Pump head assembly quality on-line detection system and method

Abstract

The application discloses a pump head assembly quality online detection system and method, and belongs to the technical field of packaging material detection, and solves the problem that the prior art is difficult to detect the internal hidden assembly defects of a pump head online. The detection system comprises a conveying module, a positioning module, a positive excitation module, a visible light imaging unit, an infrared thermal imaging unit, a micro-deformation vision unit and a processing unit; the processing unit controls the positive excitation module to apply physical excitation, and synchronously controls the infrared thermal imaging unit and the micro-deformation vision unit to collect images, extracts assembly integrity features from external structure images, extracts abnormal temperature rise patterns from temperature field change sequences, extracts dynamic response features from micro-deformation or vibration modal images, and compares the above features with preset qualified features to determine whether assembly defects exist. The application is used for online full detection of pump head assembly quality, and can detect hidden defects such as spring pre-tightening force misalignment and poor sealing.

Description

Technical Field

[0001] This invention relates to the field of packaging material testing technology. More specifically, this invention relates to an online testing system and method for pump head assembly quality. Background Technology

[0002] In the field of pump head manufacturing, pump heads are typically assembled from multiple components such as pump cores, springs, pistons, and seals. The quality of this assembly directly affects the pressing feel, sealing performance, and service life. Currently, the main methods for inspecting pump head assembly quality include manual visual inspection, traditional machine vision inspection, and offline functional sampling testing.

[0003] Manual visual inspection and traditional machine vision inspection can only identify external defects such as missing, misaligned, and scratched parts on the outer surface of the pump head, but cannot perceive the internal assembly condition of the pump head. Whether the spring preload is accurate, whether the sealing ring is installed in place, and whether the clearance between the piston and the cylinder wall is uniform cannot be judged from the external image.

[0004] While offline functional sampling tests can directly measure functional indicators such as pump head pressure and sealing performance, they only cover a very small percentage of products, posing a risk of missed inspections. When production processes malfunction, a large number of defective products may be generated between two sampling inspections, and offline testing has a long cycle, making online full inspection impossible.

[0005] Therefore, a long-standing technical problem is how to detect latent functional defects inside the pump head, such as inaccurate spring preload, poor seal fit, and sluggish movement of moving parts, in an online, full-inspection manner without damaging the pump head. Solving this problem faces multiple difficulties: the frictional heat generated during pump head operation is extremely weak (typically only a few tenths of a degree Celsius), the deformation amplitude is at the micrometer level, and the production line environment is subject to interference from airflow, vibration, and temperature fluctuations, making it difficult to reliably capture these weak signals. Furthermore, the strict requirements for inspection cycle time limit the application of complex inspection methods. Summary of the Invention

[0006] One object of the present invention is to solve at least the above-mentioned problems and to provide at least the advantages that will be described later.

[0007] Another objective of this invention is to provide an online inspection system for pump head assembly quality, which solves the problem of how to detect hidden assembly defects inside the pump head that cannot be judged by appearance in an online, non-destructive, and comprehensive manner.

[0008] To achieve these objectives and other advantages according to the present invention, an online inspection system for pump head assembly quality is provided, comprising: The conveying module is used to sequentially transport the pump heads to the testing station; A positioning module, used to position the pump head at the detection station; An active excitation module is installed at the detection station. The active excitation module is used to apply physical excitation to the pump head after positioning, simulating the actual working state of the pump head. A multispectral vision acquisition module, comprising: Visible light imaging unit, which is used to acquire images of the external structure of the pump head under the physical excitation; An infrared thermal imaging unit is used to acquire the sequence of surface temperature field changes of the pump head during physical excitation. Micro-deformation vision unit, which is used to acquire images of micro-deformation or vibration modes of the pump head caused by physical excitation; A processing unit, communicatively connected to the active excitation module and the multispectral vision acquisition module, is configured to perform the following steps: Send an excitation start command to the active excitation module; at a first predetermined time after the excitation start command is sent, send a data acquisition start command to the infrared thermal imaging unit and the micro-deformation vision unit simultaneously, and send a data acquisition stop command at a second predetermined time. The assembly integrity features are extracted from the external structure images acquired by the visible light imaging unit; the abnormal temperature rise patterns are extracted from the temperature field change sequences acquired by the infrared thermal imaging unit; and the dynamic response features are extracted from the micro-deformation or vibration mode images acquired by the micro-deformation vision unit. The assembly integrity features, abnormal temperature rise patterns, and dynamic response features are compared with preset qualified features, and the pump head is determined to have assembly defects based on the comparison results.

[0009] Preferably, the active excitation module is at least one of a mechanical actuator and a vibration exciter; wherein the mechanical actuator is used to perform a simulated pressing stroke on the pressing part of the pump head, and the vibration exciter is used to apply micro-amplitude vibrations of a specific frequency and amplitude to the pump head.

[0010] Preferably, the assembly integrity feature includes at least one of the following: missing components, misaligned components, surface scratches, or dimensional deviations of the pump head; wherein, missing components are determined by detecting whether a corresponding component exists in a predetermined area, the dimensional deviation is calculated by measuring the difference between the critical dimension and the nominal dimension of the pump head, the misaligned components are determined by comparing the actual component position of the pump head with the positional deviation of a standard template, and the surface scratches are determined by detecting abnormal grayscale gradients or abrupt texture changes on the surface of the pump head.

[0011] Preferably, the abnormal temperature rise mode is characterized by at least one of the following parameters: the temperature rise rate, the temperature drop time constant, or the peak temperature of the pump head surface; wherein the temperature rise rate is the slope of the temperature change of the pump head surface over time during the excitation process, the temperature drop time constant is the time required for the temperature to drop to 37% of the initial temperature difference after the excitation ends, and the peak temperature is the highest temperature reached by the pump head surface during the excitation process.

[0012] Preferably, the dynamic response characteristics are characterized by at least one of the following parameters: deformation amplitude, deformation recovery time, vibration frequency, or vibration damping ratio of the pump head; wherein, the deformation amplitude is the maximum displacement of the pump head during the excitation process, the deformation recovery time is the time required for the deformation to recover to 10% of the maximum deformation value after the excitation ends, and the vibration frequency and vibration damping ratio are obtained by performing spectral analysis on the vibration signal of the pump head.

[0013] Preferably, when the processing unit compares the assembly integrity feature, the abnormal temperature rise mode, and the dynamic response feature with the preset qualified features, it calculates the difference degree of each feature, and determines that the pump head has an assembly defect when the difference degree of any feature exceeds a preset threshold; wherein, the difference degree is the ratio of the deviation of each feature value from the mean of the qualified features to the standard deviation, or the distance of each feature value from the qualified feature interval, and the preset threshold is set according to the statistical distribution of the reference pump head.

[0014] Preferably, the processing unit is further configured to: The thermal response time curve is extracted from the temperature field change sequence acquired by the infrared thermal imaging unit, and the mechanical response time curve is extracted from the micro-deformation or vibration mode image acquired by the micro-deformation vision unit. Calculate the phase difference between the thermal response time curve and the mechanical response time curve; The phase difference is compared with a preset phase difference threshold; When the phase difference exceeds the threshold, it is determined that the pump head has an assembly defect, and the defect type is identified according to the sign and magnitude of the phase difference.

