Online quality control method and system in actuator production process

Abstract

The invention belongs to the technical field of production quality control, and particularly relates to an online quality control method and system in an actuator production process, and the method comprises the steps: obtaining a current sequence, a displacement deviation sequence and a real-time speed sequence of an actuator in a complete test period; obtaining a hysteresis coefficient of the actuator by combining the reference current parameter and the maximum stroke parameter; obtaining a viscous slip index in combination with a speed inhibition factor determined by a real-time speed sequence; extracting hysteresis features and degradation features of the target sample and the reference sample to obtain a distance adjustment weight, and correcting the Euclidean distance through the distance adjustment weight to obtain a working condition distance; and finally, obtaining a quality anomaly index according to the working condition distance, and determining an online quality control result. According to the method, the spatial distance is adjusted through the distance adjusting weight, the quality anomaly index is obtained, and the accuracy of hidden defect detection of the actuator under the complex working condition is improved.

Description

Technical Field

[0001] This invention relates to the field of production quality control technology. More specifically, this invention relates to an online quality control method and system for actuator manufacturing processes. Background Technology

[0002] As the core actuators of industrial control and automation systems, actuators play an important role in precision manufacturing, aerospace and other fields. During the processing, assembly and operation of actuators, physical problems such as the intrusion of internal foreign objects or interference fit of sealing rings often occur. These problems can lead to a sudden increase in local operating resistance and induce jamming of mechanical transmission components. Therefore, online quality control and defect detection of actuators are particularly important to ensure the safe and stable operation of industrial equipment.

[0003] In related technologies, methods such as judgment based on operating resistance threshold or density-based machine learning clustering are commonly used. These methods mainly collect resistance-related data during actuator operation to monitor the magnitude of the absolute value of resistance, or extract multi-dimensional operating data to construct features and calculate Euclidean distance in pure mathematical space, thereby achieving the classification and determination of normal and defective parts.

[0004] However, the aforementioned detection methods have specific physical limitations in the complex production line environment. On the one hand, when the ambient temperature of the production line changes globally, the thermal expansion and contraction and the changes in the physical properties of the internal lubricating medium will cause an overall shift in the basic friction characteristics of all actuator samples. Traditional clustering methods, which rely solely on Euclidean distance in pure mathematical space, cannot resist such global temperature disturbances and are prone to topological clustering separation failure. On the other hand, the typical physical phenomenon caused by interference from foreign objects inside the actuator or wear of the sealing ring is viscous slip, specifically manifested as a step characteristic of mechanical parts suddenly sliding after being stuck and storing force. This characteristic is easily induced under low-speed conditions, while the small fluctuations under high-speed conditions are mostly inertial noise of the system itself. Traditional technologies often only focus on the absolute value of resistance, ignoring the difference in the physical response of this step characteristic caused by frictional abrupt change with the operating speed gradient. This makes it difficult to effectively distinguish between normal tightness caused by assembly tolerances and viscous slip frictional abrupt changes that pose an early failure risk. Summary of the Invention

[0005] To address the technical problems of topological clustering failure caused by feature shift due to global temperature variations in the production line environment, and difficulty in distinguishing between normal tightness and viscous slip defects caused by operating speed gradient interference, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides an online quality control method for actuator manufacturing, comprising: acquiring a current sequence, a displacement deviation sequence, and a real-time speed sequence of the actuator during a complete test cycle; acquiring the hysteresis coefficient of the actuator at each sampling point based on the ratio of current data in the current sequence to a reference current parameter and the ratio of data in the displacement deviation sequence to a maximum stroke parameter; calculating the absolute value of the difference in hysteresis coefficient between the sampling point and the previous sampling point; acquiring a speed suppression factor based on the ratio of speed data in the real-time speed sequence to the maximum speed parameter; acquiring the viscous slip index of the actuator at each sampling point based on the absolute value of the difference and the speed suppression factor; and constructing a target by calculating the mean of the current sequence and the displacement deviation sequence. The feature vector is defined with the mean of the actuator hysteresis coefficient as the target hysteresis feature and the maximum of the actuator viscous slip index as the target degradation feature. The Euclidean distance between the target feature vector and the corresponding benchmark feature vector in the benchmark qualified sample library is obtained. The maximum value between the target hysteresis feature and the benchmark hysteresis feature of the sample is extracted, as is the maximum value between the target degradation feature and the benchmark degradation feature of the sample. The benchmark hysteresis feature is the mean of the hysteresis coefficient of the sample, and the benchmark degradation feature is the maximum value of the viscous slip index of the sample. Distance adjustment weights are obtained based on these two maximum values. The operating distance is obtained according to the Euclidean distance and the distance adjustment weights. The quality anomaly index of the actuator is obtained based on the operating distance, and online quality control of the actuator is performed.

