Airplane control surface automatic detection method and system based on virtual instrument and matrix switching

Abstract

The invention discloses an aircraft control surface automatic detection method and system based on virtual instrument and matrix switching, and relates to the technical field of steering engine detection. The aircraft control surface automatic detection method and system based on virtual instrument and matrix switching comprises the following steps: S1, obtaining control surface actuator response data, and preprocessing the control surface actuator response data; s2, a relay is driven to realize channel switching, the stability degree of the relay before and after switching is evaluated, and a stable sampling period is screened; s3, identifying an effective response starting point and a response ending point of each channel, calculating each component index in a response section, and constructing a response feature vector of the control surface actuator; and S4, evaluating the deviation degree between the current response behavior and the ideal state, identifying an abnormal response state, generating a channel-level response behavior tag, and summarizing to form a classification result. The problem that the initial section of a displacement response curve is misjudged due to the fact that transient kick caused by relay switching is not isolated by a time window is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of steering gear detection, and in particular to an automatic detection method and system for aircraft control surfaces based on virtual instruments and matrix switching. Background Art

[0002] As a crucial electric actuator in the flight control system, the rudder actuator is responsible for fine-tuning the flight attitude. Its performance directly affects the aircraft's response accuracy and system stability. In actual engineering, comprehensive testing and characteristic analysis of the rudder actuator's response behavior is a key step in system design verification and fault prediction. To improve testing efficiency, a relay matrix signal switching structure built on a virtual instrument platform is often used, combined with a high-frequency excitation signal output module. Under the coordination of the FPGA control unit, an adjustable-frequency sinusoidal control voltage is applied to multiple actuator samples or test channels in sequence to realize the dynamic response characteristic acquisition of the multi-channel rudder control system.

[0003] However, during channel switching, relay contacts are susceptible to micro-arcing at the moment of conduction, causing a brief rise in the control voltage or waveform distortion. This can lead to a sudden jump in the initial response data, manifesting as premature displacement response or response drift distortion. Existing test systems generally lack time window isolation and data stability verification mechanisms for these relay switching disturbances. This makes it impossible to eliminate or identify unstable responses when the excitation signal is first applied, which can easily lead to problems such as misidentification of the response starting point and misjudgment of the disturbance state.

[0004] Furthermore, actuators exhibit complex nonlinear dynamic response behaviors during operation, including voltage and displacement hysteresis, feedback delay, drive current fluctuations, and mechanical response overshoot. These multi-dimensional coupled characteristics are more pronounced under high-frequency excitation conditions. Traditional methods rely on empirical point selection or static threshold analysis of a single physical quantity. These methods lack the ability to jointly model dynamic processes, making it difficult to accurately identify and classify response behaviors.

[0005] Especially when faced with complex scenarios such as multi-frequency excitation, continuous switching, and parallel channel collaborative testing, extracting stable reference-value response features from multi-source dynamic data, constructing standard response segments, and accurately determining the response start and end points presents a technical bottleneck that existing systems struggle to overcome. Furthermore, abnormal response states often manifest as coordinated deviations of multi-dimensional signals. Without a unified evaluation model and feature vector construction method, automatic classification of abnormal states and generation of channel-level behavioral labels are difficult to achieve, further limiting the test system's intelligent data processing and efficient identification capabilities in rudder control system engineering applications.

[0006] Therefore, in response to the above problems, there is an urgent need for an automatic detection method and system for aircraft control surfaces based on virtual instruments and matrix switching. Summary of the Invention

[0007] Technical problems solved

[0008] In response to the shortcomings of the existing technology, the present invention provides an automatic detection method and system for aircraft control surfaces based on virtual instruments and matrix switching, which solves the problem that transient jumps caused by relay switching are not isolated by the time window, resulting in misjudgment of the initial segment of the displacement response curve.

[0009] Technical Solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: an automatic detection method and system for aircraft rudder surfaces based on virtual instruments and matrix switching, comprising the following steps: S1, dynamically collecting the entire process of rudder actuator operation, obtaining rudder actuator response data, and performing time axis repositioning, dynamic denoising, phase compensation and scale unification on the rudder actuator response data; S2, generating a continuous control excitation signal based on the preprocessed rudder actuator response data, driving a relay to achieve channel switching, and evaluating the stability before and after the relay switching, screening a stable sampling period, and constructing a stable response data set; S3, screening candidate response starting points based on the stable response data set, identifying the effective response starting point and response end point of each channel, calculating the component indicators in the response segment, and constructing a rudder actuator response feature vector; S4, extracting and constructing the response feature vector, evaluating the degree of deviation between the current response behavior and the ideal state, identifying abnormal response states based on the evaluation results, generating channel-level response behavior labels, and summarizing them to form classification results.

[0011] Furthermore, the whole process of dynamic acquisition of the rudder actuator operation is carried out to obtain the rudder actuator response data, and the specific steps of time axis repositioning, dynamic denoising, phase compensation and scale unification of the rudder actuator response data are as follows: the whole process of dynamic acquisition of the rudder actuator operation is carried out to obtain the rudder actuator response data, and the rudder actuator response data includes control excitation voltage, actuator feedback voltage, actuator mechanical displacement, actuator current, excitation frequency, control voltage phase angle, feedback voltage phase angle; dynamic window delay based on relay state synchronization mark A sampling method is used to reposition the time axis of the rudder actuator response data, correcting the transient response drift and starting displacement misjudgment caused by micro-arcing disturbances in the initial stage of relay switching; a sliding fitting method with response derivative constraints is used to dynamically denoise the rudder actuator response data to filter out high-frequency noise caused by the startup phase and inrush current; a delay identification method based on control-response phase correlation is used to identify and compensate for phase errors caused by mechanical lag and sampling offset; and amplitude normalization and response rate standard deviation scaling strategies are used to unify the scale of the rudder actuator response data.

