A fault early warning method for automatic welding equipment for miniature fan motor housings

Abstract

This invention belongs to the field of alarm system technology, specifically relating to a fault early warning method for automatic welding equipment for miniature fan motor housings. The method includes: obtaining a standardized torque sequence; constructing a trajectory matrix based on the standardized torque sequence and performing singular value decomposition; calculating the energy concentration of the principal components of the standardized torque sequence, determining the waveform oscillation factor of the eigenvector corresponding to each principal component of the standardized torque sequence; determining the true assembly stress index and fault early warning threshold of the standardized torque sequence, thereby determining the risk of motor housing springback to achieve fault early warning for the automatic welding equipment. This invention analyzes the singular values, energy values, and time-domain fluctuation characteristics of the standardized torque sequence to identify true springback under noise interference, overcoming the mode mixing problem of traditional singular spectrum analysis algorithms. It utilizes a dynamic fault early warning threshold to determine the risk of motor housing springback, improving the accuracy of fault early warning for automatic welding equipment.

Description

Technical Field

[0001] This invention relates to the field of alarm system technology. More specifically, this invention relates to a fault early warning method for an automatic welding equipment for miniature fan motor housings. Background Technology

[0002] With the development of microelectronics manufacturing technology, micro fan assembly lines generally adopt a process of servo press fitting followed by automatic welding. The stable operation of welding equipment depends on the physical input state of the parts to be welded. However, the motor housing of micro fans is mostly a thin-walled stamped part. Due to the influence of materials and tolerances, excessively tight press fitting will generate huge internal assembly stress. If this hidden stress is not identified before welding, it will directly lead to abnormal boundary conditions at the welding station. If welding is forced under this condition, it is very easy to cause process failures such as sparking and needle sticking, which will directly lead to electrode head ablation or damage to precision parts, and seriously shorten the life of welding equipment.

[0003] Existing welding current monitoring technologies are mostly post-processing detection technologies, which cannot reveal the internal stress state and are difficult to effectively prevent equipment operation risks before physical action occurs. Therefore, using torque data analysis of the pre-welding process environment from the moment the servo motor contacts the motor housing until the pressure holding ends has become the key to protecting welding equipment. Singular spectrum analysis algorithm, as a powerful tool for processing nonlinear time series, can effectively extract the real torque trend, thereby achieving more accurate real-time fault warnings for welding equipment.

[0004] However, when using singular spectrum analysis algorithms to decompose torque data during the servo holding pressure stage, grouping is typically based on the magnitude of the singular values ​​of each component. By default, larger singular values ​​correspond to the main trend components. In the actual operating environment of automatic welding equipment for micro fan motor housings, not only electronic thermal noise but also strong electromagnetic interference exists. Unlike the low energy values ​​and low singular values ​​corresponding to electronic thermal noise, the noise energy generated by strong electromagnetic interference is stronger, causing its corresponding singular values ​​to approach or even exceed the singular values ​​of the true springback trend. In this case, grouping and reconstructing solely based on the magnitude of singular values ​​will lead to mode aliasing, that is, misidentifying high-energy noise components as stress trends or filtering out genuine weak springback signals as noise. This aliasing phenomenon will cause the calculated stress indicators to be distorted, leading to false alarms or missed alarms. More seriously, if the system fails to accurately identify abnormal welding boundary conditions caused by excessive assembly stress, the welding equipment will face an extremely high risk of process failure when performing subsequent welding operations. This latent failure will not only lead to serious welding defects but will also directly cause damage to the electrode head of the welding equipment due to abnormal energy release, seriously affecting the service life of the welding equipment. Summary of the Invention

