Automatic motor assembly line and method thereof

Abstract

The invention discloses an automatic motor assembly line and a method thereof, and relates to the field of intelligent assembly.The method comprises the steps that firstly, a wire penetrates through an iron core hole site to obtain a stator after threading, secondly, the stator after threading is pressed into a shell through an iron core pressing assembly to form a stator assembly, then, the stator assembly and a rotor are subjected to quality detection and screening, and finally, the stator assembly and the rotor are assembled. The method comprises the following steps of: firstly, assembling a stator and a rotor which are qualified in quality inspection to obtain an assembled motor, and finally, packaging the assembled motor to form a motor finished product. In the process, high-precision and high-efficiency automatic assembly is realized, and the quality and the reliability of the motor are ensured. Therefore, the problems of low precision, low efficiency and unstable quality existing in traditional manual assembly are solved, automatic assembly is realized, and the stability and consistency of production are ensured.

Description

Technical Field

[0001] The present application relates to the field of intelligent assembly, and more specifically, to an automatic motor assembly line and method thereof. Background Art

[0002] Motors are widely used in various mechanical and electronic equipment in modern industry, and their performance directly affects the operating efficiency and reliability of the equipment. Motors are not only the core components of drive systems, but also play a key role in automated production lines, robots, household appliances, transportation and other fields. Efficient motors can significantly reduce energy consumption and improve production efficiency.

[0003] Chinese patent CN107186472B proposes an assembly line and method for a range hood fan motor, which starts with manually threading the iron core, and goes through multiple steps such as pressing the iron core, manually tapping the terminals, and testing the static performance of the stator, before finally completing the motor assembly, tapping the ground wire, and packaging.

[0004] The assembly of the motors in the above patents is mostly done manually. However, traditional manual assembly relies on the experience and skills of the operator, which makes it difficult to achieve precise control and easily leads to unstable product quality. In addition, operators are prone to fatigue, low efficiency and high error rate in a large number of repetitive tasks. In addition, it is difficult for manual labor to monitor and adjust production parameters in real time, and the level of intelligence is low, which affects the overall production efficiency and product quality. For example, in the process of pressing the iron core, it is difficult for manual labor to accurately control the pressure, which easily leads to the stator not being fully pressed into the shell or damaging the shell during the pressing process.

[0005] Therefore, an automated motor assembly solution is desired. Summary of the invention

[0006] In order to solve the above technical problems, this application is proposed.

[0007] According to one aspect of the present application, a method for automating the assembly of a motor is provided, which comprises: passing a wire through an iron core hole to obtain a threaded stator; using a pressed iron core assembly to press the threaded stator into a housing to obtain a stator assembly; performing quality inspection and screening on the stator assembly and the rotor to obtain a quality-qualified stator and a quality-qualified rotor; assembling the quality-qualified stator and the quality-qualified rotor to obtain an assembled motor; packaging the assembled motor to obtain a finished motor; wherein, using a pressed iron core assembly to press the threaded stator into a housing to obtain a stator assembly, the method comprises: obtaining a time queue of pressure data collected by a pressure sensor deployed on the pressed iron core assembly; performing sequence encoding on the time queue of the pressure data to obtain a sequence of local temporal correlation features of pressure, and calculating Calculate the steady-state information of the sequence of the pressure local timing association characteristics to obtain the pressure local timing mean characteristics; perform dynamic propagation based on node feature perception on the sequence of the pressure local timing association characteristics to obtain the pressure timing dynamic aggregation representation; extract search features from the sequence of the pressure local timing association characteristics, and calculate the response features between the pressure local timing mean characteristics and the pressure timing dynamic aggregation representation and the search features respectively to obtain the pressure timing baseline response association characteristics and the pressure timing dynamic response association characteristics; calculate the pressure timing search reasoning multi-dimensional representation between the pressure timing baseline response association characteristics and the pressure timing dynamic response association characteristics, and generate a control result based on the pressure timing search reasoning multi-dimensional representation, and the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder.

[0008] According to another aspect of the present application, an automated motor assembly line is provided, which includes: a threading device for passing a wire through a core hole to obtain a threaded stator; a core pressing device for pressing the threaded stator into a housing using a core pressing assembly to obtain a stator assembly; a detection device for performing quality detection and screening on the stator assembly and the rotor to obtain a quality-qualified stator and a quality-qualified rotor; an assembly device for assembling the quality-qualified stator and the quality-qualified rotor to obtain an assembled motor; a packaging device for packaging the assembled motor to obtain a finished motor; wherein the core pressing device includes: a pressure data acquisition module for acquiring a time queue of pressure data collected by a pressure sensor deployed on the core pressing assembly; a pressure data processing module for performing sequence encoding on the time queue of the pressure data to obtain a local temporal correlation feature of the pressure sequence, and calculate the steady-state information of the sequence of the pressure local timing correlation feature to obtain the pressure local timing mean feature; a pressure timing propagation module, used to perform dynamic propagation of the sequence of the pressure local timing correlation feature based on node feature perception to obtain the pressure timing dynamic aggregation representation; a pressure timing response module, used to extract search features from the sequence of the pressure local timing correlation feature, and respectively calculate the response features between the pressure local timing mean feature and the pressure timing dynamic aggregation representation and the search features to obtain the pressure timing baseline response correlation feature and the pressure timing dynamic response correlation feature; a control module, used to calculate the pressure timing search reasoning multi-dimensional representation between the pressure timing baseline response correlation feature and the pressure timing dynamic response correlation feature, and generate a control result based on the pressure timing search reasoning multi-dimensional representation, and the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder.

[0009] Compared with the prior art, the present application provides an automatic motor assembly line and method thereof, which collects the time queue of pressure data through the pressure sensor, and uses the data analysis and encoding method based on deep learning to perform time series encoding on the pressure data, and at the same time calculates the mean of each pressure local time series correlation feature as a benchmark, and then performs time series dynamic propagation aggregation on each pressure local time series correlation feature, and then calculates the response between the pressure local time series correlation feature and the mean feature and the aggregation feature at the last position, so as to intelligently judge whether to stop the downward pressure of the downward pressure cylinder according to the multi-dimensional characterization between the pressure time series benchmark response correlation feature and the pressure time series dynamic response correlation feature. In this way, the pressure change can be processed and analyzed in real time, eliminating the response delay problem in the traditional threshold mediation method. And the downward pressure can be dynamically adjusted to avoid the stator being pressed too deep or the shell being damaged, ensuring the accuracy of pressure control. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0011] Figure 1 Flowchart of a method for automatic assembly of a motor according to an embodiment of the present application.

[0012] Figure 2 The present invention is a flow chart of using a core pressing assembly to press the threaded stator into a housing to obtain a stator assembly in an automated motor assembly method according to an embodiment of the present application.

[0013] Figure 3 The data flow diagram of the stator assembly is shown in FIG. 1 , in which the threaded stator is pressed into a housing by using a pressing core assembly in the motor automated assembly method according to an embodiment of the present application to obtain the stator assembly.