[0012] Furthermore, the specific steps of generating a continuous control excitation signal based on the preprocessed rudder actuator response data and driving the relay to realize channel switching are as follows: based on the preprocessed rudder actuator response data, a digital sine sequence is generated, and a continuous control excitation voltage is output through a digital-to-analog converter; the control excitation voltage is connected to the target channel after power amplification, and the excitation frequency is set by a clock control circuit and output synchronously with the control excitation voltage; a control signal is output through a decoding circuit to drive the relay to switch channels; after the relay switching is completed, the control excitation voltage, actuator feedback voltage and actuator current in a continuous fixed period before and after the relay switching are extracted in the sampling period, and the average values ​​before and after the switching are calculated respectively.

[0013] Furthermore, the specific steps for evaluating the stability before and after the relay switching are as follows: Based on the average values ​​before and after each switching, the stability before and after the relay switching is evaluated: the absolute value of the difference between the average values ​​of the control excitation voltage before and after the relay switching is calculated to obtain the control input change term; the absolute value of the difference between the average values ​​of the actuator feedback voltage before and after the relay switching is calculated to obtain the voltage response change term; the absolute value of the difference between the average values ​​of the actuator current before and after the relay switching is calculated to obtain the current response change term; the control input change term, the voltage response change term and the current response change term are summed to obtain the switching joint disturbance evaluation value.

[0014] Furthermore, the specific steps for screening stable sampling periods and constructing a stable response data set are as follows: compare the switching joint disturbance evaluation value and the switching disturbance threshold in real time. If the switching joint disturbance evaluation value is greater than the switching disturbance threshold, it is determined that there is still a non-steady-state disturbance in the current channel, and the recording of the rudder actuator response data in the current sampling period is suspended, and the judgment is made again after re-evaluation in the next sampling period; if the switching joint disturbance evaluation value is less than or equal to the switching disturbance threshold, it is determined that the current channel is stable, and starting from the current sampling period, the rudder actuator response data in a fixed number of sampling periods are continuously recorded to construct a stable response data set.

[0015] Furthermore, the specific steps of screening candidate response starting points based on the stable response data set and identifying the effective response starting point and response end point of each channel are as follows: based on the stable response data set, extract the maximum value of the actuator current; perform first-order difference on the actuator mechanical displacement signal to obtain the displacement change rate, identify the displacement change trend mutation point, and mark the time corresponding to the trend mutation point as the candidate response starting point; extract the rudder actuator response data of the continuous fixed sampling period starting from the candidate response starting point, and construct the response starting evaluation value: calculate the absolute value of the difference between the control excitation voltage phase angle and the actuator feedback voltage phase angle, divide this absolute value by π and take its inverse relative value to obtain the phase response synergy term; calculate the ratio of the actuator current to the maximum value of the actuator current to obtain the drive strength normalization term; multiply the displacement change rate, the phase response synergy term and the drive strength normalization term in each sampling period, The product result is weightedly accumulated and averaged within a continuous fixed sampling period to obtain a response start evaluation value; the response start evaluation value and the response start threshold are compared in real time. If the response start evaluation value is greater than the response start threshold, the current candidate point is confirmed as a valid response starting point, and the rudder actuator response data after the starting point is intercepted; if the response start evaluation value is lower than the response start threshold, the candidate response starting point is slid backward and the above calculation process is repeated until the first valid response starting point that meets the response start threshold condition is identified; after the valid response starting point is determined, the sampling period is slid backward to analyze and calculate the absolute value of the first-order difference of the actuator feedback voltage, the absolute value of the first-order difference of the actuator mechanical displacement and the standard deviation of the actuator current in the sliding window. If the three indicators are all lower than the lower limit of the feedback voltage change, the lower limit of the displacement rate and the current fluctuation stability threshold within the continuous fixed sampling period, the current sampling period is determined to be the response end point.

[0016] Furthermore, the specific steps of calculating each component index in the response segment and constructing the response characteristic vector of the rudder actuator are as follows: intercepting the rudder actuator response data in the continuous sampling period from the effective response start point to the response end point, and calculating each component index in the response segment: calculating the difference between the maximum and minimum values ​​of the actuator feedback voltage in the response segment to obtain the response voltage range; statistically calculating the maximum value of the first-order difference of the actuator mechanical displacement in the response segment to obtain the maximum response displacement rate; extracting the maximum value of the difference between the control excitation voltage phase angle and the actuator feedback voltage phase angle in the response segment to obtain the response phase offset peak; calculating the standard deviation of the actuator current in the response segment, and extracting the maximum value of the actuator current in the response segment to obtain the response current standard deviation and the response current maximum value; extracting the response voltage range, the response displacement maximum rate, the response phase offset peak, the response current standard deviation and the response current maximum value, and jointly constructing the response characteristic vector.

[0017] Furthermore, the specific steps of extracting and constructing the response characteristic vector and evaluating the degree of deviation between the current response behavior and the ideal state are as follows: extracting and constructing the response characteristic vector and evaluating the degree of deviation between the current response behavior and the ideal state: dividing the response voltage range by the response voltage range plus the absolute value of the maximum response displacement rate as the amplitude change term; adding 1 to the ratio of the absolute value of the response phase offset peak to π as the phase difference amplification term; taking the ratio of the response current standard deviation to the maximum response current as the current fluctuation normalization term; multiplying the amplitude change term, the phase difference amplification term and the current fluctuation normalization term in sequence to obtain the response deviation evaluation value; comparing the calculated response deviation evaluation value with the response deviation threshold: if the response deviation evaluation value is less than or equal to the response deviation threshold, it is determined that the current The response segment is a normal response; if the response deviation evaluation value is greater than the response deviation threshold, the local anomaly analysis link is entered, and the characteristic threshold judgment is performed on the four indicators of response voltage range, response displacement maximum rate, response phase offset peak, and response current standard deviation: if the voltage response range is greater than the voltage amplitude threshold, and the maximum displacement rate is lower than the displacement rate lower limit, it is identified as a response hysteresis state; if the phase offset peak is greater than the phase difference threshold, and the duration is not less than the lower limit of the continuous sampling period, it is identified as a feedback lag state; if the maximum displacement rate is greater than the rate upper limit, and the subsequent displacement change in the response segment tends to be flat, it is identified as a displacement saturation state; if the current standard deviation ratio is higher than the current fluctuation ratio threshold, and the number of fluctuations per unit time exceeds the current fluctuation threshold, it is identified as a drive current disturbance state.

