Process state real-time correction method based on acoustic features

Abstract

The application provides a process state real-time correction method based on acoustic characteristics, and relates to the technical field of process control, and comprises the following steps: collecting acoustic signals generated by interaction between a processing medium and a processing unit when the processing unit is executing a processing task; performing real-time spectrum analysis on the acoustic signals to extract characteristic frequencies and energy distribution characteristics thereof which are related to the current physical state of the processing medium; determining whether the current physical state of the processing medium is in an abnormal critical region in real time based on changes of the characteristic frequencies and the energy distribution characteristics thereof relative to a preset reference; the abnormal critical region comprises a first precursor state in which the processing medium tends to be excessively dried or a second precursor state in which the processing medium tends to be excessively accumulated; and if it is determined that the processing medium is in the abnormal critical region, the physical state of the processing medium being currently processed is corrected in real time by adjusting at least one execution parameter which influences the physical state of the processing medium before the processing medium being currently processed reaches a target region.

Description

Technical Field

[0001] The present invention belongs to the technical field of process control, and particularly relates to a real-time correction method for process state based on acoustic characteristics. Background Art

[0002] In many industrial processes, the quality of the final product highly depends on the intermediate state of the processed medium during the process. For example, in the electrostatic spraying process, the atomization state of the coating directly determines the gloss, flatness and adhesion of the coating; in the 3D printing process, the fluidity of the molten material determines the forming accuracy and the interlayer bonding strength; in the chemical granulation process, the particle size distribution of the droplets determines the uniformity and solubility of the final particles. However, during the short transmission process of the processed medium from the processing unit to the target area, it is extremely vulnerable to the influence of environmental conditions (such as temperature, humidity, air flow disturbance) or changes in the characteristics of the medium itself (such as viscosity, concentration fluctuation), resulting in abnormal changes in its physical state, and then causing quality defects in the final product.

[0003] The existing solutions to such problems mainly fall into two categories: one is to rely on the pre-measurement and parameter compensation of the target object or environmental conditions, such as measuring the temperature of the workpiece before spraying and presetting the process parameters accordingly. Such methods belong to open-loop control, with slow response and inability to adapt to the dynamic changes in the production process. The other is to perform feedback adjustment based on the quality inspection of the final product, such as performing optical or tactile inspection on the formed coating and then inversely correcting the process parameters. Such methods belong to after-the-fact remedies. When a defect is detected, defective products have already been produced, resulting in material waste and loss of production efficiency.

[0004] The common defect of the above two types of methods is that: the nodes of monitoring and intervention lag behind the real-time changes in the medium state, and the quality control link is after or at the time when the medium reaches the target area, and it is impossible to block the generation process of defects. Therefore, how to perform real-time perception, accurate determination and active correction of the physical state of the medium within the short window period when the medium is in transit is a technical problem亟待解决 in this field. Summary of the Invention

[0005] In view of the above defects or deficiencies in the prior art, a real-time correction method for process state based on acoustic characteristics is provided, including the following steps:

[0006] Collect the acoustic signals generated by the interaction between the processed medium and the processing unit when the processing unit executes the processing task;

[0007] Perform real-time spectral analysis on the acoustic signals, and extract the characteristic frequencies and their energy distribution characteristics that are correlated with the current physical state of the processed medium;

[0008] Based on the changes in the characteristic frequency and its energy distribution characteristics relative to a preset benchmark, it is determined in real time whether the current physical state of the processed medium is in an abnormal critical region; the abnormal critical region includes a first precursor state in which the medium tends to be excessively dry or a second precursor state in which it tends to be excessively accumulated;

[0009] If it is determined to be in an abnormal critical zone, the physical state of the medium being processed is corrected in real time by adjusting at least one execution parameter that affects the physical state of the medium being processed before the medium being processed reaches the target area; wherein, the medium being processed is a subset of the medium being processed at the current time point, referring to the part of the medium that is currently flying from the processing unit to the target area.

[0010] According to the technical solution provided in this application, the step of determining in real time whether the current physical state of the processed medium is in an abnormal critical region based on the change of the characteristic frequency and its energy distribution characteristics relative to a preset benchmark includes the following steps:

[0011] The instantaneous value of the feature frequency extracted in real time is compared with a reference benchmark to identify whether the feature frequency has drifted.

[0012] When the characteristic frequency is detected to drift towards a higher frequency and the high-frequency harmonic energy is enhanced, it is determined to be the first precursor state;

[0013] When the characteristic frequency is detected to drift to a lower frequency and the spectrum is broadened or low-frequency noise appears, it is determined to be the second precursor state.

[0014] According to the technical solution provided in this application, the real-time correction of the physical state of the medium being processed by adjusting at least one execution parameter that affects the physical state of the medium being processed includes the following steps:

[0015] When the first precursor state is determined, the adjustment of reducing the parameters of the current-carrying medium is performed first, and the adjustment of increasing the medium supply rate is performed after a first predetermined delay period.

[0016] When the second precursor state is determined, the adjustment of increasing the parameters of the current-carrying medium is performed first, and the adjustment of decreasing the medium supply rate is performed after the second predetermined delay time.

[0017] The setting of the first predetermined delay time and the second predetermined delay time both ensure that the adjustment effect of the medium supply rate and the change of the medium state after the adjustment of the current-carrying medium parameters are synchronously matched when the medium arrives at the target area.

[0018] According to the technical solution provided in this application, the following steps are also included:

[0019] During the calibration phase, a step adjustment of the current-carrying medium parameters is performed, and the response process of the characteristic frequency is monitored in real time using the acoustic signal.

[0020] Record the time required from issuing the adjustment command to the characteristic frequency reaching a new steady-state value, as the state stabilization duration under this operating condition;

[0021] The transmission time is calculated based on the transmission distance and the average flight speed of the medium.

[0022] The first predetermined delay duration or the second predetermined delay duration is set as: the stable state duration minus the medium transmission duration. When the calculation result is positive, this value is taken, and when the calculation result is negative, zero is taken.

[0023] According to the technical solution provided in this application, the method further includes the following steps:

[0024] During system operation, the long-term statistical average of the characteristic frequencies is tracked and recorded;

[0025] The short-term instantaneous value of the feature frequency collected and extracted in real time is compared with the long-term statistical average value to calculate the short-term dynamic offset.

[0026] The short-term dynamic offset is used to determine in real time whether the current physical state of the processed medium is in an abnormal critical zone.

[0027] Specifically, by introducing the long-term statistical average value as a dynamic benchmark, the acoustic characteristics slowly drifting due to mechanical wear and gap changes of the processing unit are separated from the frequency drift caused by rapid changes in the medium process state.

[0028] According to the technical solution provided in this application, the reference benchmark is a benchmark that is dynamically updated based on the long-term statistical characteristics of the characteristic frequency;

[0029] The step of comparing the instantaneous value of the extracted feature frequency with a reference standard to identify whether the feature frequency has drifted includes the following steps:

[0030] During system operation, the long-term moving average of the characteristic frequency is continuously calculated and used as a reference benchmark for the dynamic update.

[0031] The difference between the instantaneous value of the feature frequency extracted in real time and the current reference benchmark is calculated as the instantaneous frequency offset;

[0032] When the instantaneous frequency offset is continuously positive and exceeds the first dynamic threshold, it is determined that the characteristic frequency has drifted to a higher frequency.

[0033] When the instantaneous frequency offset is continuously negative and exceeds the second dynamic threshold, it is determined that the characteristic frequency is drifting to a lower frequency.

[0034] According to the technical solution provided in this application, the following steps are also included:

[0035] Synchronously acquire the real-time position information of the processing unit relative to the target area;

[0036] Based on the real-time location information, it is determined whether the processing unit is located in a preset geometric feature region that is prone to flow field distortion.

[0037] The process involves adjusting at least one execution parameter that affects the physical state of the medium being processed before it reaches the target area, to perform real-time correction of the physical state of the medium being processed. This includes the following steps:

[0038] If the processing unit is not located in a preset geometric feature region that is prone to flow field distortion, real-time correction is performed.

