Water-based ink grinding end point judgment method and system based on voiceprint analysis

Abstract

The invention relates to the technical field of intelligent sensor monitoring and intelligent grinding control, and discloses a water-based ink grinding end point judgment method and system based on voiceprint analysis. The method comprises the following steps: arranging n acoustic sensors in different directions of water-based ink grinding equipment; extracting frequency domain envelope features and time domain disturbance features; performing state sequence reasoning on the composite feature sequence; generating a stage stability label vector by adopting a dynamic monitoring mechanism driven by a state stability criterion; and judging whether the grinding process reaches an end-point state or not. The technical problems that in the prior art, the grinding end point is judged mostly depending on artificial experience, and especially in the scene with multi-stage transition fuzziness and remarkable voiceprint disturbance in the water-based ink grinding and dispersing process, low-misjudgment end point recognition cannot be achieved are solved. Due to the fact that dominant acoustic channel construction and multi-state hidden Markov sequence modeling are introduced, the intelligent level of water-based ink preparation process control is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensor monitoring and intelligent grinding control technology, and in particular to a method and system for determining the grinding endpoint of water-based inks based on acoustic signature analysis. Background Technology

[0002] Currently, the grinding stage in the preparation of water-based inks has a crucial impact on the color stability, dispersion uniformity, and adhesion performance of the final product. In actual production, the grinding process usually involves high-shear mixing of pigments, resins, and additives, particle size refinement, and viscosity adjustment. The determination of the endpoint directly affects the consistency of the product and production efficiency. However, due to factors such as uneven particle breakage rate, nonlinear evolution of rheological properties, and uncertainty of grinding media wear during the grinding process, traditional endpoint determination methods that rely on manual experience or fixed time settings have significant shortcomings. For example, in the existing technology, some solutions use offline testing of ink samples, such as optical particle size analysis, viscosity testing, or color difference evaluation, to determine whether the grinding meets the standards. However, such offline analysis methods have the following problems: (1) poor real-time performance, unable to respond online to dynamic changes during the grinding process; (2) long detection cycle, easily leading to over-grinding or under-grinding; (3) high dependence on equipment and personnel, unsuitable for intelligent or unattended production scenarios.

[0003] In addition, some studies have attempted to model the process and predict the endpoint based on operating condition signals such as equipment current, voltage, and temperature. However, these methods are easily affected by physical interference and operational errors. The models have weak generalization ability and it is difficult to distinguish the acoustic behavior differences of different grinding stages. Especially under conditions of large equipment load fluctuations or significant batch differences of raw materials, traditional methods often cannot accurately reflect the true state of ink grinding.

[0004] Therefore, there is an urgent need for a new method for determining the grinding endpoint of water-based inks based on acoustic analysis. This method should be able to accurately determine the grinding endpoint of inks in dynamic environments without interfering with the grinding process, combined with highly sensitive and high-resolution online sensing methods. This would improve grinding consistency, reduce energy consumption, and enhance production flexibility. Summary of the Invention

[0005] To address the aforementioned technical shortcomings, the purpose of this invention is to propose a method for determining the grinding endpoint of water-based inks based on acoustic analysis. This method aims to solve the technical problem that existing technologies often rely on human experience to determine the grinding endpoint, especially in scenarios where the grinding and dispersion process of water-based inks involves multiple stages of ambiguity and significant acoustic disturbances, making it impossible to achieve low-false-judgment endpoint identification.

[0006] To solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a method for determining the grinding endpoint of water-based inks based on acoustic text analysis.

[0007] The method for determining the grinding endpoint of water-based inks based on acoustic analysis includes:

[0008] Step S10: Deploy n acoustic sensors at different locations on the water-based ink grinding equipment. Based on the n acoustic sensors, employ a feature cross-compensation mechanism based on the response differences during the grinding stages to construct the dominant acoustic channel and output a weighted composite acoustic signal. ;

[0009] Step S20: For the weighted composite acoustic signal A dual-domain modulation analysis mechanism is used to perform frequency domain envelope enhancement and time domain perturbation suppression processing to obtain a set of composite feature sequences. ;

[0010] Step S30: Based on the composite feature sequence set A multi-state Hidden Markov Model based on the grinding stage is used to perform the grinding state sequence reasoning task, and the grinding state sequence is output. ;

[0011] Step S40: Based on the grinding state sequence A dynamic monitoring mechanism driven by a state stability criterion is used to perform the dynamic termination determination task, and the stage stability label vector L is output. ;

[0012] Step S50: Based on the stage stability label vector, a multi-condition triggered endpoint determination mechanism is used to determine whether the grinding process has reached the endpoint state.

