Method for online monitoring of bolt forming quality based on multi-dimensional sensing data

Abstract

The application discloses a bolt forming quality online monitoring method based on multi-dimensional sensing data and relates to the technical field of cold upsetting forming quality monitoring.The method acquires mold vibration signals and forming area acoustic emission signals corresponding to each monitoring moment in the cold upsetting forming process, and collects initial wall thickness distribution data of each bolt blank before forming; the mold vibration signals and the forming area acoustic emission signals at the same monitoring moment are time-domain synchronized and aligned to generate a synchronized sensing frame, and vibration main frequency offset and acoustic emission pulse cumulative energy are extracted; a phase consistency index between vibration main frequency offset fluctuation phase and acoustic emission pulse cumulative energy release phase in a subsequence is calculated, and when the index exceeds a preset morphology threshold, a quality abnormal bolt is marked.The method can identify phase mismatch between material flow and mold vibration in cold upsetting forming, and enhance the online monitoring pertinence to local thinning defects at the thread root.

Description

Technical Field

[0001] This invention relates to the field of cold heading quality monitoring technology, specifically to an online monitoring method for bolt forming quality based on multidimensional sensor data. Background Technology

[0002] Cold heading is the main stage for forming the root wall thickness of bolt threads, and the material flow state and mold dynamic response during the heading process directly affect product quality. Currently, online monitoring of bolt heading quality often uses single physical quantity detection methods. For example, vibration sensors are installed on the mold or equipment to determine the mold state through vibration amplitude, spectral characteristics, or energy changes; or acoustic emission sensors are used to collect elastic waves released by material deformation, identifying crack initiation through acoustic emission energy or count rate. Some methods also rely on post-heading dimensional measurements, weighing, or visual inspection. The monitoring quantities obtained by these methods are mostly global statistical characteristics, lacking a description of the temporal phase relationship between mold vibration and material acoustic emission, making it difficult to reflect the instantaneous mismatch between material flow and mold dynamic response. When the bolt blank has a locally thinner wall, abnormal material flow during cold heading may not cause significant changes in vibration amplitude or total acoustic emission energy, but rather manifest as a difference in the rhythm of two types of signal fluctuations at a specific heading stage. Single signal thresholds or global energy characteristics are prone to missed detection. The initial wall thickness of bolt blanks often exhibits circumferential variations before forming. These variations, after cold heading deformation, are reflected in the wall thickness at the thread root, creating localized weak areas. Existing monitoring methods do not establish a correlation between the spatial distribution of the initial wall thickness of the blank and the temporal characteristics of the cold heading process. They typically extract uniform features over time, failing to focus on analyzing the forming stages corresponding to the weak areas. Monitoring results are affected by the averaging effect of signals from non-weak areas, resulting in insufficient ability to identify localized thinning defects at the thread root. Online monitoring of cold heading forming quality requires enhanced characterization of the phase relationship between material flow and die dynamics, and incorporating the spatial distribution of the initial wall thickness of the blank into process monitoring to improve the ability to identify localized thinning defects. Summary of the Invention

[0003] To address the shortcomings of existing online monitoring methods for cold heading forming quality, which lack characterization of the time-phase relationship between mold vibration and acoustic emission from a single sensor signal and fail to utilize the spatial distribution of the initial wall thickness of the billet for process feature extraction, this invention provides an online monitoring method for bolt forming quality based on multi-dimensional sensor data. This method extracts the vibration dominance frequency offset and the accumulated energy of the acoustic emission pulse from synchronous sensor frames, extracts process feature subsequences based on the weak wall thickness azimuth angle, calculates the phase consistency index, and marks bolts with abnormal quality online.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an online monitoring method for bolt forming quality based on multi-dimensional sensor data. This method integrates mold vibration and acoustic emission signals from the forming zone in a synchronous sensing frame, and extracts key subsequences corresponding to the weak wall thickness azimuth angle from the time series of the cold heading process. It utilizes a phase consistency index to identify the phase mismatch between material flow and mold vibration, thereby improving the accuracy and timeliness of online monitoring of bolt forming quality. The method includes the following steps: Acquire the mold vibration signal and the acoustic emission signal of the forming area at each monitoring moment during the cold heading process, and collect the initial wall thickness distribution data of each bolt blank before forming; The mold vibration signal and the acoustic emission signal of the forming area at the same monitoring time are synchronized in the time domain to generate a synchronous sensing frame at that time, and the vibration main frequency offset and the cumulative energy of the acoustic emission pulse are extracted in the synchronous sensing frame. Based on the initial wall thickness distribution data of the same bolt blank, the wall thickness at the root of the thread of the bolt after forming is spatially mapped to determine the weak azimuth angle of the wall thickness. The vibration dominant frequency offset and acoustic emission pulse cumulative energy at each monitoring moment are spliced ​​together in chronological order to form a process characteristic sequence, and the corresponding time subsequence is extracted from the process characteristic sequence based on the weak wall thickness azimuth angle. Calculate the phase consistency index between the fluctuation phase of the vibration dominant frequency offset and the energy release phase of the accumulated energy of the acoustic emission pulse in the subsequence; If the phase consistency index exceeds the preset morphology threshold, the bolt is marked as a bolt with abnormal quality.