[0018] Furthermore, the specific steps for identifying abnormal response states based on the evaluation results, generating channel-level response behavior labels and summarizing them to form classification results are as follows: the identified response states are structurally bound with the current channel number and sampling time information to generate response behavior labels; at the same time, a re-inspection flag is set for the channels identified as abnormal, and the corresponding abnormal states are automatically marked; after all channel detection tasks are completed, the response behavior labels of all channels are automatically summarized to generate response state statistical maps and abnormal distribution overviews, providing supporting data for subsequent fault location and performance change trend analysis.

[0019] The second aspect of the present invention provides an aircraft rudder automatic detection system based on virtual instruments and matrix switching, which is characterized by comprising: a rudder response data acquisition and preprocessing module, a control excitation and channel switching evaluation module, a response start and end judgment and feature extraction module, and an abnormality recognition and response behavior classification module, wherein: the rudder response data acquisition and preprocessing module is used to dynamically collect the entire process of rudder actuator operation, obtain rudder actuator response data, and perform time axis repositioning, dynamic denoising, phase compensation and scale unification on the rudder actuator response data; the control excitation and channel switching evaluation module is used to generate continuous control based on the preprocessed rudder actuator response data. The system generates an excitation signal to drive the relay to achieve channel switching, evaluates the stability before and after the relay switching, selects the stable sampling period, and constructs a stable response data set. The response start and end determination and feature extraction module is used to screen candidate response starting points based on the stable response data set, identify the effective response starting and end points of each channel, calculate the component indicators within the response segment, and construct the rudder actuator response feature vector. The abnormality identification and response behavior classification module is used to extract and construct the response feature vector, evaluate the degree of deviation between the current response behavior and the ideal state, identify the abnormal response state based on the evaluation results, generate channel-level response behavior labels, and summarize them to form a classification result.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This method and system for automatic detection of aircraft control surfaces based on virtual instruments and matrix switching integrates the displacement change rate, phase response synergy term, and drive strength normalization term, and achieves high-precision identification of the response starting point of the control surface actuator based on a weighted average strategy within a continuous fixed sampling period. It can effectively avoid the problem of misjudgment of the initial displacement caused by micro-arc disturbances in the early stage of relay switching, solve the response early and displacement drift diagnosis errors faced by traditional systems in the early stage of response, and improve the robustness and timing accuracy of response starting point identification.

[0023] (2) The aircraft control surface automatic detection method and system based on virtual instrument and matrix switching constructs a switching joint disturbance evaluation value index system. After the relay channel is switched, the control input change term, voltage response change term and current response change term are calculated and summed up respectively. The stability of the sampling period before and after the relay switching is dynamically evaluated. It can accurately screen out the sampling period of the non-steady-state channel and ensure that the stable response data set only contains the real response behavior under the controlled excitation, providing a reliable data basis for the subsequent response feature extraction.

[0024] (3) The aircraft control surface automatic detection method and system based on virtual instrument and matrix switching constructs a response feature vector, extracts five indicators including the response voltage range, the maximum response displacement rate, the response phase offset peak, the response current standard deviation and the maximum response current, and forms a multi-dimensional and multi-channel response behavior feature expression, providing comprehensive and detailed feature support for subsequent response state recognition and anomaly classification.

[0025] (4) The aircraft control surface automatic detection method and system based on virtual instrument and matrix switching evaluates the degree of deviation between the current response behavior and the ideal state by constructing a response deviation evaluation value index system. If the response deviation evaluation value exceeds the response deviation threshold, it further combines each individual index with the characteristic threshold judgment to identify the abnormal state including subdivision, and automatically generates channel-level response behavior labels, providing an accurate basis for system-level abnormality tracing. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of an automatic detection method for aircraft control surfaces based on virtual instruments and matrix switching;

[0027] Figure 2 This is the structural diagram of the aircraft control surface automatic detection system based on virtual instrument and matrix switching;

[0028] Figure 3 Schematic diagram of stimulus signal generation and response feedback;

[0029] Figure 4 A block diagram of the structure of the main control program and modular detection functions;

[0030] Figure 5 Schematic diagram of the hardware structure of the multi-channel detection device for rudder actuators based on the PC104 platform. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0032] See also Figures 1-5The embodiment of the application provides a technical scheme: an airplane control surface automatic detection method and system based on a virtual instrument and matrix switching, comprising the following steps: S1, full-process dynamic acquisition is performed on the operation of a control surface actuator, control surface actuator response data is acquired, and time axis repositioning, dynamic denoising, phase compensation and scale unification are performed on the control surface actuator response data; S2, continuous control excitation signals are generated based on the pretreated control surface actuator response data, relays are driven to realize channel switching, and the stability degrees before and after the relay switching are evaluated, a stable sampling period is screened, and a stable response data set is constructed; S3, candidate response starting points are screened based on the stable response data set, effective response starting points and response ending points of each channel are identified, component indicators in the response section are calculated, and a control surface actuator response feature vector is constructed; S4, the response feature vector is extracted, the deviation degree between the current response behavior and the ideal state is evaluated, an abnormal response state is identified based on the evaluation result, a channel-level response behavior label is generated, and a classification result is formed by summarizing.

[0033] Specifically, the specific steps of full-process dynamic acquisition on the operation of the control surface actuator, acquisition of the control surface actuator response data, and time axis repositioning, dynamic denoising, phase compensation and scale unification on the control surface actuator response data are as follows: full-process dynamic acquisition is performed on the operation of the control surface actuator, control surface actuator response data is acquired, the control surface actuator response data includes control excitation voltage, actuator feedback voltage, actuator mechanical displacement, actuator current, excitation frequency, control voltage phase angle and feedback voltage phase angle; a synchronization criterion based on signal sampling sequence stability is established in combination with dynamic correlation information of excitation frequency fluctuation characteristics and response starting delay behavior; the time axis of the control surface actuator response data is repositioned through a dynamic window delay sampling method based on a relay state synchronization marker, and transient response drift and initial displacement misjudgment caused by micro-arc disturbance at the initial stage of relay switching is repaired; the control surface actuator response data is dynamically denoised through a response derivative constraint sliding fitting method, and high-frequency noise caused by the starting stage and impact current is filtered out; a low-pass filter and a differential amplification circuit are introduced in the acquisition link, which can effectively suppress high-frequency arc interference and common-mode noise and improve the measurement accuracy of the actuator feedback voltage and current signals; a morphological smoothing function of a continuous current slope change interval is introduced to enhance the response trend stability identification ability and assist in removing non-control source driving components; a delay identification method based on control-response phase correlation is used to identify and compensate the phase error caused by mechanical hysteresis and sampling offset; the control surface actuator response data is processed through an amplitude normalization and response rate standard deviation scaling strategy, so that the response sequences of all channels have uniform order of magnitude weights in the subsequent calculation links and are conducive to the comparison and analysis between feature indicators.