[0007] This invention obtains the hysteresis coefficient by comparing the current data in the current sequence with the reference current parameter, and the deviation data in the displacement deviation sequence with the maximum stroke parameter. This captures the extra energy consumed by the actuator to overcome abnormal friction, thus providing positive amplification excitation in both electrical and mechanical dimensions during the early stages of dry friction or foreign object interference within the actuator, and identifying the risk of internal damping abrupt changes in the actuator. Furthermore, this invention obtains the viscous slip index by combining the absolute value of the difference between the hysteresis coefficient at the current sampling point and the hysteresis coefficient at the previous sampling point with a velocity suppression factor determined by the real-time velocity sequence, suppressing system slippage generated under high-speed conditions. Inertial noise amplifies the characteristics of components jamming and suddenly sliding under low-speed conditions, thus distinguishing between normal tightness caused by assembly tolerances and viscous slippage with early failure risk in complex operating environments. This invention extracts target hysteresis features and target degradation features, adjusts the Euclidean distance between the target feature vector and the reference feature vector to obtain the operating condition distance, so that defective samples with abnormal mechanical damping or viscous slippage degradation are pushed away from the distribution center of qualified samples. Thus, even when the sample features shift as a whole due to environmental temperature disturbances, the clustering and separation effect is maintained, and the accurate interception of actuator defects is achieved.

[0008] Preferably, the acquisition of the current sequence, displacement deviation sequence, and real-time speed sequence of the actuator within a complete test cycle includes: acquiring the command displacement sequence issued by the control system and acquiring the actual displacement sequence of the actuator using a displacement sensor; calculating the absolute value of the displacement difference between the command displacement sequence and the actual displacement sequence at the corresponding sampling time to construct a displacement deviation sequence; calculating the absolute value of the difference between the actual displacement data corresponding to the current sampling point and the immediately preceding sampling point in the actual displacement sequence, and dividing the absolute value of the difference by the sampling time interval to obtain the real-time speed sequence and the current sequence.

[0009] Preferably, before obtaining the hysteresis coefficient of the actuator at each sampling point based on the ratio of the current data in the current sequence to the reference current parameter and the ratio of the data in the displacement deviation sequence to the maximum stroke parameter, the method further includes: controlling the actuator to perform full-stroke no-load reciprocating motion, recording the difference between the maximum displacement extreme value and the minimum displacement extreme value output by the displacement sensor, and using the difference as the maximum stroke parameter; recording the maximum safe current allowed in the calibration process, and using the maximum safe current as the reference current parameter.

[0010] Preferably, the hysteresis coefficient satisfies the following relationship: In the formula, For the first The hysteresis coefficient at each sampling point The first in the current sequence Current data at each sampling point As the reference current parameter, The first in the displacement deviation sequence Deviation data for each sampling point This is the maximum travel parameter. For the sigmoid function, This is the sensitivity gain factor.

[0011] This invention provides amplified excitation for sudden increases in current resistance and displacement hysteresis phenomena in both electrical and mechanical dimensions, enabling the obtained hysteresis coefficient to respond more sensitively to minute changes in internal resistance, thus highlighting the potential abnormal friction risk of the actuator.

[0012] Preferably, the viscous slip index satisfies the following relationship: In the formula, For the first Viscous slip index at each sampling point For the first The hysteresis coefficient at each sampling point For the first The hysteresis coefficient at each sampling point The first in the real-time velocity sequence The speed of each sampling point For maximum speed parameters, It is an exponential function with the natural constant as its base. This is the speed modulation coefficient.