[0015] Preferably, the processing unit is further configured to: During the process of applying physical excitation by the active excitation module, the temperature field change sequence acquired by the infrared thermal imaging unit and the micro-deformation or vibration mode image sequence acquired by the micro-deformation vision unit are recorded simultaneously. The temperature field change sequence and the micro-deformation or vibration mode image sequence are decomposed in the time and frequency domain respectively to extract the thermal response energy distribution and mechanical response energy distribution corresponding to each frequency band. Based on the ratio of the thermal response energy distribution to the mechanical response energy distribution in different frequency bands, a multi-source energy distribution map of the pump head is constructed. The multi-source energy distribution map is compared with a preset defect source reference map, which includes at least one of injection molding defect reference map, assembly defect reference map, and calibration defect reference map. Based on the comparison results, the root cause of the pump head assembly defect was identified.

[0016] Preferably, the processing unit is further configured to: Statistical analysis of the defect source identification results for a continuously predetermined number of pump heads; When a predetermined number of pump heads are identified as having defects from the same source, a corresponding process parameter adjustment instruction is generated based on the source of the defect. The process parameter adjustment command is sent to the corresponding upstream process equipment, which includes at least one of an injection molding machine, an assembly machine, or a calibration device. After sending the process parameter adjustment command, the assembly quality of subsequent pump heads is continuously monitored, and the effectiveness of the process parameter adjustment command is verified based on the monitoring results.

[0017] This invention also discloses a detection method based on the aforementioned online pump head assembly quality detection system, comprising the following steps: Step A: Convey and position the pump heads sequentially at the testing station; Step B: Apply physical excitation to the positioned pump head to simulate its actual working state; Step C: During the application of the physical excitation, simultaneously acquire images of the external structure of the pump head, the sequence of surface temperature field changes, and images of micro-deformation or vibration modes; Step D: Extract assembly integrity features from the external structure image, extract abnormal temperature rise patterns from the temperature field change sequence, and extract dynamic response features from the micro-deformation or vibration mode image; Step E: Compare the assembly integrity feature, the abnormal temperature rise mode, and the dynamic response feature with the preset qualified features, and determine whether the pump head has assembly defects based on the comparison results.

[0018] The present invention has at least the following beneficial effects: First, by simultaneously acquiring information from the visible light imaging unit, the infrared thermal imaging unit, and the micro-deformation vision unit, the pump head's appearance geometry, surface temperature field spatiotemporal distribution, and micro-deformation or vibration mode information under active excitation can be obtained simultaneously. This comprehensively characterizes the assembly quality of the pump head from three physical dimensions: appearance, thermodynamics, and mechanics, providing a sufficient data foundation for identifying latent defects.

[0019] Secondly, by extracting assembly integrity features from external structural images, abnormal temperature rise patterns from temperature field change sequences, and dynamic response features from microscopic deformation or vibration modal images through the processing unit, and comparing these three types of features with preset qualified features, it is possible to identify hidden assembly defects that cannot be determined by a single sensor, including inaccurate spring preload, poor seal fit, and sluggish movement of moving parts, thus filling the detection blind spot of existing technologies that can only detect appearance defects or can only pass random inspection tests.

[0020] Third, by controlling the timing relationship between the excitation start command and the acquisition start command through the processing unit, the infrared thermal imaging unit and the micro-deformation vision unit start acquiring at the first predetermined moment after the excitation is started and stop acquiring at the second predetermined moment. This ensures that there is a definite time phase relationship between image acquisition and physical excitation, and can flexibly capture the transient response during the excitation process or the residual heat decay process after the excitation ends, thus overcoming the problem of key signal loss caused by the asynchrony between acquisition and excitation.

[0021] Fourth, the processing unit calculates the difference between each feature and the preset qualified features, and determines that there is an assembly defect when the difference of any feature exceeds the preset threshold. This realizes statistical quantitative judgment, eliminates the influence of fluctuations in qualified feature values ​​under different batches and different environmental conditions on the judgment results, and improves the consistency and reliability of the detection.

[0022] Fifth, the processing unit extracts thermal response time curves from temperature field change sequences and mechanical response time curves from microscopic deformation or vibration mode images, calculates the phase difference between the two curves, and uses the sign and magnitude of the phase difference to identify defect types. This not only determines whether defects exist, but also distinguishes specific types such as spring defects, sealing defects, or friction defects, providing clear information for process analysis and rework.

[0023] Sixth, the processing unit performs time-frequency domain decomposition on the temperature field change sequence and the micro-deformation or vibration mode image sequence, extracts the thermal response energy distribution and mechanical response energy distribution of each frequency band, and calculates the ratio of the two in different frequency bands to construct a multi-source energy distribution spectrum. This spectrum is compared with the preset injection molding defect, assembly defect or calibration defect reference spectrum, which can identify the root cause of assembly defects and solve the problem that traditional detection systems can only determine whether they are qualified but cannot trace the process in which the defects are generated.

[0024] Seventh, by statistically analyzing the defect source identification results of multiple consecutive pump heads through the processing unit, when a predetermined number of consecutive pump heads are identified as having defects from the same source, the corresponding process parameter adjustment instructions are automatically generated and sent to the injection molding machine, assembly machine, or calibration equipment. After adjustment, the assembly quality of subsequent pump heads is continuously monitored to verify the adjustment effect, thus realizing closed-loop process control based on detection results. This significantly shortens the process adjustment time after batch defects occur, and reduces downtime losses and scrap rates.

[0025] Eighth, through the coordinated work of the conveying module, positioning module, active excitation module, multispectral vision acquisition module, and processing unit, online, non-destructive, and full inspection of pump head assembly quality is achieved. The inspection process does not require damage to the pump head and can cover all products. Compared with the existing technology that separates appearance visual inspection from functional sampling inspection, it significantly improves the inspection coverage and reduces the risk of defective products entering the market.

[0026] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Detailed Implementation

[0027] The present invention will be further described in detail below with reference to embodiments, so that those skilled in the art can implement it based on the description.

[0028] This invention discloses an online inspection system for pump head assembly quality, comprising: The conveying module is used to sequentially transport the pump heads to the testing station; A positioning module, used to position the pump head at the detection station; An active excitation module is installed at the detection station. The active excitation module is used to apply physical excitation to the pump head after positioning, simulating the actual working state of the pump head. A multispectral vision acquisition module, comprising: Visible light imaging unit, which is used to acquire images of the external structure of the pump head under the physical excitation; An infrared thermal imaging unit is used to acquire the sequence of surface temperature field changes of the pump head during physical excitation. Micro-deformation vision unit, which is used to acquire images of micro-deformation or vibration modes of the pump head caused by physical excitation; A processing unit, communicatively connected to the active excitation module and the multispectral vision acquisition module, is configured to perform the following steps: Send an excitation start command to the active excitation module; at a first predetermined time after the excitation start command is sent, send a data acquisition start command to the infrared thermal imaging unit and the micro-deformation vision unit simultaneously, and send a data acquisition stop command at a second predetermined time. The assembly integrity features are extracted from the external structure images acquired by the visible light imaging unit; the abnormal temperature rise patterns are extracted from the temperature field change sequences acquired by the infrared thermal imaging unit; and the dynamic response features are extracted from the micro-deformation or vibration mode images acquired by the micro-deformation vision unit. The assembly integrity feature, the abnormal temperature rise mode, and the dynamic response feature are compared with the preset qualified features, and the assembly defects of the pump head are determined based on the comparison results. The active excitation module is at least one of a mechanical actuator and a vibration exciter; wherein, the mechanical actuator is used to perform a simulated pressing stroke on the pressing part of the pump head, and the vibration exciter is used to apply micro-amplitude vibrations of a specific frequency and amplitude to the pump head; The assembly integrity features include at least one of the following: missing components, misaligned components, surface scratches, or dimensional deviations of the pump head; wherein, missing components are determined by detecting whether a corresponding component exists in a predetermined area, dimensional deviations are calculated by measuring the difference between the critical dimension and the nominal dimension of the pump head, misaligned components are determined by comparing the actual component position of the pump head with the positional deviation of a standard template, and surface scratches are determined by detecting abnormal grayscale gradients or abrupt texture changes on the surface of the pump head.