[0005] As a technical solution of the present invention, when acquiring the mold vibration signal and the acoustic emission signal of the forming area corresponding to each monitoring moment during the cold heading process, a piezoelectric accelerometer is installed on the mold base to continuously collect the vibration acceleration of the mold base at a sampling frequency to obtain the original vibration waveform; an acoustic emission sensor is installed on the side of the forming mold to continuously collect the acoustic emission signal of the forming area at a synchronous sampling frequency to obtain the original acoustic emission waveform; the original vibration waveform and the original acoustic emission waveform are respectively bandpass filtered to filter out power frequency interference and mechanical transmission noise, and the filtered waveforms are aligned according to the monitoring moment to output the mold vibration signal and the acoustic emission signal of the forming area corresponding to each monitoring moment.

[0006] As a technical solution of the present invention, when synchronizing the mold vibration signal and the acoustic emission signal of the forming area at the same monitoring time in the time domain, the sampling time stamp of the mold vibration signal is used as a reference to find the sampling point in the acoustic emission signal of the forming area with the smallest difference from the time stamp, and the sampling point is used as the synchronization matching point of the acoustic emission signal; a vibration segment centered on the time stamp and with a duration of a preset window length is extracted from the mold vibration signal, and an acoustic emission segment with the same duration centered on the synchronization matching point is extracted from the acoustic emission signal of the forming area; the vibration segment and the acoustic emission segment are encapsulated into a synchronization sensing frame, and the corresponding monitoring time label is marked on the synchronization sensing frame.

[0007] As a technical solution of the present invention, when extracting the vibration dominant frequency offset and the cumulative energy of acoustic emission pulses within a synchronous sensing frame, a fast Fourier transform is performed on the vibration segment within the synchronous sensing frame to obtain the vibration spectrum, and the frequency point with the maximum amplitude in the vibration spectrum is located as the vibration dominant frequency at that monitoring moment; the vibration dominant frequency is subtracted from the no-load operating fundamental frequency of the cold heading forming equipment to obtain the vibration dominant frequency offset; envelope detection is performed on the acoustic emission segment within the synchronous sensing frame to obtain the acoustic emission envelope curve, and the pulse segment exceeding the trigger threshold in the acoustic emission envelope curve is integrated to obtain the cumulative energy of the acoustic emission pulses. Preferably, the envelope detection uses the Hilbert transform demodulation method to extract the envelope curve of the acoustic emission signal.

[0008] As a technical solution of the present invention, when determining the weak azimuth angle of the wall thickness based on the initial wall thickness distribution data of the same bolt blank, the initial wall thickness values ​​at multiple radial positions are collected at equal angles along the circumferential direction of the bolt blank end face to construct the initial circumferential wall thickness distribution vector; the initial circumferential wall thickness distribution vector is mapped to the wall thickness distribution at the corresponding circumferential position of the thread root after forming, and the wall thickness reduction rate at each circumferential position is calculated; the central angle corresponding to the circumferential position with the largest wall thickness reduction rate is selected as the weak azimuth angle of the wall thickness.

[0009] As a technical solution of the present invention, when the vibration dominant frequency offset and acoustic emission pulse cumulative energy at each monitoring moment are spliced ​​into a process feature sequence in chronological order, the total duration of the cold heading process is divided into multiple monitoring moments at equal time intervals, and the vibration dominant frequency offset and acoustic emission pulse cumulative energy at each moment are arranged into two time-series vectors of equal length according to the order of the monitoring moments to form the process feature sequence. Based on the forming station angle corresponding to the weak wall thickness azimuth angle, the forming stage time when the bolt blank rotates to that angle is calculated, and the start and end times of the interception are determined. Vector segments located between the start and end times of the interception are extracted from the process feature sequence as subsequences. Preferably, the start time of the interception is the monitoring moment before the forming station angle corresponding to the weak wall thickness azimuth angle enters the forming mold engagement zone, and the end time of the interception is the monitoring moment after the forming station angle leaves the forming mold engagement zone.

[0010] As a technical solution of the present invention, when calculating the phase consistency index, the vibration dominant frequency offset in the subsequence is normalized to obtain a normalized offset sequence, and a Hilbert transform is performed on the normalized offset sequence to extract the instantaneous phase angle sequence of the vibration dominant frequency offset; the acoustic emission pulse cumulative energy in the subsequence is normalized to obtain a normalized energy sequence, and a Hilbert transform is performed on the normalized energy sequence to extract the instantaneous phase angle sequence of the acoustic emission pulse cumulative energy; the phase angle difference between the instantaneous phase angle sequence of the vibration dominant frequency offset and the instantaneous phase angle sequence of the acoustic emission pulse cumulative energy at the same monitoring time is calculated, and the average value of the phase angle difference at all monitoring times is taken to obtain the phase consistency index.