[0013] Preferably, in step S10, n acoustic sensors are deployed at different locations in the water-based ink grinding equipment. Based on the n acoustic sensors, a feature cross-compensation mechanism based on the response differences of the grinding stages is used to perform the task of constructing the dominant acoustic channel, and a weighted composite acoustic signal is output. The steps specifically include:

[0014] Step S101: Obtain the current grinding stage, which includes the initial crushing stage, the intermediate dispersion stage, and the final stabilization stage; for the i-th acoustic sensor among n acoustic sensors, extract the first acoustic response vector in the initial crushing stage, the second acoustic response vector in the intermediate dispersion stage, and the third acoustic response vector in the final stabilization stage; the first acoustic response vector is used to reflect the high-frequency structural impact signal characteristics caused by the early large particle fracture; the second acoustic response vector is used to reflect the periodic turbulence and mid-frequency resonance characteristics during particle agglomeration; the third acoustic response vector is used to reflect the low-frequency viscous flow acoustic stability under the homogeneous state of the slurry; and based on the first, second, and third acoustic response vectors, perform stage decoupling analysis using the generalized distance mapping method to output the stage response difference vector;

[0015] Step S102: Based on the stage response difference vectors of n acoustic sensors, construct an acoustic channel sorting list using the maximum difference vector sorting method, and select the top channels in the acoustic channel sorting list. Each acoustic channel constructs a dominant channel candidate set, and the remaining acoustic channels in the acoustic channel sorting list are selected to construct a redundant channel candidate set.

[0016] Step S103: Based on the dominant channel candidate set, calculate the feature similarity of adjacent acoustic channels in the dominant channel candidate set using the adjacent channel symmetric cosine similarity fusion method. Furthermore, based on the dominant channel candidate set, a dynamic sliding window entropy shift criterion is used to calculate the characteristic cross-stability score of adjacent acoustic channels in the dominant channel candidate set. Based on feature similarity and feature cross stability score Perform linear weighted fusion to generate dynamic weighted coefficients. Based on dynamic weighting coefficients The first, second, and third acoustic response vectors are subjected to cross-stage weighted calibration processing, and the final output is a weighted composite acoustic signal. .

[0017] Preferably, in step S20, the weighted composite acoustic signal is... A dual-domain modulation analysis mechanism is used to perform frequency domain envelope enhancement and time domain perturbation suppression processing to obtain a set of composite feature sequences. The steps specifically include:

[0018] Step S201: First, target the weighted composite acoustic signal Short-time Fourier transform processing is performed to obtain the spectrum of the composite acoustic signal. Based on the spectrum of the composite acoustic signal, the spectral envelope energy E is calculated using the sliding band window analysis method. ;

[0019] Step S202: Then, for the weighted composite acoustic signal The wavelet domain singular value compression algorithm is used to remove disturbances and suppress low-frequency vibrations, and the reconstructed smooth signal smooth(t) is obtained.

[0020] Step S203: Based on the spectral envelope energy E The instantaneous gain features of the spectral envelope, the slope features of the spectral shape fluctuation, and the covariance features of the spectral domain energy distribution are extracted. Mel-frequency cepstral features are then extracted based on the reconstructed smooth signal smooth(t). Finally, a set of composite feature sequences is constructed using a normalized co-coding method based on the instantaneous gain features of the spectral envelope, the slope features of the spectral shape fluctuation, the covariance features of the spectral domain energy distribution, and the Mel-frequency cepstral features. .

[0021] Preferably, in step S30, based on the composite feature sequence set A multi-state Hidden Markov Model based on the grinding stage is used to perform the grinding state sequence reasoning task, and the grinding state sequence is output. The steps specifically include:

[0022] Step S301: Obtain the current grinding stage, which includes the initial crushing stage, the intermediate dispersion stage, and the final stabilization stage; construct a multi-state set S based on the three different stage categories of the initial crushing stage, the intermediate dispersion stage, and the final stabilization stage; construct a state transition matrix A based on the multi-state set S and combined with the weighted maximum likelihood estimation statistical analysis method with time constraints; the state transition matrix A is used to measure the transition probability between different grinding states in the time series.

[0023] Step S302: Pre-set a multi-state hidden Markov model, obtain the historical state transition matrix, the historical composite feature sequence set, and the historical grinding state sequence; use the historical state transition matrix and the historical composite feature sequence set as the input of the multi-state hidden Markov model, use the historical grinding state sequence as the output of the multi-state hidden Markov model, and pre-train the multi-state hidden Markov model by combining the weighted cross-entropy loss function.

[0024] Step S303: Combine the state transition matrix A and the set of composite feature sequences. Inputting the pre-trained multi-state hidden Markov model, the multi-state hidden Markov model outputs a grinding state sequence. .