[0029] The abnormal temperature rise mode is characterized by at least one of the following parameters: the temperature rise rate, the temperature drop time constant, or the peak temperature of the pump head surface; wherein, the temperature rise rate is the slope of the temperature change of the pump head surface with time during the excitation process, the temperature drop time constant is the time required for the temperature to drop to 37% of the initial temperature difference after the excitation ends, and the peak temperature is the highest temperature reached by the pump head surface during the excitation process. The dynamic response characteristics are characterized by at least one of the following parameters: deformation amplitude, deformation recovery time, vibration frequency, or vibration damping ratio of the pump head; wherein, the deformation amplitude is the maximum displacement of the pump head during the excitation process, the deformation recovery time is the time required for the deformation to recover to 10% of the maximum deformation value after the excitation ends, and the vibration frequency and vibration damping ratio are obtained by performing spectral analysis on the vibration signal of the pump head. When the processing unit compares the assembly integrity feature, the abnormal temperature rise mode, and the dynamic response feature with the preset qualified feature, it calculates the difference degree of each feature, and determines that the pump head has an assembly defect when the difference degree of any feature exceeds the preset threshold. The difference degree is the ratio of the deviation of each feature value from the mean of the qualified feature to the standard deviation, or the distance of each feature value from the qualified feature interval. The preset threshold is set according to the statistical distribution of the reference pump head.

[0030] In the above technical solution, the processing unit can be an industrial control computer, which has a built-in image acquisition card, motion control card and digital input / output interface, used to coordinate the timing control, image acquisition, feature extraction and defect judgment of the entire detection process.

[0031] The conveying module can use a belt conveyor to sequentially transport the pump heads to the inspection station. The digital output interface of the processing unit connects to the driver of the conveyor motor to control the start and stop of the conveyor belt. The positioning module may include a through-beam photoelectric sensor, which is installed at a certain distance (e.g., 50mm) in front of the inspection station, and its signal output terminal is connected to the input interface of the processing unit. There are several ways to fix the pump head at the inspection station. For example, one method is that when the processing unit receives a signal from the photoelectric sensor that the pump head is blocking the light path, the processing unit does not immediately stop the conveyor belt, but waits for a pre-calibrated delay time before sending a stop command, so that the pump head stops exactly at the center of the inspection station. The calibration method for this delay time is as follows: first, set an initial delay time (e.g., the theoretical calculated value t = L / v), then run several times to measure the deviation between the actual stopping position of the pump head and the center of the inspection station, adjust the delay time according to the deviation, and repeat several times until the stopping position of the pump head meets the accuracy requirements (e.g., ±1mm). After calibration, this delay time is stored as a fixed parameter in the processing unit for direct use in subsequent inspections. Another approach is to directly mount the photoelectric sensor at the center of the detection station. When the pump head blocks the light path, the conveyor belt stops immediately, at which point the pump head is precisely positioned at the detection station. Alternatively, a mechanical stop can be installed at the end of the detection station as a hard limit. When the pump head moves with the conveyor belt to contact the stop, it stops. The photoelectric sensor detects whether the pump head is in position, and a flexible cushioning material (such as polyurethane) can be installed on the surface of the stop. After the pump head stops, clamping cylinders clamp and fix the pump head from both sides, further correcting for minor positional deviations. Those skilled in the art can choose the most suitable method based on factors such as production line speed, positioning accuracy requirements, and cost.

[0032] The active excitation module can be installed on a lifting platform driven by a cylinder or lead screw module, enabling the entire excitation module to move vertically. Before the pump head enters the inspection station, the lifting platform is at a high position, providing space for the pump head; after the positioning module clamps and fixes the pump head, the lifting platform descends to the working position; after the excitation action is completed, the lifting platform rises again to the high position for subsequent sorting. The active excitation module can use either an electric cylinder or a vibration exciter. An electric cylinder can have a silicone indenter with a Shore A40 hardness installed at its end to simulate the action of a human hand pressing the pump head. The processing unit sends pulse signals to the servo driver of the electric cylinder via a motion control card or digital output interface to control the pressing action. The vibration exciter can be a piezoelectric ceramic vibrator with a hard contact head (e.g., stainless steel or ceramic) installed at its end. When the lifting platform descends, this contact head contacts the surface of the pump head. The processing unit drives the vibrator via an analog output interface or a power amplifier. The above two methods can be selected according to the specific type of pump head and the inspection requirements. The lifting platform can be driven by a cylinder in conjunction with a linear guide rail, or by a lead screw module driven by a stepper motor.

[0033] The multispectral vision acquisition module includes a visible light imaging unit, an infrared thermal imaging unit, and a micro-deformation vision unit. The visible light imaging unit can employ an industrial area array camera, equipped with a 16mm fixed-focus lens and a ring LED light source. The camera can be mounted on the side of the inspection station. This camera can be connected to the image acquisition card of the processing unit via Ethernet or USB interface. The processing unit can extract assembly integrity features from the external structural image: Specifically, the missing component is determined by detecting whether a corresponding component exists in a predetermined area. For example, the image to be tested is compared with a pre-stored qualified template using a normalized cross-correlation algorithm, with a similarity threshold set to 0.85. If no corresponding component responds in the predetermined area, it is determined that the component is missing. The misalignment of the component is determined by comparing the actual position of the pump head with the position deviation of the standard template. When the deviation exceeds a preset pixel distance (e.g., 5 pixels), it is determined that the component is misaligned. The appearance scratches are determined by detecting abnormal gray-level gradients or abrupt texture changes on the surface of the pump head. For example, the texture is analyzed using a gray-level co-occurrence matrix. When the contrast or entropy value exceeds a threshold, it is identified as a scratch. The size deviation is calculated by measuring the difference between the key dimensions of the pump head and the nominal dimensions. For example, the pump head pressing surface diameter and pump head height are extracted by edge detection. When the difference between the measured dimensions and the nominal dimensions exceeds 0.1 mm, it is determined that the size deviation is dimensional.