[0039] According to the technical solution provided in this application, after determining whether the processing unit is located in a preset geometric feature region prone to flow field distortion, the method further includes the following steps:

[0040] If the processing unit is located in a preset geometric feature region that is prone to flow field distortion, then from the moment the processing unit enters the geometric feature region, the complete evolution sequence of the characteristic frequency and its harmonic energy distribution characteristics as the processing unit's position or time changes is continuously recorded.

[0041] The complete evolution sequence is matched and analyzed with a pre-stored feature pattern library, wherein the pre-stored feature pattern library defines the standard evolution pattern of acoustic features caused by changes in pure flow field boundary conditions under the geometric feature region.

[0042] Based on the matching analysis results, the deviation between the complete evolutionary sequence and the standard evolutionary pattern is calculated;

[0043] If the deviation is lower than the preset deviation threshold, it is determined that the current acoustic feature change is mainly caused by flow field distortion, and the current execution parameters are kept unchanged.

[0044] According to the technical solution provided in this application, the adjustment amount for reducing or increasing the parameters of the current-carrying medium, and the adjustment amount for increasing or decreasing the medium supply rate, are determined through the following steps:

[0045] A calibration mapping relationship is established, which includes a first sub-mapping relationship and a second sub-mapping relationship. The first sub-mapping relationship is used to determine the initial adjustment amount of the current-carrying medium parameters based on the currently detected change in acoustic characteristics, and the second sub-mapping relationship is used to determine the initial adjustment amount of the medium supply rate based on the change in acoustic characteristics detected at the end of the delay time after the adjustment of the current-carrying medium parameters.

[0046] According to the technical solution provided in this application, the correction mapping relationship is established by performing the following steps during the calibration phase:

[0047] For a known first or second precursor state, different amplitudes of current-carrying medium parameter adjustments are applied and the corresponding acoustic characteristic changes are recorded to establish the first sub-mapping relationship; subsequently, based on the current-carrying medium parameter adjustments, different amplitudes of medium supply rate adjustments are applied and the corresponding acoustic characteristic changes are recorded to establish the second sub-mapping relationship.

[0048] In actual operation, the initial adjustment amount of the current-carrying medium parameters and the medium supply rate is determined by using the correction mapping relationship based on the real-time detected changes in acoustic characteristics.

[0049] After each calibration, the calibration effect is evaluated by monitoring whether the acoustic characteristics return to the normal range. If the calibration effect does not meet expectations, the adjustment amount in the calibration mapping relationship is compensated.

[0050] Compared with the prior art, the beneficial effects of this application are as follows:

[0051] I. This method enables in-process quality intervention, fundamentally shifting the quality control node upstream. By real-time acquisition of acoustic signals generated by the interaction between the processing unit and the processed medium, the physical state of the medium can be sensed and determined in real time as it travels from the processing unit to the target area. Once an abnormal critical state is identified—whether the medium is becoming excessively dry (first precursor state) or excessively accumulating (second precursor state)—the system can complete a corrective action within a short window before the medium reaches the target area. This shifts quality control from traditional post-event detection or arrival-time control to in-process intervention, fundamentally eliminating batch defects and significantly improving the yield and stability of the production process.

[0052] II. This method provides a non-contact, highly sensitive, and universal medium condition monitoring approach. It utilizes naturally generated broadband acoustic signals during processing as the information carrier, eliminating the need for complex optical or contact sensors in harsh environments such as high temperature, high humidity, and high static electricity. This allows for real-time, in-situ sensing of the medium's condition. The acoustic spectrum contains rich information on medium dynamics and is extremely sensitive to subtle changes in the medium's dryness, accumulation state, particle size, and other physical properties. It can capture early, subtle process state drifts that are undetectable by traditional methods, providing a real-time data foundation for process optimization and stable control.

[0053] Third, an adaptive, fast closed-loop control system was constructed, exhibiting high robustness to complex operating conditions. This method determines the state by analyzing the changes in characteristic frequencies and their energy distribution characteristics relative to a preset benchmark, and introduces a long-term statistical average as a dynamic benchmark. This effectively separates the slow drift caused by mechanical wear and gap changes in the processing unit from the frequency drift caused by rapid changes in the medium's process state, ensuring stable judgment sensitivity throughout the system's entire lifecycle. Simultaneously, a coordinated correction strategy of first adjusting the current-carrying medium and then adjusting the supply rate with a delay ensures that the effects of the two types of adjustments are synchronously matched when the medium reaches the target area, achieving a smooth and accurate correction process. Attached Figure Description

[0054] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0055] Figure 1 A flowchart illustrating the steps of the real-time process state correction method based on acoustic features provided in this application. Detailed Implementation

[0056] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0057] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0058] As mentioned in the background section, this application proposes a real-time process state correction method based on acoustic features, such as... Figure 1 As shown, it includes the following steps:

[0059] S1. Acoustic signals generated by the interaction between the processed medium and the processing unit when the acquisition and processing unit performs processing tasks;

[0060] S2. Perform real-time spectrum analysis on the acoustic signal to extract characteristic frequencies and their energy distribution characteristics that are correlated with the current physical state of the processed medium;

[0061] S3. Based on the changes in the characteristic frequency and its energy distribution characteristics relative to a preset benchmark, determine in real time whether the current physical state of the processed medium is in an abnormal critical region; the abnormal critical region includes a first precursor state of the medium tending to be excessively dry or a second precursor state of tending to be excessively accumulated;

[0062] S4. If it is determined that the medium is in an abnormal critical zone, the physical state of the medium being processed is corrected in real time by adjusting at least one execution parameter that affects the physical state of the medium being processed before the medium being processed reaches the target area; wherein, the medium being processed is a subset of the medium being processed at the current time point, referring to the part of the medium that is currently flying from the processing unit to the target area.

[0063] Specifically, the following detailed description uses a specific application of this method in the field of electrostatic spraying as an example. In this embodiment, the processing unit is an electrostatic spraying rotary cup, the processed medium is paint, the target area is the workpiece surface, and the execution parameters include forming air pressure and / or flow rate (as a specific form of current-carrying medium parameter) and paint output (as a specific form of medium supply rate). The physical state of the processed medium includes atomization state, a first precursor state corresponding to a precursor state tending towards dry spraying, and a second precursor state corresponding to a precursor state tending towards deposit defects. Those skilled in the art should understand that this specific embodiment is only used to explain the present invention and is not intended to limit the scope of protection thereto.

[0064] Acoustic signal acquisition: This step aims to acquire the raw sound vibration information generated by the processing unit (in this embodiment, a rotary cup) during operation. In practice, a high-sensitivity, wide-frequency response acoustic sensor, such as an environmentally resistant MEMS microphone or accelerometer, can be mounted on the molded cover, spacer, or adjacent rigid support of the rotary cup. The sensor's mounting location should effectively capture the sound field of the rotary cup's atomizing cone region while minimizing direct interference from robot vibrations or environmental noise. The acquired signal is an analog voltage signal, which needs to be converted into a digital signal by an analog-to-digital converter for subsequent processing.

[0065] Real-time spectrum analysis and feature extraction involves performing a frequency domain transformation on the acquired time-domain acoustic signal to extract characteristic frequency components reflecting the physical state of the processed medium (coating in this example). This is typically implemented using the Fast Fourier Transform (FFT) algorithm or its variants (such as the Short-Time Fourier Transform). By performing FFT calculations on continuous or overlapping time windows, the frequency-energy distribution map of the acoustic signal, i.e., the spectrum, is obtained.