[0005] To address the technical problem that traditional singular value analysis algorithms, when decomposing torque data during the servo holding pressure stage, group the data based on the magnitude of the singular values ​​of each component, making it difficult to distinguish between noise interference and the true stress trend, leading to damage to the welding equipment electrode head and severely affecting the service life of the welding equipment, this invention provides a fault early warning method for automatic welding equipment for micro fan motor housings. The method includes: acquiring the original torque sequence of the servo motor during the holding pressure stage; standardizing the original torque sequence to obtain a standardized torque sequence; constructing a trajectory matrix based on the standardized torque sequence; performing singular value decomposition on the trajectory matrix; and determining the ratio between the sum of the energy of all principal components obtained from the singular value decomposition and the sum of the energy of all components. The difference between the singular values ​​of the first principal component and the second principal component is used to determine the energy concentration of the principal components in the standardized torque sequence. Based on the sum of the absolute values ​​of the differences between adjacent time values ​​in the eigenvectors corresponding to each principal component in the standardized torque sequence, the waveform oscillation factor of the eigenvectors corresponding to each principal component in the standardized torque sequence is determined. The true assembly stress index of the standardized torque sequence is determined based on the energy concentration of the principal components and the waveform oscillation factor. A fault warning threshold for the standardized torque sequence is determined based on the true assembly stress index. Based on the true assembly stress index and the fault warning threshold, it is determined whether there is a risk of springback in the micro fan motor housing, thus achieving fault warning for the automatic welding equipment of the micro fan motor housing.

[0006] This invention eliminates interference from electronic thermal noise by calculating the energy concentration of principal components and utilizing the energy value differences between different components. By calculating the waveform oscillation factor, it addresses the mode aliasing problem that easily occurs under strong electromagnetic noise by utilizing the difference between the smoothness of the rebound trend and the high-frequency oscillation of strong electromagnetic interference, thus more accurately distinguishing between real rebound and strong electromagnetic noise. By combining the principal component energy concentration and waveform oscillation factor to calculate the real assembly stress index, it retains the energy characteristics of the real rebound trend while suppressing the component affected by strong electromagnetic noise. By determining the fault warning threshold based on the real assembly stress index, it achieves dynamic adjustment of the fault alarm threshold according to the stability of the production process. When the process stability is low, the fault alarm threshold is increased to prevent false alarms; when the process is stable, the fault alarm threshold is decreased to prevent missed alarms, thus achieving more accurate and real-time early warning for the automatic welding equipment for micro fan motor housings.

[0007] Preferably, obtaining the original torque sequence of the servo motor during the pressure holding stage includes: at the station where the motor housing is pressed into the shaft core, acquiring the original torque sequence of the servo motor from the point of contact with the motor housing until the pressure holding ends through the bus communication interface of the servo driver.

[0008] Preferably, the method for obtaining the standardized torque sequence is as follows: the mean of the original torque sequence is first subtracted from the value of each data point in the original torque sequence, and the result is then divided by the standard deviation of the original torque sequence to obtain the numerical sequence as the standardized torque sequence.

[0009] Preferably, the step of constructing a trajectory matrix based on the standardized torque sequence includes: setting the embedding dimension to one-third of the length of the standardized torque sequence, and constructing a trajectory matrix based on the standardized torque sequence.

[0010] Preferably, the singular value decomposition of the trajectory matrix includes: using a singular spectrum analysis algorithm to perform singular value decomposition on the trajectory matrix to obtain all components of the standardized torque sequence, and taking the three components with the largest singular values ​​among all components as the principal components of the standardized torque sequence.

[0011] Preferably, the energy concentration of the principal components satisfies the expression:

[0012] In the formula, For the first The degree of principal component energy concentration of a standardized torque sequence The number of principal components in the standardized torque sequence. The total number of components in the standardized torque sequence. For the first The first standardized torque sequence The singular values ​​of each principal component For the first The first standardized torque sequence The energy values ​​of each principal component For the first The first standardized torque sequence Singular values ​​of each component For the first The first standardized torque sequence The energy value of each component For the first The singular values ​​of the first principal component of a normalized torque sequence For the first The singular values ​​of the second principal component of a standardized torque sequence It is the maximum-minimum normalization function.

[0013] This invention constructs a positive correlation function relationship based on the product of the ratio between the sum of the energies of all principal components and the sum of the energies of all components, and the difference between the singular values ​​of the first principal component and the singular values ​​of the second principal component. This enables the assessment of the energy concentration of principal components. The ratio between the sum of the energies of all principal components and the sum of the energies of all components reflects the degree of energy concentration of the energy values ​​of the standardized torque sequence on the principal components. The difference between the singular values ​​of the first principal component and the singular values ​​of the second principal component highlights the difference between the first principal component and the second principal component. This allows the calculation of a larger energy concentration of principal components for a standardized torque sequence with a springback tendency, thereby mapping the monotonically decaying trend of the springback signal and providing a reliable benchmark for the calculation of the actual assembly stress index.