[0034] The infrared thermal imaging unit can use an uncooled infrared thermal imager with a thermal sensitivity of 0.05℃ and a frame rate of 50fps. This infrared thermal imager can be installed to the side or slightly above the inspection station, with the lens aimed at the pump head sidewall. Its optical axis can be at an angle of 30 to 45 degrees to the pump head axis to avoid interference with the lifting platform of the active excitation module and the pump head inlet / outlet channel. The infrared thermal imager can be connected to the processing unit via Ethernet or USB interface for continuous acquisition of the temperature field change sequence on the pump head surface. The processing unit can extract abnormal temperature rise patterns from the temperature field sequence: for example, a circular area with a diameter of 5mm can be selected on the pump head sidewall, and the curve of the average temperature change of this area over time can be calculated. The temperature rise rate can be obtained by linearly fitting the temperature data within 100ms after the excitation begins, in units of ℃ / ms. The peak temperature can be the highest temperature during the excitation process, in units of ℃. The temperature drop time constant can be obtained by fitting the temperature decay curve T(t) = T0 + (T... peak The equation -T0)·exp(-t / τ) is obtained, where: T(t) is the temperature at time t, in °C; T0 is the stable temperature after a sufficiently long time (e.g., 5s) following the excitation, in °C; T peak The peak temperature at the moment the excitation ends is expressed in °C; t is time, expressed in seconds; τ is the temperature drop time constant, expressed in seconds, reflecting the temperature drop from T... peak The rate of descent to T0. Fitting can be performed using the least squares method, applying a nonlinear regression of the experimental data points to the aforementioned exponential function to obtain an estimate of τ. This can be achieved using common numerical computation software (such as MATLAB or Python SciPy).

[0035] The micro-deformation vision unit can employ a high-speed camera, paired with a telecentric lens and a coaxial light source. The camera can be mounted on the other side of the inspection station, at a distance of 80mm from the pump head surface, with an acquisition frequency set to 200fps. This camera connects to the image acquisition card of the processing unit via a high-speed data interface (e.g., Camera Link or USB 3.0). To enhance surface texture for digital image correlation calculations, the natural surface texture formed during pump head injection molding can be used as feature points. If the natural texture is insufficient, a laser projector can be installed at the inspection station to project a random speckle pattern onto the pump head surface the instant the high-speed camera acquires the image. Neither of these methods requires spraying or physical contact with the pump head, thus not affecting product cleanliness.

[0036] To achieve pixel-level spatial correspondence between visible light images, infrared thermal images, and micro-deformation images, a joint calibration board is used to spatially calibrate the multispectral vision acquisition module. The joint calibration board is a ceramic plate with multiple high-emissivity dots (identifiable by infrared) and high-reflectivity crosshairs (identifiable by visible light) machined on its surface. The dots and crosshairs have a known spatial geometric relationship. The calibration board is placed at the pump head reference position of the detection station, and visible light images, infrared thermal images, and micro-deformation images are acquired respectively. The coordinates of feature points in each image are extracted, and the rotation and translation matrices between the coordinate systems of each camera are solved through homography transformation, establishing a pixel mapping relationship from infrared and micro-deformation images to visible light images. After calibration, the processing unit can accurately map the infrared temperature field information and micro-deformation information to the corresponding pump head region in the visible light image, achieving spatial fusion of multi-source data.

[0037] The processing unit can calculate the displacement field using digital image correlation: the subset size can be set to 31×31 pixels, the step size can be set to 5 pixels, and the zero-mean normalized cross-correlation function can be used as a similarity measure. Sub-pixel displacement accuracy of 0.05 pixels can be achieved through cubic spline interpolation. Dynamic response features can be extracted from the displacement field: for example, the deformation amplitude can be the maximum displacement at a point on the pump head sidewall during excitation, in mm; the deformation recovery time can be the time required for the displacement to recover to 10% of its maximum value after excitation, in seconds; a fast Fourier transform can be performed on the displacement-time curve to extract the dominant frequency in the spectrum as the vibration frequency, in Hz, and the vibration damping ratio can be calculated using the half-power bandwidth method, i.e., damping ratio = Δf / (2f0), where f0 is the peak frequency (Hz), Δf is the bandwidth (Hz) corresponding to a 3dB drop in peak value, and the damping ratio is dimensionless.

[0038] The processing unit connects the electric cylinder driver and each camera via a hardware synchronization line. Its timing control can be as follows: an excitation start command is sent to the electric cylinder, causing it to press down. At the first predetermined moment after the excitation start command (which can be set to 0ms, i.e., simultaneous start, or the end of pressing, e.g., 100ms), a data acquisition start command is simultaneously sent to the infrared thermal imager and high-speed camera (via a trigger signal line). At the second predetermined moment (e.g., 500ms after excitation start), a data acquisition stop command is sent. The method for establishing preset qualified features can be as follows: Take 30 manually confirmed qualified reference pump heads, and extract assembly integrity features, abnormal temperature rise mode features, and dynamic response features under the same excitation and data acquisition conditions. Calculate the arithmetic mean μ and sample standard deviation σ for each feature. The processing unit compares each feature value x of the pump head under test with the corresponding μ and σ, calculating the difference D = |x-μ| / σ. When the difference D of any feature exceeds a preset threshold, the pump head is determined to have an assembly defect. This threshold can be set according to specific product requirements; for example, it can be set to 3, meaning a defect is judged when it exceeds the mean ± 3 times the standard deviation. As an example, the measured temperature rise rate of a pump head is 0.04℃ / ms, while the acceptable benchmark has μ=0.10℃ / ms and σ=0.02℃ / ms, so D=3, which is exactly equal to the threshold. If the measured value is 0.03℃ / ms, then D=3.5, exceeding the threshold, and can be judged as a defect. Regarding the temperature drop time constant, if the acceptable benchmark has μ=1.5s and σ=0.3s, and the measured τ=2.4s, then D=3, exceeding the threshold, and can be judged as a poor seal. If the measured τ=1.8s, then D=1, not exceeding the threshold, and is judged as acceptable. It should be noted that for non-normally distributed features, they can be converted to an approximately normal distribution using Box-Cox transformation or other methods before statistical analysis. In this example, it is assumed that each feature presents an approximately normal distribution in the acceptable samples.

[0039] After completing the determination, the processing unit can output the determination result through its communication interface (such as Ethernet, RS232 or digital output port) for use or display in subsequent processes.

[0040] The technical advantages of this solution are as follows: It enables online, non-destructive, and full inspection of hidden assembly defects inside the pump head, including inaccurate spring preload, poor seal fit, and sluggish movement of moving parts. Through multispectral visual acquisition and fusion analysis, it can detect defect types that cannot be identified by a single sensor; the dynamic threshold established through statistical methods can adapt to normal fluctuations in different batches of pump heads; the inspection process does not require damage to the pump head, effectively reducing the risk of defective products flowing into subsequent processes.

[0041] In another technical solution, the processing unit is further configured as follows: The thermal response time curve is extracted from the temperature field change sequence acquired by the infrared thermal imaging unit, and the mechanical response time curve is extracted from the micro-deformation or vibration mode image acquired by the micro-deformation vision unit. Calculate the phase difference between the thermal response time curve and the mechanical response time curve; The phase difference is compared with a preset phase difference threshold; When the phase difference exceeds the threshold, it is determined that the pump head has an assembly defect, and the defect type is identified according to the sign and magnitude of the phase difference.

[0042] In the above technical solution, based on the aforementioned system, the processing unit can also calculate the phase difference between the thermal response time curve and the mechanical response time curve in the following manner. When the active excitation module uses an electric cylinder to simulate pressing, the pressing frequency can be set to 5Hz, i.e., the period T=200ms.