[0014] Figure 4 4 is a block diagram of an automated motor assembly line according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0016] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0017] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0018] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0019] Motors are widely used in various mechanical and electronic equipment in modern industry, and their performance directly affects the operating efficiency and reliability of the equipment. Motors are not only the core components of drive systems, but also play a key role in automated production lines, robots, household appliances, transportation and other fields. Efficient motors can significantly reduce energy consumption and improve production efficiency.

[0020] Chinese patent CN107186472B proposes an assembly line and method for a range hood fan motor, which starts with manually threading the iron core, and goes through multiple steps such as pressing the iron core, manually tapping the terminals, and testing the static performance of the stator, before finally completing the motor assembly, tapping the ground wire, and packaging.

[0021] The assembly of the motors in the above patents is mostly done manually. However, traditional manual assembly relies on the experience and skills of the operator, which makes it difficult to achieve precise control and easily leads to unstable product quality. In addition, operators are prone to fatigue, low efficiency and high error rate in a large number of repetitive tasks. In addition, it is difficult for manual labor to monitor and adjust production parameters in real time, and the level of intelligence is low, which affects the overall production efficiency and product quality. For example, in the process of pressing the iron core, it is difficult for manual labor to accurately control the pressure, which easily leads to the stator not being fully pressed into the shell or damaging the shell during the pressing process.

[0022] Based on this, the present application proposes a method for automated assembly of motors, which effectively solves the problems of low precision, low efficiency and unstable quality existing in traditional manual assembly, realizes automated assembly, and ensures the stability and consistency of production. Figure 1 FIG. 1 is a flow chart of a method for automating the assembly of a motor according to an embodiment of the present application. Figure 1 As shown, the motor automated assembly method specifically includes: S110, passing the wire through the core hole to obtain a stator after threading; S120, using a core pressing assembly to press the stator after threading into a housing to obtain a stator assembly; S130, performing quality inspection and screening on the stator assembly and the rotor to obtain a quality-inspected stator and a quality-inspected rotor; S140, assembling the quality-inspected stator and the quality-inspected rotor to obtain an assembled motor; S150, packaging the assembled motor to obtain a finished motor.

[0023] Specifically, in step S110, the wire is passed through the core hole to obtain the stator after threading. Specifically, this step is operated by an automatic threading system, which is mainly composed of a winding machine, a control system, a wire supply device, a tension regulator and an insulation treatment device. The winding machine is the core equipment, which is responsible for winding the wire evenly and neatly in the slot of the core. The control system includes a computer, a PLC (programmable logic controller) and a sensor for monitoring and controlling the entire threading process. The wire supply device provides a continuous supply of wire to ensure that the winding process is uninterrupted. The tension regulator adjusts the tension of the wire to ensure that the wire is not too tight or too loose during the winding process. The insulation treatment device lays insulating paper in the core slot to prevent the wire from directly contacting the core and avoiding short circuit.

[0024] In actual operation, first of all, it is necessary to prepare the required materials such as wires, iron cores and insulating paper, and ensure that the winding machine, control system and other auxiliary equipment are in good condition. The workbench should be kept clean and tidy, and ensure that there is enough space for operation. Next, use a detergent and a brush to remove the oil and dust on the surface of the iron core, and lay a layer of insulating paper in the slot of the iron core to prevent the wire from directly contacting the iron core and avoid short circuit. Then, according to the design requirements of the motor, set the parameters of the winding machine, such as winding speed, tension, etc. Fix the wire at the starting point of the winding machine and start the winding machine for automatic threading. The winding machine winds the wire evenly and neatly in the slot of the iron core through the preset path and speed. The operator monitors the winding process in real time through the control system to ensure that the wire is evenly and neatly wound on the iron core. If any abnormality is found, the parameters can be adjusted in time or the winding machine can be paused for inspection. After the winding is completed, the two ends of the wire are led out of the iron core, leaving enough length for subsequent welding or connection. Insulate again at the lead-out wire to ensure safety. Next, carefully check whether the wires are evenly distributed in the slots of the core, and whether there is any looseness or crossing. If necessary, use hand tools to make fine adjustments to ensure that the wires are arranged neatly and tightly.

[0025] Specifically, in step S120, the threaded stator is pressed into the housing by using a core pressing assembly to obtain a stator assembly. In particular, the core pressing assembly includes a frame (upper fixed plate, lower fixed plate, column), a pressing mechanism (pressing cylinder, lower pressing plate, slide rod, pressure sensor), a track and a lifting mechanism (lifting rod, top plate), and a controller. The controller is used to control the action of the pressing cylinder according to the pressure value detected by the pressure sensor. Specifically, the threaded stator is placed on the workbench of the core pressing assembly, and the stator and the housing are accurately aligned using a visual system to ensure that the center of the stator coincides with the center of the housing, and then the core pressing assembly is started, and the stator is slowly and evenly pressed into the housing by the pressing cylinder.

[0026] It should be understood that in the process of automated motor assembly, pressing the stator into the housing is a key step. This process requires precise control of pressure to ensure a tight fit between the stator and the housing, while avoiding equipment damage or product quality problems caused by excessive pressure. Traditional methods mostly use threshold mediation. However, the threshold mediation method has a response delay. When the pressure sensor detects that the pressure exceeds the preset value, the controller needs time to process the signal and issue instructions. During this time, the downward pressure cylinder may continue to apply additional pressure, causing the stator to be pressed too deep or the housing to be damaged. In addition, because the threshold mediation method relies on a fixed preset value and cannot adjust the pressure in real time, it is easy to cause the stator to be incorrectly pressed in or the pressing depth to be inconsistent, thereby affecting the performance and reliability of the motor.

[0027] Therefore, in response to the above technical problems, the technical concept of the present application is to collect the time queue of pressure data through the pressure sensor, and use the data analysis and encoding method based on deep learning to perform time series encoding on the pressure data, and at the same time calculate the mean of each pressure local time series correlation feature as a benchmark, and then perform time series dynamic propagation aggregation on each pressure local time series correlation feature, and then calculate the response between the pressure local time series correlation feature and the mean feature and the aggregation feature at the last position, so as to intelligently judge whether to stop the downward pressure of the downward pressure cylinder based on the multi-dimensional characterization between the pressure time series benchmark response correlation feature and the pressure time series dynamic response correlation feature. In this way, the pressure change can be processed and analyzed in real time, eliminating the response delay problem in the traditional threshold mediation method. And the downward pressure can be dynamically adjusted to avoid the stator being pressed too deep or the shell being damaged, ensuring the accuracy of pressure control.