[0034] In this implementation scheme, by introducing dynamic window delay sampling, response derivative sliding fitting, phase-correlated delay identification and scale normalization processing, the processing accuracy of multi-source rudder actuator response data such as control excitation voltage, actuator feedback voltage, actuator mechanical displacement and actuator current is effectively improved in terms of time synchronization, signal purity, phase consistency and dimensional comparability, providing a data foundation for subsequent response starting point identification and feature extraction.

[0035] Specifically, the following steps are used to generate a continuous control excitation signal based on the preprocessed control surface actuator response data and drive the relay to achieve channel switching: a digital sinusoidal sequence is generated based on the preprocessed control surface actuator response data, and a continuous control excitation voltage is output through a digital-to-analog converter. The excitation signal amplitude, frequency, and phase are referenced to the characteristic mean of the historical stable segment in the self-stabilizing response data set; the control excitation voltage is connected to the target channel after power amplification, and the excitation frequency is set by a clock control circuit and output synchronously with the control excitation voltage to ensure the consistency of the excitation rhythm and response alignment between channels; the frequency, phase, and waveform of the control excitation signal are precisely controlled by an FPGA module, and frequency synthesis is performed by a digital phase-locked loop device to ensure that the output excitation and response acquisition are strictly aligned on the time basis; a control signal is output through a decoding circuit to drive the relay to switch channels, and the duration of the relay control signal is set by a scheduling strategy to avoid the arc noise persistence window; after the relay switching is completed, the control excitation voltage, actuator feedback voltage, and actuator current within a continuous fixed period before and after the relay switching are extracted on the sampling period, and the mean values ​​before and after the switching are calculated. The synchronization relationship between the excitation frequency and the sampling period is recorded for subsequent response stability judgment.

[0036] like Figure 3 The figure shows the schematic diagram of excitation signal generation and response feedback, illustrating the complete process of excitation signal generation and response feedback during the control surface actuator response test. The integrated waveform generator outputs a sinusoidal signal at the target frequency, which is converted into an analog voltage signal via a D / A converter. This signal is then input to the power amplifier module for amplitude amplification, driving the actuator and simultaneously triggering the displacement sensor for response displacement detection. A computer receives the feedback data for subsequent response feature extraction and behavioral classification analysis. This structure demonstrates the continuity and high precision of excitation waveform control, signal amplitude adjustment, power output drive, and feedback sensing, providing fundamental support for the quantitative evaluation of the dynamic response behavior of control surface actuators.

[0037] In the embodiment, the continuous control excitation voltage is generated based on the pretreated rudder surface actuator response data, the excitation frequency and phase output are accurately set, the high-precision switching control of the relay is realized by combining the decoding circuit, and the mean values of the control excitation voltage, the actuator feedback voltage and the actuator current before and after switching are extracted, thereby providing a high-quality input basis with consistent rhythm, controllable response and clear switching for subsequent identification of channel stable state and construction of stable response data set.

[0038] Specifically, the specific steps for evaluating the stable degree before and after the switching of the relay are as follows: based on the mean values before and after the switching, the stable degree before and after the switching of the relay is evaluated: first, according to the sampling rhythm of the stable response data set, the absolute value of the difference between the mean values of the control excitation voltage before and after the switching of the relay is calculated to obtain the control input change item; then, the absolute value of the difference between the mean values of the actuator feedback voltage before and after the switching of the relay is calculated in combination with the response matching of the excitation signal output link and the feedback link to obtain the voltage response change item; subsequently, the absolute value of the difference between the mean values of the actuator current before and after the switching of the relay is calculated with reference to the actuator current driving characteristics to obtain the current response change item; the control input change item, the voltage response change item and the current response change item are summed to obtain the switching joint disturbance evaluation value.

[0039] The specific calculation formula of the switching joint disturbance evaluation value is as follows:

[0040]

[0041] In the formula, Q represents the switching joint disturbance evaluation value, represents the mean value of the control excitation voltage before the switching of the relay, represents the mean value of the control excitation voltage after the switching of the relay, represents the mean value of the actuator feedback voltage before the switching of the relay, represents the mean value of the actuator feedback voltage after the switching of the relay, represents the mean value of the actuator current before the switching of the relay, represents the mean value of the actuator current after the switching of the relay.

[0042] In the embodiment, by accurately extracting the change amplitudes of the control excitation voltage mean value, the actuator feedback voltage mean value and the actuator current mean value before and after the switching of the relay, the transient disturbance degree caused by the switching of the relay is effectively identified, and it is ensured that the data segment entering the subsequent sampling has sufficient voltage stability, current consistency and feedback response continuity, thereby laying a data foundation for accurate construction of the stable response data set.

[0043] Specifically, the specific steps for screening stable sampling periods and constructing a stable response data set are as follows: real-time comparison of the switching joint disturbance evaluation value and the switching disturbance threshold. The disturbance evaluation value is composed of the control input change term, the voltage response change term, and the current response change term superimposed in sequence, which can reflect the multi-source disturbance intensity during the relay switching process. If the switching joint disturbance evaluation value is greater than the switching disturbance threshold, it is determined that there is still a non-steady-state disturbance in the current channel. Such disturbances can easily cause instantaneous jumps in the actuator feedback voltage, intensified fluctuations in the actuator current, and offsets in the mechanical displacement response. The recording of the rudder actuator response data in the current sampling period is suspended, and the judgment is made again after re-evaluation in the next sampling period. If the switching joint disturbance evaluation value is less than or equal to the switching disturbance threshold, it is determined that the current channel is stable, starting from the current sampling period, and at the same time observing whether the actuator feedback voltage trend returns to the steady-state trajectory and the actuator current enters the stable driving range. The rudder actuator response data in a fixed number of sampling periods are continuously recorded to construct a stable response data set.