[0013] This invention characterizes the intensity of resistance abrupt change at adjacent sampling times by calculating the absolute value of the difference between the hysteresis coefficient of the current sampling point and the hysteresis coefficient of the previous sampling point, and extracts the ratio of the absolute value of the speed to the maximum speed parameter in the real-time speed sequence to construct a speed suppression factor, thereby distinguishing the resistance response under different operating speeds and extracting the feature of sudden sliding after the component jams and stores power under low-speed conditions.

[0014] Preferably, obtaining the distance adjustment weight includes: extracting the maximum value between the target hysteresis feature and the benchmark hysteresis feature corresponding to the sample, and the maximum value between the target degradation feature and the benchmark degradation feature corresponding to the sample; and obtaining the distance adjustment weight based on the sum of the two extracted maximum values.

[0015] Preferably, the working distance satisfies the following relationship: In the formula, For the actuator and the first in the benchmark qualified sample library The working condition distance between samples The target feature vector and the first qualified sample in the benchmark sample library The Euclidean distance between the baseline feature vectors of each sample. It is an exponential function with the natural constant as its base. It is a function with maximum value. The target is characterized by hysteresis. The first in the benchmark qualified sample library The baseline hysteresis features corresponding to each sample For the target degradation characteristics, The first in the benchmark qualified sample library The baseline degradation features corresponding to each sample This is the static distance expansion coefficient. This is the dynamic distance expansion coefficient.

[0016] This invention measures the maximum deterioration of the actuator and the benchmark qualified sample in terms of static hysteresis and degradation by extracting the maximum value between the target hysteresis feature and the benchmark hysteresis feature, and by extracting the maximum value between the target degradation feature and the benchmark degradation feature. When one of the two samples involved in the calculation exhibits abnormal mechanical resistance or viscous slip degradation, the distance adjustment weight can be used to correct the Euclidean distance upwards, thus actively pushing the sample with defect risk away from the center of the qualified cluster distribution in the sample distribution space. When the mechanical smoothness of both samples involved in the calculation remains good and there is no abrupt friction, the distance adjustment weight obtained by calculation approaches 1, thereby maintaining the original cluster distribution state, resisting feature drift caused by environmental temperature disturbances, and improving the accuracy of topological separation of abnormal samples.

[0017] Preferably, obtaining the quality anomaly index of the actuator based on the working condition distance includes: selecting a preset number of samples from the benchmark qualified sample library that are closest to the actuator in ascending order of working condition distance between the actuator and the samples in the benchmark qualified sample library to construct a nearest neighbor set; and using the mean of the working condition distances from the actuator to all samples in the nearest neighbor set as the quality anomaly index.

[0018] This invention uses the average working distance from the actuator to all samples in the nearest neighbor set as the quality anomaly index. It uses the average spatial distance within the local nearest neighbor range to reflect the relative density distribution of the actuator in the global sample space, smoothing out accidental distance jumps and improving the tolerance of anomaly judgment to local data fluctuations, thereby more stably reflecting the overall quality deviation of the actuator.

[0019] Preferably, the online quality control of the actuator includes: in response to a quality anomaly index being greater than or equal to a preset rejection threshold, determining that the actuator has an internal jamming or damping defect, and outputting an abnormal level signal to the production line control system to trigger a rejection action; in response to a quality anomaly index being less than the preset rejection threshold, determining that the actuator is a qualified product.

[0020] Secondly, the present invention provides an online quality control system for the actuator manufacturing process, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned online quality control method for the actuator manufacturing process is implemented.

[0021] By adopting the above technical solution, the online quality control method in the actuator production process described above is generated into a computer program and stored in a memory for loading and execution by a processor. This allows for the creation of a terminal device based on the memory and processor, facilitating its use.

