Full-automatic motor stator assembly control system and method

Abstract

The invention provides a full-automatic motor stator assembly control system and a full-automatic motor stator assembly control method, which relate to the field of intelligent production, and are characterized in that a time sequence of winding wire tension data acquired by a tension sensor is acquired, and local time sequence characteristic analysis and association of the winding wire tension data are carried out by adopting a data analysis technology based on deep learning; therefore, whether winding is stopped or not is automatically judged according to propagation aggregation reasoning characteristics of winding wire tension data time sequence characteristic information of each local time point on the whole time dimension. Therefore, the tension data of the winding wire can be acquired in real time, the accuracy of tension control is ensured, and the coil quality problem caused by non-uniform tension is avoided. And meanwhile, a deep learning model is used for carrying out time sequence characteristic analysis on the tension data, so that the optimal winding stop point is automatically identified, errors caused by manual judgment are reduced, and the automation level of stator winding control is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent production, and more specifically, to a fully automatic motor stator assembly control system and method. Background Art

[0002] The motor is an indispensable device in modern industry and daily life, and is widely used in various fields such as mechanical equipment, household appliances, and electric vehicles. The performance of the motor directly affects the performance and efficiency of these devices. The motor mainly consists of two parts: a stator and a rotor. The motor stator, as the heart of the motor, its assembly quality is crucial for the overall performance of the motor.

[0003] However, due to relying on manual operation, the production efficiency of the traditional motor stator assembly system is limited by the work efficiency and fatigue of workers, and it is difficult to achieve long-term continuous operation. Specifically, the accuracy of manual winding and shaping is limited by the skill level of workers and their working state on the day, which may lead to quality differences between products. In addition, due to the lack of precise automatic control, the traditional system is difficult to perform complex or delicate assembly tasks, restricting product innovation and diversity. In a high-demand and rapidly changing market environment, the traditional system is difficult to quickly adjust the production scale or adapt to new production requirements.

[0004] Therefore, an optimized motor stator assembly control system is needed. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide a fully automatic motor stator assembly control system and method, which obtain the time series of winding wire tension data collected by a tension sensor, and use deep learning-based data analysis technology to perform local time series feature analysis and correlation of the winding wire tension data. Based on this, it automatically determines whether to stop winding according to the propagation aggregation inference features of the winding wire tension data time series feature information at each local time point in the entire time dimension. In this way, it can collect the tension data of the winding wire in real time, ensure the accuracy of tension control, and avoid coil quality problems caused by uneven tension. At the same time, use a deep learning model to perform time series feature analysis on the tension data, so as to automatically identify the best winding stop point, reduce the error of manual judgment, and thus improve the automation level of stator winding control.

[0006] According to one aspect of this application, a fully automatic motor stator assembly control system is provided, which includes: A motor stator fixing module, configured to place the motor stator to be assembled on an assembly table and fix the motor stator to be assembled; A stator winding coil module for pasting insulating paper in the core slots of the stator of the motor to be assembled and winding winding wires into the core slots to obtain stator winding coils; A stator winding coil forming module for forming the stator winding coils to obtain formed stator winding coils; A motor stator forming module for trimming, welding and insulating the formed stator winding coils to obtain a formed motor stator; Among them, the stator winding coil module includes: A winding wire tension data acquisition unit for obtaining a time series of winding wire tension data collected by a tension sensor; A winding wire tension local time series encoding unit for arranging the time series of the winding wire tension data into a winding wire tension time series input vector in the time dimension and then passing it through a time series encoder to obtain a sequence of winding wire tension local time series correlation feature vectors; A winding wire tension local time series aggregation unit for passing the sequence of the winding wire tension local time series correlation feature vectors through an information forward propagation aggregation inference module based on node energy time series attenuation to obtain a winding wire tension local time series information aggregation inference feature vector; A control result generation unit for obtaining a control result based on the winding wire tension local time series information aggregation inference feature vector, where the control result is used to indicate whether to stop winding.

