A SMD device dispensing quality detection method and related equipment

Abstract

The present invention discloses a method for detecting the quality of glue dispensing of SMD devices and related equipment, the method comprising: obtaining the acoustic signal generated by the colloid material during the glue dispensing process; performing time-frequency domain transformation on the acoustic signal to extract the acoustic feature vector that characterizes the physical properties of the colloid; matching the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result; based on the matching result, performing multi-time scale trend analysis on the acoustic feature changes of multiple consecutive glue dispensings to obtain a trend analysis result; matching the trend analysis result with a preset abnormal pattern library to determine the abnormal type and degree of the colloid material, and obtaining a glue dispensing quality prediction result. The technical solution of the present invention can predict the glue dispensing quality in real time through acoustic feature analysis during the glue dispensing process, without the need to arrange additional special detection time after completing the glue dispensing process, significantly improving production efficiency, and at the same time, can detect potential quality problems in advance, reduce the rework rate, and improve product yield.

Description

Technical Field

[0001] The invention relates to the technical field of SMD device glue dispensing quality detection, and in particular to an SMD device glue dispensing quality detection method and related equipment. Background Art

[0002] SMD devices are short for surface mount devices, which are widely used components in the manufacture of modern electronic products. Compared with traditional through-hole components, SMD devices are directly soldered on the surface of printed circuit boards. This design not only significantly reduces the size of components, but also improves production efficiency. In the manufacturing process of SMD devices, the dispensing process plays a vital role. It precisely applies the colloid material to the specified position of the circuit board to fix components, conduct heat, insulate or fill gaps. As electronic products develop towards miniaturization and high integration, the requirements for dispensing quality are getting higher and higher. During the flow of colloid from the storage device to the target position, its physical properties will change dynamically. These changes directly affect the final dispensing quality, so the detection of dispensing quality becomes particularly important.

[0003] At present, the commonly used method for dispensing quality inspection in the industry is to collect images of glue dots through a visual system after dispensing is completed, and judge the dispensing quality through comparative analysis. This post-detection method requires additional special inspection time after the dispensing process is completed, which significantly reduces production efficiency. On a high-speed production line, each additional inspection process will directly affect the throughput of the entire production line. Especially in mass production scenarios, this additional inspection time will accumulate and become a key factor affecting the production cycle. With the increasingly fierce competition in the manufacturing industry, how to reduce or eliminate this additional inspection time and improve production efficiency without affecting quality assurance has become a core technical problem that needs to be solved in the field of SMD device manufacturing. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problem that the existing SMD device dispensing quality detection method needs to arrange additional special detection time after completing the dispensing process, which significantly reduces the production efficiency.

[0005] A first aspect of the present invention provides a method for detecting the quality of glue dispensing of an SMD device, the method comprising:

[0006] Acquire the acoustic signal generated by the colloid material during the dispensing process;

[0007] Performing time-frequency domain transformation on the acoustic signal to extract acoustic feature vectors representing the physical properties of the colloid;

[0008] Matching the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result;

[0009] Based on the matching results, a multi-time scale trend analysis is performed on the acoustic characteristic changes of multiple consecutive dispensing operations to obtain a trend analysis result;

[0010] According to the trend analysis result, the abnormal type and degree of the colloid material are determined by matching with a preset abnormal pattern library, and the dispensing quality prediction result is obtained.

[0011] Preferably, the step of obtaining the acoustic signal generated by the colloid material during the dispensing process includes:

[0012] Acquire raw acoustic data in the frequency range of 1Hz to 200kHz during the dispensing process;

[0013] Collect the background noise characteristics and equipment vibration characteristics of the dispensing environment and build filtering parameters;

[0014] Filtering the original acoustic data according to the filtering parameters to obtain acoustic data after noise reduction;

[0015] According to the physical characteristics of colloid flow, the noise-reduced acoustic data is divided into dispensing start-up phase data, stable flow phase data and stop phase data;

[0016] Extracting startup characteristic parameters from the data of the dispensing startup phase, extracting stability characteristic parameters from the data of the stable flow phase, and extracting termination characteristic parameters from the data of the stopping phase;

[0017] An acoustic feature sequence is constructed according to the startup feature parameter, the stability feature parameter and the termination feature parameter as an acoustic signal.

[0018] Preferably, the step of performing a time-frequency domain transformation on the acoustic signal to extract an acoustic feature vector representing the physical properties of the colloid includes:

[0019] Performing wavelet packet decomposition and Hilbert-Huang transform on the acoustic signal to obtain a multi-scale time-frequency representation and a time-varying characteristic;

[0020] Calculate the spectrum centroid, frequency band energy ratio and harmonic structure parameters of the startup phase data, the stable flow phase data and the stop phase data respectively to obtain a frequency domain feature set;

[0021] Extracting energy distribution features, key change points and phase continuity parameters from the multi-scale time-frequency representation and time-varying characteristics to obtain a time domain feature set;

[0022] Extracting parameters reflecting colloid viscosity change, bubble distribution parameters, colloid uniformity parameters and needle state parameters according to the frequency domain feature set and the time domain feature set;

[0023] The colloid viscosity variation parameter, the bubble distribution parameter, the colloid uniformity parameter and the needle state parameter are combined to obtain an acoustic characteristic vector.

[0024] Preferably, matching the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result includes:

[0025] Comparing the colloid viscosity change parameter of the acoustic feature vector with the typical viscosity characteristics stored in the acoustic fingerprint library, calculating the dispensing temperature and the colloid aging state index, and establishing a preliminary matching matrix;

[0026] Comparing the bubble distribution parameters based on the acoustic feature vector with the bubble feature template in the acoustic fingerprint library, and combining the dispensing pressure value, the colloid microstructure matching degree is obtained;

[0027] Cross-comparing the colloid uniformity parameter and the needle state parameter of the acoustic feature vector with the corresponding features in the acoustic fingerprint library, and generating a process adaptability score according to the dispensing height and the surface characteristics of the substrate;

[0028] The preliminary matching matrix, the colloid microstructure matching degree and the process adaptability score are integrated and a matching result is obtained through nonlinear weighted calculation.

[0029] Preferably, the colloid uniformity parameter and the needle state parameter of the acoustic feature vector are cross-compared with the corresponding features in the acoustic fingerprint library, and a process adaptability score is generated according to the dispensing height and the surface characteristics of the substrate, including:

[0030] Calculating the Mahalanobis distance between the colloidal uniformity parameter of the acoustic feature vector and the standard uniformity features corresponding to different substrate materials in the acoustic fingerprint library to generate a substrate fitness index;

[0031] Performing a correlation analysis on the needle state parameters of the acoustic feature vector and the state characteristics of needles of different usage time in the acoustic fingerprint library, and obtaining a needle wear degree score in combination with the accumulated usage time of the needle;

[0032] According to the substrate adaptability index and the dispensing height parameter, the expected range of the ratio of the dispensing diameter to the height is calculated to obtain the shape matching coefficient;

[0033] The needle wear degree score, the shape matching coefficient, the substrate surface temperature and the ambient humidity data are combined to generate a process adaptability score reflecting the bonding strength between the colloid and the substrate interface.