[0066] The object of feature extraction is the characteristic frequency and its energy distribution characteristics that are correlated with the current physical state of the processed medium. In this embodiment, the characteristic frequency is specifically manifested as the atomization cone resonance dominant frequency: when the electrostatic swivel cup rotates at high speed (usually tens of thousands of revolutions per minute) and atomizes the coating, due to the cup head structure, the periodic tearing of the coating liquid film, and specific aerodynamic coupling effects, a relatively stable and energy-concentrated characteristic frequency is excited, called the atomization cone resonance dominant frequency, which appears as a significant peak on the spectrum. Its extraction is carried out by performing a peak search on the spectrum to find the frequency point with the highest energy within the expected frequency band (e.g., 5kHz to 20kHz, the specific range depends on the swivel cup model and rotation speed) as the instantaneous value of the dominant frequency.

[0067] Energy distribution characteristics refer to the harmonic energy distribution at a characteristic frequency, such as the energy distribution at integer multiples of the dominant frequency (second harmonic, third harmonic, etc.). In a healthy and stable process state, the harmonic energy usually exhibits a specific attenuation pattern. In practice, it is necessary to calculate and record the energy amplitude or integrated energy near the dot-frequency points of the dominant frequency.

[0068] The abnormal critical state determination step, based on the changes in characteristic frequencies and their energy distribution characteristics relative to a preset benchmark, determines in real time whether the current physical state of the processed medium is in an abnormal critical region. Abnormal critical regions include either a first precursor state of the medium tending towards excessive dryness or a second precursor state of tending towards excessive accumulation.

[0069] In this embodiment, the first precursor state corresponds to the precursor state of dry spraying, which refers to the intermediate transition state where the atomized paint particles, during their flight towards the workpiece, experience a significant reduction in particle size and tend towards solidification due to rapid solvent evaporation, but have not yet formed an orange peel-like rough coating on the workpiece surface. The second precursor state corresponds to the precursor state of accumulation, which refers to the intermediate transition state where the atomized paint particles, due to poor atomization or slow solvent evaporation, have a larger particle size and poor flowability, exhibiting a tendency to accumulate on the surface during their flight towards the workpiece, but have not yet formed a sagging or excessively thick coating.

[0070] The specific method for determination includes: comparing the instantaneous value of the extracted characteristic frequency with a reference standard to identify whether the characteristic frequency has drifted. When a drift of the characteristic frequency to a higher frequency and an increase in high-frequency harmonic energy are detected, it is determined to be the first precursor state; when a drift of the characteristic frequency to a lower frequency and a broadened spectrum or the appearance of low-frequency noise are detected, it is determined to be the second precursor state.

[0071] A reference baseline is used for comparison to determine the drift. In an initial implementation, this can be a preset fixed frequency value, such as the nominal value of a characteristic frequency measured under standard process parameters and conditions. A more preferred implementation is a dynamic reference as described later.

[0072] When identifying frequency drift, the system continuously calculates the difference between the instantaneous value of the extracted feature frequency and the reference standard. If this difference remains positive and exceeds a small dead-zone threshold, it is determined to be a drift towards higher frequencies; if it remains negative and exceeds the dead-zone threshold, it is determined to be a drift towards lower frequencies. The determination needs to consider persistence to avoid noise interference; for example, it requires that the condition be met for multiple consecutive sampling periods.

[0073] Enhanced high-frequency harmonic energy: When determining high-frequency drift, it is necessary to verify whether the energy of high-frequency harmonics (such as the second and third harmonics) has increased relative to historical normal levels or relative to the dominant frequency energy. This can be achieved by calculating whether the energy ratio of the harmonics to the fundamental frequency exceeds a threshold.

[0074] Spectral broadening or low-frequency noise: When determining low-frequency drift, it is necessary to analyze the spectral morphology. Spectral broadening refers to a thickening of the main peak, which can be determined by calculating whether the -3dB bandwidth of the main peak increases. The presence of low-frequency noise refers to a wide-band energy boost without significant peaks in frequency bands below the dominant frequency (e.g., 1kHz-5kHz), which can be determined by calculating whether the energy integral of this low-frequency band increases abnormally.

[0075] This implementation transforms the abstract anomaly determination into a logical judgment based on explicit physical quantities (frequency direction, harmonic energy). Taking this embodiment as an example, its criteria are established based on the acoustic response understanding of the formation mechanism of dry spray and sludge defects: For the dry spray trend (first precursor state), rapid solvent evaporation leads to a decrease in droplet mass and a change in surface tension, making the atomized particles finer and lighter, and better able to follow in high-speed airflow. This changes the aerodynamic vibration characteristics of the atomizing cone, which may lead to an increase in the system's resonant frequency; at the same time, the sound wave scattering and turbulence noise generated by the finer particle group may manifest in the harmonic frequency band. For the sludge trend (second precursor state), poor atomization leads to large and unevenly distributed droplets. Large droplets have greater inertia, which may impact the air and generate stronger low-frequency disturbances, and destroy the stability of the atomizing cone, resulting in a decrease in the dominant frequency and dispersion (broadening) of the spectral energy, accompanied by more random low-frequency turbulence noise.

[0076] In this embodiment, the execution parameters specifically include the forming air pressure and / or flow rate (as the carrier medium parameter) and the paint output (as the medium supply rate). The correction action must be completed before the currently sprayed paint reaches the workpiece surface. This requires that the total time from the determination of an anomaly to the actuator completing the adjustment must be less than the flight time of the paint from the rotary cup to the workpiece.

[0077] This implementation achieves in-flight interception and real-time suppression of process defects. Traditional methods only detect and compensate for defects after they have formed on the surface of the target object (such as a workpiece), which is too late. This method, however, sets the monitoring point during the medium's flight, significantly advancing the control point and eliminating defects before they actually appear. Taking this embodiment as an example, the atomization quality of the coating (particle fineness, uniformity, velocity distribution) directly determines its leveling, adhesion, and final film formation effect on the workpiece surface. As a complex fluid dynamics and rotating mechanical process, any subtle change in the atomization state (such as changes in particle size distribution or atomization cone shape) immediately alters the resulting acoustic spectrum characteristics. Therefore, by interpreting these acoustic characteristics in real time, the health of the atomization state can be accurately perceived indirectly, in situ, and non-contactly, providing timely feedback control and forming a rapid closed loop based on direct feedback from the physical state.

[0078] In a preferred embodiment, the step of determining in real time whether the current physical state of the processed medium is in an abnormal critical region based on the change of the characteristic frequency and its energy distribution characteristics relative to a preset benchmark includes the following steps:

[0079] The instantaneous value of the feature frequency extracted in real time is compared with a reference benchmark to identify whether the feature frequency has drifted.

[0080] When the characteristic frequency is detected to drift towards a higher frequency and the high-frequency harmonic energy is enhanced, it is determined to be the first precursor state;

[0081] When the characteristic frequency is detected to drift to a lower frequency and the spectrum is broadened or low-frequency noise appears, it is determined to be the second precursor state.

[0082] Specifically, the system continuously calculates the difference between the instantaneous value of the extracted feature frequency and the reference benchmark. When this difference is consistently positive and exceeds a small dead-zone threshold, it is determined to be a drift towards higher frequencies; when it is consistently negative and exceeds the dead-zone threshold, it is determined to be a drift towards lower frequencies. The determination needs to consider persistence to avoid noise interference, for example, requiring that the condition be met for multiple consecutive sampling periods.

[0083] While determining high-frequency drift, it is necessary to verify whether the energy of high-frequency harmonics (such as second and third harmonics) has increased relative to historical normal levels or relative to the fundamental frequency energy. This can be achieved by calculating whether the energy ratio of harmonics to the fundamental frequency exceeds a threshold. When both high-frequency drift and high-frequency harmonic energy enhancement are satisfied, it is determined to be the first precursor state (corresponding to the dry injection precursor in this embodiment).

[0084] While determining low-frequency drift, it is necessary to analyze the spectral morphology. Spectral broadening refers to the main peak becoming wider and blunter, which can be determined by calculating whether the -3dB bandwidth of the main peak increases. The presence of low-frequency noise refers to a wide-band energy boost without significant peaks in frequency bands below the main frequency (e.g., 1kHz-5kHz), which can be determined by calculating whether the energy integral of this low-frequency band increases abnormally. When low-frequency drift and spectral broadening or low-frequency noise are simultaneously satisfied, it is determined to be the second precursor state (corresponding to the accumulation precursor in this embodiment).