[0022] The beneficial effects of this invention are as follows: By synchronously acquiring the current sequence, displacement deviation sequence, and real-time speed sequence of the actuator within a complete test cycle, and combining the electrical load response and mechanical motion feedback to extract the hysteresis coefficient and viscous slip index, the invention eliminates the hidden frictional resistance surge caused by interference from foreign objects inside the actuator or excessive wear of the sealing ring, and the physical step phenomenon of sudden sliding after mechanical parts are stuck and stored. Risk can be locked in the early stage when minor wear occurs inside the actuator. Simultaneously, the Euclidean distance between the target feature vector and the benchmark feature vector corresponding to the sample in the benchmark qualified sample library is modulated, so that actuator samples with mechanical damping abnormalities or viscous slip defects are actively pushed away from the center of the qualified cluster in the feature space. This effectively offsets the phenomenon of overall drift of basic operating resistance caused by changes in lubricating oil viscosity due to diurnal temperature variations in the production line environment, thereby improving the high robustness of online quality detection of actuators. This invention extracts the relative density distribution features in the global sample space based on the working condition distance to obtain the quality anomaly index, reducing the local misjudgment interference caused by isolated distance values. This invention triggers production line shutdown for re-inspection in advance before batch defective products are generated, reducing the scrap rate of industrial production and ensuring the safe and reliable operation of core execution components of industrial automation equipment. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an online quality control method in the actuator manufacturing process according to the present invention; Figure 2 This is a schematic diagram showing the comparison of the viscosity slip index between normal and abnormal samples in this invention. Detailed Implementation

[0024] 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.

[0025] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] This invention discloses an online quality control method in the actuator manufacturing process, referring to... Figure 1 This includes steps S1-S5: S1. The current sequence, displacement deviation sequence, and real-time speed sequence of the actuator are acquired through the sensors on the test bench during a complete test cycle.

[0027] Specifically, during actuator operation, the command displacement sequence issued by the control system is acquired, and the real-time current sequence is obtained using a current transformer equipped on the test bench. The actual displacement sequence of the actuator is acquired using a displacement sensor, and the absolute value of the displacement difference between the command displacement sequence and the actual displacement sequence at the corresponding sampling time is calculated to construct a displacement deviation sequence. The absolute value of the difference between the actual displacement data corresponding to the current sampling point and the immediately preceding sampling point in the actual displacement sequence is calculated, and this difference is divided by the sampling time interval to obtain the real-time velocity sequence.

[0028] S2. Based on the current sequence and displacement deviation sequence, combined with the maximum stroke parameter and the reference current parameter, obtain the hysteresis coefficient of the actuator at each sampling point.

[0029] It should be noted that the intrusion of foreign objects into the actuator or the interference fit of the sealing ring can cause a sudden increase in local operating resistance. This change in resistance will simultaneously cause an increase in the drive current and a lag in the actual displacement relative to the commanded displacement. Therefore, this invention obtains the hysteresis coefficient of the actuator at each sampling point based on the current sequence and displacement deviation sequence, combined with the maximum stroke parameter and the reference current parameter.

[0030] Specifically, the actuator is controlled to perform a full-stroke no-load reciprocating motion, and the difference between the maximum and minimum displacement extreme values ​​output by the displacement sensor is recorded. This difference is used as the maximum stroke parameter. Simultaneously, the maximum allowable safe current during the calibration process is recorded, and this maximum safe current is used as the reference current parameter. The relative ratio of current load is determined based on the current data in the current sequence and the reference current parameter. The relative ratio of displacement hysteresis is determined based on the deviation data in the displacement deviation sequence and the maximum stroke parameter. Combining the relative ratio of current load and the relative ratio of displacement hysteresis, the hysteresis coefficient is obtained.

[0031] Specifically, the hysteresis coefficient satisfies the following relationship: ; In the formula, For the first The hysteresis coefficient at each sampling point The first in the current sequence Current data at each sampling point As the reference current parameter, The first in the displacement deviation sequence Deviation data for each sampling point This is the maximum travel parameter. For the sigmoid function, This is the sensitivity gain factor. The empirical value range is [2,5]. In this embodiment... The value is 3.5, which can be determined by the implementers based on the actual situation. When the production line requires strict control over the minute wear of the actuator, this parameter can be appropriately increased to improve the amplification factor of the nonlinear mapping segment; when the electromagnetic noise in the sensor environment is large, this parameter can be appropriately decreased to prevent the noise disturbance from being excessively amplified.