[0014] Preferably, the waveform oscillation factor satisfies the expression:

[0015] In the formula, For the first In the standardized torque sequence, the first... The waveform oscillation factor of each principal component corresponds to the eigenvector. For the first The lengths of the eigenvectors corresponding to all principal components in a standardized torque sequence. For the first In the standardized torque sequence, the first... The eigenvector corresponding to the principal component is the th eigenvector. The value at each moment. For the first In the standardized torque sequence, the first... The eigenvector corresponding to the principal component is the th eigenvector. The value at each moment. For the first In the standardized torque sequence, the first... Each principal component corresponds to the maximum value in the eigenvector at all time points. For the first In the standardized torque sequence, the first... Each principal component corresponds to the minimum value in the eigenvector across all time points. To prevent hyperparameters with a denominator of zero.

[0016] This invention achieves a quantitative evaluation of the waveform oscillation factor of the eigenvector corresponding to each principal component in a standardized torque sequence by constructing a normalized function relationship between the sum of the absolute values ​​of the numerical differences between adjacent moments in the eigenvector corresponding to each principal component and the range of the numerical values ​​between adjacent moments in the eigenvector corresponding to each principal component. The sum of the absolute values ​​of the numerical differences between adjacent moments in the eigenvector corresponding to each principal component reflects the cumulative amount of high-frequency fluctuations of the eigenvector in the time domain. The range of the numerical values ​​between adjacent moments in the eigenvector corresponding to each principal component provides a normalized benchmark based on amplitude to eliminate the interference of signal energy magnitude on shape judgment. This allows the calculation of a larger waveform oscillation factor for strong electromagnetic interference components with higher energy but more volatile time domain fluctuations, while the calculation of a smaller waveform oscillation factor for rebound components with smoother time domain fluctuations. This enables the elimination of principal components affected by strong electromagnetic interference, providing a more accurate basis for subsequently constructing a true assembly stress index based on the energy concentration of the principal components.

[0017] Preferably, the actual assembly stress index satisfies the expression:

[0018] In the formula, For the first The true assembly stress index of a standardized torque sequence. For the first The degree of principal component energy concentration of a standardized torque sequence The number of principal components in the standardized torque sequence. For the first The first standardized torque sequence The singular values ​​of each principal component For the first The first standardized torque sequence The energy values ​​of each principal component For the first In the standardized torque sequence, the first... The waveform oscillation factor of each principal component corresponds to the eigenvector. It is the maximum-minimum normalization function.

[0019] This invention achieves the evaluation of the true assembly stress index by constructing a functional relationship that uses the energy concentration of principal components and the waveform oscillation factor to perform exponential decay weighting. The exponential decay term of the waveform oscillation factor suppresses the principal components subject to strong electromagnetic interference in the summation because their weights approach zero, while the principal components with weaker electromagnetic interference are effectively included because their weights are retained. This allows for the accurate removal of strong electromagnetic noise energy mixed in the principal components, outputting the true stress index of the motor housing, and providing a high signal-to-noise ratio data foundation for the adaptive calculation of subsequent fault warning thresholds.

[0020] Preferably, the fault warning threshold satisfies the expression:

[0021] In the formula, For the first Fault warning threshold for a standardized torque sequence For the first The mean of the true assembly stress exponents of the most recent standardized torque sequences preceding a given standardized torque sequence. For the first The standard deviation of the true assembly stress index of the most recent normalized torque sequences preceding a given normalized torque sequence. Based on the basic safety factor.