[0043] The processing unit extracts the thermal response time curve from the temperature field change sequence acquired by the infrared thermal imaging unit: a circular area with a diameter of 5mm is selected on the side wall of the pump head. This area can be located at the corresponding position on the outer surface of the connection between the pump core and the bottle cap. The average temperature value of all pixels in this area is calculated. The average temperature values ​​at each moment are arranged in chronological order to obtain the original thermal response time curve, and low-pass filtering (cutoff frequency 25Hz) can be performed.

[0044] The processing unit extracts the mechanical response time curve from the micro-deformation or vibration mode images acquired by the micro-deformation vision unit: a feature point on the side wall of the pump head is selected using the digital image correlation method, the displacement of the feature point in each frame image is calculated, and the displacement at each moment is arranged in chronological order to obtain the original mechanical response time curve, which is also low-pass filtered (cutoff frequency 25Hz).

[0045] The processing unit uses cross-correlation analysis to calculate the phase difference between the two curves. Let the thermal response time curve be T(t) (temperature, unit °C), where t is time, unit ms; and the mechanical response time curve be D(t) (displacement, unit mm). The cross-correlation function R(τ) = ∫T(t)·D(t+τ)dt is calculated, where τ is the time delay variable, unit ms, and the integration interval covers a complete cycle (e.g., from 50ms before excitation begins to 150ms after excitation ends). The τ value corresponding to the maximum value of the cross-correlation function is the time delay Δt, unit ms. The phase difference is calculated using the formula φ = 360°·Δt / T, unit degrees, where T is the excitation period (unit ms). The method for establishing the preset phase difference threshold is as follows: Take 30 manually verified qualified reference pump heads, calculate their phase differences under the same excitation and acquisition conditions, and obtain the arithmetic mean μ of the phase difference. φand sample standard deviation σ φ Set the acceptable range of phase difference to μ. φ ±3σ φ In this embodiment, the phase difference range of a normal pump head is typically between -3° and 3°, therefore the acceptable phase difference range can be set to -5° to 5°. When the phase difference of the pump head under test is less than -5° or greater than 5°, an assembly defect is determined to exist. Furthermore, the defect type can be identified based on the specific value of the phase difference: if the phase difference is greater than 5° and less than or equal to 15°, it is determined that the spring preload is too large; if the phase difference is greater than 15°, it is determined that the spring preload is severely excessive or stuck; if the phase difference is less than -5° and greater than or equal to -15°, it is determined that the sealing ring is too tight; if the phase difference is less than -15°, it is determined that the sealing ring is severely too tight or there is dry friction.

[0046] The technical advantages of this solution are as follows: By analyzing the phase difference between thermal and mechanical responses, hidden assembly defects that cannot be detected by visual images or single physical quantities can be identified. For example, when the spring preload deviates from the design value, the pump head may appear normal, but the mechanical response speed during pressing will change, and the timing of frictional heat generation will also change, resulting in a characteristic shift in the phase difference. This method utilizes this physical law, using the phase difference as a comprehensive evaluation index, with its magnitude and direction directly correlated to the defect type. Compared to traditional methods that only use absolute temperature values ​​or deformation amplitudes, the phase difference is less sensitive to changes in signal strength and less affected by ambient temperature fluctuations and sensor sensitivity attenuation, thus exhibiting better stability in actual production line environments. By dividing the phase difference into multiple intervals, different defect types such as inaccurate spring preload, overly tight seals, and abnormal friction in moving parts can be distinguished, providing clear feedback information for process adjustments on the production line and avoiding blind troubleshooting.

[0047] In another technical solution, the processing unit is further configured as follows: During the process of applying physical excitation by the active excitation module, the temperature field change sequence acquired by the infrared thermal imaging unit and the micro-deformation or vibration mode image sequence acquired by the micro-deformation vision unit are recorded simultaneously. The temperature field change sequence and the micro-deformation or vibration mode image sequence are decomposed in the time and frequency domain respectively to extract the thermal response energy distribution and mechanical response energy distribution corresponding to each frequency band. Based on the ratio of the thermal response energy distribution to the mechanical response energy distribution in different frequency bands, a multi-source energy distribution map of the pump head is constructed. The multi-source energy distribution map is compared with a preset defect source reference map, which includes at least one of injection molding defect reference map, assembly defect reference map, and calibration defect reference map. Based on the comparison results, the root cause of the pump head assembly defect was identified.

[0048] In the above technical solution, based on the aforementioned system, the processing unit can also identify the root cause of pump head assembly defects in the following manner. When the active excitation module uses an electric cylinder to simulate pressing, the pressing frequency is set to 5Hz and the period is 200ms. The processing unit extracts the thermal response signal H(t) from the temperature field change sequence acquired by the infrared thermal imaging unit and extracts the mechanical response signal M(t) from the deformation or vibration mode image acquired by the micro-deformation vision unit. H(t) and M(t) are respectively subjected to wavelet transform, and Morlet wavelet is selected as the basis function. After wavelet transform, the thermal response time-frequency energy spectrum P1(f,t) and the mechanical response time-frequency energy spectrum P2(f,t) are obtained. The processing unit calculates the time-averaged energy within each frequency band f, that is, the thermal response time-averaged energy E1(f)=∫P1(f,t)dt and the mechanical response time-averaged energy E2(f)=∫P2(f,t)dt, and the integration interval covers the entire excitation process and the subsequent 200ms. Then, the energy ratio R(f) = E1(f) / E2(f) for each frequency band is calculated, where E1(f) and E2(f) are dimensionless relative energy values, therefore R(f) is a dimensionless ratio. The R(f) values ​​for each frequency band are arranged in frequency order to form a multi-source energy distribution map. The frequency range is set to 10Hz to 200Hz, divided into 5 frequency bands: B1 (10Hz to 30Hz), B2 (30Hz to 60Hz), B3 (60Hz to 100Hz), B4 (100Hz to 150Hz), and B5 (150Hz to 200Hz).

[0049] The pre-defined defect source reference spectrum was obtained through experimental calibration. Thirty samples each of injection molding defects, assembly defects, and calibration defects were collected. Injection molding defect samples refer to pump heads with dimensional deviations or material inhomogeneity due to deviations in injection molding process parameters, but with normal appearance and assembly. Assembly defect samples refer to pump heads with loose components due to improper assembly. Calibration defect samples refer to pump heads with malfunctions due to spring preload deviations or improper stroke calibration. The energy ratio R(f) for each frequency band was extracted for each type of sample. The mean μ and standard deviation σ of the three types of samples in each frequency band were calculated to form three types of reference spectra. For example, for injection molding defects, the energy ratio is higher in frequency bands B4 and B5, e.g., μ=2.5, σ=0.3; for assembly defects, the energy ratio is lower in frequency bands B2 and B3, e.g., μ=0.4, σ=0.1; for calibration defects, the energy ratio peaks in frequency band B3, e.g., μ=1.8, σ=0.2, and is close to 1 in other frequency bands. Specific values ​​need to be calibrated through experiments; this example is only provided as a case study.