[0032] in, and These represent the relative ratio of current load and the relative ratio of displacement hysteresis, respectively.

[0033] The larger the value, the greater the possibility of dry friction or foreign object interference inside the actuator, thus providing a positive amplification excitation to the final hysteresis coefficient in the electrical dimension; The smaller the value, the greater the likelihood that the actuator is well-lubricated and has no additional damping, thus attenuating the hysteresis coefficient in an electrical dimension. The larger the value, the greater the possibility that the mechanical transmission components of the actuator may jam, resulting in a slow response, thus providing a positive amplification excitation to the hysteresis coefficient in the mechanical dimension; the smaller the value, the greater the possibility that the mechanical transmission components of the actuator may follow the control command more closely, thus attenuating the hysteresis coefficient in the mechanical dimension.

[0034] It represents the transient lag work intensity of the actuator at a single point in time. The larger the value, the higher the extra energy consumed by the actuator to overcome abnormal friction, resulting in a hysteresis coefficient that is closer to 1, thereby amplifying the defect response. The smaller the value, the greater the likelihood that the actuator will follow smoothly, resulting in a hysteresis coefficient that is closer to 0, thus suppressing normal operating noise.

[0035] S3. Based on the hysteresis coefficient and real-time velocity sequence, combined with the maximum velocity parameter, obtain the viscous slip index of the actuator at each sampling point.

[0036] It should be noted that the typical physical phenomenon caused by foreign objects inside the actuator or wear of the sealing ring is viscous slip, specifically manifested as a mechanical component suddenly sliding after being jammed and storing force. Existing detection methods often only focus on the absolute value of the resistance, ignoring the step characteristics caused by abrupt changes in friction, making it difficult to distinguish between normal tightness caused by assembly tolerances and abrupt changes in friction that pose an early risk of failure. In addition, viscous slip is easily induced under low-speed conditions, while the small fluctuations under high-speed conditions are mostly system inertial noise. Therefore, this invention obtains the viscous slip index of the actuator at each sampling point based on the hysteresis coefficient and real-time speed sequence, combined with the maximum speed parameter, in order to extract the adaptive abrupt change characteristics of the operating conditions.

[0037] Specifically, the maximum speed parameter of the actuator preset in the controller is retrieved, and the first speed parameter in the real-time speed sequence is extracted. The absolute value of the velocity data at the nth sampling point is divided by the maximum velocity parameter to obtain the relative velocity coefficient. The calculation of the nth... The hysteresis coefficient of the sampling point and the first sampling point The absolute value of the difference in hysteresis coefficients at each sampling point is used to obtain the first-order time-domain difference of the hysteresis coefficient. This first-order time-domain difference is then combined with the relative velocity coefficient to obtain the viscous slip index.

[0038] Specifically, the viscous slip index satisfies the following relationship: ; In the formula, For the first Viscous slip index at each sampling point For the first The hysteresis coefficient at each sampling point For the first The hysteresis coefficient at each sampling point The first in the real-time velocity sequence The speed of each sampling point For maximum speed parameters, It is an exponential function with the natural constant as its base. For speed modulation coefficients, The empirical value range is [1,3]. In this embodiment... The value is 2, and the implementers can determine the number based on the actual situation. When the rated operating speed of the actuator is low, this parameter can be appropriately increased to enhance the amplification effect of sudden changes in the low-speed domain; when the system has continuous high-frequency mechanical vibration, this parameter can be appropriately decreased to reduce the sensitivity to high-frequency random jitter.

[0039] in, This represents the intensity of the sudden change in resistance between adjacent sampling times. The larger the value, the greater the possibility of a sudden and drastic change in friction, resulting in a larger viscous slip index; the smaller the value, the less likely the possibility of a sudden and drastic change in friction, resulting in a smaller viscous slip index.

[0040] This represents the velocity inhibition factor; the lower the real-time velocity, the greater the likelihood of fatal viscous slippage in the actuator. The larger the value, the more it amplifies the correction effect on the intensity of sudden changes in resistance; the higher the real-time speed, the greater the likelihood that the fluctuation originates from system inertial noise. The closer it is to 1, the more it corrects and suppresses the intensity of sudden changes in resistance.