[0034] Preferably, based on the matching result, performing multi-time scale trend analysis on the acoustic characteristic changes of multiple consecutive dispensing operations to obtain trend analysis results includes:

[0035] According to the preliminary matching matrix, the colloid microstructure matching degree and the process adaptability score in the matching results, combined with the acoustic feature vectors of multiple consecutive dispensing, a microscopic time scale feature sequence, a mesoscopic time scale feature sequence and a macroscopic time scale feature sequence are constructed;

[0036] According to the colloid viscosity information in the preliminary matching matrix, a morphological feature detection algorithm is applied to the microscopic time scale feature sequence to identify transient abnormal points within a single dispensing, and obtain microscopic abnormal features;

[0037] Combined with the colloid microstructure matching degree, a sliding window change point detection algorithm is applied to the mesoscopic time scale feature sequence to identify the trend change of colloid performance during 2-200 consecutive dispensing times, and obtain the mesoscopic change pattern;

[0038] According to the process adaptability score, a periodic decomposition and trend extraction algorithm is applied to the macroscopic time scale feature sequence to separate the periodic fluctuation caused by environmental factors and the long-term change trend of colloid performance from 1-1000 dispensing data to obtain macroscopic trend characteristics;

[0039] The microscopic abnormal characteristics, mesoscopic change patterns and macroscopic trend characteristics are integrated at multiple scales, and the quality risk advance indicators of abnormal colloid viscosity, increased bubbles, needle blockage and colloid aging are calculated to obtain trend analysis results.

[0040] Preferably, the trend analysis result is matched with a preset abnormal pattern library to determine the abnormal type and degree of the colloid material to obtain the dispensing quality prediction result, including:

[0041] The trend analysis result is used to form a time series feature map in the feature space, and a similarity matrix is ​​calculated with the typical abnormal feature map stored in the preset abnormal pattern library to obtain the abnormal pattern matching degree;

[0042] According to the abnormal pattern matching degree and the rate of change in the trend analysis result, the abnormal type and degree of abnormality of colloid viscosity abnormality, increased bubbles, needle blockage and colloid aging are determined, and an abnormality diagnosis report is generated;

[0043] Based on the abnormal diagnosis report and historical quality data, the dispensing volume deviation, dispensing position deviation, irregularity of the glue point shape and the reduction rate of the glue point adhesion strength are calculated to obtain the predicted value of the quality parameter;

[0044] The quality parameter prediction value is combined with the product quality standard to determine the dispensing quality grade and quality risk level, and obtain the dispensing quality prediction result.

[0045] Preferably, the calculation of the dispensing volume deviation, dispensing position deviation, glue point shape irregularity and glue point adhesion strength reduction rate based on the abnormal diagnosis report and historical quality data to obtain the quality parameter prediction value includes:

[0046] Extracting the type and degree data of colloid viscosity abnormality from the abnormal diagnosis report, and combining it with the historical viscosity-volume relationship data to calculate the predicted value of dispensing volume deviation;

[0047] Extract the needle blockage abnormality type and degree data from the abnormal diagnosis report, construct a flow deviation index based on the needle status assessment result, and calculate the dispensing position deviation prediction value;

[0048] Extract the data of the type and degree of abnormal bubble distribution from the abnormal diagnosis report, and calculate the predicted value of irregularity of the shape of the glue point in combination with the surface tension characteristics of the colloid;

[0049] Extract the type and degree of colloid aging abnormality data from the abnormality diagnosis report, and calculate the predicted value of the glue point adhesion strength reduction rate in combination with the substrate surface characteristics and environmental condition parameters;

[0050] The quality parameter prediction value is obtained by combining the predicted value of glue dispensing volume deviation, the predicted value of glue dispensing position deviation, the predicted value of glue dot shape irregularity and the predicted value of glue dot adhesion strength reduction rate.

[0051] A second aspect of the present invention provides a SMD device dispensing quality detection device, the SMD device dispensing quality detection device comprising:

[0052] An acoustic signal acquisition module is used to obtain the acoustic signal generated by the colloid material during the dispensing process;

[0053] A time-frequency analysis module, used for performing time-frequency domain transformation on the acoustic signal and extracting acoustic feature vectors representing the physical properties of the colloid;

[0054] A fingerprint matching module, used to match the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result;

[0055] A trend analysis module, used for performing multi-time scale trend analysis on the acoustic characteristic changes of multiple consecutive dispensing operations based on the matching results to obtain trend analysis results;

[0056] The quality prediction module is used to match the trend analysis result with a preset abnormal pattern library to determine the abnormal type and degree of the colloid material and obtain the dispensing quality prediction result.

[0057] The third aspect of the present invention provides an SMD device dispensing quality detection device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected through lines; the at least one processor calls the instructions in the memory so that the SMD device dispensing quality detection device performs the steps of the above-mentioned SMD device dispensing quality detection method.

[0058] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the above-mentioned SMD device dispensing quality detection method.

[0059] The technical solution provided in the embodiment of the present application performs quality inspection around the real-time collection and multi-dimensional analysis of acoustic signals, thereby avoiding the need to arrange additional special inspection processes after the dispensing is completed. First, by obtaining the acoustic signals generated by the colloid material during the dispensing process, key factors such as the colloid flow state, viscosity changes, and bubble distribution can be grasped in a timely manner. Since these acoustic information are closely related to the internal structure of the colloid, potential abnormal signs can be captured before the dispensing is completed. For example, once an acoustic waveform or spectral feature that deviates significantly from the normal flow pattern is detected, the system can immediately give a warning or correction instruction on the production line, thereby reducing the chance of large-scale re-inspection or rework in the later stage.

[0060] After the acoustic signal is collected, it is transformed in the time-frequency domain and the characteristic parameters representing the physical properties of the colloid are extracted. Through wavelet packet decomposition or other time-frequency analysis methods, the originally complex acoustic data can be refined and layered, so that the acoustic information at different stages can be mapped to the high-frequency area, low-frequency area and various intermediate scales. In this process, the colloid viscosity fluctuation, bubble appearance, needle state change and other situations can be accurately identified. After obtaining these acoustic characteristics related to the health of the colloid, they are compared with the acoustic fingerprint library established in advance to quickly determine whether the current dispensing process matches the normal process range. Once the viscosity is found to be too high, the bubble distribution is abnormal, or the needle wear is too large, the system can directly trigger an early warning or adaptive process adjustment.

[0061] The above steps enable a deep integration of the production process and quality inspection. Combined with multi-time scale trend analysis, it can cover the transient anomalies within a single dispensing, the fluctuations between several batches, and the aging evolution during long-term operation. In other words, the production line can collect and evaluate process indicators in real time while dispensing is in progress, without having to wait until all components are dispensed before performing video or manual inspection. This not only saves additional inspection time, but also helps to shorten the exposure cycle of the problem. Once the risk is detected, remedial or adjustment measures can be taken immediately, thereby preventing the problem from being infinitely magnified in subsequent batch production. Since it is no longer necessary to reserve a large number of process buffers on the entire production line to wait for the test results, the overall production efficiency is also improved. Through these steps, the bottleneck caused by the additional arrangement of special inspections can be resolved, and the quality monitoring in the production process is more flexible, providing an effective way to further optimize the production rhythm in a large-scale manufacturing environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0063] Figure 1 A schematic diagram of an embodiment of a method for detecting the dispensing quality of an SMD device in an embodiment of the present invention;

[0064] Figure 2 A schematic diagram of an embodiment of a device for detecting the quality of dispensing glue of an SMD device according to an embodiment of the present invention;

[0065] Figure 3 It is a schematic diagram of an embodiment of an SMD device dispensing quality inspection device in an embodiment of the present invention.