[0025] Preferably, in step S40, based on the grinding state sequence The steps for executing the dynamic termination determination task using a dynamic monitoring mechanism driven by the state stability criterion and outputting the stage stability label vector specifically include:

[0026] Step S401: Obtain the grinding state sequence , ,in, This represents the first grinding state index value in the grinding state sequence. This represents the second grinding state index value in the grinding state sequence. This represents the T-th grinding state index value corresponding to time t in the grinding state sequence;

[0027] Step S402: For the grinding state sequence The ADF unit root test method was used to calculate the grinding condition test statistic. And for the grinding state sequence The first-order autocorrelation residuals are calculated using a partial autocorrelation function corrected by the Bartlett window. Based on grinding condition test statistics and first-order autocorrelation residuals Constructing a joint stationarity flag function ;

[0028] Step S403: Grinding state sequence Set a fixed-size time window w, and define a stationarity flag function for the fixed-size time window w. A trend aggregation operation is performed using the trend cumulative integral moving average method, outputting the stable state trigger score corresponding to the time window w. When the stable state trigger score is greater than the preset stable state trigger score threshold, the time window w is determined to be a stationary window. The grinding state subsequences corresponding to the stationary windows are obtained. For all grinding state subsequences within the stationary windows, the Dynamic Time Warping (DTW) clustering algorithm is used to construct and output the stage stability label vector L. .

[0029] Preferably, in step S40, the grinding state test statistic... The formula is expressed as:

[0030] ;

[0031] in, For grinding state sequence The estimated value of the first-order lag coefficient, The basic regression model was calculated using the ADF test; The standard error of the first-order lag coefficient estimate;

[0032] First-order autocorrelation residuals The formula is expressed as:

[0033] ;

[0034] in, Grinding state sequence The ADF residual sequence value corresponding to the i-th item in the sequence; Grinding state sequence The ADF residual sequence value corresponding to the (i+1)th term in the sequence; The first-order weighting coefficients are corrected using the Bartlett window function; T is the grinding state sequence. The corresponding sequence length.

[0035] Preferably, step S50, which uses a multi-condition triggered endpoint determination mechanism based on the stage stability label vector to determine whether the grinding process has reached the endpoint state, specifically includes:

[0036] Step S501: Preset trigger step size Based on trigger step size Extract sub-stage stability label vectors from the stage stability label vectors, and perform stage label similarity calculation using a sequence matching similarity method based on Hamming distance based on the sub-stage stability label vectors and the preset stability template vector set, and output the stability index flag;

[0037] Step S502: Construct a stationarity index sequence based on all stationarity index flags within the stage stability label vector;

[0038] Step S503: Based on the stationarity index sequence, a trend inversion detection mechanism based on a moving average window is used to perform trend analysis, outputting a trend score factor and a trend volatility factor; when the trend score factor is greater than a preset trend score threshold... Furthermore, the trend volatility factor is greater than the preset trend volatility threshold. If the grinding process reaches its endpoint, it is determined that the grinding process has reached its endpoint; otherwise, it is determined that the grinding process has not reached its endpoint.

[0039] This invention also provides a water-based ink grinding endpoint determination system based on acoustic signature analysis, comprising:

[0040] The dominant acoustic channel construction module is used to deploy n acoustic sensors at different locations in the water-based ink grinding equipment. Based on the n acoustic sensors, a feature cross-compensation mechanism based on the response differences of the grinding stages is used to perform the dominant acoustic channel construction task, and output a weighted composite acoustic signal. ;

[0041] The dual-domain acoustic feature analysis module is used for weighted composite acoustic signals. A dual-domain modulation analysis mechanism is used to perform frequency domain envelope enhancement and time domain perturbation suppression processing to obtain a set of composite feature sequences. ;

[0042] The grinding stage sequence reasoning module is used for processing composite feature sequence sets. A multi-state Hidden Markov Model based on the grinding stage is used to perform the grinding state sequence reasoning task, and the grinding state sequence is output. ;

[0043] Dynamic stability monitoring module, used for grinding state sequence A dynamic monitoring mechanism driven by a state stability criterion is used to perform the dynamic termination determination task, and the stage stability label vector L is output. ;

[0044] The multi-condition endpoint determination module is used to determine whether the grinding process has reached the endpoint state based on the stage stability label vector and a multi-condition triggered endpoint determination mechanism.

[0045] The present invention also provides a device for determining the grinding endpoint of water-based ink based on acoustic fingerprint analysis, comprising: a memory, a processor, and a program for determining the grinding endpoint of water-based ink based on acoustic fingerprint analysis stored in the memory and executable on the processor. When the program for determining the grinding endpoint of water-based ink based on acoustic fingerprint analysis is executed by the processor, a method for determining the grinding endpoint of water-based ink based on acoustic fingerprint analysis is implemented.

[0046] The present invention also provides a computer program product, including a water-based ink grinding endpoint determination program based on acoustic analysis, wherein the water-based ink grinding endpoint determination program based on acoustic analysis implements the water-based ink grinding endpoint determination method based on acoustic analysis when executed by a processor.

[0047] The beneficial effects of this invention are as follows: By constructing a dominant acoustic channel and introducing a dual-domain modulation analysis mechanism, this invention significantly enhances the frequency domain envelope features and suppresses temporal random disturbances. Combined with the grinding state sequence reasoning and dynamic stationarity label generation mechanism based on a multi-state hidden Markov model, it effectively avoids misjudgment problems caused by environmental noise, sensor anomalies, or blurred transitions in the grinding stage, thereby significantly improving the accuracy of grinding endpoint determination.