[0085] In a preferred embodiment, the real-time correction of the physical state of the medium being processed by adjusting at least one execution parameter that affects the physical state of the medium being processed includes the following steps:

[0086] When the first precursor state is determined, the adjustment of reducing the parameters of the current-carrying medium is performed first, and the adjustment of increasing the medium supply rate is performed after a first predetermined delay period.

[0087] When the second precursor state is determined, the adjustment of increasing the parameters of the current-carrying medium is performed first, and the adjustment of decreasing the medium supply rate is performed after the second predetermined delay time.

[0088] The setting of the first predetermined delay time and the second predetermined delay time both ensure that the adjustment effect of the medium supply rate and the change of the medium state after the adjustment of the current-carrying medium parameters are synchronously matched when the medium arrives at the target area.

[0089] Specifically, the fluid-carrying medium parameters are manifested as the forming air pressure and / or flow rate, and the medium supply rate is manifested as the amount of paint ejected. Prioritizing adjustments to lower / increase the forming air: Forming air refers to the airflow ejected from the annular gap of the rotary cup forming hood, used to shape and constrain the cloud shape of the atomized paint. Adjustments are made by regulating the proportional valve or high-speed switching valve on the air path to change its pressure or flow rate. Prioritizing this means that after identifying early signs of defects, the adjustment of the forming air should be initiated first and immediately; this is the first step in the correction sequence.

[0090] For the initial signs of dry spraying (first warning stage): Reduce the pressure / flow rate of the forming air. In this case, the controller sends a command to the forming air valve to reduce its setpoint by a preset amount (e.g., 10%). The aim is to reduce the shearing and cooling effect of the airflow on the paint particles and slow down the evaporation rate of the solvent during flight.

[0091] For early signs of buildup (second precursor state): Increase the pressure / flow rate of the forming air. In practice, increase the setpoint by a preset margin (e.g., 15%). The aim is to enhance shear force, promote atomization, produce finer particles, and accelerate airflow to remove more solvent.

[0092] After a delay, the medium supply rate (paint output) is adjusted: Paint output refers to the volume of paint supplied from the center of the rotary cup per unit time, usually controlled by a gear pump or servo plunger pump. Adjustment is achieved by changing the pump's speed or stroke. The first / second predetermined delay time is a key timing parameter; it is not a fixed value but is calculated or calibrated based on system dynamics and transmission time. Its purpose is to ensure that the effects of the two adjustments work synchronously in the target area (workpiece surface).

[0093] When making adjustments, the system issues a command to adjust the paint output after a corresponding delay. For signs of dry spraying: Increase the paint output (e.g., by 5%) to increase the amount of wet paint per unit area as the particles dry, compensating for excessive solvent evaporation and ensuring leveling. For signs of buildup: Reduce the paint output (e.g., by 8%) to directly reduce the amount of paint deposited per unit time, mitigating the risk of excessive buildup at the source.

[0094] Synchronous matching is the core control objective. This means that by carefully setting the delay time, the newly sprayed medium (new coating) after the medium supply rate adjustment and the medium cloud (atomized coating cloud) that has reached a new stable state after the carrier medium parameters are adjusted arrive at the same location in the target area (workpiece surface) at exactly the same time. In this way, the target area receives the medium supplied at the new supply rate under the new, improved medium state, and the combined effects of both achieve optimal synergistic correction.

[0095] This implementation provides a step-by-step, coordinated parameter correction strategy, solving the problem of how to simultaneously adjust multiple parameters to address complex defects. Its sequential design of adjusting the carrier medium first and then adjusting the material (air first, material later) avoids system oscillations or mutual cancellation of effects that may be caused by simultaneous parameter abrupt changes, achieving a smooth and accurate correction process. The underlying technical principle is as follows: adjusting the carrier medium parameters (such as forming air) directly affects the aerodynamic environment of the medium field (shear force, temperature field, flow field), and its effect acts on all medium particles already in flight, with a relatively fast response (rapid airflow changes). Adjusting the medium supply rate (such as the amount of paint dispensed) changes the supply rate of the medium to be sprayed subsequently; it requires pumps and pipelines, has a relatively slower response, and its effect is mainly reflected in the subsequent medium. If both are adjusted simultaneously, the previous batch of medium, under the influence of the new airflow, will still have the old supply rate, while the subsequent batch will have a new supply rate but may not yet have adapted to the new flow field, causing inconsistent effects. Therefore, prioritizing the adjustment of the carrier medium aims to create an improved environment for the medium flying in the air; delaying the adjustment of the supply rate is to match this new environment and match the appropriate supply amount for the medium that arrives next, so that the material state that finally arrives at the target area is consistent.

[0096] In a preferred embodiment, the following steps are also included:

[0097] During the calibration phase, a step adjustment of the current-carrying medium parameters is performed, and the response process of the characteristic frequency is monitored in real time using the acoustic signal.

[0098] Record the time required from issuing the adjustment command to the characteristic frequency reaching a new steady-state value, as the state stabilization duration under this operating condition;

[0099] The transmission time is calculated based on the transmission distance and the average flight speed of the medium.

[0100] The first predetermined delay duration or the second predetermined delay duration is set as: the stable state duration minus the medium transmission duration. When the calculation result is positive, this value is taken, and when the calculation result is negative, zero is taken.

[0101] Specifically, the calibration phase refers to a series of characteristic tests conducted in commissioning mode before the system is officially put into production, or after changing important components or coating types. This includes performing a step adjustment of the carrier medium parameter (forming air): During the calibration phase, the pressure or flow rate of the forming air is controlled to suddenly change from one stable value to another (e.g., an increase of 20% from the standard value) and held. This step signal is used to stimulate the system's dynamic response.

[0102] The response process of monitoring the characteristic frequency (the main resonant frequency of the atomized cone): While performing step adjustment, the system continuously collects acoustic signals at a high sampling rate and calculates the instantaneous value of the characteristic frequency in real time, thereby obtaining a curve of the characteristic frequency changing with time.

[0103] Determination of the stable state duration (in this embodiment, the stable atomization state duration): The new steady-state value refers to the new average value of the characteristic frequency curve after the step adjustment. The system precisely times the period from the moment the adjustment command is issued until the moment the characteristic frequency value enters and remains within a small tolerance band centered on the new steady-state value. This period is the stable state duration (T_stable). This is essentially the time required for the medium field (atomization field) to transition from one stable aerodynamic state to another stable state.

[0104] Calculation of medium transmission time (paint flight time in this embodiment): Transmission distance refers to the preset typical distance from the processing unit to the target area (spraying distance from the nozzle of the rotary cup to the workpiece surface in this embodiment), which can be obtained from the robot program or process file. Average medium flight speed (average paint flight speed in this embodiment) is a parameter that needs to be calibrated or estimated. It can be obtained by high-speed camera measurement or by theoretical estimation based on parameters such as the velocity of the current-carrying medium and the electrostatic field strength. Calculation: Medium transmission time (T_flight) = Transmission distance / Average medium flight speed.

[0105] The formula for setting the delay duration is: Set the delay duration (T_delay) to T_stable - T_flight. A positive result indicates that the time required for the state to stabilize is longer than the media transmission time. This means that even if the media supply rate is adjusted immediately, the media state may not be stable when the newly supplied media arrives at the target area. Therefore, a period of time (T_delay) is needed until the media state is basically stable before adjusting the supply rate to ensure that both effects arrive synchronously. A negative or zero result indicates that the state stabilizes very quickly, long before the media transmission time. In this case, the media supply rate can be adjusted immediately (or with only a very short delay) after adjusting the current-carrying medium parameters to achieve synchronization. Setting it to 0 indicates that no additional delay is needed; the second adjustment step should be performed immediately or as soon as possible.

