear indices with the original parameters.
8. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The multimodal fusion wear assessment module includes a feature vector construction unit, a dual-channel deep learning network unit, and a wear level and life prediction unit.
9. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The environmental adaptive early warning decision module includes an environmental parameter mapping unit, an early warning threshold dynamic adjustment unit, and a graded early warning instruction generation unit.
10. The UAV rotating nozzle wear condition monitoring and early warning system as described in claim 1, characterized in that: The human-computer interaction module includes a ground station data visualization unit, a data storage and management unit, and a remote interaction interface unit.