[0048] This invention employs a distributed acoustic sensor deployment and a status tag-driven endpoint determination strategy, applicable to various models of water-based ink grinding equipment. It requires no modification to the equipment itself, and all judgment logic is based on a lightweight statistical model and threshold rules, offering advantages such as low computational load and high real-time performance. It can be directly integrated into industrial edge gateways or PLC control units for deployment and operation, improving the level of intelligence and application efficiency. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the first embodiment of a method for determining the grinding endpoint of water-based inks based on acoustic analysis according to the present invention.

[0051] Figure 2 This is a schematic diagram of the spectral response of each sensor during the fine grinding stage of the first embodiment of the water-based ink grinding endpoint determination method based on acoustic analysis of the present invention.

[0052] Figure 3 This is a schematic diagram of the dominant channel composite acoustic signal of a first embodiment of a water-based ink grinding endpoint determination method based on acoustic text analysis according to the present invention.

[0053] Figure 4 This is a schematic diagram of the device for determining the grinding endpoint of water-based ink based on acoustic text analysis, according to the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Example 1: As Figure 1 The diagram shown is a flowchart of the first embodiment of the water-based ink grinding endpoint determination method based on acoustic text analysis of the present invention. The first embodiment of the water-based ink grinding endpoint determination method based on acoustic text analysis of the present invention is presented.

[0056] In the first embodiment, the method for determining the grinding endpoint of water-based inks based on acoustic signature analysis includes:

[0057] Step S10: Deploy n acoustic sensors at different locations on the water-based ink grinding equipment. Based on the n acoustic sensors, employ a feature cross-compensation mechanism based on the response differences during the grinding stages to construct the dominant acoustic channel and output a weighted composite acoustic signal. ;

[0058] It should be noted that the "feature cross-compensation mechanism" refers to analyzing the differences in amplitude-frequency characteristics of the acoustic response of each channel in different grinding stages (such as pre-dispersion stage, medium-speed grinding stage, and high-efficiency fine grinding stage) from the raw signals collected by n deployed acoustic sensors (such as capacitive microphones, piezoelectric acoustic sensors, etc.) to construct a weighted model that includes the channel-specific weights of each sensor. This model not only considers the physical deployment location of the sensors but also integrates dynamic features such as multi-time period acoustic energy density, spectral envelope change rate, and short-time energy variance, thereby performing adaptive compensation for low signal-to-noise ratio channels and improving the consistency and sensitivity of the overall acoustic signature representation.

[0059] Understandably, constructing a weighted composite acoustic signal can significantly improve the discriminative power and stability of acoustic features in subsequent grinding state identification. Compared to single-channel or average-channel processing methods, this approach can dynamically adapt to the dominant acoustic feature channels at different stages, enhancing the feature's responsiveness to key process nodes (such as particle size breakage critical points and agglomerate dissociation critical points), thereby providing a more reliable feature basis for subsequent state modeling and endpoint determination.

[0060] It should be understood that traditional methods typically use fixed sensors or fixed weighting to extract acoustic signals, which cannot take into account spatial distribution differences and dynamic changes in stage response, easily leading to the loss of some key acoustic features or signal weakening. The "feature cross-compensation mechanism" adopted in this invention realizes real-time evaluation and weighting adjustment of the feature contribution between channels, effectively avoiding feature offset problems caused by channel signal-to-noise differences or mechanical interference.

[0061] For example, such as Figure 2 As shown, the raw acoustic spectra collected by six sensors during the "high-efficiency fine grinding stage" are displayed. It is evident that differences in response amplitude and localized abrupt changes exist among the different sensors in certain frequency bands; for example... Figure 3 As shown, the weighted composite acoustic spectrum of each stage after fusion based on the "feature cross-compensation mechanism" can be seen. It can be seen that the signal after weighted fusion exhibits a clearer spectral envelope structure in the target frequency band (about 2~6 kHz), highlighting the energy concentration area in the grinding process, which helps the subsequent model to accurately identify the critical change point and endpoint trend of particle size.

[0062] Step S20: For the weighted composite acoustic signal A dual-domain modulation analysis mechanism is used to perform frequency domain envelope enhancement and time domain perturbation suppression processing to obtain a set of composite feature sequences. ;

[0063] It should be noted that the "dual-domain modulation analysis mechanism" simultaneously performs differentiated feature processing tasks in two domains and fuses the results to improve overall signal quality. In the frequency domain, an improved Hilbert envelope extraction algorithm combined with a wavelet subband decomposition strategy is used to model the energy variation trend of the acoustic signature signal in different frequency bands with fine granularity, and the ability to distinguish subtle envelope fluctuations is improved by normalizing the dynamic envelope ratio. In the time domain, a disturbance detection threshold model is constructed based on intra-frame disturbance interval statistics (such as short-time energy range, second derivative of acceleration, etc.) to perform edge suppression and dynamic smoothing of abnormal vibrations and environmental reflection noise. Finally, the processing results from the two domains are fused in a weighted manner to obtain a set of composite feature sequences with a clearer feature structure.