[0106] The principle of this method lies in identifying and quantifying two key time constants in the correction process: system response time (duration of state stabilization) and process transmission time (duration of medium transmission). Ideally, coordinated control requires that the effective time of the second control action (adjusting the supply rate) should be one system response time later than the effective time of the first control action (adjusting the current-carrying medium), but it must ensure that its effect reaches the endpoint simultaneously with the effect of the first action. Here, the endpoint is the target area (workpiece surface). Therefore, the triggering time of the second action should be later than the first action by (system response time - process transmission time). If this difference is positive, then the system needs to wait for that duration; if it is negative, it indicates that the process transmission time itself is sufficient for the system to complete the response, therefore no waiting is necessary.

[0107] In a preferred embodiment, the method further includes the following steps:

[0108] During system operation, the long-term statistical average of the characteristic frequencies is tracked and recorded;

[0109] The short-term instantaneous value of the feature frequency collected and extracted in real time is compared with the long-term statistical average value to calculate the short-term dynamic offset.

[0110] The short-term dynamic offset is used to determine in real time whether the current physical state of the processed medium is in an abnormal critical zone.

[0111] Specifically, by introducing the long-term statistical average value as a dynamic benchmark, the acoustic characteristics slowly drifting due to mechanical wear and gap changes of the processing unit are separated from the frequency drift caused by rapid changes in the medium process state.

[0112] Specifically, the long-term statistical average refers to the result of statistically averaging the instantaneous values ​​of characteristic frequencies over a sufficiently long time window. The key to its implementation lies in the definition of "long-term," which should be much longer than the process fluctuation cycle (e.g., tens of minutes, hours, or a production shift). In software implementation, an exponentially weighted moving average algorithm or a fixed-length sliding window averaging method can be used. For example, a window containing tens of thousands of sampling points can be set, and the arithmetic mean of the characteristic frequency data within the window can be continuously calculated. This average aims to characterize the inherent or background frequency characteristics of the processing unit (in this embodiment, a rotating cup system) under its current mechanical state (e.g., bearing wear, dynamic balance). It changes slowly, primarily responding to long-term, slowly varying factors such as mechanical wear and thermal expansion caused by temperature.

[0113] Short-term instantaneous values ​​refer to the characteristic frequency values ​​extracted from acoustic signals within the current moment or an extremely short time window (such as tens of milliseconds) through real-time spectrum analysis. They reflect the instantaneous process state and are directly affected by rapidly changing factors such as medium characteristics (such as coating viscosity), instantaneous fluctuations in the supply rate, and environmental disturbances (such as air disturbances).

[0114] Short-term dynamic offset calculation: The system performs the following calculation in each processing cycle (e.g., every 10 milliseconds): Short-term dynamic offset = Current short-term instantaneous value - Current long-term statistical average. This offset is a positive or negative value, and its physical meaning is the frequency change caused purely by the current process state after removing the mechanical background frequency. A positive value indicates that the process state causes the instantaneous frequency to be higher than the mechanical reference, while a negative value indicates that it is lower than the reference.

[0115] In the decision-making logic, the system no longer directly compares short-term instantaneous values ​​with a fixed threshold, but instead uses short-term dynamic offsets as the decision input. For example, a threshold for the offset can be preset: when the positive offset continuously exceeds a certain threshold, it indicates that the process is trending towards the first precursor state (corresponding to the spraying trend in this embodiment); when the negative offset continuously exceeds another threshold, it indicates that the process is trending towards the second precursor state (corresponding to the silting trend in this embodiment). In this way, even if mechanical wear causes the long-term statistical average to slowly drift from the initial 30kHz to 29.5kHz, as long as the positive offset caused by the process state reaches +500Hz, the system can still reliably identify the first precursor state, because the decision is based on the relative change relative to the current mechanical reference, rather than the absolute value.

[0116] This implementation significantly improves the robustness and reliability of the acoustic monitoring system during long-term continuous operation. It fundamentally eliminates the problem of slow drift in monitoring benchmarks caused by normal wear and aging of rotating components such as the rotary cup and bearings. This eliminates the need for frequent recalibration, allowing the system to maintain consistent defect identification thresholds for weeks or even months, greatly reducing maintenance requirements and operating costs, and ensuring long-term product quality stability. The technical principle of this solution is based on the precise understanding and separation of the signal source. The acoustic signal collected during the spraying process is a mixture of multiple physical processes: first, the mechanical acoustic background determined by the inherent vibration characteristics of the rotating mechanical system (motor, bearing, cup head); and second, the process acoustic modulation generated by the coupling of the fluid dynamics process of paint atomization with air and mechanical structures. The former changes slowly (large time constant), while the latter changes rapidly (small time constant). Through dual-timescale analysis (long-term averaging to capture the slowly changing mechanical background, and short-term instantaneous reflection of the rapidly changing process state), and calculating the difference (offset) between the two, a high-pass filtering or background subtraction operation is essentially completed. It extracts and purifies the fast-changing components containing process information from the mixed signal, so that subsequent defect judgment is only sensitive to the actual changes in process conditions, while ignoring slow mechanical performance degradation.

[0117] In a preferred embodiment, the reference benchmark is a benchmark that is dynamically updated based on the long-term statistical characteristics of the characteristic frequency;

[0118] The step of comparing the instantaneous value of the extracted feature frequency with a reference standard to identify whether the feature frequency has drifted includes the following steps:

[0119] During system operation, the long-term moving average of the characteristic frequency is continuously calculated and used as a reference benchmark for the dynamic update.

[0120] The difference between the instantaneous value of the feature frequency extracted in real time and the current reference benchmark is calculated as the instantaneous frequency offset;

[0121] When the instantaneous frequency offset is continuously positive and exceeds the first dynamic threshold, it is determined that the characteristic frequency has drifted to a higher frequency.

[0122] When the instantaneous frequency offset is continuously negative and exceeds the second dynamic threshold, it is determined that the characteristic frequency is drifting to a lower frequency.

[0123] Specifically, the dynamically updated benchmark here refers to a reference benchmark used for drift determination that is not fixed but automatically updated as the system runs. Its implementation typically employs a long-term moving average. A specific algorithm could be a sliding window average: maintaining a queue of fixed length (e.g., covering data from the past 5 minutes), each time a new instantaneous value of a feature frequency enters the queue, the oldest value is removed, and the average of all values ​​in the queue is recalculated as the new benchmark. Alternatively, it could be a forgetting factor average: new benchmark = α × old benchmark + (1-α) × new instantaneous value, where α is a factor close to 1 (e.g., 0.999), causing the influence of the old value to decay slowly. This method has low computational cost and can continuously track changes.

[0124] Instantaneous frequency offset calculation: In each control cycle, the system executes: Instantaneous frequency offset = Current instantaneous value of characteristic frequency - Current long-term moving average (i.e., dynamic baseline). This step is essentially the same as the aforementioned short-term dynamic offset calculation, and is a process of real-time removal of mechanical background.

[0125] The logic for determining drift is refined: Drift towards higher frequencies and drift towards lower frequencies are given clear and quantifiable criteria. A consistently positive value means that the instantaneous frequency offset needs to remain positive for multiple consecutive sampling periods (e.g., 10 periods, corresponding to approximately 100 milliseconds) to exclude random noise spikes. Exceeding the first dynamic threshold: This threshold (Δf_high) is not a fixed value. It can be adaptively set based on the historical instantaneous frequency offset statistical variance under the current dynamic benchmark. For example, it can be set to 3 times the historical standard deviation σ (3σ criterion), so that the threshold can adaptively adjust with the volatility of the process itself, being more sensitive in stable conditions and more robust to disturbances when there are large fluctuations.

[0126] The system determines a high-frequency drift as follows: A statistically significant high-frequency drift event occurs only when both conditions—a sustained positive frequency deviation and an amplitude exceeding the threshold—are met simultaneously. This triggers the subsequent first precursor state determination process (which requires consideration of conditions such as enhanced high-frequency harmonic energy). The determination logic for low-frequency drift is symmetrical: a low-frequency drift is determined when the instantaneous frequency deviation is consistently negative and its absolute value exceeds the second dynamic threshold.