[0066] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0069] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, which must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0070] An embodiment of the present application provides a method for detecting the dispensing quality of an SMD device. Figure 1 A flow chart of a method for detecting the dispensing quality of SMD devices provided in an embodiment of the present application. In this embodiment, the method includes:

[0071] See also Figure 1 , obtain the acoustic signal generated by the colloid material during the dispensing process;

[0072] In one embodiment of the present invention, the step of obtaining the acoustic signal generated by the colloid material during the dispensing process includes:

[0073] Acquire raw acoustic data in the frequency range of 1Hz to 200kHz during the dispensing process;

[0074] Collect the background noise characteristics and equipment vibration characteristics of the dispensing environment and build filtering parameters;

[0075] Filtering the original acoustic data according to the filtering parameters to obtain acoustic data after noise reduction;

[0076] According to the physical characteristics of colloid flow, the noise-reduced acoustic data is divided into dispensing start-up phase data, stable flow phase data and stop phase data;

[0077] Extracting startup characteristic parameters from the data of the dispensing startup phase, extracting stability characteristic parameters from the data of the stable flow phase, and extracting termination characteristic parameters from the data of the stopping phase;

[0078] An acoustic feature sequence is constructed according to the startup feature parameter, the stability feature parameter and the termination feature parameter as an acoustic signal.

[0079] The following is a detailed description of the steps involved in the above embodiment:

[0080] In the first step, during the specific implementation, a sensor with broadband capture capability is arranged at the acoustic collection end, and the signal is collected in real time during the dispensing operation. The range of 1Hz to 200kHz can take into account the mechanical vibration components in the low-frequency band and the tiny bubble bursting sound and colloid flow friction sound in the high-frequency band, so as to effectively collect complete colloid acoustic information. The consideration of adopting this range is that during the dispensing process, both relatively low-frequency overall flow and equipment vibration will be generated, and high-frequency collisions and micro-cracks between colloid molecules will also appear. If the sampling bandwidth is insufficient, features that have an important impact on quality may be missed. By obtaining the original acoustic data in this range, subsequent analysis can be more comprehensive and accurate.

[0081] In the second step, before the acoustic data is formally collected, the surrounding environment is scanned, including the continuous or pulse noise that may be generated by the air compressor, transmission mechanism and exhaust system, and the spectral distribution and amplitude of these noises are measured. At the same time, the vibration sensor is used to record the vibration characteristic curve of the equipment during operation. If the external interference in the dispensing area is not identified, it will cover or confuse the subtle changes in the colloid acoustic signal. After obtaining these characteristic data, set parameters such as filter bandwidth, noise threshold and suppression coefficient to make the subsequent noise reduction more targeted.

[0082] In the third step, the fixed noise and stable vibration interference in the original acoustic signal are digitally filtered, band-stop filtered, or wavelet decomposition threshold processed using the filter bandwidth and threshold obtained in the previous step, and the high-fidelity colloid acoustic information is retained. The effective signal during dispensing usually has obvious characteristics in both the low-frequency and high-frequency sections. If broadband filtering is used directly, unnecessary noise will inevitably remain, and if the sampling bandwidth is over-contracted, the target signal will be lost. Therefore, with the help of the vibration distribution and noise energy peak determined in the previous step, the redundant frequency bands can be suppressed, and the clutter interference can be reduced while retaining the essential acoustic characteristics. The noise reduction data obtained in this way has higher reliability for accurately extracting the characteristics of colloid flow and viscosity changes.

[0083] In the fourth step, in the noise reduction results, the data is divided into three periods: start-up, stability, and stop, based on the start-up time of the glue discharge, the pressure sensor feedback, and the change in acoustic amplitude. The viscosity and flow rate of the glue will gradually increase from zero at the start-up, and will show relatively constant acoustic characteristics during stable flow, while in the stop stage, it will show rapid attenuation of the sound source and residual sound of glue cutting. By distinguishing these three stages, on the one hand, it is possible to better identify the unique characteristics of each stage, and on the other hand, it avoids confusing the transient peaks of the start-up or stop process with the continuous data of the stable stage.

[0084] In the fifth step, in the three divided stages, the acoustic pulse amplitude and rise rate at the initial stage of dispensing are extracted in the startup stage to measure the quality of the transition from static to smooth flow; the spectrum centroid, harmonic components and time-varying energy are extracted in the stable stage to reflect the consistency of the colloid flow and the viscosity stability; the attenuation gradient and residual vibration of the acoustic waveform are observed in the stop stage to evaluate the cleanliness of the final glue cutting. The setting of these parameters is based on the fluid dynamics performance at different stages of the dispensing process: in the startup stage, more attention is paid to the smoothness of flow establishment, in the stable stage, more attention is paid to the overall continuity, and in the stop stage, the focus is on the completeness of the end.

[0085] In the sixth step, the parameters obtained in the above three stages are combined in chronological order or feature correlation to construct a comprehensive sequence containing start, stabilization and stop information. The dispensing process is a dynamic behavior that runs through the entire process. It is difficult to accurately predict the overall quality by extracting information only in a single stage. After packaging the features of the three stages, comprehensive acoustic features can be presented in one sequence, which is convenient for subsequent in-depth analysis of dimensions such as colloid viscosity, flow uniformity, and needle wear. This final sequence not only has a data basis for noise reduction, but also reflects the differentiated acoustic performance of different stages, which is of great reference value for determining the overall level of dispensing quality.

[0086] Please continue reading Figure 1 , performing time-frequency domain transformation on the acoustic signal to extract acoustic feature vectors representing the physical properties of the colloid;

[0087] In one embodiment of the present invention, the step of performing a time-frequency domain transformation on the acoustic signal to extract an acoustic feature vector representing the physical properties of the colloid includes:

[0088] Performing wavelet packet decomposition and Hilbert-Huang transform on the acoustic signal to obtain a multi-scale time-frequency representation and a time-varying characteristic;

[0089] Calculate the spectrum centroid, frequency band energy ratio and harmonic structure parameters of the startup phase data, the stable flow phase data and the stop phase data respectively to obtain a frequency domain feature set;

[0090] Extracting energy distribution features, key change points and phase continuity parameters from the multi-scale time-frequency representation and time-varying characteristics to obtain a time domain feature set;

[0091] Extracting parameters reflecting colloid viscosity change, bubble distribution parameters, colloid uniformity parameters and needle state parameters according to the frequency domain feature set and the time domain feature set;

[0092] The colloid viscosity variation parameter, the bubble distribution parameter, the colloid uniformity parameter and the needle state parameter are combined to obtain an acoustic characteristic vector.

[0093] The following is a detailed description of the steps involved in the above embodiment:

[0094] In the first step, a digital signal processing module is deployed in the collected acoustic signal, and wavelet packet decomposition is used to perform multi-level splitting of the energy of the low-frequency band and the high-frequency band, and then each sub-band is finely reconstructed to retain the fluctuations of different frequency components at various scales. After the processing is completed, the Hilbert-Huang transform can be combined to perform instantaneous frequency and instantaneous energy operations on each layer of components to reveal the non-stationary characteristics of the acoustic waveform on the time axis. In some cases, multi-core parallel computing devices can also be used to increase the overall analysis rate, and other joint time-frequency analysis techniques can also be selected, but wavelet packet decomposition and Hilbert-Huang transform are more advantageous in adaptability to nonlinear and non-stationary signals. Through this process, details such as slight frequency jumps and sudden changes in pulse amplitude of the dispensing sound can be accurately presented, laying the foundation for identifying colloid performance.