[0064] Understandably, the dual-domain modulation analysis mechanism can simultaneously consider the spectral structure characteristics and temporal evolution of the signal, making it particularly suitable for complex signal scenarios involving overlapping frequency bands, frequency drift, and short-period disturbances during water-based ink grinding. Frequency domain enhancement focuses on the mid-to-high frequency feature envelope generated by particle breakage, while temporal disturbance suppression effectively mitigates transient anomalies introduced by mechanical instability or fluid disturbances. Compared to simple filtering or noise reduction, this mechanism is more sensitive to physical processes and more specific to stage responses, accurately characterizing the key acoustic signature changes at each grinding stage.

[0065] For example, in an actual grinding experiment, a certain type of water-based blue ink enters a stage of rapid particle size reduction around the 9th minute of high-speed grinding. At this time, the original acoustic signal exhibits severe energy disturbance and densely interspersed frequency spectrum lines. If traditional Hilbert transform is used for frequency domain analysis, the envelope curve is prone to breakage and discontinuity due to short-term fluctuations. However, the dual-domain modulation mechanism of this invention, through multi-scale wavelet enhancement in the frequency domain and local perturbation suppression in the time domain, makes the envelope curve corresponding to this stage more stable at the peak position and smoother in the transition segment. Its slope change shows a high linear correlation with the particle size detection result (Pearson correlation coefficient increased to 0.86), significantly improving the discriminability of the endpoint state.

[0066] Step S30: Based on the composite feature sequence set A multi-state Hidden Markov Model based on the grinding stage is used to perform the grinding state sequence reasoning task, and the grinding state sequence is output. ;

[0067] It should be noted that the "multi-state hidden Markov model based on grinding stages" refers to introducing the state transition probability matrix and observation probability density function corresponding to different stages in the grinding process (such as premixing and dispersion, coarse grinding, fine grinding, and near the endpoint) to achieve stage-wise modeling of acoustic feature sequences in the time dimension. This model obtains a set of parameters through offline training, including the state transition matrix, observation probability matrix, and initial state distribution. The observed sequence is a composite feature sequence (such as frequency domain envelope peak distribution, temporal abrupt change rate, etc.), and a multi-state configuration (≥4 hidden states) is specifically designed to ensure accurate characterization of the sound pattern evolution at different stages of the grinding process.

[0068] It should be understood that traditional grinding process identification often relies on setting a single feature threshold or empirical rules, which is difficult to cope with nonlinear feature perturbations and the ambiguity of transitions between stages. The "multi-state hidden Markov model based on grinding stages" adopted in this invention introduces explicit constraints on temporal continuity and stage dependence through state transition probability modeling, effectively mitigating the misjudgment phenomenon caused by feature mutations.

[0069] For example, in a batch of water-based ink grinding tests, the collected composite acoustic feature sequence exhibited a typical four-segment fluctuation trend within the 0-600 second range, corresponding to the premixing start-up stage (high-frequency perturbation and low envelope), the mid-stage stabilization stage (mid-frequency dominance and stable envelope), the final fine grinding stage (high-frequency energy decline and perturbation enhancement), and the endpoint stage (weakening feature changes and spectral stabilization). After inputting this feature sequence into a trained four-state HMM model, the output state sequence successfully identified the above four-stage structure, with an average error of less than 8 seconds compared to the manual reference markers for the stage switching time points.

[0070] Step S40: Based on the grinding state sequence A dynamic monitoring mechanism driven by a state stability criterion is used to perform the dynamic termination determination task, and the stage stability label vector L is output. ;

[0071] It should be noted that the "state stationarity criterion-driven dynamic monitoring mechanism" refers to defining a set of statistical criteria for state switching frequency, state dwell time, and inter-state transition probability in the grinding state sequence output by the Hidden Markov Model (HMM) to determine whether the current state sequence has entered a stable phase. This mechanism extracts nearest-neighbor state sequence segments through a sliding time window, combines the stability analysis of the state transition matrix with the autocorrelation structure of the time series, and calculates a set of state stationarity indices (such as state variance, state transition probability entropy, and dwell time offset coefficient). Based on these indices, stability labels are assigned to each time segment.

[0072] Understandably, by introducing a dynamic monitoring mechanism, compared to traditional methods that rely on fixed time thresholds or empirical particle size determination, this approach can identify earlier and more accurately when the grinding process has entered the convergence range of state fluctuations, effectively capturing the characteristics of the state stability stage and providing a solid basis for subsequent endpoint determination. This method significantly enhances the ability to perceive the stable recovery trend under abnormal oscillation segments or operating condition disturbances, reducing the risk of misjudgment and missed judgment.