[0050] When comparing the multi-source energy distribution spectrum of the pump head under test with three types of reference spectra, Euclidean distance is used. The distances M1 between the tested spectrum and the injection molding defect reference spectrum, M2 between the tested spectrum and the assembly defect reference spectrum, and M3 between the tested spectrum and the calibration defect reference spectrum are calculated. The distance formula is as follows: Where i is the frequency band number, k=1,2,3 correspond to injection molding defects, assembly defects, and calibration defects respectively, and R i R represents the energy ratio of the multi-source energy distribution spectrum of the pump head under test in frequency band i. ki M is the mean value of the reference spectrum for the k-th type of defect in frequency band i. k The Euclidean distance between the multi-source energy distribution spectrum of the pump head under test and the reference spectrum of the k-th type of defect is dimensionless. The defect source corresponding to the smallest distance is taken as the identification result. For example, if M1 is the smallest, the defect source is determined to be an injection molding defect; if M2 is the smallest, it is determined to be an assembly defect; if M3 is the smallest, it is determined to be a calibration defect. When the minimum distance is greater than a preset threshold (e.g., twice the standard deviation), it can be determined to be a defect of unknown source. After the processing unit completes the identification, it outputs the defect source information through the communication interface.

[0051] The technical effects of this solution are as follows: By analyzing the energy ratio of thermal and mechanical responses at different frequency bands, the root causes of pump head assembly defects (injection molding defects, assembly defects, or calibration defects) can be distinguished. Injection molding defects mainly affect material internal friction, with energy concentrated in the mid-to-high frequency range; assembly defects mainly generate mechanical collisions, with energy concentrated in the mid-to-low frequency range; calibration defects alter the storage of elastic potential energy, with energy peaking at specific frequency bands. This method requires no additional hardware and is entirely based on existing sensor data, providing a basis for targeted improvements to the production process. For example, it prompts the inspection of injection molding machine parameters when injection molding defects occur continuously, and prompts the inspection of the assembly machine when assembly defects occur.

[0052] In another technical solution, the processing unit is further configured as follows: Statistical analysis of the defect source identification results for a continuously predetermined number of pump heads; When a predetermined number of pump heads are identified as having defects from the same source, a corresponding process parameter adjustment instruction is generated based on the source of the defect. The process parameter adjustment command is sent to the corresponding upstream process equipment, which includes at least one of an injection molding machine, an assembly machine, or a calibration device. After sending the process parameter adjustment command, the assembly quality of subsequent pump heads is continuously monitored, and the effectiveness of the process parameter adjustment command is verified based on the monitoring results.

[0053] In the above technical solution, based on the aforementioned system, the processing unit can also achieve closed-loop control of process parameters in the following manner: The processing unit statistically analyzes the defect source identification results of continuous pump heads, with a preset number of consecutive defects of up to three. When the processing unit continuously identifies that the defect source of all three pump heads is an injection molding defect, the processing unit generates an injection molding machine adjustment command based on a pre-stored process parameter mapping table. In this mapping table, the adjustment commands that can be set for injection molding defects include at least one of the following: increasing injection pressure by 5%, extending holding time by 0.1s, increasing barrel temperature by 2°C, and increasing cooling time by 0.2s. The above adjustment range can be set based on experience, and each adjustment range shall not exceed 10% of the total adjustment range. The processing unit sends the adjustment command to the injection molding machine controller via an industrial Ethernet or RS485 communication interface. When the defect source of all three pump heads is continuously identified as an assembly defect, the processing unit generates an assembly machine adjustment command, which may include at least one of the following: increasing assembly pressing force by 5%, adjusting pick-and-place position offset by 0.1mm, and extending pressing time by 0.1s. When three consecutive pump head defects are identified as originating from calibration defects, the processing unit generates calibration equipment adjustment instructions, which may include at least one of the following: increasing spring preload by 5%, increasing calibration stroke by 0.2 mm, or increasing calibration speed by 10%. The specific selection of these instructions can be pre-configured in the system, and those skilled in the art can set them according to the actual equipment and process.

[0054] After sending the adjustment command, the processing unit continues to monitor the defect source identification results of subsequent pump heads, counting the number of defects from the same source in 20 consecutive pump heads. For example, if an adjustment was made for injection molding defects, the number of injection molding defects identified in the subsequent 20 pump heads is counted. If this number is less than or equal to 2, the adjustment is confirmed to be effective, and the system resumes normal monitoring. If the number is greater than or equal to 5, the adjustment is deemed insufficient, and the processing unit generates a readjustment command, which can increase the original adjustment by 50% (e.g., if the original adjustment was a 5% increase in pressure, it is adjusted to a 7.5% increase in pressure), and the above monitoring steps are repeated. If, after three consecutive adjustments, the number of defects from the same source in the subsequent 20 pump heads is still greater than or equal to 5, the processing unit issues an alarm signal through the human-machine interface, prompting the operator to intervene and check. The processing unit can also record the parameters of each adjustment and the subsequent changes in the number of defects, forming a closed-loop log for process analysis.

[0055] This technical solution employs closed-loop control to automatically adjust upstream process parameters when batch defects occur, keeping the defect rate within an acceptable range. A threshold for the number of consecutive defects (3 triggers adjustment, no more than 2 out of 20 is considered valid, and more than 5 indicates insufficient adjustment) provides a clear criterion, preventing erroneous adjustments due to random fluctuations. Simultaneously, the incremental adjustment strategy and alarm mechanism ensure system stability and maintainability.

[0056] The present invention also provides a detection method based on the aforementioned online pump head assembly quality detection system, comprising the following steps: Step A: Convey and position the pump heads sequentially at the testing station; Step B: Apply physical excitation to the positioned pump head to simulate its actual working state; Step C: During the application of the physical excitation, simultaneously acquire images of the external structure of the pump head, the sequence of surface temperature field changes, and images of micro-deformation or vibration modes; Step D: Extract assembly integrity features from the external structure image, extract abnormal temperature rise patterns from the temperature field change sequence, and extract dynamic response features from the micro-deformation or vibration mode image; Step E: Compare the assembly integrity feature, the abnormal temperature rise mode, and the dynamic response feature with the preset qualified features, and determine whether the pump head has assembly defects based on the comparison results; In the above technical solution, the system can perform online inspection of pump head assembly quality according to the following steps. Step A: The pump head is sequentially conveyed to the inspection station by the conveying module, and the pump head is positioned at the inspection station by the positioning module. The conveyor belt speed can be set to 0.2 m / s. When the photoelectric sensor detects the pump head, the processing unit waits for a pre-calibrated delay time (e.g., 0.25 s) and then stops the conveyor belt, so that the pump head stops at the center of the inspection station. Then, the clamping cylinder clamps and fixes the pump head with a pressure of 0.2 MPa.

[0057] Step B involves applying physical excitation to the positioned pump head using an active excitation module to simulate its actual working state. The active excitation module can be an electric cylinder with a stroke of 10mm, a speed of 30mm / s, and a silicone pressure head with a Shore A40 hardness installed at the end, with a pressing frequency of 5Hz; alternatively, it can be a vibration exciter with a frequency of 20Hz, an amplitude of 0.05mm, and a hard contact head installed at the end.

[0058] Step C: During the application of physical excitation, the external structure image of the pump head is simultaneously acquired by the visible light imaging unit, the surface temperature field change sequence of the pump head is simultaneously acquired by the infrared thermal imaging unit, and the micro-deformation or vibration mode image of the pump head is simultaneously acquired by the micro-deformation vision unit. The visible light camera resolution can be set to 1280×1024 pixels, with a frame rate of 30fps; the infrared thermal imager frame rate is set to 100fps, with a thermal sensitivity of 0.05℃; the high-speed camera frame rate is set to 200fps, and with a telecentric lens, the pump head surface can be sprayed with water-based acrylic speckle paint to enhance the texture. The acquisition timing is as follows: the processing unit sends an excitation start command to the electric cylinder, simultaneously sends an acquisition start command to the infrared thermal imager and the high-speed camera at the first predetermined time (e.g., 0ms), and sends an acquisition stop command at the second predetermined time (e.g., 300ms after excitation start).