[0044] In this implementation scheme, by comparing the switching joint disturbance assessment value with the switching disturbance threshold, the channel stability state after the relay switching is dynamically identified, and the multi-source changes of the control excitation voltage, actuator feedback voltage and actuator current are combined to realize the quantitative evaluation of the disturbance intensity. It can effectively shield the transient interference influence in the non-steady-state stage, and start sampling and recording only under the conditions of interference attenuation and response recovery stability, thereby ensuring that the subsequently constructed stable response data set has high reliability and consistency, laying an accurate data foundation for the dynamic response analysis of the rudder actuator.

[0045] Specifically, the specific steps of screening candidate response starting points based on the stable response data set and identifying the effective response starting point and response end point of each channel are as follows: based on the stable response data set, extract the maximum value of the actuator current; perform first-order difference on the actuator mechanical displacement signal to obtain the displacement change rate, identify the displacement change trend mutation point, and mark the time corresponding to the trend mutation point as the candidate response starting point; extract the rudder actuator response data of the continuous fixed sampling period starting from the candidate response starting point, and construct the response starting evaluation value: calculate the absolute value of the difference between the control excitation voltage phase angle and the actuator feedback voltage phase angle, divide this absolute value by π and take its inverse relative value to obtain the phase response synergy term; calculate the ratio of the actuator current to the maximum value of the actuator current to obtain the drive strength normalization term; multiply the displacement change rate, the phase response synergy term and the drive strength normalization term in each sampling period, and The product results are weighted and accumulated within a continuous fixed sampling period and then averaged to obtain a response start evaluation value; the response start evaluation value and the response start threshold are compared in real time. If the response start evaluation value is greater than the response start threshold, the current candidate point is confirmed as a valid response starting point, and the rudder actuator response data after the starting point is intercepted; if the response start evaluation value is lower than the response start threshold, the candidate response starting point is slid backward and the above calculation process is repeated until the first valid response starting point that meets the response start threshold condition is identified; after the valid response starting point is determined, the sampling period is slid backward to analyze and calculate the absolute value of the first-order difference of the actuator feedback voltage, the absolute value of the first-order difference of the actuator mechanical displacement, and the standard deviation of the actuator current within the sliding window. If the three indicators are all lower than the lower limit of the feedback voltage change, the lower limit of the displacement rate, and the current fluctuation stability threshold within the continuous fixed sampling period, the current sampling period is determined to be the response end point.

[0046] The specific calculation formula for the response initial evaluation value is:

[0047]

[0048] Where X represents the response start evaluation value, T represents the fixed sampling period, t0 represents the candidate response start point, S represents the mechanical displacement of the actuator, represents the displacement change rate, θ c represents the control excitation voltage phase angle, θ f represents the actuator feedback voltage phase angle, I represents the actuator current, and I m Indicates the maximum actuator current.

[0049] In this implementation scheme, the maximum value of the actuator current, the rate of change of the actuator mechanical displacement, the phase angle difference of the control excitation voltage and the phase angle difference of the actuator feedback voltage are comprehensively utilized to construct a response starting evaluation value, and the dynamic characteristics within the continuous sampling period are calculated to accurately determine the starting point of the effective response; on this basis, the absolute value of the first-order difference of the feedback voltage, the absolute value of the first-order difference of the mechanical displacement and the standard deviation of the actuator current are combined to construct an end point judgment mechanism to ensure the accuracy and stability of the start and end judgment of the response segment, provide a precise interval basis for the subsequent response feature extraction, and improve the accuracy and robustness of the overall response behavior recognition.

[0050] Specifically, the specific steps for calculating each component index in the response segment and constructing the response characteristic vector of the rudder actuator are as follows: intercepting the rudder actuator response data in the continuous sampling period from the effective response start point to the response end point, and calculating each component index in the response segment: calculating the difference between the maximum and minimum values ​​of the actuator feedback voltage in the response segment, and obtaining the response voltage range, which is used to reflect the fluctuation amplitude change characteristics of the feedback voltage signal in the response process; statistically calculating the maximum value of the first-order difference of the actuator mechanical displacement in the response segment, and obtaining the maximum response displacement rate, which reflects the response sensitivity and mutation characteristics of the mechanical displacement; extracting the control excitation in the response segment; and obtaining the maximum value of the first-order difference of the actuator mechanical displacement in the response segment. The maximum value of the difference between the excitation voltage phase angle and the actuator feedback voltage phase angle is used to obtain the response phase offset peak, which is used to evaluate the synchronization error in the phase transmission process; the standard deviation of the actuator current in the response segment is calculated, and the maximum value of the actuator current in the response segment is extracted to obtain the response current standard deviation and the response current maximum value, which are used to characterize the stability and peak load capacity of the current driving process; the response voltage range, the maximum response displacement rate, the response phase offset peak, the response current standard deviation and the response current maximum value are extracted to jointly construct the response feature vector, providing multi-dimensional numerical support for subsequent anomaly identification and channel behavior analysis.

[0051] In this implementation plan, by carefully quantifying the response data of the rudder actuator between the effective response start point and the response end point, the system extracts multiple indicators such as the feedback voltage range, the maximum rate of mechanical displacement, the phase angle offset peak, the current standard deviation and the current maximum value, and constructs a response feature vector covering four types of characteristics: amplitude fluctuation, displacement change, phase response, and current drive. This effectively enhances the structured description capability of the response behavior of the rudder actuator and provides a high-resolution parameter basis for abnormality identification.