[0073] It should be understood that traditional endpoint determination methods mostly rely on sound intensity change thresholds, particle size testing intervals, or human experience, making it difficult to address state hysteresis and false triggering caused by factors such as equipment aging, load fluctuations, and solvent viscosity changes, resulting in significant uncertainty. In contrast, the state stability criterion-driven mechanism proposed in this invention possesses the advantages of adaptability, statistical robustness, and dynamic response, enabling intelligent identification of process stability intervals solely based on acoustic state sequences without relying on particle size detection equipment.

[0074] Step S50: Based on the stage stability label vector, a multi-condition triggered endpoint determination mechanism is used to determine whether the grinding process has reached the endpoint state.

[0075] It should be noted that the "multi-condition triggered endpoint determination mechanism" in this step refers to, based on the obtained stage stability label vector, further introducing three types of criteria: time persistence constraint, fluctuation boundary constraint, and acoustic energy stabilization constraint, to construct a joint triggering condition set for endpoint determination. The grinding process will be considered to have reached the endpoint state when the following conditions are simultaneously met.

[0076] It is understandable that factors such as batch differences in materials and background noise may affect the output of a single acoustic feature or model state. Therefore, adopting a multi-condition joint judgment mechanism can effectively resist the interference of individual anomalies and improve the accuracy of endpoint determination.

[0077] It should be understood that traditional methods for determining the grinding endpoint rely on manual experience or single-variable monitoring (such as fixed grinding time, particle size detection, etc.), which cannot dynamically adapt to differences in different batches or equipment operating conditions, leading to problems such as misjudgment of the endpoint or over-processing. In contrast, this invention, through the construction of an endpoint determination strategy that combines state label vectors with multi-dimensional feature criteria, possesses strong versatility and generalization ability, does not depend on specific sensors or model structures, and is applicable to various types of water-based ink grinding processes.

[0078] Example 2: Furthermore, the present invention provides a water-based ink grinding endpoint determination system based on acoustic fingerprint analysis, employing a water-based ink grinding endpoint determination method based on acoustic fingerprint analysis as described in the above embodiments, which can solve the technical problem of determining the grinding endpoint of water-based inks based on acoustic fingerprint analysis. Compared with the prior art, the beneficial effects of the water-based ink grinding endpoint determination system based on acoustic fingerprint analysis provided by the present invention are the same as the beneficial effects of the water-based ink grinding endpoint determination method based on acoustic fingerprint analysis provided in the above embodiments, and other technical features of the water-based ink grinding endpoint determination system based on acoustic fingerprint analysis are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0079] Example 3: This invention provides a device for determining the grinding endpoint of water-based inks based on acoustic signature analysis. Please refer to... Figure 4A device for determining the grinding endpoint of water-based ink based on voiceprint analysis includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the method for determining the grinding endpoint of water-based ink based on voiceprint analysis described in Embodiment 1 above. The device for determining the grinding endpoint of water-based ink based on voiceprint analysis in this embodiment of the invention may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. This device for determining the grinding endpoint of water-based ink based on voiceprint analysis is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the invention. A device for determining the grinding endpoint of water-based ink based on voiceprint analysis may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An I / O interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows a water-based ink grinding endpoint determination device based on acoustic signature analysis to communicate wirelessly or wiredly with other devices to exchange data. While the figure shows one water-based ink grinding endpoint determination device based on acoustic signature analysis with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0080] Example 4: This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for determining the grinding endpoint of water-based inks based on voiceprint analysis. The computer program product provided by this invention can solve the technical problem of determining the grinding endpoint of water-based inks based on voiceprint analysis. Compared with the prior art, the beneficial effects of the computer program product provided by this invention are the same as those of the method for determining the grinding endpoint of water-based inks based on voiceprint analysis provided in the above embodiments, and will not be repeated here.

[0081] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this invention.

[0082] It should be understood that the various parts disclosed in this invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0083] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for determining the grinding endpoint of water-based inks based on acoustic signature analysis, characterized in that, The methods include: Step S10: Deploy n acoustic sensors at different locations on the water-based ink grinding equipment. Based on the n acoustic sensors, employ a feature cross-compensation mechanism based on the response differences during the grinding stages to construct the dominant acoustic channel and output a weighted composite acoustic signal. ; Step S20: For the weighted composite acoustic signal A dual-domain modulation analysis mechanism is used to perform frequency domain envelope enhancement and time domain perturbation suppression processing to obtain a set of composite feature sequences. ; Step S30: Based on the composite feature sequence set A multi-state Hidden Markov Model based on the grinding stage is used to perform the grinding state sequence reasoning task, and the grinding state sequence is output. ; Step S40: Based on the grinding state sequence A dynamic monitoring mechanism driven by a state stability criterion is used to perform the dynamic termination determination task, and the stage stability label vector L is output. ; Step S50: Based on the stage stability label vector, a multi-condition triggered endpoint determination mechanism is used to determine whether the grinding process has reached the endpoint state.