[0028] Figure 2 The present invention is a flow chart of using a core pressing assembly to press the threaded stator into a housing to obtain a stator assembly in an automated motor assembly method according to an embodiment of the present application. Figure 3 The data flow diagram of the motor automatic assembly method according to the embodiment of the present application is to use the iron core assembly to press the threaded stator into the housing to obtain the stator assembly. Figure 2 and Figure 3As shown, according to the embodiment of the present application, the method for automatic assembly of a motor includes: S210, obtaining a time queue of pressure data collected by a pressure sensor deployed on the pressure core assembly; S220, performing sequence encoding on the time queue of the pressure data to obtain a sequence of local pressure timing correlation features, and calculating the steady-state information of the sequence of local pressure timing correlation features to obtain a local pressure timing mean feature; S230, performing dynamic propagation based on node feature perception on the sequence of local pressure timing correlation features to obtain a pressure timing dynamic aggregation representation; S240, extracting search features from the sequence of local pressure timing correlation features, and calculating the response features between the local pressure timing mean feature and the pressure timing dynamic aggregation representation and the search features respectively to obtain a pressure timing baseline response correlation feature and a pressure timing dynamic response correlation feature; S250, calculating a pressure timing search reasoning multidimensional representation between the pressure timing baseline response correlation feature and the pressure timing dynamic response correlation feature, and generating a control result based on the pressure timing search reasoning multidimensional representation, wherein the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder.

[0029] In step S210, a time queue of pressure data collected by a pressure sensor deployed in the core pressing assembly is obtained. It should be understood that the pressure data refers to data reflecting the force applied by the downward pressure cylinder collected in real time by the pressure sensor during the process of pressing the stator into the shell. These data are usually recorded in the form of a time series, that is, each time point corresponds to a pressure value. Based on this, in order to better understand and analyze the time series changes of the force during the pressing process, so as to control the stability and accuracy of the pressing process, in the technical solution of the present application, a time queue of pressure data collected by the pressure sensor is obtained. In particular, the pressure sensor can be installed at a key position of the core pressing assembly, such as the area where the pressure head contacts the core, to ensure that the sensor can accurately measure the pressure.

[0030] In step S220, the time queue of the pressure data is sequence-encoded to obtain a sequence of pressure local time series correlation features, and the steady-state information of the sequence of pressure local time series correlation features is calculated to obtain the pressure local time series mean features. Specifically, in an embodiment of the present application, the time queue of the pressure data is sequence-encoded to obtain a sequence of pressure local time series correlation features, and the steady-state information of the sequence of pressure local time series correlation features is calculated to obtain the pressure local time series mean features, including: inputting the time queue of the pressure data into a sequence encoder based on 1D-CNN to obtain a sequence of pressure local time series correlation feature vectors as the sequence of pressure local time series correlation features; calculating the positional mean vector of the sequence of pressure local time series correlation feature vectors to obtain the pressure local time series mean vector as the pressure local time series mean features. Accordingly, considering that the time series of the pressure data contains rich local information, such as the trend of pressure changes, short-term fluctuations, etc. And considering that 1D-CNN slides on the time series through the convolution kernel, it can effectively capture local time features. For example, the convolution kernel can capture the trend and amplitude of pressure changes in a short period of time. Therefore, in the technical solution of the present application, the time queue of the pressure data is input into a sequence encoder based on 1D-CNN to capture and refine the local time series characteristics and dynamic time series changes of pressure in different local time periods, and obtain a sequence of pressure local time series associated feature vectors. Then, in order to generate a steady-state information representing the overall pressure characteristics and lay the foundation for subsequent characteristic response calculations, in the technical solution of the present application, the mean vector of the sequence of the pressure local time series associated feature vectors is calculated to obtain the pressure local time series mean vector. In other words, the mean vector provides a global perspective, reflecting the average steady-state pressure characteristics of the entire pressing process, which helps to understand the overall behavior of the entire process.

[0031] In step S230, the sequence of the local temporal correlation features of pressure is dynamically propagated based on node feature perception to obtain a dynamic aggregation representation of the pressure time series. Furthermore, considering that each local temporal correlation feature of pressure in the sequence of local temporal correlation features of pressure expresses the local features and change patterns of pressure data at a specific time point, and as time goes by, the features of the current time point are more strongly correlated with the surrounding time points, while the correlation with earlier time points gradually decreases. This means that the features of the current time point not only represent the pressure level at the current moment, but are also mainly affected by the pressure changes and patterns of the nearby time points, while the influence of the earlier time points becomes less significant. In order to capture and represent the correlation and influence of pressure data at different times, so as to better reflect the complex change patterns of pressure data in time, so as to obtain a comprehensive and integrated representation, in the technical solution of the present application, the sequence of the local temporal correlation features of pressure is dynamically propagated based on node feature perception to obtain a dynamic aggregation representation of the pressure time series.

[0032] Specifically, the sequence of pressure local temporal association features is dynamically propagated based on node feature perception to obtain a pressure temporal dynamic aggregation representation, including: extracting a current pressure local temporal association feature vector from the sequence of pressure local temporal association feature vectors; using the current pressure local temporal association feature vector as a feature prompt gate, calculating the feature association strength prompt value of each pressure local temporal association feature vector in the sequence of pressure local temporal association feature vectors relative to the current pressure local temporal association feature vector to obtain a sequence of pressure local temporal feature association strength prompt values; and according to the sequence of pressure local temporal feature association strength prompt values, performing gated modulation dynamic aggregation on the sequence of pressure local temporal association feature vectors to obtain the pressure temporal dynamic aggregation representation.

[0033] Specifically, in an embodiment of the present application, the current pressure local time series associated feature vector is extracted from the sequence of the pressure local time series associated feature vector. That is to say, in the distribution of the feature sequence, each node has unique attributes or characteristics, and these feature vectors reflect the information of the node itself. When extracting the feature vector of a specific node, it is actually to determine a reference point and use the node as the reference center to evaluate the relative importance of other nodes. This can be analogous to selecting a core point in the feature sequence, and then building a sequence attention framework with this core point as the center.

[0034] Specifically, in the embodiment of the present application, the current pressure local temporal association feature vector is used as a feature prompt gate, and the feature association strength prompt value of each pressure local temporal association feature vector in the sequence of the pressure local temporal association feature vector is calculated relative to the current pressure local temporal association feature vector to obtain a sequence of pressure local temporal association strength prompt values; this process can be expressed by the formula: ; in, is the sequence of the pressure local time series associated characteristic vectors, and are the first, second, and third pressure local time series correlation feature vectors in the sequence. and The local time series correlation feature vector of pressure, It is The characteristic value of each position in the pressure local time series correlation characteristic vector, It is The characteristic value of each position in the pressure local time series correlation characteristic vector, Represents the logarithmic function value with base 2 is the number of eigenvalues ​​in each pressure local time series correlation eigenvector, represents the exponential function value with the natural constant e as the base, yes The corresponding pressure local temporal feature correlation strength prompt value. That is, the purpose of this process is to evaluate the strength of the relationship between each pressure local temporal correlation feature relative to the current node. In particular, this is essentially an application of the attention mechanism, in which the attention score is calculated based on the similarity of the features between nodes.

[0035] Specifically, according to the sequence of the pressure local temporal feature association strength prompt values, the sequence of the pressure local temporal association feature vectors is gated modulated and dynamically aggregated to obtain the pressure temporal dynamic aggregation representation, including: extracting the median of the sequence of the pressure local temporal feature association strength prompt values ​​as the pressure local temporal feature association strength gate threshold; using the pressure local temporal feature association strength gate threshold as the switch threshold of the gating unit, inputting the sequence of the pressure local temporal association feature vectors and the sequence of the pressure local temporal feature association strength prompt values ​​into the gating unit to obtain a sequence of modulated pressure local temporal association feature vectors; inputting the sequence of modulated pressure local temporal association feature vectors into a feature dynamic propagation module based on a forward LSTM model to obtain a pressure temporal dynamic aggregation representation vector as the pressure temporal dynamic aggregation representation.