[0095] In the second step, after obtaining the multi-scale time-frequency representation, the audio clips of the startup phase, stable flow phase and stop phase are processed separately. For the decomposed subbands of each phase, the main concentrated position of the subband on the spectrum, that is, the spectrum centroid, is calculated respectively, and the distribution ratio of different energy segments is statistically calculated to obtain the band energy ratio, and then the balance of the acoustic characteristics of each phase is evaluated. The harmonic structure parameters can be obtained by detecting the integer and fractional frequency components of the spectrum, which are used to determine whether there are significant resonance characteristics when the colloid flows or the needle works. Some implementation schemes can also use a high-precision spectrum analyzer with a digital processing platform to complete the above operations, or use a mathematical software library to directly call the fast Fourier transform and bandwidth integration module. Through these operations, the key indicators of each stage in the frequency domain can be obtained, which can be used to observe the stability of the colloid flow and abnormal resonance.

[0096] In the third step, based on the multi-scale decomposition results and instantaneous features generated in the previous step, the energy accumulation and local fluctuations of each layer of decomposition components in the time direction are further counted to obtain the energy distribution characteristics related to the glue discharging process. Combined with the observation of instantaneous frequency, key change points such as burst pulses or sudden pauses can be calibrated to evaluate abnormal conditions such as bubble formation or discontinuous glue discharging. In addition, the instantaneous phase continuity of each component can be calculated to confirm whether the colloid flow state is disturbed or blocked at the point where the phase suddenly jumps. If there are more abundant instrument resources, high-speed data acquisition cards can also be used in conjunction with a graphical programming environment to perform real-time phase tracking for each time segment. In this way, a set of feature information that is closely integrated with the time axis can be obtained to measure whether the glue discharging process is smooth.

[0097] In the fourth step, after the above statistics of frequency domain and time domain indicators are completed, they are mapped to possible physical meanings according to the degree of influence of each indicator on the flow behavior of the colloid. For example, the drift of the spectrum center of mass is often strongly correlated with the fluctuation of colloid viscosity, the bubble distribution can be detected by the pulse of instantaneous energy, and the time domain energy distribution and phase continuity can reveal the uniform changes in the internal structure of the colloid. The combination of harmonic structure parameters and key change points can reflect whether the needle has signs of wear or blockage. Some implementations will match the statistical range of each feature with the corresponding physical phenomenon according to the pre-established acoustic database, so as to quickly generate these four types of parameters. The parameter system obtained in this way is relatively comprehensive, covering the main sources of abnormalities and differences in operating conditions during the dispensing process.

[0098] In the fifth step, after obtaining the above four parameters, they are integrated into a complete feature vector through pre-set vector assembly rules or splicing methods based on high-dimensional data structures. Some solutions can attach this vector to the real-time monitoring system for comparison with subsequent fingerprint libraries or trend analysis modules, or directly connect to the production control device for early warning. By packaging information from multiple dimensions into one, the overall state of dispensing can be more clearly displayed, and further diagnosis and prediction can be facilitated. This operation enables the acoustic performance of the colloid throughout the dispensing cycle to be uniformly summarized.

[0099] Please continue reading Figure 1 , matching the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result;

[0100] In one embodiment of the present invention, matching the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result includes:

[0101] Comparing the colloid viscosity change parameter of the acoustic feature vector with the typical viscosity characteristics stored in the acoustic fingerprint library, calculating the dispensing temperature and the colloid aging state index, and establishing a preliminary matching matrix;

[0102] Comparing the bubble distribution parameters based on the acoustic feature vector with the bubble feature template in the acoustic fingerprint library, and combining the dispensing pressure value, the colloid microstructure matching degree is obtained;

[0103] Cross-comparing the colloid uniformity parameter and the needle state parameter of the acoustic feature vector with the corresponding features in the acoustic fingerprint library, and generating a process adaptability score according to the dispensing height and the surface characteristics of the substrate;

[0104] The preliminary matching matrix, the colloid microstructure matching degree and the process adaptability score are integrated and a matching result is obtained through nonlinear weighted calculation.

[0105] The following is a detailed description of the steps involved in the above embodiment:

[0106] In the first step, when processing the viscosity change parameters of the colloid in the acoustic feature vector, the typical viscosity characteristics stored in the acoustic fingerprint library are first used as a reference, and the actual flow viscosity level of the colloid is estimated by comparing the difference between the two. In order to achieve this process, a set of standard parameter curves can be loaded into the industrial computer. These curves are obtained by combining the sound sampling accumulated during multiple batches of normal dispensing processes with the viscosity experimental data. According to the comparison results, the real-time dispensing temperature information obtained by the temperature sensor is further combined, and the aging state index of the colloid under the current temperature conditions is obtained by using a numerical fitting algorithm or an interpolation calculation method. The aging state index is usually presented in a digital form, directly reflecting the degree of activity attenuation of the colloid, and is paired with the corresponding fingerprint library entries in a matrix form to form a preliminary matching matrix. The matching matrix formed in this step provides judgment support at the viscosity level for subsequent analysis, which enables subsequent steps to better assess whether the colloid has a significant attenuation risk.

[0107] In the second step, after obtaining the bubble distribution parameters, the bubble feature template specially recorded in the acoustic fingerprint library is retrieved to find the record entry that is closest to the measured bubble feature. The bubble feature template often includes the number of bubbles in the dispensing process, the distribution area, the frequency range corresponding to the generation and rupture, etc. Some production workshops will arrange pressure sensors near the dispensing equipment to measure the dispensing pressure, and compare the measured bubble distribution with the template in the fingerprint library to determine whether the colloid maintains a relatively uniform internal structure. For example, if the system detects higher pressure fluctuations and the corresponding high-density bubble features, it will be inferred that the colloid has a disordered microstructure when it is discharged, thereby obtaining the colloid microstructure matching degree. Through this comparison and numerical fusion, the potential impact of internal bubble abnormalities on dispensing quality can be prompted at an early stage.

[0108] In the third step, after the colloid uniformity parameters and needle state parameters are sorted out, they are cross-compared with various standard features in the acoustic fingerprint library. The acoustic fingerprint library may store the acoustic performance corresponding to different substrate materials and needle usage time. Combined with the dispensing height and substrate surface characteristics, the measured parameters can be mapped to a process adaptability score. To improve real-time performance, a set of fast correlation analysis modules can be set up in the data acquisition software to evaluate whether the spread of the colloid in the dispensing area is balanced by comparing the uniformity curve with the standard distribution. If the needle state parameters match the acoustic characteristics of wear after multiple batches of use, it indicates that the needle is currently in a high loss stage. Combined with the surface roughness or material properties of the substrate, a comprehensive score value can be obtained to clarify the overall adaptability of the dispensing process.

[0109] In the fourth step, after the analysis of the above three parts is completed, the preliminary matching matrix, the colloid microstructure matching degree and the process adaptability score are input into the nonlinear weighted calculation module. The module can use the radial basis kernel function or the neural network weighting strategy to comprehensively evaluate the three results according to the pre-set weight distribution method to obtain the matching result. The result is usually expressed as a quantitative coefficient or grade in numerical terms, which is used to reflect the overall fit between the dispensing state and the reference fingerprint library. By integrating parameters of different dimensions, the interaction of factors such as viscosity, bubbles and uniformity in the current production environment can be revealed, providing a more refined basis for judgment for monitoring personnel or automatic control systems, and triggering corresponding abnormal alarms or process adjustment strategies when necessary.