[0011] As a technical solution of the present invention, when judging a quality anomaly, the phase consistency index is compared with a preset morphological threshold, which is an upper control limit determined by statistically analyzing the distribution of phase consistency indices of qualified bolt samples; when the phase consistency index is greater than the upper control limit, it is determined that there is a phase mismatch between the material flow and the mold vibration during the forming process of the bolt; a quality anomaly mark is generated for the bolt, and the corresponding wall thickness weak azimuth angle and phase consistency index value of the bolt are recorded.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: Simultaneously acquiring mold vibration signals and acoustic emission signals from the forming area at various monitoring moments during the cold heading process, the two types of signals at the same monitoring moment are synchronized in the time domain to generate a synchronous sensing frame. Within the synchronous sensing frame, the vibration dominant frequency offset and the accumulated energy of the acoustic emission pulses are extracted and spliced ​​in chronological order to form a process feature sequence. The vibration dominant frequency offset in the sub-sequence is normalized and subjected to Hilbert transform to extract the instantaneous phase angle sequence. Similarly, the accumulated energy of the acoustic emission pulses is normalized and subjected to Hilbert transform to extract the instantaneous phase angle sequence. The phase angle difference at the same monitoring moment is calculated and averaged to obtain the phase consistency index. The mold vibration dominant frequency offset reflects the forming load and changes in the mold's dynamic stiffness, while the accumulated energy of the acoustic emission pulses reflects the energy released by material plastic deformation and microcracks. Incorporating both into the same time coordinate and comparing the instantaneous phase can capture the phase mismatch between material flow and mold vibration. Compared to judging by a single signal amplitude or total energy threshold, the phase consistency index is more sensitive to early local material flow anomalies, reducing missed detections due to insignificant global energy changes and lowering false alarms caused by operating noise. Initial wall thickness distribution data of bolt blanks before forming is collected. Initial wall thickness values ​​at multiple radial positions are obtained at equal angles along the circumference of the end face. An initial wall thickness circumferential distribution vector is constructed, and this vector is mapped to the wall thickness distribution at the corresponding circumferential position of the thread root after forming. The wall thickness reduction rate at each circumferential position is calculated, and the central angle corresponding to the circumferential position with the largest wall thickness reduction rate is selected as the weak wall thickness azimuth angle. Based on the forming station angle corresponding to the weak wall thickness azimuth angle, the forming stage time when the bolt blank rotates to that angle is calculated, and the start and end times of the interception are determined. Vector segments located between the start and end times are extracted from the process feature sequence as subsequences. By correlating the initial wall thickness spatial distribution of the blank with the cold heading forming time process, the phase consistency analysis is focused on the period when the wall thickness weak area actually participates in deformation, avoiding the average dilution of features by signals from non-weak areas, and enhancing the targeted identification of local thinning defects at the thread root. Before and after the thin-walled azimuth angle enters the forming mold engagement zone, the subsequence covers the stage where the material flow is most intense in this area, and the extracted phase consistency index is directly related to the local forming state of the root. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0014] Figure 1 This is a flowchart of a bolt forming quality online monitoring method based on multidimensional sensor data; Figure 2This is a flowchart of the synchronous sensing frame generation and feature extraction of vibration and acoustic emission signals for cold heading equipment; Figure 3 These are the timing curves of the vibration dominant frequency and the cumulative energy of the acoustic emission pulse during the cold heading process; Figure 4 It is the time-domain variation curve of the vibration dominant frequency shift and the cumulative energy subsequence of acoustic emission pulses; Figure 5 This is a schematic diagram of the acoustic emission envelope curve, trigger threshold, and shaped occlusion zone. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0016] See Figure 1 This invention provides an online monitoring method for bolt forming quality based on multi-dimensional sensor data. During cold heading, mold vibration signals and acoustic emission signals from the forming area are acquired at monitoring times, and initial wall thickness distribution data of each bolt blank before forming are collected. The mold vibration signals and acoustic emission signals from the forming area at the same monitoring time are synchronized in the time domain to generate a synchronous sensing frame for that time. The vibration dominant frequency offset and acoustic emission pulse cumulative energy are extracted within the synchronous sensing frame. Based on the initial wall thickness distribution data of the same bolt blank, the wall thickness at the thread root of the bolt after forming is spatially mapped to determine the weak wall thickness azimuth angle. The vibration dominant frequency offset and acoustic emission pulse cumulative energy at each monitoring time are concatenated in chronological order to form a process feature sequence, and a subsequence corresponding to the time of the weak wall thickness azimuth angle is extracted from the process feature sequence. The phase consistency index between the fluctuation phase of the vibration dominant frequency offset and the energy release phase of the acoustic emission pulse cumulative energy in the subsequence is calculated. When the phase consistency index exceeds a preset morphological threshold, the bolt is marked as a bolt with abnormal quality.