[0036] Specifically, in an embodiment of the present application, the median of the sequence of the pressure local temporal feature association strength prompt values ​​is extracted as the pressure local temporal feature association strength gate threshold.

[0037] The process can be expressed as: ;in, and , are the first, second and third values ​​in the sequence of the pressure local temporal feature association strength prompt value. The correlation strength prompt value of the local time series characteristics of pressure, To get the median of a sequence, is the threshold of the correlation strength gate of the local time series feature of pressure. Here, one advantage of using the median instead of the mean is that the median is not affected by extreme values ​​and can more accurately capture the center position of the local time series dataset of pressure. This method helps to maintain the balance of feature transfer in the network and prevent the negative impact on the overall performance due to the emergence of individual extreme values.

[0038] Specifically, in an embodiment of the present application, the pressure local temporal feature association strength gate threshold is used as the switch threshold of the gating unit, and the sequence of the pressure local temporal feature association strength feature vectors and the sequence of the pressure local temporal feature association strength prompt values ​​are input into the gating unit to obtain a sequence of modulated pressure local temporal feature vectors, including: comparing each pressure local temporal feature association strength prompt value in the sequence of the pressure local temporal feature association strength prompt value with the pressure local temporal feature association strength gate threshold to obtain a sequence of weights; using each value in the sequence of weights as a weight to perform weighted modulation on the sequence of the pressure local temporal feature vectors to obtain a sequence of modulated pressure local temporal association feature vectors; wherein, in response to the pressure local temporal feature association strength prompt value being greater than or equal to the pressure local temporal feature association strength gate threshold, the value obtained by dividing the pressure local temporal feature association strength prompt value and the pressure local temporal feature association strength gate threshold is used as the weight; in response to the pressure local temporal feature association strength prompt value being less than the pressure local temporal feature association strength gate threshold, the weight is reset to zero.

[0039] The process can be expressed as: ; in, is the first in the sequence of the pressure local time series correlation feature vector The local time series correlation feature vector of pressure, yes The corresponding pressure local time series feature correlation strength prompt value, is the threshold of the correlation strength gate of the local temporal characteristics of pressure, For gated mask operation, for The local temporal correlation eigenvector of the modulated pressure after weighted modulation, and are the first, second and third in the sequence of the local temporal correlation feature vector of the pressure after modulation. The local temporal correlation feature vector of the pressure after modulation, It is The pressure local time series correlation feature vector, is the sequence of the modulated pressure local time series correlation feature vectors. That is, by setting the switch threshold of the gating unit to the median of the feature correlation strength, the features closely related to the current pressure local time series features can be effectively screened out. This method allows the focus of attention on other node data to be dynamically adjusted, providing accurate data support for subsequent feature propagation.

[0040] Specifically, in an embodiment of the present application, the sequence of the modulated pressure local time series associated feature vectors is input into a feature dynamic propagation module based on a forward LSTM model to obtain a pressure time series dynamic aggregation representation vector as the pressure time series dynamic aggregation representation.

[0041] The process can be expressed as: ;in, is the sequence of the local temporal correlation characteristic vectors of the pressure after modulation, Encode for LSTM, The pressure time series dynamic aggregation representation vector is obtained. Therefore, the sequence of the modulated pressure local time series correlation feature vectors is input into the feature dynamic propagation module based on the forward LSTM model to identify and extract the long-term mutual relationship between node features at different time points, so as to deeply understand the pressure time series change characteristics in different time periods and obtain the pressure time series dynamic aggregation representation vector.

[0042] In step S240, the search feature is extracted from the sequence of the pressure local time series association features, and the response features between the pressure local time series mean feature and the pressure time series dynamic aggregation representation and the search feature are calculated to obtain the pressure time series baseline response association feature and the pressure time series dynamic response association feature. Specifically, in an embodiment of the present application, a search feature is extracted from the sequence of the pressure local timing association feature, and the response features between the pressure local timing mean feature and the pressure timing dynamic aggregation representation and the search feature are calculated to obtain a pressure timing baseline response association feature and a pressure timing dynamic response association feature, including: taking the pressure local timing mean vector as a reference template, taking the pressure timing dynamic aggregation representation vector as a dynamic template, and taking the pressure local timing association feature vector at the last position in the sequence of the pressure local timing association feature vector as the search feature, and respectively calculating the response feature vector between the reference template and the search feature and the response feature vector between the dynamic template and the search feature to obtain a pressure timing baseline response association feature vector as the pressure timing baseline response association feature and a pressure timing dynamic response association feature vector as the pressure timing dynamic response association feature. Accordingly, in order to more accurately evaluate the characteristics of the pressure value at the current time point, so as to intelligently judge the state of the pressing process, in the technical solution of the present application, the pressure local time series mean vector is used as the reference template, the pressure time series dynamic aggregation representation vector is used as the dynamic template, and the pressure local time series associated feature vector at the last position in the sequence of the pressure local time series associated feature vector is used as the search feature, and the response feature vector between the reference template and the search feature and the response feature vector between the dynamic template and the search feature are calculated respectively to obtain the pressure time series reference response associated feature vector and the pressure time series dynamic response associated feature vector. In other words, by calculating the response feature vector between the reference template (pressure local time series mean vector) and the search feature (pressure local time series associated feature vector at the last position), the similarity between the pressure feature at the current time point and the overall pressure characteristic can be evaluated, which helps to determine whether the pressure at the current time point meets the expected normal range. By calculating the response feature vector between the dynamic template (pressure time series dynamic aggregation representation vector) and the search feature, the similarity between the pressure characteristics at the current time point and the global pressure dynamic characteristics in the time series can be evaluated, so as to more accurately determine whether the pressure at the current time point is related to the dynamic changes in the historical data.

[0043] More specifically, in an embodiment of the present application, the pressure local time series mean vector is used as a reference template, the pressure time series dynamic aggregation representation vector is used as a dynamic template, and the pressure local time series associated feature vector at the last position in the sequence of the pressure local time series associated feature vectors is used as the search feature, and the response feature vector between the reference template and the search feature and the response feature vector between the dynamic template and the search feature are calculated respectively to obtain a pressure time series reference response associated feature vector as the pressure time series reference response associated feature and a pressure time series dynamic response associated feature vector as the pressure time series dynamic response associated feature, including: calculating the positional dot product of the pressure local time series associated feature vector at the last position and the pressure local time series mean vector, and then adding the obtained pressure reference dot product vector to the reference bias vector to obtain the pressure time series reference response associated feature vector; calculating the positional dot product of the pressure local time series associated feature vector at the last position and the pressure time series dynamic aggregation representation vector, and then adding the obtained pressure dynamic dot product vector to the dynamic bias vector to obtain the pressure time series dynamic response associated feature vector.