[0110] In one embodiment of the present invention, the colloid uniformity parameter and the needle state parameter of the acoustic feature vector are cross-compared with the corresponding features in the acoustic fingerprint library, and a process adaptability score is generated according to the dispensing height and the substrate surface characteristics, including:

[0111] Calculating the Mahalanobis distance between the colloidal uniformity parameter of the acoustic feature vector and the standard uniformity features corresponding to different substrate materials in the acoustic fingerprint library to generate a substrate fitness index;

[0112] Performing a correlation analysis on the needle state parameters of the acoustic feature vector and the state characteristics of needles of different usage time in the acoustic fingerprint library, and obtaining a needle wear degree score in combination with the accumulated usage time of the needle;

[0113] According to the substrate adaptability index and the dispensing height parameter, the expected range of the ratio of the dispensing diameter to the height is calculated to obtain the shape matching coefficient;

[0114] The needle wear degree score, the shape matching coefficient, the substrate surface temperature and the ambient humidity data are combined to generate a process adaptability score reflecting the bonding strength between the colloid and the substrate interface.

[0115] The following is a detailed description of the steps involved in the above embodiment:

[0116] In the first step, after extracting the colloid uniformity parameters, first retrieve the standard uniformity features recorded for different substrate materials in the acoustic fingerprint library, and then use the Mahalanobis distance to calculate the difference between the two. During implementation, the standard features in the fingerprint library can be regarded as multidimensional vectors of uniformity distribution, and the corresponding covariance matrix can be established; the uniformity parameters in the acoustic feature vector are projected in the same dimension, and then the Mahalanobis distance formula is used to quantify the similarity between the two in the overall distribution morphology. Combining the material properties and common thicknesses of various types of substrates, an index value reflecting the degree of adaptation between the substrate and the colloid uniformity can be obtained. The smaller the value, the closer the uniformity parameter is to the standard features of the target substrate, which also means that the probability of unevenness in the colloid spreading process is lower.

[0117] In the second step, after obtaining the needle status parameters, they are compared with the acoustic characteristics of the needles for different usage cycles or usage durations in the acoustic fingerprint library, and correlation calculations are performed to determine whether the current needle condition matches a certain wear level recorded in the library. If waveform and spectral features with high similarity are detected, the wear level score of the needle can be obtained in conjunction with the accumulated usage time information of the equipment. Some implementation scenarios will be equipped with a robotic arm or automatic identification system to monitor the replacement frequency and actual usage hours of the needle, and then summarize the results with the acoustic comparison to more intuitively evaluate the contribution of the needle to the stability of glue discharge. The higher the score, the closer the needle is to being scrapped or severely worn, and special attention needs to be paid to the subsequent dispensing accuracy.

[0118] In the third step, after obtaining the substrate fit index and dispensing height parameters, the reasonable ratio between the diameter and height of the glue dot can be estimated based on the numerical relationship between the two, thereby calculating the shape matching coefficient. Here, a geometric modeling algorithm is usually used to input the substrate type, dispensing height and colloid characteristics into the model to simulate the condensation shape of the colloid at different heights, and calculate the expected diameter range that is close to the actual requirements. The shape matching coefficient can be obtained by comparing the deviation between the measured value and the theoretical range. Some implementation methods may use a high-resolution imager to obtain the dispensing cross-section or surface profile, and verify it with numerical simulation to ensure that the ratio is consistent with the actual product morphology. If the coefficient is in a reasonable range, it means that the morphology of the glue dot after dispensing is within the design requirements; if the deviation is too large, it indicates that there is a risk in process consistency.

[0119] In the fourth step, based on the needle wear score and shape matching coefficient, and by incorporating external parameters such as substrate surface temperature and ambient humidity into the evaluation model, a process adaptability score that reflects the bonding strength between the colloid and the substrate interface can be synthesized. During implementation, a set of multi-factor fusion algorithms, such as weighted summation or multidimensional regression, is usually preset in the monitoring platform to numerically superimpose the wear score, shape coefficient, and temperature and humidity data to obtain a comprehensive index that characterizes the adhesion stability between the colloid and the substrate in the current environment. A high score level indicates that the dispensing interface has a more reliable bonding effect under the existing equipment and environmental conditions; when the score is lowered, it indicates that there may be problems such as inappropriate temperature and humidity, severe needle wear, or limited colloid spreadability, and it is necessary to adjust the process or maintain the equipment in a timely manner.

[0120] Please continue reading Figure 1 , based on the matching results, performing multi-time scale trend analysis on the acoustic characteristic changes of multiple consecutive dispensing operations to obtain trend analysis results;

[0121] In one embodiment of the present invention, based on the matching result, multi-time scale trend analysis is performed on the acoustic characteristic changes of multiple consecutive dispensing operations to obtain trend analysis results, including:

[0122] According to the preliminary matching matrix, the colloid microstructure matching degree and the process adaptability score in the matching results, combined with the acoustic feature vectors of multiple consecutive dispensing, a microscopic time scale feature sequence, a mesoscopic time scale feature sequence and a macroscopic time scale feature sequence are constructed;

[0123] According to the colloid viscosity information in the preliminary matching matrix, a morphological feature detection algorithm is applied to the microscopic time scale feature sequence to identify transient abnormal points within a single dispensing, and obtain microscopic abnormal features;

[0124] Combined with the colloid microstructure matching degree, a sliding window change point detection algorithm is applied to the mesoscopic time scale feature sequence to identify the trend change of colloid performance during 2-200 consecutive dispensing times, and obtain the mesoscopic change pattern;

[0125] According to the process adaptability score, a periodic decomposition and trend extraction algorithm is applied to the macroscopic time scale feature sequence to separate the periodic fluctuation caused by environmental factors and the long-term change trend of colloid performance from 1-1000 dispensing data to obtain macroscopic trend characteristics;

[0126] The microscopic abnormal characteristics, mesoscopic change patterns and macroscopic trend characteristics are integrated at multiple scales, and the quality risk advance indicators of abnormal colloid viscosity, increased bubbles, needle blockage and colloid aging are calculated to obtain trend analysis results.

[0127] The following is a detailed description of the steps involved in the above embodiment:

[0128] In the first step, after obtaining the preliminary matching matrix, the colloid microstructure matching degree and the process adaptability score, the acoustic feature vectors obtained from multiple dispensing can be sorted in chronological order to construct a multi-level time scale feature sequence. The microscopic time scale feature sequence mainly focuses on the changes in the acoustic signal at each moment in a single dispensing process, the mesoscopic time scale feature sequence covers the continuous evolution of several dispensings, and the macroscopic time scale feature sequence extends to data of larger batches and longer time spans. Some implementation methods can enable multi-threaded processing in the industrial computer, split and store acoustic data of different batches or different time periods, and then segment and mark the data according to the microscopic, mesoscopic and macroscopic time window parameters, so as to have a more intuitive hierarchical structure in subsequent analysis. This operation can provide a basic index for the subsequent identification of instantaneous anomalies, trend fluctuations and long-term changes. The combination of different time scales can reflect the complete picture of the dispensing process from details to the whole.

[0129] In the second step, when analyzing the microscopic time scale feature sequence, a morphological feature detection algorithm is introduced to identify transient anomalies within a single dispensing. This algorithm can use the set structural elements to perform multiple expansion and corrosion operations on the acoustic waveform to find pulse-shaped or step-shaped abnormal points. If the system detects a surge or sharp drop in the instantaneous amplitude, it can be determined that there are abnormalities such as microcracks, sudden bubbles, or needle jumps at this point. For data processing methods, the morphological operation module can be called in the digital signal processing platform or a custom filter can be written to complete the operation. In this way, small instability factors can be accurately located within a single dispensing cycle, which can help further correct the process or screen out defective products.