[0127] This implementation first automatically compensates for the effects of equipment aging, ensuring that the judgment criteria are up-to-date. Second, it filters out transient interference pulses through continuous judgment. Finally, through statistically based dynamic threshold settings, the system possesses the ability to adapt to the inherent fluctuation levels of the process, capturing minor anomalies during calm periods and avoiding false alarms during periods of fluctuation, significantly improving the signal-to-noise ratio and accuracy of defect identification. Its principle is based on the theories of statistical process control and adaptive filtering. Using the long-term moving average as a dynamic benchmark is equivalent to estimating and eliminating the trend or gradual variation components in time series analysis, focusing the analysis on the fluctuation term. Performing persistence and significance tests on the fluctuation term (i.e., instantaneous frequency deviation) is a standard method for anomaly detection. The persistence test utilizes the characteristic that process anomalies usually have a certain time continuity, distinguishing them from the randomness of white noise. The significance test applies the idea of ​​statistical inference; only when the observed frequency deviation is so large that it is unlikely to be caused by normal fluctuations is a substantial change considered to have occurred.

[0128] In a preferred embodiment, the following steps are also included:

[0129] Synchronously acquire the real-time position information of the processing unit relative to the target area;

[0130] Based on the real-time location information, it is determined whether the processing unit is located in a preset geometric feature region that is prone to flow field distortion.

[0131] The process involves adjusting at least one execution parameter that affects the physical state of the medium being processed before it reaches the target area, to perform real-time correction of the physical state of the medium being processed. This includes the following steps:

[0132] If the processing unit is not located in the preset geometric feature region that is prone to flow field distortion, real-time correction is performed.

[0133] Specifically, real-time position information refers to the three-dimensional spatial coordinates (X, Y, Z) and azimuth angles (usually represented by Euler angles or quaternions) of the end effector (mounted rotary cup) of the painting robot in the workpiece coordinate system. This information is provided in real time by the robot controller and transmitted to the processing system of this method via fieldbus or communication interface.

[0134] The pre-defined geometric feature regions prone to flow field distortion are a series of spatial regions predefined during offline programming or system learning. During implementation, mathematical models of these regions need to be established. Taking this embodiment as an example, common geometric features include:

[0135] Edge / Edge Area: Defined as a strip of space extending outward at a certain distance (e.g., 20mm) from the boundary between the workpiece's solid surface and free space. When the rotary cup approaches the edge, one side is blocked by the workpiece, while the other side is open space, resulting in a severely asymmetrical airflow field during forming.

[0136] Concave / Inner Cavity Region: Defined as an inwardly recessed corner or cavity interior space. Flow fields may generate eddies and vortices here, resulting in complex pressure distribution.

[0137] Convex corner / protrusion area: defined as a sharp corner or the area around a column that protrudes outward. Streamlines are sharply bent and separated here.

[0138] The area near the orifice / slit: defined as a certain range around the small-diameter feature. Air may be drawn in or expelled from the orifice, interfering with the main atomizing cone.

[0139] These areas can be predefined by performing geometric analysis on the CAD model of the workpiece and linked to the robot trajectory program to form a risk map.

[0140] During real-time operation, the system performs rapid collision detection or distance calculation on the received real-time position and attitude information of the spinning cup and the pre-stored geometric feature region model. If the current reference point of the spinning cup (such as the center of the cup rim) enters the boundary range of any predefined region, it is determined to be within that region.

[0141] If the processing unit is not located in the preset geometric feature region that is prone to flow field distortion (this is the normal case), the system trusts the acoustic judgment result and directly performs real-time correction. If it is located in the distortion region, it proceeds to the next stage of processing (as described below).

[0142] This implementation significantly enhances the intelligence and reliability of the entire system when actually spraying complex three-dimensional workpieces by integrating the robot's spatial context information. It effectively prevents false positive alarms and malfunctions caused by physical distortions of the flow field at workpiece edges, corners, and other special locations. This upgrades the acoustic monitoring method from being applicable to simple flat workpieces to a universal solution capable of reliably handling real industrial parts with complex geometries, thus broadening its application scope.

[0143] In a preferred embodiment, after determining whether the processing unit is located in a preset geometric feature region prone to flow field distortion, the method further includes the following steps:

[0144] If the processing unit is located in a preset geometric feature region that is prone to flow field distortion, then from the moment the processing unit enters the geometric feature region, the complete evolution sequence of the characteristic frequency and its harmonic energy distribution characteristics as the processing unit's position or time changes is continuously recorded.

[0145] The complete evolution sequence is matched and analyzed with a pre-stored feature pattern library, wherein the pre-stored feature pattern library defines the standard evolution pattern of acoustic features caused by changes in pure flow field boundary conditions under the geometric feature region.

[0146] Based on the matching analysis results, the deviation between the complete evolutionary sequence and the standard evolutionary pattern is calculated;

[0147] If the deviation is lower than the preset deviation threshold, it is determined that the current acoustic feature change is mainly caused by flow field distortion, and the current execution parameters are kept unchanged.

[0148] Specifically, the complete evolution sequence refers to the data sequence of characteristic frequencies and their harmonic energy distribution characteristics recorded in chronological order from the moment the processing unit (in this embodiment, a rotary cup) enters the boundary of a certain geometric feature region (such as the edge of a workpiece, a concave corner, etc.) until the current moment (or leaving the region). This includes not only the instantaneous values ​​of the features, but also dynamic information such as their trajectory, slope, and fluctuation patterns as they change with time or position.

[0149] The pre-stored feature pattern library is a data knowledge base established during system debugging or self-learning. For each predefined geometric feature region (such as a sharp right edge or a deep concave corner), the library stores one or more standard evolution pattern curves. These standard curves are obtained through historical data accumulation: under the premise of ensuring normal process operation, the robot is repeatedly allowed to spray through the specific region, recording the typical changes in acoustic characteristics, and then averaging or extracting representative patterns. This standard curve depicts the repeatable acoustic characteristic changes that are inevitably caused by changes in the flow field due to geometric boundaries. For example, when spraying the edge of a workpiece, the characteristic frequency may first slightly increase and then decrease, and the harmonic energy ratio may exhibit a specific waveform.

[0150] Matching analysis and deviation calculation: The system aligns and compares the real-time collected complete evolutionary sequence with the standard evolutionary pattern corresponding to the current region. The comparison can employ various methods:

[0151] Dynamic time warping: an algorithm commonly used to compare two time series of different lengths and speeds. It can calculate an optimal matching path and cumulative distance, which can be used as the deviation.

[0152] Correlation coefficient or cosine similarity: Calculates the similarity of the shapes of two sequences. 1 minus the similarity can be used as the deviation.

[0153] Feature point comparison: Extract key points of the sequence (such as peaks, valleys, and zero crossings) and compare them.

[0154] The calculated deviation is a quantitative indicator that represents the difference between the acoustic evolution currently observed and the acoustic evolution that should theoretically be caused by pure flow field distortion.

[0155] The preset deviation threshold is an empirically or statistically determined threshold value. It defines the degree of difference that can still be considered normal flow field distortion.

[0156] If the deviation is below the threshold, it indicates that the observed acoustic change closely matches the known pure flow field distortion pattern. Therefore, it is determined that the current anomaly is mainly caused by predictable locational factors (flow field distortion) rather than a deterioration of the medium's process state itself (i.e., a non-coating atomization process anomaly in this embodiment). In this case, the system decides to keep the current spraying process parameters unchanged. This means that the system chooses to ignore this acoustic anomaly and not trigger correction, because it is determined to be a false alarm.

[0157] If the deviation exceeds the threshold, it indicates that the acoustic change exceeds the range that pure flow field distortion can explain, and is likely superimposed with real process anomalies (such as ongoing dry spraying or siltation). In this case, the system should trust the acoustic judgment and trigger the corresponding correction action.