[0017] In specific implementation, please refer to Figure 2The acquisition process of mold vibration signals and acoustic emission signals from the forming area is as follows: A piezoelectric accelerometer is installed on the mold base of the cold heading forming equipment. The piezoelectric accelerometer collects the vibration acceleration of the mold base at a preset sampling frequency to generate the original vibration waveform. An acoustic emission sensor is installed on the side of the forming mold. The acoustic emission sensor collects the acoustic emission signal from the forming area at a synchronous sampling frequency to generate the original acoustic emission waveform. The original vibration waveform and the original acoustic emission waveform are filtered by a second-order Butterworth bandpass filter. The lower cutoff frequency of the bandpass filter is set to 1kHz, and the upper cutoff frequency is set to 200kHz to filter out power frequency interference and mechanical transmission noise. The filtered mold vibration signal and the filtered acoustic emission signal from the forming area are aligned according to the monitoring time, and the mold vibration signal and the acoustic emission signal from the forming area corresponding to each monitoring time are output.

[0018] The generation process of the synchronous sensing frame is as follows: Using the sampling timestamp of the mold vibration signal as a reference, find the sampling point in the acoustic emission signal of the molding area with the smallest difference from the sampling timestamp of the mold vibration signal. This sampling point with the smallest difference from the sampling timestamp of the mold vibration signal is used as the synchronization matching point of the acoustic emission signal. A vibration segment centered on the sampling timestamp of the mold vibration signal and with a duration of a preset window length is extracted from the mold vibration signal. Similarly, an acoustic emission segment centered on the synchronization matching point of the acoustic emission signal and with the same preset window length is extracted from the acoustic emission signal of the molding area. The vibration segment and the acoustic emission segment are encapsulated into a synchronous sensing frame, and the corresponding monitoring time label is assigned to the synchronous sensing frame. The preset window length is set to 5 milliseconds.

[0019] The process of extracting the vibration dominant frequency offset is as follows: Perform a Fast Fourier Transform (FFT) on the vibration segment within the synchronous sensing frame to obtain the vibration spectrum. The calculation formula for the FFT is: in, Represents the first in the vibration spectrum Complex amplitude values ​​at each frequency point Indicates the first segment of the vibration segment Vibration acceleration values ​​at each sampling point This indicates the total number of sampling points within the vibration segment. Represents the imaginary unit. Indicates the frequency point number. The range of values ​​is , Indicates the time-domain sampling point number. The range of values ​​is The frequency point with the maximum amplitude in the vibration spectrum is located as the vibration master frequency at the monitoring time. The vibration master frequency is subtracted from the no-load operating fundamental frequency of the cold heading forming equipment to obtain the vibration master frequency offset. The no-load operating fundamental frequency of the cold heading forming equipment is determined by locating the frequency point with the maximum amplitude after the cold heading forming equipment is running in a no-load state without blank and collecting the mold vibration signal.

[0020] The process of extracting the accumulated energy of acoustic emission pulses is as follows: The acoustic emission segments within the synchronous sensing frame are demodulated using a Hilbert transform to obtain the acoustic emission envelope curve. The Hilbert transform demodulation process involves constructing an analytical signal for the acoustic emission segment. The real part of the analytical signal is the acoustic emission segment itself, and the imaginary part is the Hilbert transform result of the acoustic emission segment. The magnitude of the analytical signal is the acoustic emission envelope curve. The pulse segments in the acoustic emission envelope curve that exceed the trigger threshold are integrated to obtain the accumulated energy of the acoustic emission pulses. The trigger threshold is set to 5 times the root mean square value of the background noise in the non-plastic deformation stage of the acoustic emission envelope curve. The root mean square value of the background noise is determined by collecting the acoustic emission signal during the no-load operation of the cold heading forming equipment and calculating the root mean square value of the amplitude of the acoustic emission envelope curve.

[0021] See Figure 3 In the figure, the horizontal axis represents the monitoring time sequence number, the left side of the vertical axis represents the frequency value (in Hz) of the dominant vibration frequency and the frequency offset, and the right side of the vertical axis represents the cumulative energy value of the acoustic emission pulse. The blue solid line in the figure represents the dominant vibration frequency of the mold vibration signal, the red dashed line represents the no-load operating fundamental frequency of the cold heading forming equipment, the orange dashed line represents the frequency offset (i.e., the difference between the dominant vibration frequency and the no-load fundamental frequency), and the green solid line represents the cumulative energy of the acoustic emission pulse.

[0022] As shown in the figure, the dominant vibration frequency fluctuates around the no-load operating fundamental frequency, and shifts upwards between approximately 150 and 400 Hz. The amplitude of this shift is relatively large during this period, reaching a peak before gradually decreasing to near zero. The acoustic emission pulse accumulation energy curve shows a monotonically increasing trend, especially with a significantly faster energy accumulation rate between 150 and 400 Hz, indicating enhanced acoustic emission activity in the forming zone during this phase.