[0044] The process can be expressed as: ; in, is the local temporal correlation characteristic vector of the pressure at the last position, It is the point multiplication by position. is the local time series mean vector of the pressure, is the reference bias vector, is the pressure time series reference response associated feature vector, is the pressure time series dynamic aggregation representation vector, is the dynamic bias vector, is the associated characteristic vector of the pressure time series dynamic response.

[0045] In step S150, based on the multimodal joint representation of the state of the first monitoring node, a state monitoring result is obtained, and it is determined whether to transmit the state monitoring result to other monitoring nodes. Specifically, in the embodiment of the present application, the pressure timing search and reasoning multidimensional representation between the pressure timing reference response associated feature and the pressure timing dynamic response associated feature is calculated, and a control result is generated based on the pressure timing search and reasoning multidimensional representation, and the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder, including: fusing the pressure timing reference response associated feature vector and the pressure timing dynamic response associated feature vector to obtain the pressure timing search and reasoning multidimensional representation vector as the pressure timing search and reasoning multidimensional representation; based on the pressure timing search and reasoning multidimensional representation, the control result is obtained, and the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder. More specifically, in the embodiment of the present application, based on the pressure timing search and reasoning multidimensional representation, the control result is obtained, including: passing the pressure timing search and reasoning multidimensional representation vector through a downward pressure controller based on a classifier to obtain the control result. Afterwards, it is considered that the pressure time series reference response associated feature vector and the pressure time series dynamic response associated feature vector respectively reflect the corresponding relationship between the pressure characteristics at the current time point relative to the overall average pressure characteristics and the dynamic characteristics in the historical data. Based on this, in order to be able to comprehensively and multi-dimensionally reflect the pressure characteristics at the current time point, in the technical solution of the present application, the pressure time series reference response associated feature vector and the pressure time series dynamic response associated feature vector are fused to obtain a pressure time series search reasoning multi-dimensional representation vector to provide support for subsequent downward pressure control. In particular, in a specific embodiment of the present application, the pressure time series search reasoning multi-dimensional representation vector can be obtained by calculating the position-weighted sum of the pressure time series reference response associated feature vector and the pressure time series dynamic response associated feature vector. That is, the pressure time series search reasoning multi-dimensional representation obtained by fusing the pressure time series reference response associated feature vector and the pressure time series dynamic response associated feature vector is used for classification processing to intelligently determine whether to stop the downward pressure of the downward pressure cylinder. In this way, the pressure change can be processed and analyzed in real time, eliminating the response delay problem in the traditional threshold mediation method. It can also dynamically adjust the downward pressure to prevent the stator from being pressed too deep or the housing from being damaged, thus ensuring the accuracy of pressure control.

[0046] In particular, in a specific embodiment of the present application, the pressure timing search reasoning multidimensional representation vector is passed through a classifier-based pressure controller to obtain a control result, including: using the fully connected layer of the classifier-based pressure controller to fully connect the pressure timing search reasoning multidimensional representation vector to obtain a pressure timing search reasoning multidimensional representation fully connected encoded classification feature vector; inputting the pressure timing search reasoning multidimensional representation fully connected encoded classification feature vector into the Softmax classification function of the classifier-based pressure controller to obtain the probability value of the pressure timing search reasoning multidimensional representation vector belonging to each classification label, the classification label including the classification labels for indicating stopping the pressure cylinder and not stopping the pressure cylinder; and determining the classification label corresponding to the largest of the probability values ​​as the control result.

[0047] It should be understood that when the pressure time series benchmark response associated feature vector and the pressure time series dynamic response associated feature vector respectively represent the response characteristics of the near-time domain time series associated features of the pressure data relative to the full-time domain benchmark time series mean distribution and the dynamic time series aggregate distribution, when performing feature fusion, the pressure time series search and reasoning multi-dimensional representation vector in the local time domain space relative to the response representation of different time series dimensions in the full time domain will also have a selective time domain feature response state space representation in the distribution field, resulting in an imbalance in the regression understanding of the long-range response feature representation relative to the source time domain time series semantics, affecting the accuracy of the classification results.

[0048] Therefore, the present application considers optimizing the pressure time series search reasoning multidimensional characterization vector when the pressure time series search reasoning multidimensional characterization vector is passed through a classifier-based pressure controller to obtain a control result. The optimization process includes: determining the number of super-distributed eigenvalues ​​in the pressure time series search reasoning multidimensional characterization vector whose difference from the characteristic mean is greater than a threshold, and respectively calculating the reciprocal of the logarithm of the number of super-distributed eigenvalues ​​with base two and the exponential value of the reciprocal of the number of super-distributed eigenvalues ​​with base natural constants to obtain a first pressure time series search reasoning multidimensional nonlinear trajectory representation value and the second pressure time series search inference multidimensional nonlinear trajectory representation value : ; in, represents the multi-dimensional representation vector of the pressure time series search reasoning, The first representation vector of the multidimensional representation vector of the pressure time series search reasoning eigenvalues, represents the mean of all eigenvalues ​​of the multidimensional representation vector of the pressure time series search reasoning, Representation calculation The number of is the number of super-distributed eigenvalues ​​in the multi-dimensional representation vector of the pressure time series search inference, and is the threshold hyperparameter.

[0049] Calculate the hyperbolic sine function value of the sum of squares of all eigenvalues ​​of the pressure time series search reasoning multidimensional characterization vector: ;in, represents the length of the multi-dimensional representation vector of the pressure time series search reasoning, represents the hyperbolic sine function, Represents the value of the hyperbolic sine function.

[0050] And calculate its exponential value with the natural constant as the base and then divide it by the square of the length of the pressure time series search reasoning multidimensional representation vector to obtain the pressure time series search reasoning multidimensional potential manifold value ;in, represents a natural constant, Representing stress temporal search inference on multidimensional latent manifold values.

[0051] Calculate the pressure time series search inference multidimensional potential manifold value Relative to the first pressure time series search inference multidimensional nonlinear trajectory representation value and the second pressure time series search inference multidimensional nonlinear trajectory representation value The pressure time series search inference multidimensional covariance integration value: ;in, Represents the multi-dimensional covariance integration value of the pressure time series search inference.

[0052] Calculate the power function feature vector of the pressure time series search reasoning multidimensional characterization vector with the pressure time series search reasoning multidimensional covariance integration value as the inverse, and calculate the autocorrelation matrix of the power function feature vector to obtain the pressure time series search reasoning multidimensional adaptive characterization matrix, that is: ;in, represents a power function feature vector whose inverse is the integrated value of the multi-dimensional covariance of the pressure time series search reasoning for calculating the multi-dimensional characterization vector of the pressure time series search reasoning, represents matrix multiplication, represents the transpose of a vector, Representing stress temporal search inference with multidimensional adaptive representation matrices.

[0053] The pressure time series search reasoning multidimensional representation vector is matrix-multiplied with the pressure time series search reasoning multidimensional adaptive representation matrix to obtain an optimized pressure time series search reasoning multidimensional representation vector, that is: Here, the pressure time series search reasoning multidimensional representation vector is a row vector, Representing optimized stress-series search reasoning multidimensional representation vector.