[0022] This invention is achieved by means of the first The average of the true assembly stress exponents of the most recent standardized moment sequences preceding the first standardized moment sequence is used as the benchmark, combined with the first... The standard deviation of the true assembly stress exponent of the most recent standardized moment sequences preceding the first standardized moment sequence, and the standard deviation of the first standardized moment sequence. The ratio of the mean values ​​of the true assembly stress exponents of the most recent standardized torque sequences preceding the first standardized torque sequence is used to adaptively adjust the functional relationship, thereby achieving adaptive setting of the fault warning threshold. The mean of the true assembly stress exponents of the most recent standardized torque sequences preceding the first standardized torque sequence reflects the normal fluctuation boundary of the production process. The standard deviation of the true assembly stress exponent of the most recent standardized moment sequences preceding the first standardized moment sequence, and the standard deviation of the first standardized moment sequence. The ratio of the mean of the true assembly stress index of the most recent standardized torque sequences before a standardized torque sequence reflects the stability of the current production process. This allows for the automatic widening of the alarm threshold when a larger fault warning threshold is calculated under conditions of severe process fluctuations or environmental interference, and the automatic tightening of the alarm threshold when a smaller fault warning threshold is calculated under stable process conditions. This avoids false alarms caused by overly tight fixed thresholds and prevents missed fault detections caused by overly loose thresholds, thus providing a logical guarantee for the safe operation of the automatic welding equipment for micro fan motor housings.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention solves the technical problem that traditional singular spectrum analysis algorithms cannot distinguish between strong background noise and real springback due to their reliance on singular value size grouping, thus leading to false alarms or missed alarms in welding equipment fault warnings. This is achieved by introducing a real assembly stress evaluation mechanism based on the fusion of singular value energy characteristics and eigenvector waveform morphology.

[0025] This invention establishes the intrinsic mapping relationship between the energy concentration of principal components, the cumulative energy ratio of principal components, and the difference between the first two singular values ​​by analyzing the characteristics of the singular values ​​and energy values ​​of the standardized torque sequence. On this basis, a waveform oscillation factor is further introduced to couple the index reflecting the energy value characteristics of the signal with the index reflecting the time domain fluctuation of the signal, thereby realizing the identification of the physical properties of principal components under strong electromagnetic interference environment.

[0026] This invention can calculate the waveform oscillation factor of the time-domain smoothness for each principal component. By constructing a weighted exponential decay model, it achieves precise matching between energy statistics and time-domain fluctuation states. For principal components with high energy values ​​but violent oscillations, extremely low weights are applied to suppress them, while for principal components with true rebound signals, their energy weights are retained for accurate calculation. Simultaneously, a fault warning threshold is constructed. Under conditions of severe process fluctuations or environmental interference, the fault warning threshold is increased to prevent false alarms, while it is decreased when the process is stable to prevent missed alarms. Ultimately, this invention improves the accuracy and anti-interference capability of assembly stress detection for micro fan motor housings, enhances the robustness of the fault warning system, and can timely and accurately identify welding boundary anomalies caused by excessively tight assembly dimensions. It effectively avoids process faults such as sparking and needle sticking, providing a solid technical guarantee for the long-term healthy operation of automated welding equipment. Attached Figure Description

[0027] Figure 1 The flowchart illustrates a fault early warning method for an automatic welding equipment for a miniature fan motor housing according to the present invention.

[0028] Figure 2 A schematic diagram illustrating fault warnings for automatic welding equipment based on fixed warning thresholds in existing technology;

[0029] Figure 3 This is a schematic diagram illustrating the adaptive fault warning threshold of the present invention for fault warning of automatic welding equipment. Detailed Implementation

[0030] 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, not all, of the embodiments of the present invention. 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.

[0031] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] This invention discloses a fault early warning method for automatic welding equipment for miniature fan motor housings, referring to... Figure 1 This includes steps S001 to S005:

[0033] S001: Obtain the original torque sequence of the servo motor during the pressure holding phase, and standardize the original torque sequence to obtain the standardized torque sequence.

[0034] Specifically, at the station where the motor housing is pressed into the shaft core, the original torque sequence of the servo motor from the moment it contacts the motor housing until the pressure holding ends is collected through the bus communication interface of the servo driver. The value of each data point in the original torque sequence is first subtracted from the mean of the original torque sequence, and the result is then divided by the standard deviation of the original torque sequence to obtain the numerical sequence as the standardized torque sequence. In this embodiment, the acquisition frequency of the original torque sequence is 1000Hz. In other embodiments, the implementer can set the acquisition frequency of the original torque sequence according to the actual implementation situation.