[0158] This implementation enables the system to distinguish between benign environmental disturbances and severe process failures in complex, interference-filled painting environments. It significantly reduces false alarms and malfunctions at workpiece edges, corners, and other critical locations, preventing the introduction of new coating quality issues in these areas due to incorrect corrections. Simultaneously, it does not sacrifice the ability to detect genuine defects; when process anomalies are sufficiently strong, their acoustic characteristics will break through standard flow field distortion patterns, allowing the system to detect them.

[0159] In a preferred embodiment, the adjustment amount for decreasing or increasing the parameters of the current-carrying medium, and the adjustment amount for increasing or decreasing the medium supply rate, are determined through the following steps:

[0160] A calibration mapping relationship is established, which includes a first sub-mapping relationship and a second sub-mapping relationship. The first sub-mapping relationship is used to determine the initial adjustment amount of the current-carrying medium parameters based on the currently detected change in acoustic characteristics, and the second sub-mapping relationship is used to determine the initial adjustment amount of the medium supply rate based on the change in acoustic characteristics detected at the end of the delay time after the adjustment of the current-carrying medium parameters.

[0161] Specifically, the correction mapping relation is a mathematical model or data lookup table that maps the observed degree of acoustic feature anomaly to the required process parameter adjustment. Essentially, it is a control rule or function that takes the change in acoustic features as input and the suggested process parameter adjustment as output. It is further subdivided into two sub-relationships to match the step-by-step adjustment timing.

[0162] The first sub-mapping relation is specifically used to determine the initial adjustment amount of the first-step flow-carrying medium parameter (forming air in this embodiment). Its input is the currently detected change in acoustic characteristics. For example, the instantaneous frequency offset (Δf_now) of the characteristic frequency drifting to higher frequencies or the increase in the high-frequency harmonic energy ratio relative to a reference (ΔE_high) can be input. Its output is the percentage adjustment of the forming air pressure or flow rate (e.g., ΔP_air = F1(Δf_now, ΔE_high)). This function F1 can be a linear function (e.g., ΔP_air = -K1×Δf_now, the negative sign indicating that the pressure needs to be reduced when drifting to higher frequencies), a piecewise linear function, or a nonlinear mapping based on a neural network.

[0163] The second sub-mapping relationship is used to determine the initial adjustment amount of the second-step medium supply rate (paint output in this embodiment). Its input is the change in acoustic characteristics detected at the end of the first (or second) predetermined delay time after the adjustment of the current-carrying medium parameters. The key here is that the input characteristics are measured after the first-step adjustment takes effect. For example, after waiting for T_delay, the system measures the characteristic frequency or harmonic characteristics again and calculates its remaining change relative to the initial state before the first-step adjustment (Δf_after_air). Its output is the percentage adjustment of the paint output (e.g., ΔQ_paint = F2(Δf_after_air)). The design logic of function F2 is: if the first-step air adjustment has brought the acoustic characteristics back to normal (Δf_after_air is small), the second-step paint adjustment amount can be small or even zero; if there are still significant anomalies after the first step, a larger paint adjustment is needed for supplementary correction.

[0164] Once an anomaly is detected and its type (first precursor state or second precursor state) is identified during system operation, the system executes:

[0165] Step 1: Based on the real-time measurement value (Δf_now, etc.), query the first sub-mapping relationship to obtain ΔP_air, and immediately perform the adjustment of the carrier medium parameter (forming air).

[0166] Step 2: Wait for the preset delay time T_delay. When T_delay is reached, acquire new acoustic signals, extract features, and calculate the change Δf_after_air.

[0167] Step 3: Input Δf_after_air into the second sub-mapping relationship to obtain ΔQ_paint, and perform medium supply rate (paint output) adjustment.

[0168] In a preferred embodiment, the correction mapping relationship is established by performing the following steps during the calibration phase:

[0169] For a known first or second precursor state, different amplitudes of current-carrying medium parameter adjustments are applied and the corresponding acoustic characteristic changes are recorded to establish the first sub-mapping relationship; subsequently, based on the current-carrying medium parameter adjustments, different amplitudes of medium supply rate adjustments are applied and the corresponding acoustic characteristic changes are recorded to establish the second sub-mapping relationship.

[0170] In actual operation, the initial adjustment amount of the current-carrying medium parameters and the medium supply rate is determined by using the correction mapping relationship based on the real-time detected changes in acoustic characteristics.

[0171] After each calibration, the calibration effect is evaluated by monitoring whether the acoustic characteristics return to the normal range. If the calibration effect does not meet expectations, the adjustment amount in the calibration mapping relationship is compensated.

[0172] Taking this embodiment as an example: The calibration stage establishes the mapping relationship under controlled conditions, which is the process of actively stimulating the process and collecting data to construct a mathematical model.

[0173] The specific steps are as follows:

[0174] Simulated precursor states: In the laboratory or commissioning area, by changing the surface temperature of the workpiece (e.g., using a heating plate or cooling plate) or by deliberately setting inappropriate process parameters, a stable first precursor state (dry spraying tendency) or a second precursor state (sludge accumulation tendency) is artificially induced.

[0175] Establish the first sub-mapping relationship: Keeping the medium supply rate (coating output) constant, for the known first precursor state (or second precursor state), the system applies a series of adjustments to the current-carrying medium parameters (forming air) with different amplitudes (e.g., -15%, -10%, -5%, +5%, +10%, +15%) in a step or frequency sweep manner. For each adjustment, record two key data: the applied adjustment amount (ΔP_air_i) and the change in acoustic characteristics when the system stabilizes after adjustment (ΔS_i, such as Δf or ΔE_high). After collecting multiple sets of (ΔP_air_i, ΔS_i) data, the functional relationship F1 from the desired change in acoustic characteristics (input) to the required adjustment amount of the current-carrying medium parameters (output) can be obtained through curve fitting (e.g., linear regression, polynomial fitting), which is the first sub-mapping relationship. For the first and second precursor states, since the directions of adjustment are opposite, two sets of mapping relationships should be established separately.

[0176] Establish the second sub-mapping relationship: After completing the above-mentioned adjustment of the carrier medium parameters (e.g., for the first precursor state, the pressure ΔP_air has been reduced), keep these carrier medium parameters unchanged. Then, based on this, apply a series of medium supply rate adjustments (ΔQ_paint_j) of different magnitudes. Similarly, record the change in acoustic characteristics (relative to the state after adjusting only the carrier medium) ΔS'_j after each adjustment. Collect multiple sets of (ΔQ_paint_j, ΔS'_j) data and fit the function F2, i.e., the second sub-mapping relationship. This reflects the further influence of the medium supply rate on the acoustic characteristics under a specific carrier medium parameter setting.

[0177] Online application and initial adjustment determination: In actual spraying, when the acoustic characteristic change ΔS (input) is detected, the system substitutes it into the calibrated correspondences F1 and F2 to calculate the recommended initial adjustment amounts ΔP_air and ΔQ_paint.

[0178] Evaluation of correction effectiveness and compensatory adjustments:

[0179] Evaluation: After a complete calibration operation (performing ΔP_air and ΔQ_paint adjustments sequentially), the system continues to monitor acoustic characteristics over a subsequent period (e.g., twice the media transmission time). It calculates whether these characteristics stably return to the preset normal range.

[0180] Compensatory Correction: If the correction effect is not as expected (e.g., features fail to revert to normal, or drift out quickly after reversion), it indicates a deviation between the actual process response and the mapping relationship F1 or F2 established during the calibration phase. The system will then initiate correction logic. For example, suppose that for a certain high-frequency drift ΔS, the system calculates and executes an adjustment amount ΔP_air based on F1, but the effect is insufficient. The system will record that the actual required adjustment amount may be greater than ΔP_air. Then, it will fine-tune the coefficients in the corresponding region of F1 (e.g., increase the gain coefficient K1) so that the calculated ΔP_air will be larger when encountering a similar ΔS next time. This correction can be implemented based on adaptive algorithms such as iterative learning control or recursive least squares, gradually reducing the error between the model prediction and the actual response.