[0054] That is, for the semantic feature set of the pressure time series search reasoning multidimensional representation vector under the preset semantic space trajectory distribution condition, the cross-domain semantic positioning efficiency attenuation phenomenon is mainly due to the long-range action distance beyond the local association window. By adopting the nonlinear manifold potential unit representation method based on covariance tensor integration, the multimodal association topological structure implicit in its parameter space is analyzed, and by constructing a nonlinear manifold potential unit model with scale adaptability, the adaptive association representation mechanism between the parameters of the real-time interactive scene time series association mapping matrix is ​​re-established. Therefore, through this reconstruction strategy of the physical state evolution trajectory, the continuous state migration modeling under the long-range association of parameters is effectively realized, and the semantic field reconstruction efficiency is significantly improved. At the same time, the original encoding representation characteristics of the pressure time series search reasoning multidimensional representation vector are maintained, and the accuracy of the control result obtained by the classifier-based pressure controller is improved. In this way, the pressure change can be processed and analyzed in real time, eliminating the response delay problem in the traditional threshold mediation method. And the pressure can be dynamically adjusted to avoid the stator from being pressed too deep or the shell from being damaged, ensuring the accuracy of pressure control.

[0055] In summary, the stator after threading is pressed into the shell by using the iron core assembly to obtain the stator assembly, which is explained by collecting the time queue of pressure data by the pressure sensor, and using the data analysis and encoding method based on deep learning to perform time series encoding on the pressure data, and at the same time calculating the mean of each pressure local time series correlation feature as a benchmark, and then performing time series dynamic propagation aggregation on each pressure local time series correlation feature, and then calculating the response between the pressure local time series correlation feature and the mean feature and the aggregation feature at the last position, so as to intelligently judge whether to stop the downward pressure of the downward pressure cylinder according to the multi-dimensional characterization between the pressure time series benchmark response correlation feature and the pressure time series dynamic response correlation feature. In this way, the pressure change can be processed and analyzed in real time, eliminating the response delay problem in the traditional threshold mediation method. And the downward pressure can be dynamically adjusted to avoid the stator from being pressed too deep or the shell from being damaged, ensuring the accuracy of pressure control.

[0056] Specifically, in step S130, the stator assembly and the rotor are quality inspected and screened to obtain a quality-inspected stator and a quality-inspected rotor. Specifically, first, it is necessary to ensure that all quality inspection equipment (such as visual inspection systems, electrical testing equipment, vibration testing equipment, etc.) are in good condition and calibrated to ensure the accuracy of the measurement. At the same time, according to the design specifications and quality standards of the motor, set the inspection parameters and qualification standards. These standards may include electrical performance, mechanical properties, appearance quality and other aspects. Establish a quality inspection data management system to record and analyze the inspection results for subsequent quality traceability and improvement.

[0057] During the quality inspection of the stator assembly, the appearance inspection is first carried out. Using high-resolution cameras and image processing software, the appearance defects of the stator assembly, such as scratches, cracks, loose wires, etc., are automatically detected. Next, the electrical performance test is carried out, including measuring the resistance value of the stator winding to ensure that it is within the specified range; using a high-voltage tester to detect the insulation performance of the stator to ensure that there is no leakage; and performing a withstand voltage test on the stator to verify its insulation performance under high voltage. Finally, the mechanical performance test is carried out, using a vibration test equipment to detect the vibration of the stator in the working state to ensure its stability and reliability, and using a three-coordinate measuring machine (CMM) or other high-precision measuring equipment to detect the dimensional accuracy of the stator to ensure that it meets the design requirements.

[0058] The quality inspection of the rotor also starts with the appearance inspection. Using high-resolution cameras and image processing software, the appearance defects of the rotor, such as scratches, cracks, and magnetic steel shedding, are automatically detected. Then, the electrical performance test is carried out, including measuring the resistance value of the rotor winding to ensure that it is within the specified range; using a high-voltage tester to detect the insulation performance of the rotor to ensure that there is no leakage; and performing a withstand voltage test on the rotor to verify its insulation performance under high voltage. Finally, the mechanical performance test is carried out, using a dynamic balancing machine to detect the dynamic balancing performance of the rotor to ensure its stability at high-speed rotation; using a three-coordinate measuring machine (CMM) or other high-precision measuring equipment to detect the dimensional accuracy of the rotor to ensure that it meets the design requirements. After completing all the tests, all the test results are recorded in the quality inspection data management system and a detailed test report is generated. According to the set qualified standards, each stator component and rotor is automatically determined to be qualified. Qualified parts are marked as "Qualified Inspection Passed" and unqualified parts are marked as "To be processed". Using automated sorting equipment, the stator components and rotors that have passed the quality inspection are sent to the next assembly link, and the unqualified parts are sent to the repair or scrap area. Unqualified parts will be returned for repair and re-inspected for quality until they pass the test.

[0059] Specifically, in step S140, the stator and the rotor that have passed the quality inspection are assembled to obtain an assembled motor. Specifically, first, it is necessary to ensure that all assembly equipment (such as automated assembly lines, manipulators, torque wrenches, etc.) are in good condition and calibrated to ensure the accuracy of the operation. At the same time, ensure that the stators and rotors that have passed the quality inspection are ready and placed in designated locations for easy access. Keep the workbench clean and tidy, and ensure that there is enough space for operation.

[0060] During the assembly process, the stator and rotor are first pre-treated. Use compressed air or a special cleaning agent to clean the surface of the stator and rotor to ensure that there are no impurities such as dust and oil. Check the appearance and internal structure of the stator and rotor again to ensure that there are no obvious defects or damage. Next, use a positioning device to fix the stator on the assembly table to ensure that its position is accurate. Use an optical alignment system or a mechanical alignment device to ensure that the axis of the rotor and the stator are aligned. During the alignment process, a high-precision sensor can be used to monitor the alignment in real time to ensure the concentricity of the rotor and stator. Use a manipulator or an automated device to slowly and smoothly insert the rotor into the stator. During the insertion process, ensure that there is no friction or jamming between the rotor and the stator. Use a torque wrench or an electric screwdriver to tighten the rotor's fixing bolts to the specified torque value to ensure that the rotor is firmly fixed in the stator. Subsequently, the lead wires of the stator and rotor are connected to the terminal blocks of the motor. Use special welding equipment or crimping tools to ensure that the connection is firm and reliable, and insulate the connection parts to ensure that there is no risk of short circuit or leakage. Then install the bearings on the front and rear end covers of the motor, ensure that the bearings are installed in the correct position, and apply an appropriate amount of lubricating oil or grease on the contact surface between the bearing and the rotor to ensure smooth operation and reduce wear of the motor. Next, install the front and rear end covers at both ends of the motor and tighten the fixing bolts to the specified torque value using a torque wrench. If the motor has a fan, install the fan on the front cover of the motor and ensure that the fan blades are aligned with the motor axis. After completing the above steps, perform a final inspection to check the appearance of the motor to ensure that there are no obvious defects or damage; connect the motor to the power supply, perform a no-load running test, check the start-up and operation of the motor, and ensure that there are no abnormal sounds or vibrations; use special test equipment to detect the electrical and mechanical properties of the motor to ensure that it meets the design requirements.