[0035] S002: Construct a trajectory matrix based on the standardized torque sequence, perform singular value decomposition on the trajectory matrix, and determine the degree of energy concentration of the principal components of the standardized torque sequence based on the ratio between the sum of energy of all principal components obtained by singular value decomposition and the sum of energy of all components, as well as the difference between the singular values ​​of the first principal component and the singular values ​​of the second principal component.

[0036] It should be noted that when using singular spectrum analysis to process normalized torque sequences, it is difficult to distinguish between springback signals and electronic thermal noise based solely on the absolute magnitude of singular values. This leads to deviations in the assessment of assembly stress trend intensity and makes it impossible to accurately determine the dominant component of the signal. According to signal processing theory, the springback signal of the motor housing is a smooth, monotonically decaying trend, and its corresponding trajectory matrix has a very low rank. The energy will be highly concentrated on the first principal component, and there is usually a significant difference between the singular values ​​of the first principal component and the second principal component. In contrast, the energy distribution of electronic thermal noise signals is relatively flat and dispersed. Therefore, this invention determines the degree of energy concentration of the principal components of the normalized torque sequence to characterize the degree of energy aggregation of the springback trend in the normalized torque sequence, thus avoiding the misleading extraction of the springback trend by electronic thermal noise.

[0037] Specifically, a trajectory matrix is ​​constructed based on the standardized moment sequence. The singular value decomposition of the trajectory matrix is ​​performed using the singular spectrum analysis algorithm to obtain all components of the standardized moment sequence. The components with the largest singular values ​​among all components are taken as the principal components of the standardized moment sequence. In this embodiment, the embedding dimension in the singular spectrum analysis algorithm is one-third of the length of the standardized moment sequence, and the three components with the largest singular values ​​among all components are taken as the principal components of the standardized moment sequence. In other embodiments, implementers can set the embedding dimension and the selection method of principal components according to the actual implementation situation.

[0038] Specifically, the energy concentration of the principal components satisfies the expression:

[0039] ;

[0040] In the formula, For the first The degree of principal component energy concentration of a standardized torque sequence The number of principal components in the standardized torque sequence. The total number of components in the standardized torque sequence. For the first The first standardized torque sequence The singular values ​​of each principal component For the first The first standardized torque sequence The energy values ​​of each principal component For the first The first standardized torque sequence Singular values ​​of each component For the first The first standardized torque sequence The energy value of each component For the first The singular values ​​of the first principal component of a normalized torque sequence For the first The singular values ​​of the second principal component of a standardized torque sequence It is the maximum-minimum normalization function.

[0041] in, The larger it is, the more likely it is to be the first The more energy of a standardized moment sequence is concentrated in the principal components, the more it indicates that the... The more pronounced the monotonic decay trend of the standardized torque sequence, the less it is affected by electronic thermal noise, therefore the... The greater the energy concentration of the principal components of a standardized torque sequence, the greater the concentration of energy. The larger it is, the more likely it is to be the first The larger the difference between the singular values ​​of the first principal component and the second principal component of a standardized moment sequence, the stronger the difference. The more pronounced the monotonic decay trend of the standardized torque sequence, the less interference from electronic thermal noise it exhibits, and the greater its reliability. Therefore, the... The greater the energy concentration of the principal components of a standardized torque sequence, the greater the concentration of energy.

[0042] S003: Based on the sum of the absolute values ​​of the numerical differences between adjacent moments in the eigenvectors corresponding to each principal component in the standardized torque sequence, determine the waveform oscillation factor of the eigenvectors corresponding to each principal component in the standardized torque sequence. Based on the energy concentration of the principal components and the waveform oscillation factor, determine the true assembly stress index of the standardized torque sequence.