[0181] This implementation method enables the calibration mapping relationship to automatically adapt to time-varying factors such as batch differences in paint, changes in environmental temperature and humidity, and slow degradation of equipment component performance, ensuring that the calibration accuracy remains at a high level throughout long-term use.

[0182] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A real-time process state correction method based on acoustic features, characterized in that, Includes the following steps: Acoustic signals generated by the interaction between the processed medium and the processing unit when the acquisition and processing unit performs processing tasks; Real-time spectrum analysis is performed on the acoustic signal to extract characteristic frequencies and their energy distribution characteristics that are correlated with the current physical state of the processed medium. Based on the changes in the characteristic frequency and its energy distribution characteristics relative to a preset benchmark, it is determined in real time whether the current physical state of the processed medium is in an abnormal critical region; the abnormal critical region includes a first precursor state in which the medium tends to be excessively dry or a second precursor state in which it tends to be excessively accumulated; If it is determined to be in an abnormal critical zone, the physical state of the medium being processed is corrected in real time by adjusting at least one execution parameter that affects the physical state of the medium being processed before the medium being processed reaches the target area; wherein, the medium being processed is a subset of the medium being processed at the current time point, referring to the part of the medium that is currently flying from the processing unit to the target area. The method of determining in real time whether the current physical state of the processed medium is in an abnormal critical region based on the changes in the characteristic frequency and its energy distribution characteristics relative to a preset benchmark includes the following steps: The instantaneous value of the feature frequency extracted in real time is compared with a reference benchmark to identify whether the feature frequency has drifted. When the characteristic frequency is detected to drift towards a higher frequency and the high-frequency harmonic energy is enhanced, it is determined to be the first precursor state; When the characteristic frequency is detected to drift to a lower frequency and the spectrum is broadened or low-frequency noise appears, it is determined to be the second precursor state.

2. The real-time process state correction method based on acoustic features according to claim 1, characterized in that, The real-time correction of the physical state of the medium being processed by adjusting at least one execution parameter that affects the physical state of the medium being processed includes the following steps: When the first precursor state is determined, the adjustment of reducing the parameters of the current-carrying medium is performed first, and the adjustment of increasing the medium supply rate is performed after a first predetermined delay period. When the second precursor state is determined, the adjustment of increasing the parameters of the current-carrying medium is performed first, and the adjustment of decreasing the medium supply rate is performed after the second predetermined delay time. The setting of the first predetermined delay time and the second predetermined delay time both ensure that the adjustment effect of the medium supply rate and the change of the medium state after the adjustment of the current-carrying medium parameters are synchronously matched when the medium arrives at the target area.

3. The real-time process state correction method based on acoustic features according to claim 2, characterized in that, It also includes the following steps: During the calibration phase, a step adjustment of the current-carrying medium parameters is performed, and the response process of the characteristic frequency is monitored in real time using the acoustic signal. Record the time required from issuing the adjustment command to the characteristic frequency reaching a new steady-state value, as the state stabilization duration under this operating condition; The transmission time is calculated based on the transmission distance and the average flight speed of the medium. The first predetermined delay duration or the second predetermined delay duration is set as: the stable state duration minus the medium transmission duration. When the calculation result is positive, this value is taken, and when the calculation result is negative, zero is taken.

4. The real-time process state correction method based on acoustic features according to claim 1, characterized in that, The method also includes the following steps: During system operation, the long-term statistical average of the characteristic frequencies is tracked and recorded; The short-term instantaneous value of the feature frequency collected and extracted in real time is compared with the long-term statistical average value to calculate the short-term dynamic offset. The short-term dynamic offset is used to determine in real time whether the current physical state of the processed medium is in an abnormal critical zone. Specifically, by introducing the long-term statistical average value as a dynamic benchmark, the acoustic characteristics slowly drifting due to mechanical wear and gap changes of the processing unit are separated from the frequency drift caused by rapid changes in the medium process state.

5. The real-time process state correction method based on acoustic features according to claim 1, characterized in that, The reference benchmark is a benchmark that is dynamically updated based on the long-term statistical characteristics of the characteristic frequencies; The step of comparing the instantaneous value of the extracted feature frequency with a reference standard to identify whether the feature frequency has drifted includes the following steps: During system operation, the long-term moving average of the characteristic frequency is continuously calculated and used as a reference benchmark for the dynamic update. The difference between the instantaneous value of the feature frequency extracted in real time and the current reference benchmark is calculated as the instantaneous frequency offset; When the instantaneous frequency offset is continuously positive and exceeds the first dynamic threshold, it is determined that the characteristic frequency has drifted to a higher frequency. When the instantaneous frequency offset is continuously negative and exceeds the second dynamic threshold, it is determined that the characteristic frequency is drifting to a lower frequency.

6. The real-time process state correction method based on acoustic features according to claim 1, characterized in that, It also includes the following steps: Synchronously acquire the real-time position information of the processing unit relative to the target area; Based on the real-time location information, it is determined whether the processing unit is located in a preset geometric feature region that is prone to flow field distortion. The process involves adjusting at least one execution parameter that affects the physical state of the medium being processed before it reaches the target area, to perform real-time correction of the physical state of the medium being processed. This includes the following steps: If the processing unit is not located in the preset geometric feature region that is prone to flow field distortion, real-time correction is performed.

7. The real-time process state correction method based on acoustic features according to claim 6, characterized in that, After determining whether the processing unit is located in a preset geometric feature region prone to flow field distortion, the method further includes the following steps: If the processing unit is located in a preset geometric feature region that is prone to flow field distortion, then from the moment the processing unit enters the geometric feature region, the complete evolution sequence of the characteristic frequency and its harmonic energy distribution characteristics as the processing unit's position or time changes is continuously recorded. The complete evolution sequence is matched and analyzed with a pre-stored feature pattern library, wherein the pre-stored feature pattern library defines the standard evolution pattern of acoustic features caused by changes in pure flow field boundary conditions under the geometric feature region. Based on the matching analysis results, the deviation between the complete evolutionary sequence and the standard evolutionary pattern is calculated; If the deviation is lower than the preset deviation threshold, it is determined that the current acoustic feature change is mainly caused by flow field distortion, and the current execution parameters are kept unchanged.

8. The real-time process state correction method based on acoustic features according to claim 2, characterized in that, The adjustment amounts for decreasing or increasing the parameters of the current-carrying medium, and the adjustment amounts for increasing or decreasing the medium supply rate, are determined through the following steps: A calibration mapping relationship is established, which includes a first sub-mapping relationship and a second sub-mapping relationship. The first sub-mapping relationship is used to determine the initial adjustment amount of the current-carrying medium parameters based on the currently detected change in acoustic characteristics, and the second sub-mapping relationship is used to determine the initial adjustment amount of the medium supply rate based on the change in acoustic characteristics detected at the end of the delay time after the adjustment of the current-carrying medium parameters.

9. The real-time process state correction method based on acoustic features according to claim 8, characterized in that, The correction mapping relationship is established by performing the following steps during the calibration phase: For a known first or second precursor state, different amplitudes of current-carrying medium parameter adjustments are applied and the corresponding acoustic characteristic changes are recorded to establish the first sub-mapping relationship; Subsequently, based on the adjustment of the current-carrying medium parameters, the medium supply rate of different magnitudes was adjusted and the corresponding acoustic characteristic changes were recorded to establish the second sub-mapping relationship; In actual operation, the initial adjustment amount of the current-carrying medium parameters and the medium supply rate is determined by using the correction mapping relationship based on the real-time detected changes in acoustic characteristics. After each calibration, the calibration effect is evaluated by monitoring whether the acoustic characteristics return to the normal range. If the calibration effect does not meet expectations, the adjustment amount in the calibration mapping relationship is compensated.

Images