[0052] Specifically, the specific steps of extracting and constructing the response characteristic vector and evaluating the degree of deviation between the current response behavior and the ideal state are as follows: Extracting and constructing the response characteristic vector and evaluating the degree of deviation between the current response behavior and the ideal state: dividing the response voltage range by the response voltage range plus the absolute value of the maximum response displacement rate as the amplitude change term; adding 1 to the ratio of the absolute value of the response phase offset peak to π as the phase difference amplification term; taking the ratio of the response current standard deviation to the maximum response current as the current fluctuation normalization term; multiplying the amplitude change term, the phase difference amplification term and the current fluctuation normalization term in sequence to obtain the response deviation evaluation value; comparing the calculated response deviation evaluation value with the response deviation threshold: if the response deviation evaluation value is less than or equal to the response deviation threshold, it is determined that the current response The response segment is a normal response; if the response deviation evaluation value is greater than the response deviation threshold, the local anomaly analysis link is entered, and the characteristic threshold judgment is performed on the four indicators of response voltage range, response displacement maximum rate, response phase offset peak, and response current standard deviation: if the voltage response range is greater than the voltage amplitude threshold, and the maximum displacement rate is lower than the displacement rate lower limit, it is identified as a response hysteresis state; if the phase offset peak is greater than the phase difference threshold, and the duration is not less than the lower limit of the continuous sampling period, it is identified as a feedback lag state; if the maximum displacement rate is greater than the rate upper limit, and the subsequent displacement change in the response segment tends to be flat, it is identified as a displacement saturation state; if the current standard deviation ratio is higher than the current fluctuation ratio threshold, and the number of fluctuations per unit time exceeds the current fluctuation threshold, it is identified as a drive current disturbance state.

[0053] The specific calculation formula for the response deviation evaluation value is:

[0054]

[0055] Where P represents the response deviation evaluation value, ΔV represents the response voltage range, ΔS represents the maximum response displacement rate, Δθ represents the response phase offset peak value, and σ I represents the standard deviation of the response current, Indicates the maximum value of the response current.

[0056] In this implementation, multiple normalized discrimination indicators are constructed by using the response voltage extreme difference, maximum response displacement rate, response phase offset peak, response current standard deviation and response current maximum value in the response characteristic vector to quantify the degree of deviation between the response behavior and the ideal state. Combined with the comparison results of the response deviation evaluation value and the preset threshold, the response hysteresis state, feedback lag state, displacement saturation state and drive current disturbance state are accurately identified, effectively improving the hierarchical identification capability of abnormal response behavior of the rudder actuator.

[0057] Specifically, the specific steps for identifying abnormal response states based on the evaluation results, generating channel-level response behavior labels and summarizing them to form classification results are as follows: the identified response states are structurally bound with the current channel number and sampling time information to generate response behavior labels, and the label content includes the abnormality type, response deviation evaluation value, threshold judgment result and corresponding component indicators; at the same time, a re-inspection flag is set for the channel identified as abnormal, and the abnormality duration, start and end sampling period and abnormal peak value are attached to the flag, and the corresponding abnormal state is automatically marked to track the consistent evolution of subsequent control instructions and feedback behaviors; after all channel detection tasks are completed, the response behavior labels of all channels are automatically summarized to generate a response state statistical map and an abnormality distribution overview. The statistical map presents the channel distribution density and response characteristic indicator mean distribution of various abnormal states, providing quantitative support data and structured reference basis for subsequent fault location based on time series evolution characteristics and rudder response performance change trend analysis.

[0058] In this implementation scheme, by performing structured binding and label generation on the identified response states, not only the quantitative classification of response behaviors and the precise positioning of sampling times are achieved, but also the tracking capability of abnormal response channels is improved by adding response deviation evaluation values, abnormal duration, response segment start and end sampling periods, and abnormal peak information, further enhancing the recognition efficiency and accuracy of abnormal response behavior characteristics in rudder control system detection. The resulting statistical map of response states and overview of abnormal distribution can serve as the basis for identifying rudder response performance trends and fault location analysis, thereby enhancing the comprehensive perception and classification decision-making capabilities of dynamic behavior changes of actuators.

[0059] like Figure 2As shown, the second aspect of the present invention provides an aircraft rudder automatic detection system based on virtual instruments and matrix switching, which is characterized by comprising: a rudder response data acquisition preprocessing module, a control excitation and channel switching evaluation module, a response start and end judgment and feature extraction module, and an abnormality recognition and response behavior classification module, wherein: the rudder response data acquisition preprocessing module is used to dynamically collect the entire process of rudder actuator operation, obtain rudder actuator response data, and perform time axis repositioning, dynamic denoising, phase compensation and scale unification on the rudder actuator response data; the control excitation and channel switching evaluation module is used to generate continuous rudder actuator response data based on the preprocessed rudder actuator response data. The control module controls the excitation signal, drives the relay to achieve channel switching, and evaluates the stability before and after relay switching, selects stable sampling periods, and constructs a stable response dataset. The response start and end determination and feature extraction module is used to select candidate response starting points based on the stable response dataset, identify the effective response starting and end points for each channel, calculate the component indicators within the response segment, and construct the control surface actuator response feature vector. The anomaly identification and response behavior classification module extracts and constructs the response feature vector, evaluates the deviation between the current response behavior and the ideal state, identifies abnormal response states based on the evaluation results, generates channel-level response behavior labels, and summarizes the classification results. During the integration of these functional modules, the main control program serves as the core scheduling unit of the detection platform, coordinating tasks such as control surface response data acquisition and preprocessing, excitation signal generation and channel switching control, response behavior analysis, and classification output. It provides unified management support for data flow and control logic between system modules. Furthermore, the aircraft control surface automatic detection system based on virtual instruments and matrix switching is built on a PC104-based virtual instrument platform. It features high integration and embedded processing capabilities, making it suitable for multi-channel, multi-frequency parallel testing environments.