[0041] For example, Figure 2This is a comparison chart of the viscous slip index of normal and abnormal samples in this invention. As can be seen from the chart, the data curve of the normal sample remains stable within a low value range close to 0, indicating smooth actuator operation and effective suppression of normal background noise. In contrast, the abnormal sample's data curve shows a sharp, step-like increase within the 4-4.5 second range, with the peak value significantly higher than the normal background baseline. The viscous slip index extracted in this invention can pinpoint and amplify the subtle local jamming and abrupt friction characteristics within the actuator. While filtering out interference from normal operation, it also enhances the distinguishability between normal tolerances and early failure risks, providing reliable data support for subsequent highly robust online quality assessment.

[0042] S4. Based on the hysteresis coefficient, viscous slip index and target feature vector, and combined with the benchmark qualified sample library, obtain the working distance of the actuator.

[0043] It should be noted that density-based machine learning algorithms for quality classification typically rely on Euclidean distance in a purely mathematical space. When the production line environment temperature undergoes global changes, the features of all samples shift overall, and conventional methods are prone to topological clustering failure due to global environmental disturbances. Therefore, this invention does not operate directly in the pure mathematical space. Instead, it utilizes the aforementioned static hysteresis and dynamic mutation intensity to perform dual nonlinear modulation of the mathematical distance, thereby actively pushing away samples with inherent defect risks from the topological space.

[0044] Specifically, the mean values ​​of the current sequence and displacement deviation sequence of the current actuator within a complete test cycle are statistically analyzed to construct a target feature vector. A pre-trained benchmark qualified sample library stored in the database is retrieved; this library contains benchmark feature vectors of historically qualified actuators. The mean value of the hysteresis coefficients at all sampling points of the actuator is calculated to determine the target hysteresis feature, and the maximum value of the viscous slip index at all sampling points of the actuator is extracted to determine the target degradation feature. Similarly, the mean value of the hysteresis coefficients of the samples in the benchmark qualified sample library is calculated as the benchmark hysteresis feature, and the maximum value of the viscous slip index of the samples is extracted as the benchmark degradation feature.

[0045] Based on the target feature vector, target hysteresis feature, and target degradation feature, as well as the benchmark feature vector, benchmark hysteresis feature, and benchmark degradation feature corresponding to the samples in the benchmark qualified sample library, the working condition distance between the actuator and the samples in the benchmark qualified sample library is obtained.

[0046] Specifically, the working distance satisfies the following relationship: ; In the formula, For the actuator and the first in the benchmark qualified sample library The working condition distance between samples The target feature vector and the first qualified sample in the benchmark sample library The Euclidean distance between the baseline feature vectors of each sample. It is an exponential function with the natural constant as its base. It is a function with maximum value. The target is characterized by hysteresis. The first in the benchmark qualified sample library The baseline hysteresis features corresponding to each sample For the target degradation characteristics, The first in the benchmark qualified sample library The baseline degradation features corresponding to each sample This is the static distance expansion coefficient. This is the dynamic distance expansion coefficient. The empirical value range is [0.5, 2]. The empirical value range is [1, 3.5]. In this embodiment... It is 1.2. The value is 2, and the implementers can determine the number based on the actual situation. and When the distribution of qualified samples is relatively concentrated, the amount can be appropriately reduced. To prevent normal samples from becoming detached; when it is necessary to strictly remove products with edge-critical defects, the size can be appropriately increased. To enhance the separation effect; when the production line has extremely low tolerance for sudden jamming defects, the separation capacity can be appropriately increased. The distance to the polarization anomaly sample; when the actuator is operating normally and there are intermittent physical setbacks within the allowable range, the distance can be appropriately reduced. This is to prevent over-inspection.