[0130] In the third step, within the mesoscopic time scale, the sliding window change point detection algorithm is used to evaluate the dynamic changes of the colloid microstructure matching degree with multiple dispensing, so as to determine whether there is a more continuous trend deviation. The sliding window change point detection can count the average eigenvalue and variance on a certain number of dispensing samples. When a statistical property mutation occurs in the new window, it is regarded as a significant change in the performance of a certain colloid. In some implementation schemes, this detection logic will be linked with the rhythm control system of the production line. Once the number of change points exceeds the threshold, a prompt will be issued to the operator. In this way, performance degradation or abnormal trends caused by increased bubbles, decreased fluidity, etc. can be captured in time, providing an earlier basis for optimizing process parameters or replacing wearing parts.

[0131] In the fourth step, at the macro time scale level, the process adaptability score can be used to refer to the periodic decomposition and trend extraction of a large amount of dispensing data, and the periodic fluctuations caused by environmental factors can be distinguished from the aging or viscosity attenuation of the colloid itself. The common means of implementation is to use harmonic analysis or multivariate regression methods to perform an overall scan of the data sequence of 1 to 1000 dispensing times, identify vibration modes with fixed periods, and performance degradation that slowly presents over time. In some scenarios, the fluctuation curves of workshop temperature and humidity will be collected synchronously, and the periodic fluctuations caused by environmental interference will be removed by comparing the signals at the same frequency. This can more accurately judge the trend of the health status of the colloid and provide a reference for maintenance planning or material preparation plans for long-term production.

[0132] In the fifth step, after the micro-abnormal features, meso-change patterns and macro-trend features are extracted, a multi-scale fusion strategy can be used to comprehensively calculate the data at these three levels, and then calculate the risk indicators such as abnormal colloid viscosity, increased bubbles, needle blockage and colloid aging. The specific approach can be to use nonlinear weighting or neural network frameworks to layer and superimpose abnormal values ​​or change rates at different scales, assign corresponding weights according to the severity of various potential faults, and finally output the quality risk advance amount. In actual use, if the risk indicator reaches a critical value, it can trigger an automatic alarm or scheduling program to promptly prompt maintenance personnel or let the automation system execute measures such as slowing down production, thereby avoiding further expansion of dispensing defects and causing more rework in the later stage.

[0133] Please continue reading Figure 1 According to the trend analysis result, the abnormal type and degree of the colloid material are determined by matching with the preset abnormal pattern library, and the dispensing quality prediction result is obtained.

[0134] In one embodiment of the present invention, the trend analysis result is matched with a preset abnormal pattern library to determine the abnormal type and degree of the colloid material, and obtain the dispensing quality prediction result, including:

[0135] The trend analysis result is used to form a time series feature map in the feature space, and a similarity matrix is ​​calculated with the typical abnormal feature map stored in the preset abnormal pattern library to obtain the abnormal pattern matching degree;

[0136] According to the abnormal pattern matching degree and the rate of change in the trend analysis result, the abnormal type and degree of abnormality of colloid viscosity abnormality, increased bubbles, needle blockage and colloid aging are determined, and an abnormality diagnosis report is generated;

[0137] Based on the abnormal diagnosis report and historical quality data, the dispensing volume deviation, dispensing position deviation, irregularity of the glue point shape and the reduction rate of the glue point adhesion strength are calculated to obtain the predicted value of the quality parameter;

[0138] The quality parameter prediction value is combined with the product quality standard to determine the dispensing quality grade and quality risk level, and obtain the dispensing quality prediction result.

[0139] The following is a detailed description of the steps involved in the above embodiment:

[0140] In the first step, after obtaining the trend analysis results, the multi-scale feature data contained therein can be mapped into a multi-dimensional feature space, and the corresponding time series feature maps can be drawn in chronological order. This process can use a digital signal processing platform or a data visualization engine to project microscopic abnormal features, mesoscopic change patterns, and macroscopic trend features into time series curves or distribution diagrams. Subsequently, these maps are compared with the typical abnormal feature maps retained in the abnormal pattern library, and a similarity matrix is ​​generated based on the overlap or similarity calculation formula in the feature space. If there is a corresponding fast matching requirement, it is also possible to consider using a sparse representation method or a multi-scale convolution comparison model to increase the matching speed. Through these processing methods, the obtained abnormal pattern matching degree can quantify the degree of conformity between the current dispensing acoustic data and the existing abnormal pattern, providing an important basis for subsequent abnormal type identification.

[0141] In the second step, after obtaining the abnormal pattern matching degree, combined with the multi-time scale change rate in the trend analysis results, the type and severity of problems such as abnormal colloid viscosity, increased bubbles, needle blockage and colloid aging are comprehensively judged. In order to achieve automatic identification, a set of rule-based or machine learning-based judgment modules can be built in the system, and the similarity threshold and change rate threshold are input into the judgment logic. Each type of abnormal phenomenon is assigned a score, and finally an abnormal diagnosis report is output. If more precision is required in the judgment process, a multivariate judgment method or hierarchical analysis method can be additionally introduced to perform a weighted comparison of the dynamic evolution of each abnormality. The diagnostic report formed in this way can classify the abnormality type and give a numerical degree evaluation, so as to facilitate timely corresponding decisions at the production site.

[0142] In the third step, based on the abnormality type and degree information provided by the diagnostic report, and in conjunction with historical quality data, the dispensing volume deviation, dispensing position deviation, irregularity of the glue dot shape, and the reduction rate of the glue dot adhesion strength are calculated. This process is usually completed by the statistical analysis module or industrial database in the industrial computer: by selecting production data of similar substrates, similar processes and environmental conditions from historical records, and comparing the quantitative impact range of the current abnormal situation in past cases, the estimated value of the corresponding quality parameter is inferred. If higher accuracy is required, a model based on regression prediction or time series prediction can also be used, using the abnormality category and magnitude listed in the report as input variables, and outputting the quality parameter value at the time of prediction. This method can directly give an estimated conclusion that is more in line with the actual production situation.

[0143] In the fourth step, after the predicted values ​​of the above quality parameters are obtained, these data are compared with the product quality standards to obtain the classification of dispensing quality and the risk level assessment, thereby forming the final dispensing quality prediction result. When implemented, the result can be presented in the form of a visual panel, digital instrument or alarm signal: if the degree of deviation from the allowable standard is detected to be too high, the system will prompt that process adjustments or shutdown maintenance are required. An alternative way is to synchronize the quality prediction results to the production execution system, allowing it to automatically assign different inspection processes or select alternative colloid materials according to the risk level. By quantifying the quality grade and risk level, the production management of the dispensing process can be made more controllable, ensuring efficient output while taking into account product quality and safety.

[0144] In one embodiment of the present invention, the calculation of the dispensing volume deviation, dispensing position deviation, irregularity of the shape of the glue dots, and the reduction rate of the glue dot adhesion strength based on the abnormal diagnosis report and the historical quality data to obtain the quality parameter prediction value includes:

[0145] Extracting the type and degree data of colloid viscosity abnormality from the abnormal diagnosis report, and combining it with the historical viscosity-volume relationship data to calculate the predicted value of dispensing volume deviation;

[0146] Extract the needle blockage abnormality type and degree data from the abnormal diagnosis report, construct a flow deviation index based on the needle status assessment result, and calculate the dispensing position deviation prediction value;

[0147] Extract the data of the type and degree of abnormal bubble distribution from the abnormal diagnosis report, and calculate the predicted value of irregularity of the shape of the glue point in combination with the surface tension characteristics of the colloid;

[0148] Extract the type and degree of colloid aging abnormality data from the abnormality diagnosis report, and calculate the predicted value of the glue point adhesion strength reduction rate in combination with the substrate surface characteristics and environmental condition parameters;

[0149] The quality parameter prediction value is obtained by combining the predicted value of glue dispensing volume deviation, the predicted value of glue dispensing position deviation, the predicted value of glue dot shape irregularity and the predicted value of glue dot adhesion strength reduction rate.