[0060] like Figure 4The following diagram shows the structural composition of the main control program and modular testing functions, comprehensively illustrating the functional layering and process logic of an aircraft control surface automatic testing system based on virtual instruments and matrix switching. The main control program, as the system's core scheduling unit, coordinates four major modules: hardware control, test tasks, database operations, and result processing. The hardware control module is responsible for initializing the system and configuring related equipment, flexibly switching between multi-channel test paths through a matrix switch. The test module supports test item selection, performance testing, and system self-tests, ensuring the integrity and accuracy of data acquisition. The core acquisition and excitation control functions rely on D / A and A / D chips for high-precision analog signal processing. The database module is responsible for interrupt data transfer and structured storage of sampling results. The result processing module provides real-time display and long-term archiving of test results, improving the traceability of the testing process and the utilization of results. The overall architecture is highly integrated with clear division of labor, supporting multi-channel concurrent testing and dynamic response behavior classification and analysis. It is a key technical support for the full-process detection of control surface actuator response feature extraction and anomaly identification.

[0061] like Figure 5 Figure 2 shows the hardware architecture of a multi-channel detection device for rudder actuators based on the PC104 platform. Built on the PC104 embedded platform, the device's core modules include a signal generator, an A / D board, a D / A board, and an I / O board. These boards connect to the rudder control box and actuators via a relay matrix switch. A mouse panel and LCD display are also included for interaction and status visualization. The hardware architecture of this multi-channel detection device effectively supports the functional modules described in this implementation, including rudder actuator response data acquisition, excitation signal generation, relay channel switching, and feedback data processing, forming a complete closed-loop detection process.

[0062] In this implementation plan, by integrating continuous processes such as rudder response data acquisition and preprocessing, control excitation signal generation and relay channel switching, automatic identification of response start and end and feature vector construction, response behavior deviation evaluation and classification labeling, the dynamic response data of the rudder actuator in the multi-channel and multi-cycle operation process is accurately acquired, effective start and end interception, key feature extraction and abnormal behavior identification are achieved, which significantly improves the degree of automation of response data processing and the accuracy of behavior classification, and constructs a response feature extraction and abnormal classification mechanism suitable for the entire process of rudder control system detection, providing complete process support for rudder actuator performance stability analysis and rapid screening of response anomalies.

[0063] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0064] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An automatic detection method for aircraft control surfaces based on virtual instruments and matrix switching, characterized in that: The following steps are involved: S1, dynamically collects the entire process of the rudder actuator operation, obtains the rudder actuator response data, and performs time axis repositioning, dynamic denoising, phase compensation and scale unification on the rudder actuator response data; S2, based on the pre-processed control surface actuator response data, generates a continuous control excitation signal to drive the relay to achieve channel switching. It also evaluates the stability of the relay before and after switching, selects the stable sampling period, and constructs a stable response data set. S3, based on the stable response data set, screens candidate response starting points, identifies the effective response starting points and response ending points of each channel, calculates the component indicators within the response segment, and constructs the response feature vector of the rudder actuator; S4, extracts and constructs the response feature vector, evaluates the deviation between the current response behavior and the ideal state, identifies the abnormal response state based on the evaluation results, generates channel-level response behavior labels and summarizes them to form classification results.

2. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1, characterized in that: The specific steps of dynamically collecting the entire process of the rudder actuator operation, obtaining the rudder actuator response data, and performing time axis repositioning, dynamic denoising, phase compensation, and scale unification on the rudder actuator response data are as follows: Perform dynamic data collection of the entire rudder actuator operation process to obtain rudder actuator response data, which includes control excitation voltage, actuator feedback voltage, actuator mechanical displacement, actuator current, excitation frequency, control voltage phase angle, and feedback voltage phase angle; The dynamic window delay sampling method based on relay state synchronization marks is used to reposition the time axis of the rudder actuator response data, thereby correcting the transient response drift and starting displacement misjudgment caused by micro-arc disturbances in the initial stage of relay switching. The rudder actuator response data is dynamically denoised using a sliding fitting method constrained by response derivatives to filter out high-frequency noise caused by the startup phase and inrush current. A delay identification method based on control-response phase correlation is used to identify and compensate for phase errors caused by mechanical lag and sampling offset. The rudder actuator response data is scaled uniformly using amplitude normalization and response rate standard deviation scaling strategies.

3. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1 is characterized in that: The specific steps of generating a continuous control excitation signal based on the pre-processed control surface actuator response data and driving the relay to realize channel switching are as follows: Based on the pre-processed control surface actuator response data, a digital sine sequence is generated, and a continuous control excitation voltage is output through a digital-to-analog converter. The control excitation voltage is then power-amplified and connected to the target channel. The excitation frequency is set by a clock control circuit and output synchronously with the control excitation voltage. The control signal is output through the decoding circuit to drive the relay switching channel. After the relay switching is completed, the control excitation voltage, actuator feedback voltage and actuator current in the continuous fixed period before and after the relay switching are extracted in the sampling period, and the average values ​​before and after the switching are calculated respectively.

4. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1 is characterized in that: The specific steps of evaluating the stability of the relay before and after switching are as follows: The stability of the relay before and after switching is evaluated based on the average values ​​before and after switching. The absolute value of the difference between the average values ​​of the control excitation voltage before and after the relay switching is calculated to obtain the control input change term. The absolute value of the difference between the average values ​​of the actuator feedback voltage before and after the relay switching is calculated to obtain the voltage response change term. The absolute value of the difference between the average values ​​of the actuator current before and after the relay switching is calculated to obtain the current response change term. The switching joint disturbance evaluation value is obtained by summing the control input change term, voltage response change term and current response change term.

5. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1 is characterized in that: The specific steps of screening the stable sampling period and constructing the stable response data set are as follows: The switching joint disturbance evaluation value and the switching disturbance threshold are compared in real time. If the switching joint disturbance evaluation value is greater than the switching disturbance threshold, it is determined that there is still non-steady-state disturbance in the current channel, and the recording of the rudder actuator response data in the current sampling period is suspended until the next sampling period is re-evaluated and the judgment is made again. If the switching joint disturbance evaluation value is less than or equal to the switching disturbance threshold, it is determined that the current channel is stable, and starting from the current sampling period, the rudder actuator response data for a fixed number of sampling periods are continuously recorded to construct a stable response data set.

6. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1, characterized in that: The specific steps of screening candidate response starting points based on the stable response data set and identifying the effective response starting points and response ending points of each channel are as follows: Based on the stable response data set, the maximum value of the actuator current is extracted; the first-order difference of the actuator mechanical displacement signal is performed to obtain the displacement change rate, the displacement change trend mutation point is identified, and the time corresponding to the trend mutation point is marked as the candidate response starting point; Extract the rudder actuator response data for consecutive fixed sampling periods starting from the candidate response starting point to construct a response start evaluation value: calculate the absolute value of the difference between the control excitation voltage phase angle and the actuator feedback voltage phase angle, divide this absolute value by π and take its inverse relative value to obtain the phase response synergy term; calculate the ratio of the actuator current to the maximum actuator current to obtain the drive strength normalization term; multiply the displacement change rate, the phase response synergy term, and the drive strength normalization term in each sampling period, and perform a weighted accumulation and average of the product results in consecutive fixed sampling periods to obtain the response start evaluation value; The response start evaluation value and the response start threshold are compared in real time. If the response start evaluation value is greater than the response start threshold, the current candidate point is confirmed as a valid response start point, and the control surface actuator response data after the start point is intercepted. If the response start evaluation value is lower than the response start threshold, the candidate response start point is slid backward and the above calculation process is repeated until the first valid response start point that meets the response start threshold condition is identified. After the effective response starting point is determined, the sampling period is analyzed by sliding backward to calculate the absolute value of the first-order difference of the actuator feedback voltage, the absolute value of the first-order difference of the actuator mechanical displacement, and the standard deviation of the actuator current within the sliding window. If these three indicators are all lower than the lower limit of the feedback voltage change, the lower limit of the displacement rate, and the current fluctuation stability threshold within a continuous fixed sampling period, the current sampling period is determined to be the response end point.

7. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1, characterized in that: The specific steps of calculating the component indices in the response segment and constructing the response characteristic vector of the rudder actuator are as follows: The control surface actuator response data within the continuous sampling period from the effective response start point to the response end point is intercepted, and the component indicators within the response segment are calculated: the difference between the maximum and minimum values ​​of the actuator feedback voltage within the response segment is calculated to obtain the response voltage range; the maximum value of the first-order difference of the actuator mechanical displacement within the response segment is statistically calculated to obtain the maximum response displacement rate; the maximum value of the difference between the control excitation voltage phase angle and the actuator feedback voltage phase angle within the response segment is extracted to obtain the response phase offset peak; the standard deviation of the actuator current within the response segment is calculated, and the maximum value of the actuator current within the response segment is extracted to obtain the response current standard deviation and the response current maximum value; The response voltage range, response displacement maximum rate, response phase shift peak, response current standard deviation and response current maximum are extracted and the response feature vector is jointly constructed.

8. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1, characterized in that: The specific steps of extracting and constructing the response feature vector and evaluating the degree of deviation between the current response behavior and the ideal state are as follows: Extract and construct the response feature vector to evaluate the degree of deviation between the current response behavior and the ideal state: divide the response voltage range by the absolute value of the response voltage range plus the maximum response displacement rate as the amplitude change term; The ratio of the absolute value of the response phase offset peak to π plus 1 is used as the phase difference amplification term; the ratio of the response current standard deviation to the maximum response current is used as the current fluctuation normalization term; the amplitude change term, the phase difference amplification term, and the current fluctuation normalization term are multiplied in sequence to obtain the response deviation evaluation value; Compare the calculated response deviation evaluation value with the response deviation threshold: if the response deviation evaluation value is less than or equal to the response deviation threshold, the current response segment is determined to be a normal response; If the response deviation evaluation value is greater than the response deviation threshold, the local anomaly analysis phase is entered, and characteristic threshold judgments are performed on the four indicators of response voltage range, response displacement maximum rate, response phase offset peak, and response current standard deviation. If the voltage response range is greater than the voltage amplitude threshold and the maximum displacement rate is lower than the displacement rate lower limit, it is identified as a response hysteresis state. If the phase shift peak is greater than the phase difference threshold and the duration is not less than the lower limit of the continuous sampling period, it is identified as a feedback lag state; if the maximum displacement rate is greater than the upper rate limit and the subsequent displacement change in the response segment tends to be flat, it is identified as a displacement saturation state; If the current standard deviation ratio is higher than the current fluctuation ratio threshold, and the number of fluctuations per unit time exceeds the current fluctuation threshold, it is identified as a driving current disturbance state.

9. The method for automatically detecting aircraft control surfaces based on virtual instruments and matrix switching according to claim 1, characterized in that: The specific steps of identifying abnormal response states based on the evaluation results, generating channel-level response behavior labels, and summarizing them to form classification results are as follows: The identified response status is structurally bound to the current channel number and sampling time information to generate a response behavior label. At the same time, a re-inspection flag is set for the channel identified as abnormal, and the corresponding abnormal status is automatically marked. After all channel detection tasks are completed, the response behavior labels of all channels are automatically summarized to generate response status statistical maps and abnormal distribution overviews, providing supporting data for subsequent fault location and performance change trend analysis.

10. An automatic detection system for aircraft control surfaces based on virtual instruments and matrix switching, characterized by: include: The rudder response data acquisition and preprocessing module, the control excitation and channel switching evaluation module, the response start and end determination and feature extraction module, and the anomaly recognition and response behavior classification module include: The rudder surface response data acquisition and preprocessing module is used to dynamically collect the entire process of rudder surface actuator operation, obtain rudder surface actuator response data, and perform time axis repositioning, dynamic denoising, phase compensation and scale unification on the rudder surface actuator response data; The control excitation and channel switching evaluation module is used to generate a continuous control excitation signal based on the preprocessed control surface actuator response data, drive the relay to achieve channel switching, evaluate the stability of the relay before and after switching, select the stable sampling period, and construct a stable response data set; The response start and end determination and feature extraction module is used to screen candidate response start points based on the stable response data set, identify the effective response start and end points of each channel, calculate the component indicators within the response segment, and construct the control surface actuator response feature vector; The anomaly identification and response behavior classification module is used to extract and construct response feature vectors, evaluate the degree of deviation between the current response behavior and the ideal state, identify abnormal response states based on the evaluation results, generate channel-level response behavior labels and summarize them to form classification results.

Images