[0047] in, and These represent the maximum physical degradation of the actuator and the benchmark qualified sample in terms of static hysteresis and dynamic deterioration, respectively. Since in topological clustering algorithms, if either party involved in the distance calculation exhibits significant mechanical performance degradation, their topological association should be weakened, a maximum value function is used to extract the maximum value of the features of both parties for calculating the working condition distance. This represents the distance-adjusted weight based on the deterioration state. A larger value indicates a greater likelihood that at least one of the two samples involved in the calculation exhibits significant mechanical resistance anomalies or viscous slip deterioration, leading to a greater influence on its distance adjustment. The greater the upward correction, the faster the target sample is pushed away from the center of the qualified cluster in the feature space; the smaller the value of this term, the greater the possibility that the mechanical smoothness of both sides is well maintained and there is no sudden friction, causing the exponential function term to approach 1, thereby maintaining its original topological clustering state.

[0048] S5. Obtain the quality anomaly index of the actuator based on the working condition distance, and determine the online quality control result of the actuator.

[0049] It should be noted that isolated distance values ​​are insufficient to reflect the density distribution of data in the global sample space. To achieve highly robust anomaly detection, this invention obtains the actuator's quality anomaly index based on the operating condition distance and determines the actuator's online quality control results.

[0050] Specifically, the front actuator closest to the working condition is selected in ascending order of working condition distance. A nearest neighbor set is constructed from samples in a baseline qualified sample library. The mean operating distance from the actuator to all samples in the nearest neighbor set is calculated, and this mean is labeled as the quality anomaly index.

[0051] In one embodiment, in response to a quality anomaly index being greater than or equal to a preset rejection threshold, it is determined that the actuator has an internal jamming or damping defect, and an abnormal level signal is output to the production line control system to trigger a rejection action; in response to a quality anomaly index being less than the preset rejection threshold, the actuator is determined to be a qualified product.

[0052] In another embodiment, the construction includes continuous Historical detection queues for each test period, statistically analyzing consecutive... The moving variance of the quality anomaly index within a test cycle, in response to the continuous increase of the moving variance and its being greater than or equal to the preset trend alarm threshold, indicates a greater possibility of systematic deterioration in the processing technology of the same batch of actuators, and triggers a production line shutdown and re-inspection warning instruction.

[0053] For example, The value is 10. The value is set to 50. The elimination threshold and trend alarm threshold are set by the implementers based on the actual yield tolerance.

[0054] This invention also discloses an online quality control system for actuator manufacturing, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an online quality control method for actuator manufacturing according to the present invention.

[0055] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. An online quality control method for actuator manufacturing process, characterized in that, include: Acquire the current sequence, displacement deviation sequence, and real-time velocity sequence of the actuator during the complete test cycle; The hysteresis coefficient of the actuator at each sampling point is obtained by comparing the current data in the current sequence with the reference current parameter and the data in the displacement deviation sequence with the maximum stroke parameter. Calculate the absolute value of the difference in hysteresis coefficient between the sampling point and the previous sampling point, obtain the velocity suppression factor based on the ratio of velocity data in the real-time velocity sequence to the maximum velocity parameter, and obtain the viscous slip index of the actuator at each sampling point based on the absolute value of the difference and the velocity suppression factor. The mean values ​​of the current sequence and the displacement deviation sequence are calculated to construct the target feature vector. The mean value of the actuator hysteresis coefficient is used as the target hysteresis feature, and the maximum value of the actuator viscous slip index is used as the target degradation feature. Obtain the Euclidean distance between the target feature vector and the corresponding benchmark feature vector of the sample in the benchmark qualified sample library; extract the maximum value between the target hysteresis feature and the benchmark hysteresis feature of the sample; and extract the maximum value between the target degradation feature and the benchmark degradation feature of the sample. The benchmark hysteresis feature is the mean value of the hysteresis coefficient of the sample, and the benchmark degradation feature is the maximum value of the viscous slip index of the sample. Based on the above two maximum values, obtain the distance adjustment weight; obtain the working condition distance according to the Euclidean distance and the distance adjustment weight; obtain the quality anomaly index of the actuator according to the working condition distance, and perform online quality control of the actuator.