[0007] According to another aspect of the present application, a full-automatic motor stator assembly control method is provided, which includes: Placing the stator of the motor to be assembled on an assembly table and fixing the stator of the motor to be assembled; Pasting insulating paper in the core slots of the stator of the motor to be assembled and winding winding wires into the core slots to obtain stator winding coils; Forming the stator winding coils to obtain formed stator winding coils; Trimming, welding and insulating the formed stator winding coils to obtain a formed motor stator; Among them, pasting insulating paper in the core slots of the stator of the motor to be assembled and winding winding wires into the core slots to obtain stator winding coils includes: Obtaining a time series of winding wire tension data collected by a tension sensor; Arranging the time series of the winding wire tension data into a winding wire tension time series input vector in the time dimension and then passing it through a time series encoder to obtain a sequence of winding wire tension local time series correlation feature vectors; Pass the sequence of the local time - series correlation feature vectors of the winding wire tension through an information forward - propagation aggregation inference module based on node - energy time - series attenuation to obtain the local time - series information aggregation inference feature vectors of the winding wire tension; Based on the local time - series information aggregation inference feature vectors of the winding wire tension, obtain a control result, where the control result is used to indicate whether to stop winding.

[0008] Compared with the prior art, the full - automatic motor stator assembly control system and method provided by this application acquire the time series of the winding wire tension data collected by a tension sensor, and adopt deep - learning - based data - analysis techniques to perform local time - series feature analysis and correlation of the winding wire tension data. Thus, it automatically determines whether to stop winding according to the propagation aggregation inference features of the winding wire tension data time - series feature information at each local time point in the entire time dimension. In this way, it can collect the tension data of the winding wire in real time, ensure the accuracy of tension control, and avoid coil quality problems caused by uneven tension. At the same time, it uses a deep - learning model to perform time - series feature analysis on the tension data to automatically identify the optimal winding - stop point, reduce the error of human judgment, and thus improve the automation level of stator winding control. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following - described drawings are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings: Figure 1 FIG. is a system block diagram of a full - automatic motor stator assembly control system according to an embodiment of the present application.

[0010] Figure 2 FIG. is a block diagram of a stator winding coil module of a full - automatic motor stator assembly control system according to an embodiment of the present application.

[0011] Figure 3 FIG. is a schematic diagram of data flow of a stator winding coil module in a full - automatic motor stator assembly control system according to an embodiment of the present application.

[0012] Figure 4 FIG. is a block diagram of a local time - series aggregation unit of winding wire tension in a full - automatic motor stator assembly control system according to an embodiment of the present application.

[0013] Figure 5 FIG. is a flowchart of a full - automatic motor stator assembly control method according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0015] As the cornerstone of modern industry and daily life, motors are widely used in various fields such as various mechanical equipment, household appliances, and electric vehicles. The performance of motors is directly related to the effectiveness and efficiency of these devices. The core components of a motor include a stator and a rotor. Among them, the stator, as the "heart" of the motor, its assembly quality has a decisive impact on the overall performance of the motor.

[0016] However, the traditional motor stator assembly process mainly relies on manual operation, which restricts the production efficiency due to the work efficiency and fatigue level of workers, and it is difficult to achieve continuous high-efficiency operation. Specifically, the accuracy of manual winding and shaping is affected by the skill level and working state of workers, which may lead to inconsistencies in product quality. In addition, due to the lack of fine automation control, traditional systems have limitations in performing complex or precise assembly tasks, which restricts the innovation and diversity of products. In an environment with high and rapidly changing market demands, traditional systems are difficult to flexibly adjust the production scale or quickly adapt to new production requirements.

[0017] Based on this, the present application proposes a fully automatic motor stator assembly control system. Figure 1 FIG. is a system block diagram of a fully automatic motor stator assembly control system according to an embodiment of the present application. As Figure 1 shown, in the fully automatic motor stator assembly control system 100, it includes: a motor stator fixing module 110, configured to place the motor stator to be assembled on an assembly table and fix the motor stator to be assembled; a stator winding coil module 120, configured to paste insulating paper in the iron core slots of the motor stator to be assembled and wind winding wires into the iron core slots to obtain stator winding coils; a stator winding coil shaping module 130, configured to perform shaping processing on the stator winding coils to obtain shaped stator winding coils; and a motor stator shaping module 140, configured to trim, weld, and perform insulation processing on the shaped stator winding coils to obtain a shaped motor stator.