[0061] Specifically, in step S150, the assembled motor is packaged to obtain a finished motor. Specifically, first, ensure that all packaging equipment (such as automatic packaging machines, carton sealing machines, label printers, etc.) are in good condition and calibrated to ensure the accuracy of the operation. At the same time, prepare necessary packaging materials, such as cartons, foam pads, shockproof materials, plastic films, labels, etc. Keep the packaging area clean and tidy, and ensure that there is enough space for operation.

[0062] Choose appropriate anti-vibration materials according to the size and weight of the motor. Common anti-vibration materials include foam pads, bubble film, pearl cotton, etc. Place anti-vibration materials at the bottom and around the motor to ensure that the motor will not be affected by vibration or impact during transportation. Place the motor in the carton and ensure that the motor is firmly positioned in the carton. You can use foam pads or fixed frames to fix the motor in the carton to prevent it from moving during transportation. Use an automatic carton sealer to seal the opening of the carton. Make sure the seal is firm and there are no gaps. If necessary, you can wrap a layer of plastic film on the outside of the carton to increase the waterproof and dustproof effect. Use a label printer to print labels, which should contain important information such as the motor model, production date, number, customer information, etc. Attach the label to a conspicuous position on the carton to ensure that the label is clearly visible for subsequent warehousing management and transportation. Use a barcode scanner to scan the barcode of each motor and enter the packaging information into the quality management system. This helps to track the production batch and quality status of the motor. Put the packaged motors into the warehouse and store them according to the customer's order and delivery schedule. Ensure that the warehouse environment is dry and ventilated to prevent the motor from getting damp or damaged. Record the warehousing information of each motor, including the warehousing date, quantity, storage location, etc., to facilitate subsequent inventory management and delivery.

[0063] In summary, the motor automatic assembly method based on the embodiment of the present application is explained, which first passes the wire through the core hole to obtain the stator after threading, and then uses the iron core assembly to press the stator after threading into the housing to form a stator assembly, and then the stator assembly and rotor are quality inspected and screened to ensure that only qualified stators and rotors enter the next process, and then the stator and rotor that pass the quality inspection are assembled to obtain the assembled motor, and finally, the assembled motor is packaged to form a finished motor. This process realizes high-precision and high-efficiency automated assembly, ensuring the quality and reliability of the motor. In this way, the problems of low precision, low efficiency and unstable quality in traditional manual assembly are solved, automated assembly is realized, and the stability and consistency of production are ensured.

[0064] Figure 4 FIG. 1 is a block diagram of a motor automated assembly line according to an embodiment of the present application. Figure 4As shown, according to the embodiment of the present application, the motor automated assembly line 100 includes: a threading device 110, which is used to pass the wire through the core hole to obtain a threaded stator; a core pressing device 120, which is used to use the core pressing assembly to press the threaded stator into the shell to obtain a stator assembly; a detection device 130, which is used to perform quality inspection and screening on the stator assembly and the rotor to obtain a quality-qualified stator and a quality-qualified rotor; an assembly device 140, which is used to assemble the quality-qualified stator and the quality-qualified rotor to obtain an assembled motor; a packaging device 150, which is used to package the assembled motor to obtain a finished motor; wherein the core pressing device includes: a pressure data acquisition module, which is used to obtain a time queue of pressure data collected by a pressure sensor deployed on the core pressing assembly; a pressure data processing module, which is used to sequence the time queue of the pressure data to obtain a pressure local A sequence of timing-related features, and calculates the steady-state information of the sequence of pressure local timing-related features to obtain the pressure local timing mean feature; a pressure timing propagation module, used to dynamically propagate the sequence of pressure local timing-related features based on node feature perception to obtain the pressure timing dynamic aggregation representation; a pressure timing response module, used to extract search features from the sequence of pressure local timing-related features, and respectively calculate the response features between the pressure local timing mean feature and the pressure timing dynamic aggregation representation and the search features to obtain the pressure timing baseline response-related feature and the pressure timing dynamic response-related feature; a control module, used to calculate the pressure timing search reasoning multi-dimensional representation between the pressure timing baseline response-related feature and the pressure timing dynamic response-related feature, and generate a control result based on the pressure timing search reasoning multi-dimensional representation, wherein the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder.

[0065] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned motor automated assembly line have been described in detail above. Figures 1 to 3 The motor automated assembly method has been described in detail in the description, and therefore, its repeated description will be omitted.

[0066] As described above, the motor automation assembly line 100 according to the embodiment of the present disclosure can be implemented in various wireless terminals, such as a server with a motor automation assembly algorithm. In a possible implementation, the motor automation assembly line 100 according to the embodiment of the present disclosure can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the motor automation assembly line 100 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the motor automation assembly line 100 can also be one of the many hardware modules of the wireless terminal.

Claims

1. A method for automating the assembly of a motor, characterized in that: include: Pass the wire through the core hole to obtain the stator after threading; Using a pressed iron core assembly to press the threaded stator into a housing to obtain a stator assembly; Performing quality inspection and screening on the stator assembly and the rotor to obtain a stator and a rotor that pass the quality inspection; assembling the stator and the rotor that pass the quality inspection to obtain an assembled motor; The assembled motor is packaged to obtain a finished motor; wherein, the threaded stator is pressed into a housing by using a pressed iron core assembly to obtain a stator assembly, including: obtaining a time queue of pressure data collected by a pressure sensor deployed on the pressed iron core assembly; performing sequence encoding on the time queue of the pressure data to obtain a sequence of local pressure time series correlation features, and calculating steady-state information of the sequence of local pressure time series correlation features to obtain a local pressure time series mean feature; performing dynamic propagation based on node feature perception on the sequence of local pressure time series correlation features to obtain a pressure time series dynamic aggregation representation; extracting search features from the sequence of local pressure time series correlation features, and calculating response features between the local pressure time series mean feature and the pressure time series dynamic aggregation representation and the search features respectively to obtain a pressure time series baseline response correlation feature and a pressure time series dynamic response correlation feature; calculating a pressure time series search reasoning multidimensional representation between the pressure time series baseline response correlation feature and the pressure time series dynamic response correlation feature, and generating a control result based on the pressure time series search reasoning multidimensional representation, the control result being used to indicate whether to stop the downward pressure of the downward pressure cylinder.

2. The motor automated assembly method according to claim 1, characterized in that: The time queue of the pressure data is sequence-encoded to obtain a sequence of pressure local time series correlation features, and the steady-state information of the sequence of pressure local time series correlation features is calculated to obtain a pressure local time series mean feature, including: inputting the time queue of the pressure data into a sequence encoder based on 1D-CNN to obtain a sequence of pressure local time series correlation feature vectors as the sequence of pressure local time series correlation features; calculating the positional mean vector of the sequence of pressure local time series correlation feature vectors to obtain a pressure local time series mean vector as the pressure local time series mean feature.