[0023] Based on the synchronous sensing frame generation and feature extraction process described in the above embodiment, the synchronous change trend of the vibration dominant frequency offset and the accumulated energy of the acoustic emission pulse in the figure reflects the coupling characteristics of the mold vibration and acoustic emission signal during the cold heading process. This region corresponds to the critical forming stage of the forming mold engagement zone. The synchronous change of vibration frequency offset and acoustic emission energy provides a data basis for extracting subsequences and calculating phase consistency indices in subsequent embodiments.

[0024] In practical implementation, the process of determining the weak azimuth angle of the wall thickness is as follows: Initial wall thickness values ​​are collected at multiple radial positions along the circumferential angle of the bolt blank end face. The total number of positions collected at the circumferential angle of the bolt blank end face is... , No. The initial wall thickness value collected at each circumferential position is denoted as... , The range of values ​​is All The initial wall thickness values ​​are arranged in circumferential position order to construct an initial wall thickness circumferential distribution vector. This initial wall thickness circumferential distribution vector is then mapped to the wall thickness distribution at the corresponding circumferential position of the thread root after forming. The mapping method is as follows: for the first... At each circumferential position, obtain the wall thickness value at the corresponding circumferential position of the thread root after forming, denoted as . Calculate the wall thickness reduction rate at each circumferential location. The formula for calculating the wall thickness reduction rate is: in, Indicates the first Wall thickness reduction rate at each circumferential location, Indicates the first The initial wall thickness value was collected at each circumferential position. Indicates the first The wall thickness value at the root of the formed thread corresponding to each circumferential position. This represents the total number of locations where data was collected at equal angles along the circumference of the bolt blank end face. Indicates the circumferential position number. The range of values ​​is The central angle corresponding to the circumferential position with the largest wall thickness reduction rate is selected as the weak wall thickness azimuth angle. The central angle corresponding to the circumferential position with the largest wall thickness reduction rate is the angle value of the circumferential position with the largest wall thickness reduction rate in the circumferential coordinate system of the bolt blank end face.

[0025] The process of constructing the process feature sequence is as follows: the total duration of the cold heading process is divided into multiple monitoring times at equal time intervals, and the total number of monitoring times is... , No. Each monitoring moment is recorded as , The range of values ​​is The vibration dominant frequency offsets at each monitoring time are arranged in chronological order to form a vibration dominant frequency offset time vector. Similarly, the acoustic emission pulse cumulative energy at each monitoring time is arranged in chronological order to form an acoustic emission pulse cumulative energy time vector. Both the vibration dominant frequency offset time vector and the acoustic emission pulse cumulative energy time vector have a length of [missing information]. The equal-length time-series vectors, vibration dominance frequency offset time-series vectors, and acoustic emission pulse cumulative energy time-series vectors together constitute the process feature sequence. Based on the forming station angle corresponding to the weak wall thickness azimuth, the forming stage time when the bolt blank rotates to the forming station angle corresponding to the weak wall thickness azimuth during cold heading is calculated. The starting time of the interception is determined as the monitoring time before the forming station angle corresponding to the weak wall thickness azimuth enters the forming mold engagement zone, and the ending time of the interception is determined as the monitoring time after the forming station angle corresponding to the weak wall thickness azimuth leaves the forming mold engagement zone. The vibration dominance frequency offset time-series vector segment and the acoustic emission pulse cumulative energy time-series vector segment located between the starting and ending times of the interception are extracted from the process feature sequence, and these extracted segments are used as subsequences.

[0026] See Figure 4 In the figure, the horizontal axis represents the monitoring time sequence number, the left side of the vertical axis represents the vibration dominant frequency offset in Hertz (Hz), and the right side of the vertical axis represents the cumulative energy of the acoustic emission pulse. The blue curve represents the time-series vector of the vibration dominant frequency offset, and the green curve represents the time-series vector of the cumulative energy of the acoustic emission pulse. The figure uses a red dashed line to mark the start time, a purple dashed line to mark the end time, and a brown dashed line to mark the time corresponding to the forming stage at the weak wall thickness azimuth angle.

[0027] From the time-series vector of the vibration dominant frequency offset in the figure, before the initial intercept (approximately the 150th monitoring time), the vibration dominant frequency offset was low and the fluctuation amplitude was small, indicating that cold heading had not yet entered the forming station corresponding to the weak wall thickness azimuth angle. Subsequently, the vibration dominant frequency offset rose rapidly, reaching a peak near the forming stage corresponding to the weak wall thickness azimuth angle (approximately the 300th monitoring time), exhibiting strong fluctuation characteristics, reflecting that the vibration dominant frequency of the mold during the forming process was significantly affected by material flow. Afterward, the vibration dominant frequency offset gradually decreased, returning to a low level by the end of the intercept (approximately the 430th monitoring time), indicating that the corresponding forming stage had ended.

[0028] The time-series vector of accumulated acoustic emission pulse energy generally shows an increasing trend, with a significant increase rate around the forming stage corresponding to the azimuth angle of the thin wall thickness, indicating that the acoustic emission energy generated by plastic deformation during the forming process is enhanced. The time-series characteristic segments of vibration dominance frequency offset and accumulated acoustic emission pulse energy from the start to the end of the interception together constitute a subsequence, reflecting the dynamic behavior of the forming region corresponding to the azimuth angle of the thin wall thickness.