[0018] It is worth mentioning that in the above fully automatic motor stator assembly control system, the assembly process of the motor stator is completed by a motor stator assembly robot. Among them, the motor stator fixing module, the stator winding coil module, the stator winding coil shaping module, and the motor stator shaping module are all functional modules of the motor stator assembly robot, used to implement different functions in the motor stator assembly process.

[0019] In particular, winding control plays a crucial role in the process of motor stator assembly, as it directly affects the quality and performance of the stator winding coils. That is to say, if the winding control is improper, it may lead to loose, broken or too tight coils, thus affecting the operation effect and lifespan of the motor. However, traditional winding control methods mainly rely on manual experience and simple mechanical devices. Specifically, the manual winding process is easily affected by skill levels and work fatigue, which results in uneven quality of the produced motor stator winding coils. At the same time, traditional mechanical devices often lack fine control and rapid response capabilities for tension changes and cannot adapt to the tension adjustments required by the winding wire at different stages, increasing the risk of coil breakage or uneven winding.

[0020] Correspondingly, in the above-mentioned stator winding coil module, the technical concept of this application is to obtain the time series of the winding wire tension data collected by the tension sensor and use deep learning-based data analysis and processing techniques to perform local temporal feature analysis and correlation of the winding wire tension data, so as to automatically determine whether to stop winding according to the propagation aggregation inference features of the temporal feature information of the winding wire tension data at each local time point in the entire time dimension. In this way, the tension data of the winding wire can be collected in real time, ensuring the accuracy of tension control and avoiding coil quality problems caused by uneven tension. At the same time, the deep learning model is used to perform temporal feature analysis on the tension data to automatically identify the optimal winding stop point, reducing the error of human judgment, thereby improving the automation level of stator winding control.

[0021] Figure 2 FIG. is a block diagram of a stator winding coil module of a full-automatic motor stator assembly control system according to an embodiment of the present application. Figure 3 FIG. is a schematic diagram of data flow in a stator winding coil module of a full-automatic motor stator assembly control system according to an embodiment of the present application. As Figure 2 and Figure 3 shown, the stator winding coil module 120 includes: a winding wire tension data acquisition unit 121 for obtaining the time series of the winding wire tension data collected by the tension sensor; a winding wire tension local temporal encoding unit 122 for arranging the time series of the winding wire tension data in the time dimension into a winding wire tension temporal input vector and then obtaining a sequence of winding wire tension local temporal correlation feature vectors through a temporal encoder; a winding wire tension local temporal aggregation unit 123 for obtaining a winding wire tension local temporal information aggregation inference feature vector by passing the sequence of the winding wire tension local temporal correlation feature vectors through an information forward propagation aggregation inference module based on node energy temporal decay; and a control result generation unit 124 for obtaining a control result based on the winding wire tension local temporal information aggregation inference feature vector, where the control result is used to indicate whether to stop winding.

[0022] In the embodiment of the present application, the winding wire tension data acquisition unit 121 is configured to obtain the time series of the winding wire tension data collected by the tension sensor. It should be understood that the time series of the winding wire tension data can directly reflect the force condition of the coil during the winding process, and can also reflect the fluctuation condition of the tension value, such as whether the tension is maintained within a stable range or shows periodic fluctuations. This mode can help identify abnormal conditions, such as uneven winding of the coil or coil breakage. When such abnormal conditions occur, the winding needs to be stopped in time. Generally speaking, the time series of the winding wire tension data contains rich information, which can not only reflect the real-time state of the winding process, but also provide an important basis for judging whether to stop the winding.

[0023] In the embodiment of the present application, the winding wire tension local time series encoding unit 122 is configured to arrange the time series of the winding wire tension data in the time dimension as a winding wire tension time series input vector, and then obtain a sequence of winding wire tension local time series correlation feature vectors through a time series encoder. Specifically, in the embodiment of the present application, the winding wire tension local time series encoding unit is configured to: after arranging the time series of the winding wire tension data in the time dimension as the winding wire tension time series input vector, pass the winding wire tension time series input vector through a time series encoder based on a deep convolutional neural network to obtain a sequence of the winding wire tension local time series correlation feature vectors. It should be understood that the winding wire tension data changes continuously over time, that is, there is certain time series feature information in the winding wire tension data in the time dimension. Based on this, in the technical solution of the present application, arranging the time series of the winding wire tension data in the time dimension as a winding wire tension time series input vector can better capture the dynamic changes and fluctuations of the winding wire tension in the time series. Further, considering that 1D-CNN has good perception and capture capabilities for local time series features in time series data, therefore, in the technical solution of the present application, after arranging the time series of the winding wire tension data in the time dimension as a winding wire tension time series input vector, pass the winding wire tension time series input vector through a time series encoder based on 1D-CNN to obtain a sequence of winding wire tension local time series correlation feature vectors.