2. The method for determining the grinding endpoint of water-based inks based on acoustic signature analysis as described in claim 1, characterized in that, In step S10, n acoustic sensors are deployed at different locations in the water-based ink grinding equipment. Based on the n acoustic sensors, a feature cross-compensation mechanism based on the response differences of the grinding stages is used to perform the task of constructing the dominant acoustic channel, and a weighted composite acoustic signal is output. The steps specifically include: Step S101: Obtain the current grinding stage, which includes the initial crushing stage, the intermediate dispersion stage, and the final stabilization stage; for the i-th acoustic sensor among n acoustic sensors, extract the first acoustic response vector in the initial crushing stage, the second acoustic response vector in the intermediate dispersion stage, and the third acoustic response vector in the final stabilization stage; the first acoustic response vector is used to reflect the high-frequency structural impact signal characteristics caused by the early large particle fracture; the second acoustic response vector is used to reflect the periodic turbulence and mid-frequency resonance characteristics during particle agglomeration; the third acoustic response vector is used to reflect the low-frequency viscous flow acoustic stability under the homogeneous state of the slurry; and based on the first, second, and third acoustic response vectors, perform stage decoupling analysis using the generalized distance mapping method to output the stage response difference vector; Step S102: Based on the stage response difference vectors of n acoustic sensors, construct an acoustic channel sorting list using the maximum difference vector sorting method, and select the top channels in the acoustic channel sorting list. Each acoustic channel constructs a dominant channel candidate set, and the remaining acoustic channels in the acoustic channel sorting list are selected to construct a redundant channel candidate set. Step S103: Based on the dominant channel candidate set, calculate the feature similarity of adjacent acoustic channels in the dominant channel candidate set using the adjacent channel symmetric cosine similarity fusion method. Furthermore, based on the dominant channel candidate set, a dynamic sliding window entropy shift criterion is used to calculate the characteristic cross-stability score of adjacent acoustic channels in the dominant channel candidate set. Based on feature similarity and feature cross stability score Perform linear weighted fusion to generate dynamic weighted coefficients. Based on dynamic weighting coefficients The first, second, and third acoustic response vectors are subjected to cross-stage weighted calibration processing, and the final output is a weighted composite acoustic signal. .

3. The method for determining the grinding endpoint of water-based inks based on acoustic signature analysis as described in claim 1, characterized in that, In step S20, for the weighted composite acoustic signal A dual-domain modulation analysis mechanism is used to perform frequency domain envelope enhancement and time domain perturbation suppression processing to obtain a set of composite feature sequences. The steps specifically include: Step S201: First, target the weighted composite acoustic signal Short-time Fourier transform processing is performed to obtain the spectrum of the composite acoustic signal. Based on the spectrum of the composite acoustic signal, the spectral envelope energy E is calculated using the sliding band window analysis method. ; Step S202: Then, for the weighted composite acoustic signal The wavelet domain singular value compression algorithm is used to remove disturbances and suppress low-frequency vibrations, and the reconstructed smooth signal smooth(t) is obtained. Step S203: Based on the spectral envelope energy E The instantaneous gain features of the spectral envelope, the slope features of the spectral shape fluctuation, and the covariance features of the spectral domain energy distribution are extracted. Mel-frequency cepstral features are then extracted based on the reconstructed smooth signal smooth(t). Finally, a set of composite feature sequences is constructed using a normalized co-coding method based on the instantaneous gain features of the spectral envelope, the slope features of the spectral shape fluctuation, the covariance features of the spectral domain energy distribution, and the Mel-frequency cepstral features. .

4. The method for determining the grinding endpoint of water-based inks based on acoustic signature analysis as described in claim 1, characterized in that, In step S30, based on the composite feature sequence set A multi-state Hidden Markov Model based on the grinding stage is used to perform the grinding state sequence reasoning task, and the grinding state sequence is output. The steps specifically include: Step S301: Obtain the current grinding stage, which includes the initial crushing stage, the intermediate dispersion stage, and the final stabilization stage; construct a multi-state set S based on the three different stage categories of the initial crushing stage, the intermediate dispersion stage, and the final stabilization stage; construct a state transition matrix A based on the multi-state set S and combined with the weighted maximum likelihood estimation statistical analysis method with time constraints; the state transition matrix A is used to measure the transition probability between different grinding states in the time series. Step S302: Pre-set a multi-state hidden Markov model, obtain the historical state transition matrix, the historical composite feature sequence set, and the historical grinding state sequence; use the historical state transition matrix and the historical composite feature sequence set as the input of the multi-state hidden Markov model, use the historical grinding state sequence as the output of the multi-state hidden Markov model, and pre-train the multi-state hidden Markov model by combining the weighted cross-entropy loss function. Step S303: Combine the state transition matrix A and the set of composite feature sequences. Inputting the pre-trained multi-state hidden Markov model, the multi-state hidden Markov model outputs a grinding state sequence. .