[0059] Step D involves extracting assembly integrity features from the external structural image, abnormal temperature rise patterns from the temperature field change sequence, and dynamic response features from the micro-deformation or vibration modal images using the processing unit. Assembly integrity features can include missing parts, misalignment, scratches, and dimensional deviations, extracted using methods such as normalized cross-correlation and edge detection. A similarity threshold of 0.85 and a dimensional deviation threshold of 0.1 mm can be set. Abnormal temperature rise patterns can include the temperature rise rate, peak temperature, and temperature drop time constant. The temperature rise rate is obtained by linearly fitting the temperature data within 100 ms after the excitation begins, in °C / ms; the peak temperature is the highest temperature during the excitation process, in °C; and the temperature drop time constant is obtained by analyzing the temperature decay curve T(t) = T0 + (T...) after the excitation ends. peak The value is obtained by nonlinear regression of -T0)·exp(-t / τ), with units of seconds, where T0 is the steady-state temperature, T peak The peak temperature is given. Dynamic response characteristics may include deformation amplitude, deformation recovery time, vibration frequency, and damping ratio. Deformation amplitude is the maximum displacement at a point on the pump head sidewall during excitation, in mm; deformation recovery time is the time required for the displacement to recover to 10% of its maximum value after excitation, in seconds; vibration frequency and damping ratio are calculated by performing a fast Fourier transform and half-power bandwidth method on the displacement-time curve. Step E: The processing unit compares the assembly integrity characteristics, abnormal temperature rise mode, and dynamic response characteristics with preset qualified characteristics, and determines whether the pump head has assembly defects based on the comparison results. The method for establishing preset qualified characteristics is as follows: Take 30 benchmark pump heads that have been manually confirmed as qualified, extract each characteristic under the same conditions, and calculate the arithmetic mean μ and the sample standard deviation σ. The difference D = |x - μ| / σ. When the D of any characteristic exceeds the threshold of 3, it is determined that there is an assembly defect.

[0060] The technical effects achieved by this solution are as follows: Through the above steps, hidden assembly defects such as inaccurate spring preload, poor sealing fit, and sluggish movement inside the pump head can be inspected online, non-destructively, and comprehensively. The inspection cycle is approximately 2.5 seconds per pump, achieving full inspection coverage. The judgment results can be output through the communication interface, providing a basis for subsequent quality control.

[0061] Example 1 This embodiment provides an online inspection system for pump head assembly quality, used to detect hidden assembly defects in cosmetic pump heads. The system includes a delivery module, a positioning module, an active excitation module, a multispectral vision acquisition module, a processing unit, and a result output interface.

[0062] The conveyor module uses a belt conveyor line with a speed set to 0.2 m / s. The pump heads move sequentially with the conveyor belt. A through-beam photoelectric sensor is installed 50 mm in front of the detection station, and the sensor output is connected to the processing unit. When the pump head blocks the light path, the processing unit waits for a pre-calibrated delay time (e.g., 0.25 s) and then stops the conveyor belt, so that the pump head stops exactly at the center of the detection station. Subsequently, the processing unit controls a clamping cylinder to clamp and fix the pump head from both sides with a pneumatic pressure of 0.2 MPa, so that the central axis of the pump head coincides with the force application axis of the subsequent excitation module.

[0063] The active excitation module uses an electric cylinder mounted on a lifting platform. The lifting platform is driven by a cylinder and initially positioned at a high position. After the pump head is fixed, the lifting platform descends to the working position, and the silicone pressure head (Shore A40 hardness) at the end of the electric cylinder contacts the top of the pump head. The processing unit sends a pulse signal to the electric cylinder driver, causing the electric cylinder to press down 10mm at a speed of 30mm / s, with a pressing frequency of 5Hz (200ms period), simulating a human hand pressing action.

[0064] The multispectral vision acquisition module includes a visible light camera, an infrared thermal imager, and a high-speed camera. The visible light camera is mounted on the side of the inspection station, with its lens optical axis at a 20-degree angle to the pump head axis, acquiring images of the external structure. The infrared thermal imager (uncooled type, thermal sensitivity 0.05℃) is mounted on the front side, with its lens aimed at the pump head sidewall at a 30-degree angle, operating at a frame rate of 100fps. The high-speed camera is mounted on the other side, equipped with a telecentric lens, 80mm away from the pump head, operating at a frame rate of 200fps. The processing unit controls the timing via a hardware synchronization line: simultaneously sending an excitation start command to the electric cylinder, it also simultaneously sends acquisition start commands to the infrared thermal imager and the high-speed camera; after 300ms of acquisition, a stop command is sent.

[0065] The processing unit extracts assembly integrity features from the visible light image: It compares templates using normalized cross-correlation, with a similarity threshold of 0.85; values ​​below this threshold indicate missing or misaligned components. Edge detection measures the pump head height and pressing surface diameter; differences exceeding 0.1mm from the nominal dimensions indicate dimensional deviation. From the infrared temperature field sequence, a circular region with a 5mm diameter on the pump head sidewall is selected, and the average temperature curve is calculated. The temperature rise rate is obtained by linearly fitting data within 100ms after the excitation begins, in °C / ms; the peak temperature is the highest temperature; the temperature drop time constant τ is obtained by fitting the decay curve T(t) = T0 + (T... peak The displacement field is calculated from high-speed camera images using digital image correlation (31×31 pixels subset, 5-pixel step size), and the displacement of feature points is extracted. The deformation amplitude is the maximum displacement (mm), and the deformation recovery time is the time (s) for the displacement to decrease to 10% of the maximum value. An FFT is performed on the displacement curve to obtain the vibration frequency (Hz) and damping ratio (half-power bandwidth method).

[0066] The preset qualified features are extracted from 30 qualified reference pump heads under the same conditions, and the mean μ and standard deviation σ are calculated. For each feature value x of the pump head under test, the difference D = |x-μ| / σ is calculated. When the D of any feature exceeds 3, an assembly defect is determined to exist. In addition, the processing unit also calculates the phase difference from the thermal response curve and the mechanical response curve. Taking the data of a complete cycle (200ms), the cross-correlation function R(τ) = ∫T(t)·D(t+τ)dt is calculated to obtain the time delay Δt, and the phase difference φ = 360°·Δt / 200 (degrees). The phase difference of qualified pump heads is usually between -3° and 3°. In this embodiment, the threshold is set to an absolute value greater than 5°. If φ>5° and ≤15°, the spring preload is determined to be too large; φ>15° is determined to be severely too large or stuck; φ<-5° and ≥-15° is determined to be too tight; φ<-15° is determined to be severely too tight or dry friction.

[0067] After the processing unit completes the judgment, it outputs the results (pass / fail and defect type) through the Ethernet interface. If three consecutive pump heads are identified as having defects from the same source (injection molding, assembly, or calibration), the processing unit sends adjustment commands to the upstream equipment through the RS485 interface, such as increasing the injection pressure by 5% or increasing the assembly pressing force by 5%, and continues to monitor the number of defects in the subsequent 20 pump heads. If the number of defects is ≤2, it is confirmed as valid; if it is ≥5, it is adjusted again; if it is invalid after three consecutive attempts, an alarm is triggered.