[0043] It should be noted that, due to the complex operating environment of the automatic welding equipment for micro fan motor housings, strong electromagnetic interference at specific frequencies often exhibits high-energy characteristics similar to springback signals, making it difficult to distinguish whether the high-energy component originates from springback or strong electromagnetic interference. The actual springback of the motor housing is a low-frequency smoothing process of material stress release, while the eigenvector corresponding to the interference signal will be accompanied by severe oscillations in the time domain. Therefore, this invention determines a waveform oscillation factor to characterize the intensity of time-domain oscillations of the eigenvector corresponding to each principal component in the standardized torque sequence. Furthermore, by combining the energy concentration of the principal components with the waveform oscillation factor, the smoothness of the high-energy component is comprehensively evaluated, thereby determining the true assembly stress index and avoiding strong electromagnetic interference from misleading the assessment of stress springback trends.

[0044] Specifically, the waveform oscillation factor satisfies the following expression:

[0045] ;

[0046] In the formula, For the first In the standardized torque sequence, the first... The waveform oscillation factor of each principal component corresponds to the eigenvector. For the first The lengths of the eigenvectors corresponding to all principal components in a standardized torque sequence. For the first In the standardized torque sequence, the first... The eigenvector corresponding to the principal component is the th eigenvector. The value at each moment. For the first In the standardized torque sequence, the first... The eigenvector corresponding to the principal component is the th eigenvector. The value at each moment. For the first In the standardized torque sequence, the first... Each principal component corresponds to the maximum value in the eigenvector at all time points. For the first In the standardized torque sequence, the first... Each principal component corresponds to the minimum value in the eigenvector across all time points. To prevent hyperparameters with a denominator of zero, the range of hyperparameter values ​​is [range to be specified]. In this embodiment, the hyperparameter for preventing the denominator from being zero is set to 0.001. In other embodiments, implementers can set it according to the actual implementation situation.

[0047] In the formula, The larger it is, the more likely it is to be the first The larger the total amplitude of the jumps between adjacent time points in the eigenvector corresponding to the _th standardized torque sequence, the more it indicates that the _th In the standardized torque sequence, the first... The larger the waveform oscillation factor of the eigenvector corresponding to each principal component, the greater the oscillation factor. Used for normalization of the first In the standardized torque sequence, the first... Each principal component corresponds to the total amplitude of the jumps between adjacent time points in the eigenvector.

[0048] Furthermore, the actual assembly stress index satisfies the following expression:

[0049] ;

[0050] In the formula, For the first The true assembly stress index of a standardized torque sequence. For the first The degree of principal component energy concentration of a standardized torque sequence The number of principal components in the standardized torque sequence. For the first The first standardized torque sequence The singular values ​​of each principal component For the first The first standardized torque sequence The energy values ​​of each principal component For the first In the standardized torque sequence, the first... The waveform oscillation factor of each principal component corresponds to the eigenvector. It is the maximum-minimum normalization function.

[0051] In the formula, The larger it is, the more likely it is to be the first The more pronounced the monotonic decay trend of the standardized torque sequence, the less it is affected by electronic thermal noise, therefore the... The smaller the true assembly stress exponent of a standardized torque sequence. The larger it is, the more likely it is to be the first In the standardized torque sequence, the first... The stronger the electromagnetic interference experienced by each principal component, the greater the... In the standardized torque sequence, the first... The more the energy value of each principal component is weakened, the more the energy value of the first principal component is reduced. The smaller the true assembly stress exponent of a standardized torque sequence.

[0052] S004: Determine the fault warning threshold of the standardized torque sequence based on the actual assembly stress index.

[0053] It should be noted that after obtaining the true assembly stress index of the standardized torque sequence, this invention will determine the fault warning threshold of the standardized torque sequence based on the true assembly stress index. Traditional fault warning methods usually set a fixed static constant as the warning threshold. This invention, however, uses the historical statistical characteristics of the true assembly stress index to determine the fault warning threshold of the standardized torque sequence. The alarm threshold can be adjusted in real time according to different working conditions, which can capture abnormal welding boundary conditions in complex industrial environments and effectively avoid false alarms and missed alarms caused by fixed warning thresholds.