5. The method for determining the grinding endpoint of water-based inks based on acoustic signature analysis as described in claim 1, characterized in that, In step S40, based on the grinding state sequence The steps for executing the dynamic termination determination task using a dynamic monitoring mechanism driven by the state stability criterion and outputting the stage stability label vector specifically include: Step S401: Obtain the grinding state sequence , ,in, This represents the first grinding state index value in the grinding state sequence. This represents the second grinding state index value in the grinding state sequence. This represents the T-th grinding state index value corresponding to time t in the grinding state sequence; Step S402: For the grinding state sequence The ADF unit root test method was used to calculate the grinding condition test statistic. And for the grinding state sequence The first-order autocorrelation residuals are calculated using a partial autocorrelation function corrected by the Bartlett window. Based on grinding condition test statistics and first-order autocorrelation residuals Constructing a joint stationarity flag function ; Step S403: Grinding state sequence Set a fixed-size time window w, and define a stationarity flag function for the fixed-size time window w. A trend aggregation operation is performed using the trend cumulative integral moving average method, outputting the stable state trigger score corresponding to the time window w. When the stable state trigger score is greater than the preset stable state trigger score threshold, the time window w is determined to be a stationary window. The grinding state subsequences corresponding to the stationary windows are obtained. For all grinding state subsequences within the stationary windows, the Dynamic Time Warping (DTW) clustering algorithm is used to construct and output the stage stability label vector L. .

6. The method for determining the grinding endpoint of water-based inks based on acoustic signature analysis as described in claim 5, characterized in that, In step S40, the grinding condition test statistic... The formula is expressed as: ； in, For grinding state sequence The estimated value of the first-order lag coefficient, The basic regression model was calculated using the ADF test; The standard error of the first-order lag coefficient estimate; First-order autocorrelation residuals The formula is expressed as: ； in, Grinding state sequence The ADF residual sequence value corresponding to the i-th item in the sequence; Grinding state sequence The ADF residual sequence value corresponding to the (i+1)th term in the sequence; The first-order weighting coefficients are corrected using the Bartlett window function; T is the grinding state sequence. The corresponding sequence length.

7. The method for determining the grinding endpoint of water-based inks based on acoustic signature analysis as described in claim 1, characterized in that, Step S50, which involves determining whether the grinding process has reached its endpoint using a multi-condition triggered endpoint determination mechanism based on the stage stability label vector, specifically includes: Step S501: Preset trigger step size Based on trigger step size Extract sub-stage stability label vectors from the stage stability label vectors, and perform stage label similarity calculation using a sequence matching similarity method based on Hamming distance based on the sub-stage stability label vectors and the preset stability template vector set, and output the stability index flag; Step S502: Construct a stationarity index sequence based on all stationarity index flags within the stage stability label vector; Step S503: Based on the stationarity index sequence, a trend inversion detection mechanism based on a moving average window is used to perform trend analysis, outputting a trend score factor and a trend volatility factor; when the trend score factor is greater than a preset trend score threshold... Furthermore, the trend volatility factor is greater than the preset trend volatility threshold. If the grinding process reaches its endpoint, it is determined that the grinding process has reached its endpoint; otherwise, it is determined that the grinding process has not reached its endpoint.

8. A system for determining the grinding endpoint of water-based inks based on acoustic signature analysis, applied to the method for determining the grinding endpoint of water-based inks based on acoustic signature analysis as described in any one of claims 1 to 7, characterized in that, The water-based ink grinding endpoint determination system based on acoustic analysis includes: The dominant acoustic channel construction module is used to deploy n acoustic sensors at different locations in the water-based ink grinding equipment. Based on the n acoustic sensors, a feature cross-compensation mechanism based on the response differences of the grinding stages is used to perform the dominant acoustic channel construction task, and output a weighted composite acoustic signal. ; The dual-domain acoustic feature analysis module is used for weighted composite acoustic signals. A dual-domain modulation analysis mechanism is used to perform frequency domain envelope enhancement and time domain perturbation suppression processing to obtain a set of composite feature sequences. ; The grinding stage sequence reasoning module is used for processing composite feature sequence sets. A multi-state Hidden Markov Model based on the grinding stage is used to perform the grinding state sequence reasoning task, and the grinding state sequence is output. ; Dynamic stability monitoring module, used for grinding state sequence A dynamic monitoring mechanism driven by a state stability criterion is used to perform the dynamic termination determination task, and the stage stability label vector L is output. ; The multi-condition endpoint determination module is used to determine whether the grinding process has reached the endpoint state based on the stage stability label vector and a multi-condition triggered endpoint determination mechanism.

9. A device for determining the grinding endpoint of water-based inks based on acoustic signature analysis, characterized in that, The water-based ink grinding endpoint determination device based on acoustic analysis includes: a memory, a processor, and a water-based ink grinding endpoint determination program based on acoustic analysis stored in the memory and executable on the processor. When the water-based ink grinding endpoint determination program based on acoustic analysis is executed by the processor, it implements a water-based ink grinding endpoint determination method based on acoustic analysis according to any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a water-based ink grinding endpoint determination program based on acoustic analysis. When the water-based ink grinding endpoint determination program based on acoustic analysis is executed by the processor, it implements a water-based ink grinding endpoint determination method based on acoustic analysis according to any one of claims 1 to 7.

Images