[0068] The beneficial effects of this embodiment are: it can perform online, non-destructive, and full inspection of hidden defects inside the pump head, with a detection cycle of about 2.5 seconds per pump head, and can distinguish the defect type through phase difference analysis. It can also automatically adjust process parameters through closed-loop control, thereby reducing scrap rate and manual intervention.

[0069] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. An online inspection system for pump head assembly quality, characterized in that, include: The conveying module is used to sequentially transport the pump heads to the testing station; A positioning module, used to position the pump head at the detection station; An active excitation module is installed at the detection station. The active excitation module is used to apply physical excitation to the pump head after positioning, simulating the actual working state of the pump head. A multispectral vision acquisition module, comprising: Visible light imaging unit, which is used to acquire images of the external structure of the pump head under the physical excitation; An infrared thermal imaging unit is used to acquire the sequence of surface temperature field changes of the pump head during physical excitation. Micro-deformation vision unit, which is used to acquire images of micro-deformation or vibration modes of the pump head caused by physical excitation; A processing unit, communicatively connected to the active excitation module and the multispectral vision acquisition module, is configured to perform the following steps: sending an excitation start command to the active excitation module; simultaneously sending an acquisition start command to the infrared thermal imaging unit and the micro-deformation vision unit at a first predetermined time after the excitation start command is issued, and sending an acquisition stop command at a second predetermined time; extracting assembly integrity features from the external structure image acquired by the visible light imaging unit, extracting abnormal temperature rise patterns from the temperature field change sequence acquired by the infrared thermal imaging unit, and extracting dynamic response features from the micro-deformation or vibration mode images acquired by the micro-deformation vision unit; comparing the assembly integrity features, the abnormal temperature rise patterns, and the dynamic response features with preset qualified features, and determining whether the pump head has assembly defects based on the comparison results.

2. The online pump head assembly quality inspection system as described in claim 1, characterized in that, The active excitation module is at least one of a mechanical actuator and a vibration exciter; wherein the mechanical actuator is used to perform a simulated pressing stroke on the pressing part of the pump head, and the vibration exciter is used to apply micro-amplitude vibrations of a specific frequency and amplitude to the pump head.

3. The online pump head assembly quality inspection system as described in claim 1, characterized in that, The assembly integrity features include at least one of the following: missing components, misaligned components, surface scratches, or dimensional deviations of the pump head; wherein, missing components are determined by detecting whether a corresponding component exists in a predetermined area, dimensional deviations are calculated by measuring the difference between the critical dimension and the nominal dimension of the pump head, misaligned components are determined by comparing the actual component position of the pump head with the positional deviation of a standard template, and surface scratches are determined by detecting abnormal grayscale gradients or abrupt texture changes on the surface of the pump head.

4. The online pump head assembly quality inspection system as described in claim 3, characterized in that, The abnormal temperature rise mode is characterized by at least one of the following parameters: the temperature rise rate, the temperature drop time constant, or the peak temperature of the pump head surface; wherein, the temperature rise rate is the slope of the temperature change of the pump head surface over time during the excitation process, the temperature drop time constant is the time required for the temperature to drop to 37% of the initial temperature difference after the excitation ends, and the peak temperature is the highest temperature reached by the pump head surface during the excitation process.

5. The online pump head assembly quality inspection system as described in claim 3, characterized in that, The dynamic response characteristics are characterized by at least one of the following parameters: deformation amplitude, deformation recovery time, vibration frequency, or vibration damping ratio of the pump head; wherein, the deformation amplitude is the maximum displacement of the pump head during the excitation process, the deformation recovery time is the time required for the deformation to recover to 10% of the maximum deformation value after the excitation ends, and the vibration frequency and vibration damping ratio are obtained by performing spectral analysis on the vibration signal of the pump head.

6. The online pump head assembly quality inspection system as described in claim 1, characterized in that, When the processing unit compares the assembly integrity feature, the abnormal temperature rise mode, and the dynamic response feature with the preset qualified features, it calculates the difference degree of each feature, and determines that the pump head has an assembly defect when the difference degree of any feature exceeds the preset threshold. The difference degree is the ratio of the deviation of each feature value from the mean of the qualified features to the standard deviation, or the distance of each feature value from the qualified feature interval. The preset threshold is set according to the statistical distribution of the reference pump head.

7. The online pump head assembly quality inspection system as described in claim 1, characterized in that, The processing unit is further configured to: The thermal response time curve is extracted from the temperature field change sequence acquired by the infrared thermal imaging unit, and the mechanical response time curve is extracted from the micro-deformation or vibration mode image acquired by the micro-deformation vision unit. Calculate the phase difference between the thermal response time curve and the mechanical response time curve; The phase difference is compared with a preset phase difference threshold; When the phase difference exceeds the threshold, it is determined that the pump head has an assembly defect.

8. The online pump head assembly quality inspection system as described in claim 1, characterized in that, The processing unit is further configured to: During the process of applying physical excitation by the active excitation module, the temperature field change sequence acquired by the infrared thermal imaging unit and the micro-deformation or vibration mode image sequence acquired by the micro-deformation vision unit are recorded simultaneously. The temperature field change sequence and the micro-deformation or vibration mode image sequence are decomposed in the time and frequency domain respectively to extract the thermal response energy distribution and mechanical response energy distribution corresponding to each frequency band. Based on the ratio of the thermal response energy distribution to the mechanical response energy distribution in different frequency bands, a multi-source energy distribution map of the pump head is constructed. The multi-source energy distribution map is compared with a preset defect source reference map, which includes at least one of injection molding defect reference map, assembly defect reference map, and calibration defect reference map. Based on the comparison results, the root cause of the pump head assembly defect was identified.

9. The online pump head assembly quality inspection system as described in claim 8, characterized in that, The processing unit is further configured to: Statistical analysis of the defect source identification results for a continuously predetermined number of pump heads; When a predetermined number of pump heads are identified as having defects from the same source, a corresponding process parameter adjustment instruction is generated based on the source of the defect. The process parameter adjustment command is sent to the corresponding upstream process equipment, which includes at least one of an injection molding machine, an assembly machine, or a calibration device. After sending the process parameter adjustment command, the assembly quality of subsequent pump heads is continuously monitored, and the effectiveness of the process parameter adjustment command is verified based on the monitoring results.

10. The detection method based on the online pump head assembly quality detection system as described in claim 1, characterized in that, Includes the following steps: Step A: Convey and position the pump heads sequentially at the testing station; Step B: Apply physical excitation to the positioned pump head to simulate its actual working state; Step C: During the application of the physical excitation, simultaneously acquire images of the external structure of the pump head, the sequence of surface temperature field changes, and images of micro-deformation or vibration modes; Step D: Extract assembly integrity features from the external structure image, extract abnormal temperature rise patterns from the temperature field change sequence, and extract dynamic response features from the micro-deformation or vibration mode image; Step E: Compare the assembly integrity feature, the abnormal temperature rise mode, and the dynamic response feature with the preset qualified features, and determine whether the pump head has assembly defects based on the comparison results.