[0150] The following is a detailed description of the steps involved in the above embodiment:

[0151] In the first step, when extracting the type and degree of abnormal colloid viscosity information, first read the quantitative indicators related to viscosity from the abnormal diagnosis report, including real-time temperature, consistency change coefficient, and viscosity aging rate. Compare this information with the viscosity-volume relationship curve in the historical database, and obtain the predicted value of the dispensing volume deviation through numerical fitting or interpolation algorithm. When implemented, in some occasions, a multidimensional regression analysis module can be integrated into the digital signal processing platform to make an estimate based on the actual volume deviation corresponding to the increase or decrease in viscosity. This can not only reflect the change in flow rate caused by changes in the colloid over time or the environment, but also connect with the formula data of the production process, which helps to understand whether the dispensing volume exceeds the expected range.

[0152] In the second step, for the abnormal type and degree of needle blockage, the frequency of blockage and duration of blockage can be locked in the report, and a flow deviation index can be constructed in combination with the previously recorded needle status assessment results. This index can be calculated based on the pressure fluctuation detected by the pressure sensor and the estimated value of the needle flow area, and then combined with the acoustic reflection information during the glue discharge process for an overall assessment. In the process of calculating the dispensing position deviation, if the flow deviation index shows that the colloid flow direction deviates significantly from the main axis, the position deviation value will be further increased. Some implementation solutions can deploy a laser rangefinder on the production line to fine-tune and record the area near the needle so as to more accurately map the relationship between the degree of blockage and the glue discharge offset in the positioning model. This can form a more flexible blockage impact analysis method, which can provide a reference for equipment operation and maintenance or needle replacement at an early stage.

[0153] In the third step, when analyzing the abnormal type and degree of bubble distribution, the report usually includes the segment where the bubbles appear, the pulse shape and the distribution concentration. Combined with the colloid surface tension parameters, the degree of geometric distortion of the actual glue dots during the coating process can be inferred by modeling, thereby obtaining a predicted value of the irregularity of the glue dot shape. When implemented, some data visualization tools can use three-dimensional simulation to substitute the size, number and surface tension value of the bubbles into the fluid model to evaluate the possible trend of the colloid spreading or curling on the substrate surface. Through this process, it is more intuitive to judge whether the colloid will form edge collapse or internal voids, so as to timely warn of deformation risks during mass production.

[0154] In the fourth step, for data on abnormal colloid aging, the rate of reduction in adhesive strength can be estimated through the chemical property decay rate, temperature drift, and substrate surface characteristics recorded in the report. At this time, if the substrate material and the empirical curve of the previous aging experiment are known, the aging rate can be combined with environmental parameters (such as workshop humidity and gas composition) using relevant algorithms to infer the adhesion decay of the colloid when it continues to be used in this environment. If an infrared thermometer and humidity monitoring device are installed on the production line, the currently measured environmental data can also be used as an input factor, so that the calculation process fits the on-site situation and more effectively reflects the degree of decay of the bonding performance between the colloid and the substrate.

[0155] In the fifth step, after obtaining the above prediction values, a multi-dimensional fusion method can be adopted to weight or layer the predicted values ​​of dispensing volume deviation, dispensing position deviation, irregularity of glue point shape, and reduction rate of glue point adhesion strength, and finally generate the predicted values ​​of quality parameters. In specific implementation, a multivariate comprehensive judgment algorithm can be run in the industrial computer, and a numerical value or segmented grade reflecting the overall quality level can be output by setting thresholds or normalization processing. If this result shows that the value is too high, it means that the deviations in multiple dimensions are more significant, and it may be necessary to adjust the dispensing process or repair equipment in time to eliminate hidden dangers and maintain process consistency before mass production.

[0156] The above describes the SMD device dispensing quality detection method in the embodiment of the present invention. The following describes the SMD device dispensing quality detection device in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an SMD device dispensing quality detection device comprises:

[0157] The acoustic signal acquisition module 101 is used to acquire the acoustic signal generated by the colloid material during the dispensing process;

[0158] A time-frequency analysis module 102 is used to perform time-frequency domain transformation on the acoustic signal to extract acoustic feature vectors representing the physical properties of the colloid;

[0159] The fingerprint matching module 103 is used to match the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result;

[0160] A trend analysis module 104 is used to perform multi-time scale trend analysis on the acoustic characteristic changes of multiple consecutive dispensing operations based on the matching results to obtain trend analysis results;

[0161] The quality prediction module 105 is used to match the trend analysis result with a preset abnormal pattern library to determine the abnormal type and degree of the colloid material and obtain a dispensing quality prediction result.

[0162] above Figure 2The SMD device dispensing quality detection device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The SMD device dispensing quality detection device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0163] Figure 3 2 is a structural schematic diagram of a SMD device dispensing quality detection device provided by an embodiment of the present invention. The SMD device dispensing quality detection device 200 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 210 (for example, one or more processors) and a memory 220, and one or more storage media 230 (for example, one or more mass storage device terminals) storing application programs 233 or data 232. Among them, the memory 220 and the storage medium 230 can be short-term storage or permanent storage. The program stored in the storage medium 230 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the SMD device dispensing quality detection device 200. Furthermore, the processor 210 can be configured to communicate with the storage medium 230, and execute a series of instruction operations in the storage medium 230 on the SMD device dispensing quality detection device 200 to implement the steps of the above-mentioned SMD device dispensing quality detection method.

[0164] The SMD device dispensing quality inspection device 200 may also include one or more power supplies 240, one or more wired or wireless network interfaces 250, one or more input and output interfaces 260, and / or one or more operating systems 231, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The structure of the SMD device dispensing quality inspection device shown does not constitute a limitation on the SMD device dispensing quality inspection device provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0165] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the SMD device dispensing quality detection method.

[0166] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0167] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0168] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A method for detecting the quality of SMD device dispensing, characterized in that: include: Acquire the acoustic signal generated by the colloid material during the dispensing process; Performing time-frequency domain transformation on the acoustic signal to extract acoustic feature vectors representing the physical properties of the colloid; Matching the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result; Based on the matching results, a multi-time scale trend analysis is performed on the acoustic characteristic changes of multiple consecutive dispensing operations to obtain a trend analysis result; According to the trend analysis result, the abnormal type and degree of the colloid material are determined by matching with a preset abnormal pattern library, and the dispensing quality prediction result is obtained.

2. The SMD device dispensing quality detection method according to claim 1, characterized in that: The step of obtaining the acoustic signal generated by the colloid material during the dispensing process includes: Acquire raw acoustic data in the frequency range of 1Hz to 200kHz during the dispensing process; Collect the background noise characteristics and equipment vibration characteristics of the dispensing environment and build filtering parameters; Filtering the original acoustic data according to the filtering parameters to obtain denoised acoustic data; According to the physical characteristics of colloid flow, the noise-reduced acoustic data is divided into dispensing start-up phase data, stable flow phase data and stop phase data; Extracting startup characteristic parameters from the data of the dispensing startup phase, extracting stability characteristic parameters from the data of the stable flow phase, and extracting termination characteristic parameters from the data of the stopping phase; An acoustic feature sequence is constructed according to the startup feature parameter, the stability feature parameter and the termination feature parameter as an acoustic signal.