[0029] In practical implementation, the calculation process of the phase consistency index is as follows: the vibration dominant frequency offset in the subsequence is normalized to obtain a normalized offset sequence. The normalization process adopts the minimum-maximum normalization method, specifically: the maximum and minimum values ​​of the vibration dominant frequency offset in the subsequence are calculated; the minimum value of the vibration dominant frequency offset is subtracted from each vibration dominant frequency offset value in the subsequence to obtain the vibration dominant frequency offset difference; then, the vibration dominant frequency offset difference is divided by the difference between the maximum and minimum values ​​of the vibration dominant frequency offset to obtain the normalized offset sequence. A Discrete Hilbert Transform is performed on the normalized offset sequence to construct the analytical signal of the dominant vibration frequency offset. The real part of the analytical signal of the dominant vibration frequency offset is the normalized offset sequence, and the imaginary part is the result of the Discrete Hilbert Transform of the normalized offset sequence. Dividing the imaginary part of the analytical signal of the dominant vibration frequency offset by the real part and then taking the arctangent yields the instantaneous phase angle sequence of the dominant vibration frequency offset. The th... The instantaneous phase angle at each monitoring moment is denoted as .

[0030] The accumulated energy of acoustic emission pulses in the subsequence is normalized to obtain a normalized energy sequence. The normalization process uses a minimum-maximum normalization method, specifically: calculating the maximum and minimum accumulated energy values ​​of acoustic emission pulses in the subsequence; subtracting the minimum accumulated energy value from each accumulated energy value in the subsequence to obtain the difference in accumulated energy; and then dividing this difference by the difference between the maximum and minimum accumulated energy values ​​to obtain the normalized energy sequence. A discrete Hilbert transform is then performed on the normalized energy sequence to construct an analytic signal of the accumulated acoustic emission pulse energy. The real part of this analytic signal represents the normalized energy sequence, and the imaginary part represents the result of the discrete Hilbert transform. Dividing the imaginary part of the analytic signal by its real part and taking the arctangent yields the instantaneous phase angle sequence of the accumulated acoustic emission pulse energy. The instantaneous phase angle at each monitoring moment is denoted as .

[0031] The phase angle difference between the instantaneous phase angle sequence of the vibration dominant frequency offset and the instantaneous phase angle sequence of the acoustic emission pulse cumulative energy at the same monitoring moment is calculated, and the phase angle difference is denoted as... , The phase angle difference is unwrapped as follows: when the phase angle difference is greater than π, the phase angle difference is subtracted by 2π; when the phase angle difference is less than −π, the phase angle difference is added by 2π; the phase unwrapping process is repeated until the phase angle difference is within the range specified by the time interval. Within the interval, the average of the phase angle differences at all monitoring times is taken to obtain the phase consistency index. The formula for calculating the phase consistency index is: in, This indicates the phase consistency index. This represents the total number of monitoring times in the subsequence. Indicates the monitoring time sequence number. The range of values ​​is , Indicates the first The phase angle difference at each monitoring moment after phase unwrapping processing Indicates the first The instantaneous phase angle of the vibration dominant frequency offset at each monitoring moment Indicates the first The instantaneous phase angle of the accumulated energy of the acoustic emission pulse at each monitoring moment.

[0032] In practical implementation, the process of determining the preset morphological threshold is as follows: the number of samples collected is... For each qualified bolt sample, the phase consistency index calculation process is performed to obtain the phase consistency index set of qualified bolt samples. This indicates the total number of qualified bolt samples. The range of values ​​is Calculate the mean of the phase consistency index set of qualified bolt samples, denoted as . Calculate the standard deviation of the phase consistency index set of qualified bolt samples, denoted as . The preset morphological threshold adopts an upper control limit form, and the calculation formula for the upper control limit is: in, This indicates the upper limit of the preset morphological threshold control. This represents the mean of the set of phase consistency indexes for qualified bolt samples. The standard deviation represents the set of phase consistency indicators for qualified bolt samples. The coefficient 3 is the control limit coefficient. The basis for taking the control limit coefficient of 3 is that in statistical process control, the mean plus 3 times the standard deviation covers 99.73% of the normal fluctuation range. Values ​​exceeding the range of the mean plus 3 times the standard deviation are judged as abnormal fluctuations.

[0033] The phase consistency index of the bolt to be judged is compared with the upper limit of the preset morphological threshold. When the phase consistency index of the bolt to be judged is greater than the upper limit of the preset morphological threshold, it is determined that there is a phase mismatch between the material flow and the mold vibration during the forming process. When the phase consistency index of the bolt to be judged is less than or equal to the upper limit of the preset morphological threshold, no quality anomaly mark is generated.