3. The motor automated assembly method according to claim 2, characterized in that: The sequence of the pressure local temporal association features is dynamically propagated based on node feature perception to obtain a pressure temporal dynamic aggregation representation, including: extracting a current pressure local temporal association feature vector from the sequence of the pressure local temporal association feature vectors; using the current pressure local temporal association feature vector as a feature prompt gate, calculating the feature association strength prompt value of each pressure local temporal association feature vector in the sequence of the pressure local temporal association feature vectors relative to the current pressure local temporal association feature vector to obtain a sequence of pressure local temporal feature association strength prompt values; and according to the sequence of pressure local temporal feature association strength prompt values, performing gated modulation dynamic aggregation on the sequence of the pressure local temporal association feature vectors to obtain the pressure temporal dynamic aggregation representation.

4. The motor automated assembly method according to claim 3, characterized in that: According to the sequence of the pressure local temporal feature association strength prompt values, the sequence of the pressure local temporal association feature vectors is gated modulated and dynamically aggregated to obtain the pressure temporal dynamic aggregation representation, including: extracting the median of the sequence of the pressure local temporal feature association strength prompt values ​​as the pressure local temporal feature association strength gate threshold; using the pressure local temporal feature association strength gate threshold as the switch threshold of the gating unit, inputting the sequence of the pressure local temporal association feature vectors and the sequence of the pressure local temporal feature association strength prompt values ​​into the gating unit to obtain a sequence of modulated pressure local temporal association feature vectors; inputting the sequence of modulated pressure local temporal association feature vectors into a feature dynamic propagation module based on a forward LSTM model to obtain a pressure temporal dynamic aggregation representation vector as the pressure temporal dynamic aggregation representation.

5. The motor automated assembly method according to claim 4, characterized in that: The pressure local temporal feature association strength gate threshold is used as the switch threshold of the gating unit, and the sequence of the pressure local temporal feature association feature vectors and the sequence of the pressure local temporal feature association strength prompt values ​​are input into the gating unit to obtain the sequence of modulated pressure local temporal feature vectors, including: comparing each pressure local temporal feature association strength prompt value in the sequence of the pressure local temporal feature association strength prompt value with the pressure local temporal feature association strength gate threshold to obtain a sequence of weights; using each value in the sequence of weights as a weight to perform weighted modulation on the sequence of the pressure local temporal feature association vectors to obtain the sequence of modulated pressure local temporal feature vectors; wherein, in response to the pressure local temporal feature association strength prompt value being greater than or equal to the pressure local temporal feature association strength gate threshold, the value obtained by dividing the pressure local temporal feature association strength prompt value and the pressure local temporal feature association strength gate threshold is used as the weight; in response to the pressure local temporal feature association strength prompt value being less than the pressure local temporal feature association strength gate threshold, the weight is reset to zero.

6. The motor automated assembly method according to claim 5, characterized in that: Extracting search features from the sequence of pressure local time series association features, and calculating the response features between the pressure local time series mean feature and the pressure time series dynamic aggregation representation and the search features respectively to obtain pressure time series benchmark response association features and pressure time series dynamic response association features, including: taking the pressure local time series mean vector as a benchmark template, taking the pressure time series dynamic aggregation representation vector as a dynamic template, and taking the pressure local time series association feature vector at the last position in the sequence of the pressure local time series association feature vector as the search feature, respectively calculating the response feature vector between the benchmark template and the search feature and the response feature vector between the dynamic template and the search feature to obtain a pressure time series benchmark response association feature vector as the pressure time series benchmark response association feature and a pressure time series dynamic response association feature vector as the pressure time series dynamic response association feature.

7. The motor automated assembly method according to claim 6, characterized in that: Taking the pressure local time series mean vector as the reference template, taking the pressure time series dynamic aggregation representation vector as the dynamic template, and taking the pressure local time series associated feature vector at the last position in the sequence of the pressure local time series associated feature vectors as the search feature, respectively calculate the response feature vector between the reference template and the search feature and the response feature vector between the dynamic template and the search feature to obtain the pressure time series reference response associated feature vector as the pressure time series reference response associated feature and the pressure time series dynamic response associated feature vector as the pressure time series dynamic response associated feature, including: calculating the positional dot product of the pressure local time series associated feature vector at the last position and the pressure local time series mean vector, and then adding the obtained pressure reference dot product vector to the reference bias vector to obtain the pressure time series reference response associated feature vector; calculating the positional dot product of the pressure local time series associated feature vector at the last position and the pressure time series dynamic aggregation representation vector, and then adding the obtained pressure dynamic dot product vector to the dynamic bias vector to obtain the pressure time series dynamic response associated feature vector.

8. The motor automated assembly method according to claim 7, characterized in that: Calculate the pressure timing search and reasoning multi-dimensional representation between the pressure timing baseline response associated characteristics and the pressure timing dynamic response associated characteristics, and generate a control result based on the pressure timing search and reasoning multi-dimensional representation, and the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder, including: fusing the pressure timing baseline response associated characteristic vector and the pressure timing dynamic response associated characteristic vector to obtain a pressure timing search and reasoning multi-dimensional representation vector as the pressure timing search and reasoning multi-dimensional representation; based on the pressure timing search and reasoning multi-dimensional representation, obtain the control result, and the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder.

9. The motor automated assembly method according to claim 8, characterized in that: The control result is obtained based on the pressure time series search and reasoning multi-dimensional representation, including: passing the pressure time series search and reasoning multi-dimensional representation vector through a classifier-based pressure controller to obtain the control result.

10. An automatic motor assembly line, characterized in that: include: Threading equipment, used to pass the wire through the core hole to obtain the threaded stator; A core pressing device is used to use a core pressing assembly to press the threaded stator into a shell to obtain a stator assembly; a detection device is used to perform quality detection and screening on the stator assembly and the rotor to obtain a quality-qualified stator and a quality-qualified rotor; an assembly device is used to assemble the quality-qualified stator and the quality-qualified rotor to obtain an assembled motor; a packaging device is used to package the assembled motor to obtain a finished motor; wherein the core pressing device includes: a pressure data acquisition module, which is used to acquire a time queue of pressure data collected by a pressure sensor deployed on the core pressing assembly; a pressure data processing module, which is used to sequence encode the time queue of the pressure data to obtain a sequence of local pressure time series correlation features, and calculate the steady-state information of the sequence of local pressure time series correlation features to obtain to the pressure local time series mean feature; a pressure time series propagation module, used to dynamically propagate the sequence of the pressure local time series associated features based on node feature perception to obtain the pressure time series dynamic aggregation representation; a pressure time series response module, used to extract the search feature from the sequence of the pressure local time series associated features, and respectively calculate the response features between the pressure local time series mean feature and the pressure time series dynamic aggregation representation and the search feature to obtain the pressure time series baseline response associated feature and the pressure time series dynamic response associated feature; a control module, used to calculate the pressure time series search reasoning multi-dimensional representation between the pressure time series baseline response associated feature and the pressure time series dynamic response associated feature, and generate a control result based on the pressure time series search reasoning multi-dimensional representation, wherein the control result is used to indicate whether to stop the downward pressure of the downward pressure cylinder.

Images