[0054] Specifically, the fault warning threshold satisfies the expression:

[0055] ;

[0056] In the formula, For the first Fault warning threshold for a standardized torque sequence For the first The mean of the true assembly stress exponents of the most recent standardized torque sequences preceding a given standardized torque sequence. For the first In this embodiment, the standard deviation of the true assembly stress index of the most recent several standardized torque sequences preceding the first standardized torque sequence is selected. The standard deviation of the true assembly stress index is calculated using the 50 most recent standardized torque sequences preceding the first standardized torque sequence. In other embodiments, the implementer can set the standard deviation based on the actual implementation situation. The number of times the most recent normalized moment sequence is selected before the current normalized moment sequence. The basic safety factor is used to define the baseline width of the fault warning threshold, ensuring that the fault warning threshold has a sufficient statistical confidence interval to cover random errors. The empirical range of the basic safety factor is as follows: In this embodiment, the basic safety factor is set to 3, which corresponds to the control limit of 3 times the standard deviation in statistics. Under normal production conditions with stable welding processes, the historical data of the actual assembly stress index usually approximately follows a normal distribution, and the probability of it falling within the range of mean ± 3 times the standard deviation is about 99.73%, which almost covers all normal welding process fluctuations. This ensures high-sensitivity early warning while reducing the risk of false alarms. Even if there is a slight deviation in the data distribution, this adaptive mechanism can still effectively suppress false alarms. In other embodiments, the implementer can set the basic safety factor according to the actual implementation situation. For example, when the subsequent welding process of the micro fan motor housing has extremely high requirements for assembly accuracy and does not allow slight springback, the basic safety factor can be appropriately reduced to tighten the alarm limit. When the environmental noise at the production site is large, the basic safety factor can be appropriately increased to reduce the false alarm rate.

[0057] In the formula, The larger it is, the more likely it is to be the first The more unstable the production process is within the time frame corresponding to the 50 most recent standardized torque sequences before the first standardized torque sequence, the higher the fault warning threshold should be to prevent false alarms when the process is unstable. The larger the fault warning threshold of a standardized torque sequence, the better.

[0058] S005: Based on the actual assembly stress index and the fault warning threshold, determine whether there is a risk of springback in the micro fan motor housing, and realize the fault warning of the automatic welding equipment for micro fan motor housing.

[0059] Specifically, determining whether there is a risk of springback in the miniature fan motor housing includes:

[0060] The first The true assembly stress index of the first standardized torque sequence and the first The fault warning threshold of the standardized torque sequence is compared, in response to the fault warning threshold of the first standardized torque sequence. The true assembly stress exponent of the first standardized torque sequence is greater than that of the second. The fault warning threshold of a standardized torque sequence indicates that the assembly stress of the micro fan motor housing is too high, posing a risk of springback. The abnormal welding boundary is determined, a welding prohibition command is immediately sent, and an alarm is triggered.

[0061] like Figure 2 and Figure 3 As shown, Figure 2 The results of traditional singular spectrum analysis algorithms, which monitor based solely on fixed warning thresholds and principal component energy percentages, show that they cannot distinguish between the high-energy characteristics of strong electromagnetic interference and the true rebound trend. In both the noise interference range and the rebound fault range, they exhibit high values ​​and exceed the fixed thresholds. This confirms the modal aliasing defects of existing technologies, which cannot effectively distinguish noise interference, thus leading to false alarms in the production process. Figure 3 The results of this invention, which combines an adaptive fault warning threshold with the actual assembly stress index, are demonstrated. Thanks to the noise interference suppression mechanism, the actual assembly stress index is successfully suppressed to a low level in the noise interference range without triggering false alarms. However, in the springback fault range, the actual assembly stress index rapidly rises and exceeds the adaptive fault warning threshold. This confirms that this invention can accurately isolate strong electromagnetic interference and achieve highly robust fault warning.