[0034] When it is determined that there is a phase mismatch between the material flow and the mold vibration during the forming process of the bolt to be judged, a quality abnormality mark is generated for the bolt to be judged, and the weak wall thickness azimuth angle and the phase consistency index value of the bolt to be judged are recorded. The weak wall thickness azimuth angle and the phase consistency index value are written into the quality abnormality record entry of the bolt to be judged.

[0035] See Figure 5 In the graph, the horizontal axis represents the serial number of the qualified bolt samples, ranging from 0 to 500, and the vertical axis represents the value of the phase consistency index, ranging from 0 to 0.4. The blue scatter dots represent the phase consistency index values ​​of each qualified bolt sample, exhibiting a discrete distribution, with most values ​​clustered between 0.0 and 0.25, reflecting the fluctuation range of the phase consistency index for qualified samples. The green solid line represents the mean of the phase consistency index set of qualified bolt samples, which is approximately 0.12, representing the central trend of the phase consistency index for qualified samples. The red dashed line represents the upper control limit of the preset morphological threshold, calculated based on the mean of the phase consistency index set of qualified bolt samples plus three standard deviations. The control limit is approximately 0.24, indicating the upper limit of the normal fluctuation range of phase consistency. The red triangle in the upper right corner of the graph represents the phase consistency index value of the bolt to be judged, which is significantly higher than the upper control limit of the preset morphological threshold, reaching approximately 0.37, indicating that the phase consistency index of this bolt exceeds the normal fluctuation range and exhibits abnormal fluctuation.

[0036] In practical implementation, the envelope detection of the acoustic emission segment uses the Hilbert transform demodulation method to extract the envelope curve of the acoustic emission signal. For the first... Perform a Discrete Hilbert Transform (DHFT) on the amplitude at each sampling point to obtain the DHFT result of the acoustic emission segment. The formula for calculating the acoustic emission envelope curve is: in, The acoustic emission envelope curve represents the first... The envelope amplitude at each sampling point Indicating the first sound emission segment The amplitude of each sampling point Indicates the first sound emission segment The result of performing a discrete Hilbert transform on the amplitude of each sampling point. Indicates the sampling point number. The range of values ​​is , This indicates the total number of sampling points within an acoustic emission segment.

[0037] In practice, the process for determining the start and end times of the interception is as follows: In the monitoring time sequence of the cold heading process, the moment when the forming station angle corresponding to the weak wall thickness azimuth angle enters the forming mold engagement zone is recorded as the entry into engagement zone moment. The monitoring moment located before and adjacent to the entry into engagement zone moment is determined as the interception start time. The moment when the forming station angle corresponding to the weak wall thickness azimuth angle leaves the forming mold engagement zone is recorded as the exit from engagement zone moment. The monitoring moment located after and adjacent to the exit from engagement zone moment is determined as the interception end time. All monitoring moments between the interception start and end times constitute the time range corresponding to the subsequence.

[0038] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for online monitoring of bolt forming quality based on multidimensional sensor data, characterized in that, include: Acquire the mold vibration signal and the acoustic emission signal of the forming area at each monitoring moment during the cold heading process, and collect the initial wall thickness distribution data of each bolt blank before forming; The mold vibration signal and the acoustic emission signal of the forming area at the same monitoring time are synchronized in the time domain to generate a synchronous sensing frame at that time, and the vibration main frequency offset and the cumulative energy of the acoustic emission pulse are extracted in the synchronous sensing frame. Based on the initial wall thickness distribution data of the same bolt blank, the wall thickness at the root of the thread of the bolt after forming is spatially mapped to determine the weak azimuth angle of the wall thickness. The vibration dominant frequency offset and acoustic emission pulse cumulative energy at each monitoring moment are spliced ​​together in chronological order to form a process characteristic sequence, and the corresponding time subsequence is extracted from the process characteristic sequence based on the weak wall thickness azimuth angle. Calculate the phase consistency index between the fluctuation phase of the vibration dominant frequency offset and the energy release phase of the accumulated energy of the acoustic emission pulse in the subsequence; If the phase consistency index exceeds the preset morphology threshold, the bolt is marked as a bolt with abnormal quality.

2. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 1, characterized in that, The acquisition of mold vibration signals and acoustic emission signals from the forming area at various monitoring moments during the cold heading process specifically includes: A piezoelectric accelerometer is installed on the mold base to continuously collect the vibration acceleration of the mold base at a sampling frequency, thereby obtaining the original vibration waveform; An acoustic emission sensor is installed on the side of the molding die to continuously collect the acoustic emission signal of the molding area at a synchronous sampling frequency, thereby obtaining the original acoustic emission waveform. The original vibration waveform and the original acoustic emission waveform are bandpass filtered to remove power frequency interference and mechanical transmission noise. The filtered waveforms are then aligned according to the monitoring time, and the mold vibration signal and the acoustic emission signal of the molding area corresponding to each monitoring time are output.

3. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 1, characterized in that, The step of synchronizing and aligning the mold vibration signal and the acoustic emission signal of the molding area at the same monitoring time in the time domain to generate a synchronous sensing frame at that time specifically includes: Using the sampling timestamp of the mold vibration signal as a reference, find the sampling point in the acoustic emission signal of the molding area with the smallest difference from the timestamp, and use this sampling point as the synchronization matching point of the acoustic emission signal; A vibration segment centered on the timestamp and with a duration of a preset window length is extracted from the mold vibration signal, and an acoustic emission segment with the same duration centered on the synchronization matching point is extracted from the acoustic emission signal of the molding area. The vibration segment and the acoustic emission segment are encapsulated into a synchronous sensing frame, and the synchronous sensing frame is labeled with the corresponding monitoring time tag.

4. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 1, characterized in that, The extraction of the vibration dominant frequency offset and the accumulated energy of the acoustic emission pulse within the synchronous sensing frame specifically includes: Perform a fast Fourier transform on the vibration segment within the synchronous sensing frame to obtain the vibration spectrum, and locate the frequency point with the maximum amplitude from the vibration spectrum as the dominant vibration frequency at that monitoring moment; The vibration frequency offset is obtained by subtracting the no-load operating fundamental frequency of the cold heading forming equipment from the vibration frequency. Envelope detection is performed on the acoustic emission segments within the synchronous sensing frame to obtain the acoustic emission envelope curve. The pulse segments in the acoustic emission envelope curve that exceed the trigger threshold are integrated to obtain the cumulative energy of the acoustic emission pulse.

5. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 1, characterized in that, The step of spatially mapping the thread root wall thickness of the formed bolt based on the initial wall thickness distribution data of the same bolt blank to determine the weak wall thickness azimuth angle specifically includes: Initial wall thickness values ​​at multiple radial positions are collected at equal angles along the circumferential direction of the bolt blank end face to construct an initial wall thickness circumferential distribution vector; The initial circumferential wall thickness distribution vector is mapped to the wall thickness distribution at the corresponding circumferential position of the thread root after forming, and the wall thickness reduction rate at each circumferential position is calculated. The central angle corresponding to the circumferential position with the largest wall thickness reduction rate is selected as the weak wall thickness azimuth angle.

6. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 1, characterized in that, The process involves arranging the vibration dominant frequency offset and acoustic emission pulse cumulative energy at each monitoring moment in chronological order to form a process characteristic sequence, and then extracting a subsequence corresponding to the moment from the process characteristic sequence based on the weak wall thickness azimuth angle. Specifically, this includes: The total duration of the cold heading process is divided into multiple monitoring moments at equal time intervals. The vibration dominant frequency offset and the cumulative energy of the acoustic emission pulse at each moment are arranged into two time-series vectors of equal length according to the order of the monitoring moments, thus forming a process characteristic sequence. Based on the forming station angle corresponding to the weak wall thickness azimuth angle, calculate the forming stage time when the bolt blank rotates to that angle, and determine the start and end times of cutting. Extract the vector segment located between the start and end times of the truncation from the process feature sequence, and use it as a subsequence.

7. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 1, characterized in that, The phase consistency index between the fluctuation phase of the vibration dominant frequency offset and the energy release phase of the acoustic emission pulse accumulated energy in the calculated subsequence specifically includes: The vibration dominant frequency offset in the subsequence is normalized to obtain a normalized offset sequence. Then, Hilbert transform is performed on the normalized offset sequence to extract the instantaneous phase angle sequence of the vibration dominant frequency offset. The accumulated energy of acoustic emission pulses in the subsequence is normalized to obtain a normalized energy sequence. Then, a Hilbert transform is performed on the normalized energy sequence to extract the instantaneous phase angle sequence of the accumulated energy of the acoustic emission pulses. The phase angle difference between the instantaneous phase angle sequence of the vibration dominant frequency offset and the instantaneous phase angle sequence of the acoustic emission pulse cumulative energy at the same monitoring time is calculated, and the average value of the phase angle difference at all monitoring times is taken to obtain the phase consistency index.

8. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 1, characterized in that, If the phase consistency index exceeds a preset morphological threshold, the bolt is marked as a quality defect bolt, specifically including: The phase consistency index is compared with a preset morphological threshold, which is an upper control limit determined by statistically analyzing the distribution of the phase consistency index of qualified bolt samples. When the phase consistency index is greater than the upper control limit, it is determined that there is a phase mismatch between the material flow and the mold vibration during the forming process of the bolt. Generate a quality anomaly marker for the bolt and record the corresponding weak wall thickness azimuth angle and phase consistency index value.

9. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 4, characterized in that, The envelope detection of the acoustic emission segment within the synchronous sensing frame is specifically performed using the Hilbert transform demodulation method to extract the envelope curve of the acoustic emission signal.

10. The online monitoring method for bolt forming quality based on multi-dimensional sensor data according to claim 6, characterized in that, The start time of the interception is the monitoring time before the forming station angle corresponding to the weak wall thickness azimuth angle enters the forming mold engagement zone, and the end time of the interception is the monitoring time after the forming station angle leaves the forming mold engagement zone.

Images