3. The SMD device dispensing quality detection method according to claim 2, characterized in that: The step of performing a time-frequency domain transformation on the acoustic signal to extract an acoustic feature vector representing the physical properties of the colloid includes: Performing wavelet packet decomposition and Hilbert-Huang transform on the acoustic signal to obtain a multi-scale time-frequency representation and a time-varying characteristic; Calculate the spectrum centroid, frequency band energy ratio and harmonic structure parameters of the startup phase data, the stable flow phase data and the stop phase data respectively to obtain a frequency domain feature set; Extracting energy distribution features, key change points and phase continuity parameters from the multi-scale time-frequency representation and time-varying characteristics to obtain a time domain feature set; Extracting parameters reflecting colloid viscosity change, bubble distribution parameters, colloid uniformity parameters and needle state parameters according to the frequency domain feature set and the time domain feature set; The colloid viscosity variation parameter, the bubble distribution parameter, the colloid uniformity parameter and the needle state parameter are combined to obtain an acoustic characteristic vector.

4. The SMD device dispensing quality detection method according to claim 3 is characterized in that: The step of matching the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result includes: Comparing the colloid viscosity change parameter of the acoustic feature vector with the typical viscosity characteristics stored in the acoustic fingerprint library, calculating the dispensing temperature and the colloid aging state index, and establishing a preliminary matching matrix; Comparing the bubble distribution parameters based on the acoustic feature vector with the bubble feature template in the acoustic fingerprint library, and combining the dispensing pressure value, the colloid microstructure matching degree is obtained; Cross-comparing the colloid uniformity parameter and the needle state parameter of the acoustic feature vector with the corresponding features in the acoustic fingerprint library, and generating a process adaptability score according to the dispensing height and the surface characteristics of the substrate; The preliminary matching matrix, the colloid microstructure matching degree and the process adaptability score are integrated and a matching result is obtained through nonlinear weighted calculation.

5. The SMD device dispensing quality detection method according to claim 4, characterized in that: The colloid uniformity parameter and needle state parameter of the acoustic feature vector are cross-compared with the corresponding features in the acoustic fingerprint library, and a process adaptability score is generated according to the dispensing height and substrate surface characteristics, including: Calculating the Mahalanobis distance between the colloidal uniformity parameter of the acoustic feature vector and the standard uniformity features corresponding to different substrate materials in the acoustic fingerprint library to generate a substrate fitness index; Performing a correlation analysis on the needle state parameters of the acoustic feature vector and the state characteristics of needles of different usage time in the acoustic fingerprint library, and obtaining a needle wear degree score in combination with the accumulated usage time of the needle; According to the substrate adaptability index and the dispensing height parameter, the expected range of the ratio of the dispensing diameter to the height is calculated to obtain the shape matching coefficient; The needle wear degree score, the shape matching coefficient, the substrate surface temperature and the ambient humidity data are combined to generate a process adaptability score reflecting the bonding strength between the colloid and the substrate interface.

6. The SMD device dispensing quality detection method according to claim 4, characterized in that: Based on the matching result, a multi-time scale trend analysis is performed on the acoustic characteristic changes of multiple consecutive dispensings to obtain trend analysis results, including: According to the preliminary matching matrix, colloid microstructure matching degree and process adaptability score in the matching results, combined with the acoustic feature vectors of multiple consecutive dispensing, a microscopic time scale feature sequence, a mesoscopic time scale feature sequence and a macroscopic time scale feature sequence are constructed; According to the colloid viscosity information in the preliminary matching matrix, a morphological feature detection algorithm is applied to the microscopic time scale feature sequence to identify transient abnormal points within a single dispensing, and obtain microscopic abnormal features; Combined with the colloid microstructure matching degree, a sliding window change point detection algorithm is applied to the mesoscopic time scale feature sequence to identify the trend change of colloid performance during 2-200 consecutive dispensing times, and obtain the mesoscopic change pattern; According to the process adaptability score, a periodic decomposition and trend extraction algorithm is applied to the macroscopic time scale feature sequence to separate the periodic fluctuation caused by environmental factors and the long-term change trend of colloid performance from 1-1000 dispensing data to obtain macroscopic trend characteristics; The microscopic abnormal characteristics, mesoscopic change patterns and macroscopic trend characteristics are integrated at multiple scales, and the quality risk advance indicators of abnormal colloid viscosity, increased bubbles, needle blockage and colloid aging are calculated to obtain trend analysis results.

7. The SMD device dispensing quality detection method according to claim 3 is characterized in that: The trend analysis result is matched with a preset abnormal pattern library to determine the abnormal type and degree of the colloid material, and obtain the dispensing quality prediction result, including: The trend analysis result is used to form a time series feature map in the feature space, and a similarity matrix is ​​calculated with the typical abnormal feature map stored in the preset abnormal pattern library to obtain the abnormal pattern matching degree; According to the abnormal pattern matching degree and the rate of change in the trend analysis result, the abnormal type and degree of abnormality of colloid viscosity abnormality, increased bubbles, needle blockage and colloid aging are determined, and an abnormality diagnosis report is generated; Based on the abnormal diagnosis report and historical quality data, the dispensing volume deviation, dispensing position deviation, irregularity of the glue point shape and the reduction rate of the glue point adhesion strength are calculated to obtain the predicted value of the quality parameter; The quality parameter prediction value is combined with the product quality standard to determine the dispensing quality grade and quality risk level, and obtain the dispensing quality prediction result.

8. The SMD device dispensing quality detection method according to claim 7, characterized in that: The method of calculating the dispensing volume deviation, dispensing position deviation, glue point shape irregularity and glue point adhesion strength reduction rate based on the abnormal diagnosis report and historical quality data to obtain the quality parameter prediction value includes: Extracting the type and degree data of colloid viscosity abnormality from the abnormal diagnosis report, and combining it with the historical viscosity-volume relationship data to calculate the predicted value of dispensing volume deviation; Extract the needle blockage abnormality type and degree data from the abnormal diagnosis report, construct a flow deviation index based on the needle status assessment result, and calculate the dispensing position deviation prediction value; Extract the data of the type and degree of abnormal bubble distribution from the abnormal diagnosis report, and calculate the predicted value of irregularity of the shape of the glue point in combination with the surface tension characteristics of the colloid; Extract the type and degree of colloid aging abnormality data from the abnormality diagnosis report, and calculate the predicted value of the glue point adhesion strength reduction rate in combination with the substrate surface characteristics and environmental condition parameters; The quality parameter prediction value is obtained by combining the predicted value of glue dispensing volume deviation, the predicted value of glue dispensing position deviation, the predicted value of glue dot shape irregularity and the predicted value of glue dot adhesion strength reduction rate.

9. A SMD device dispensing quality detection device, characterized in that: The SMD device dispensing quality detection device adopts the SMD device dispensing quality detection method according to any one of claims 1 to 8, and the SMD device dispensing quality detection device comprises: An acoustic signal acquisition module is used to obtain the acoustic signal generated by the colloid material during the dispensing process; A time-frequency analysis module, used for performing time-frequency domain transformation on the acoustic signal and extracting acoustic feature vectors representing the physical properties of the colloid; A fingerprint matching module, used to match the acoustic feature vector with a preset acoustic fingerprint library to obtain a matching result; A trend analysis module, used for performing multi-time scale trend analysis on the acoustic characteristic changes of multiple consecutive dispensing operations based on the matching results to obtain trend analysis results; The quality prediction module is used to match the trend analysis result with a preset abnormal pattern library to determine the abnormal type and degree of the colloid material and obtain the dispensing quality prediction result.

10. An SMD device dispensing quality inspection device, characterized in that: The SMD device dispensing quality detection device comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory to enable the SMD device dispensing quality detection device to perform the steps of the SMD device dispensing quality detection method as described in any one of claims 1-8.

Images