Claims

1. A fault early warning method for an automatic welding equipment for miniature fan motor housings, characterized in that, include: Obtain the original torque sequence of the servo motor during the pressure holding phase, and standardize the original torque sequence to obtain the standardized torque sequence; A trajectory matrix is ​​constructed based on the standardized moment sequence. Singular value decomposition (SVD) is performed on the trajectory matrix. The energy concentration of the principal components of the standardized moment sequence is determined by the ratio between the sum of the energies of all principal components and the sum of the energies of all components, and the difference between the singular values ​​of the first and second principal components, satisfying the following conditions: ； For the first The principal component energy concentration of a standardized torque sequence ， 、 These represent the number of principal components and the total number of components in the standardized torque sequence, respectively. ， 、 The first The first standardized torque sequence Singular values ​​and energy values ​​of principal components ， 、 The first The first standardized torque sequence Singularity and energy value of each component ， 、 The first The singular values ​​of the first principal component and the singular values ​​of the second principal component of a standardized torque sequence. ， Minimum normalization function ； Based on the sum of the absolute values ​​of the numerical differences between adjacent time steps in the eigenvector corresponding to each principal component of the standardized torque sequence, the waveform oscillation factor of the eigenvector corresponding to each principal component of the standardized torque sequence is determined, satisfying: ； For the first In the standardized torque sequence, the first... Waveform oscillation factor of each principal component corresponding to the eigenvector ， For the first The length of the eigenvectors corresponding to all principal components in a standardized torque sequence ， 、 、 、 The first In the standardized torque sequence, the first... The eigenvector corresponding to the principal component is the th eigenvector. Value at each moment 、 No. The value at each moment, the maximum value among all moments, and the minimum value among all moments. ， To prevent hyperparameters with a denominator of zero; The true assembly stress index of the standardized torque sequence is determined based on the principal component energy concentration and waveform oscillation factor. ， satisfy ： ； The fault warning threshold of the standardized torque sequence is determined based on the actual assembly stress index; Based on the actual assembly stress index and fault warning threshold, it is determined whether there is a risk of springback in the micro fan motor housing, thus realizing fault warning of the automatic welding equipment for micro fan motor housing.

2. The fault early warning method for an automatic welding equipment for a miniature fan motor housing according to claim 1, characterized in that, The acquisition of the original torque sequence of the servo motor during the pressure holding stage includes: at the station where the motor housing is pressed into the shaft core, the original torque sequence of the servo motor from the point of contact with the motor housing until the pressure holding ends is acquired through the bus communication interface of the servo driver.

3. The fault early warning method for an automatic welding equipment for a miniature fan motor housing according to claim 1, characterized in that, The method for obtaining the standardized torque sequence is as follows: first, subtract the mean of the original torque sequence from the value of each data point in the original torque sequence, and then divide the result by the standard deviation of the original torque sequence to obtain the numerical sequence as the standardized torque sequence.

4. The fault early warning method for an automatic welding equipment for a miniature fan motor housing according to claim 1, characterized in that, The method of constructing a trajectory matrix based on a standardized torque sequence includes setting the embedding dimension of the trajectory matrix to one-third of the length of the standardized torque sequence.

5. The fault early warning method for an automatic welding equipment for a miniature fan motor housing according to claim 1, characterized in that, The singular value decomposition of the trajectory matrix includes: using a singular spectrum analysis algorithm to perform singular value decomposition on the trajectory matrix to obtain all components of the standardized torque sequence, and taking the three components with the largest singular values ​​among all components as the principal components of the standardized torque sequence.

6. The fault early warning method for an automatic welding equipment for a miniature fan motor housing according to claim 1, characterized in that, The fault warning threshold satisfies the expression: ； In the formula, For the first Fault warning threshold for a standardized torque sequence For the first The mean of the true assembly stress exponents of the most recent standardized torque sequences preceding a given standardized torque sequence. For the first The standard deviation of the true assembly stress index of the most recent normalized torque sequences preceding a given normalized torque sequence. Based on the basic safety factor.

7. The fault early warning method for an automatic welding equipment for a miniature fan motor housing according to claim 1, characterized in that, The determination of whether the miniature fan motor housing has a risk of rebound includes: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The true assembly stress index of the first standardized torque sequence and the first The fault warning threshold of the standardized torque sequence is compared, in response to the fault warning threshold of the first standardized torque sequence. The true assembly stress exponent of the first standardized torque sequence is greater than that of the second. The fault warning threshold of a standardized torque sequence indicates that the assembly stress of the micro fan motor housing is too high, posing a risk of springback. The abnormal welding boundary is determined, a welding prohibition command is immediately sent, and an alarm is triggered.

Images