2. The online quality control method for the actuator manufacturing process according to claim 1, characterized in that, The acquisition of the actuator's current sequence, displacement deviation sequence, and real-time velocity sequence within a complete test cycle includes: acquiring the command displacement sequence issued by the control system and acquiring the actual displacement sequence of the actuator using a displacement sensor; calculating the absolute value of the displacement difference between the command displacement sequence and the actual displacement sequence at the corresponding sampling time to construct a displacement deviation sequence; calculating the absolute value of the difference between the actual displacement data corresponding to the current sampling point and the immediately preceding sampling point in the actual displacement sequence, and dividing the absolute value of the difference by the sampling time interval to obtain the real-time velocity sequence and the current sequence.

3. The online quality control method for the actuator manufacturing process according to claim 1, characterized in that, Before obtaining the hysteresis coefficient of the actuator at each sampling point based on the ratio of the current data in the current sequence to the reference current parameter and the ratio of the data in the displacement deviation sequence to the maximum stroke parameter, the method further includes: controlling the actuator to perform full-stroke no-load reciprocating motion, recording the difference between the maximum displacement extreme value and the minimum displacement extreme value output by the displacement sensor, and using the difference as the maximum stroke parameter; recording the maximum safe current allowed in the calibration process, and using the maximum safe current as the reference current parameter.

4. The online quality control method for the actuator manufacturing process according to claim 1, characterized in that, The hysteresis coefficient satisfies the following relationship: ； In the formula, For the first The hysteresis coefficient at each sampling point The first in the current sequence Current data at each sampling point As the reference current parameter, The first in the displacement deviation sequence Deviation data for each sampling point This is the maximum travel parameter. For the sigmoid function, This is the sensitivity gain factor.

5. The online quality control method for the actuator manufacturing process according to claim 1, characterized in that, The viscous slip index satisfies the following relationship: ； In the formula, For the first Viscous slip index at each sampling point For the first The hysteresis coefficient at each sampling point For the first The hysteresis coefficient at each sampling point The first in the real-time velocity sequence The speed of each sampling point For maximum speed parameters, It is an exponential function with the natural constant as its base. This is the speed modulation coefficient.

6. The online quality control method for the actuator manufacturing process according to claim 1, characterized in that, The method for obtaining the distance adjustment weight includes: extracting the maximum value between the target hysteresis feature and the baseline hysteresis feature corresponding to the sample, and the maximum value between the target degradation feature and the baseline degradation feature corresponding to the sample; and obtaining the distance adjustment weight based on the sum of the two extracted maximum values.

7. The online quality control method for the actuator manufacturing process according to claim 1, characterized in that, The working distance satisfies the following relationship: ； In the formula, For the actuator and the first in the benchmark qualified sample library The working condition distance between samples The target feature vector and the first qualified sample in the benchmark sample library The Euclidean distance between the baseline feature vectors of each sample. It is an exponential function with the natural constant as its base. It is a function with maximum value. The target is characterized by hysteresis. The first in the benchmark qualified sample library The baseline hysteresis features corresponding to each sample For the target degradation characteristics, The first in the benchmark qualified sample library The baseline degradation features corresponding to each sample This is the static distance expansion coefficient. This is the dynamic distance expansion coefficient.

8. The online quality control method for the actuator manufacturing process according to claim 1, characterized in that, The process of obtaining the quality anomaly index of the actuator based on the working condition distance includes: selecting a preset number of samples from the benchmark qualified sample library that are closest to the actuator in ascending order of working condition distance between the actuator and the samples in the benchmark qualified sample library to construct a nearest neighbor set; and using the mean of the working condition distances from the actuator to all samples in the nearest neighbor set as the quality anomaly index.

9. An online quality control method for actuator manufacturing process according to claim 1 or 8, characterized in that, The online quality control of the actuator includes: in response to a quality anomaly index being greater than or equal to a preset rejection threshold, determining that the actuator has an internal jamming or damping defect, and outputting an abnormal level signal to the production line control system to trigger a rejection action; in response to a quality anomaly index being less than the preset rejection threshold, determining that the actuator is a qualified product.

10. An online quality control system for actuator manufacturing process, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement an online quality control method for an actuator manufacturing process according to any one of claims 1-9.

Images