[0024] In the embodiment of the present application, the winding wire tension local time-series aggregation unit 123 is configured to obtain a winding wire tension local time-series information aggregation inference feature vector by passing the sequence of the winding wire tension local time-series correlation feature vectors through an information forward propagation aggregation inference module based on node energy time-series attenuation. Correspondingly, considering that each winding wire tension local time-series correlation feature in the sequence of the winding wire tension local time-series correlation feature vectors represents the local feature information of the winding wire tension at a time point, in order to aggregate the local features of the winding wire tension at each time point to capture the dynamic change characteristics of the winding wire tension in the entire time dimension, so as to accurately infer the time-series pattern and change trend at the next time point. In the technical solution of the present application, the sequence of the winding wire tension local time-series correlation feature vectors is passed through an information forward propagation aggregation inference module based on node energy time-series attenuation to obtain a winding wire tension local time-series information aggregation inference feature vector. Specifically, the information forward propagation aggregation inference module based on node energy time-series attenuation introduces an energy attenuation mechanism and adopts the method of forward propagation of time-point information to correspondingly strengthen the features at each time point, so as to simulate the propagation mechanism of the winding wire tension in the time series, thereby enhancing the multi-representation of the entire time feature and providing better data support for the winding wire tension at the next time point. That is, the node energy time-series attenuation simulates the propagation process of information in the time series. As time goes by, the importance of early information gradually weakens, while the influence of recent information is more significant. This mechanism helps the model pay more attention to the latest data, so as to better capture the latest change trend.

[0025] Specifically, first, an energy descriptor for each feature is calculated based on the mean and standard deviation of the local temporal correlation feature vectors of each winding wire tension, so as to quantify the "energy" of the fluctuation or change of the local temporal correlation feature of each winding wire tension, that is, the dynamic range and change intensity of the data. Then, the timestamp information of each feature is extracted to obtain a sequence of winding wire tension timestamps, that is, the time point information of each feature, which helps to understand the pattern and trend of the winding wire tension changing over time. Then, the local temporal correlation feature vector of the winding wire tension at the current time point in the sequence is regarded as the current node feature vector, and the others are regarded as historical node feature vectors, and an energy decay factor for each node is calculated based on the current node feature and the historical node feature, which is used to simulate and predict the evolution trend of the feature over time. Furthermore, the historical winding wire tension propagation aggregation representation feature vector is obtained by weighted summation of the historical node features with the energy significant descriptor and the energy decay factor, so as to enhance the winding wire tension features with higher energy significance, and these features may be more important for understanding the control of the motor stator winding coil, providing rich support for the subsequent inference of the tension change at the next time point. Finally, the historical winding wire tension propagation aggregation representation feature vector and the current node feature vector are fused to obtain the winding wire tension local temporal information aggregation inference feature vector. In this way, the information from different time points can be integrated to more comprehensively understand the tension state of the winding wire and provide help for the subsequent winding control.

[0026] Specifically, Figure 4 is a block diagram of the local temporal aggregation unit of the winding wire tension in the full-automatic motor stator assembly control system according to an embodiment of the present application. As Figure 4As shown, the winding wire tension local timing aggregation unit 123 includes: a winding wire tension local timing energy significant descriptor calculation sub-unit 1231, configured to calculate an energy significant descriptor for each winding wire tension local timing correlation feature vector in the sequence of the winding wire tension local timing correlation feature vectors based on the mean and standard deviation of each winding wire tension local timing correlation feature vector in the sequence, so as to obtain a sequence of winding wire tension local timing energy significant descriptors; a winding wire tension timestamp extraction sub-unit 1232, configured to extract the timestamps of each winding wire tension local timing correlation feature vector in the sequence of the winding wire tension local timing correlation feature vectors to obtain a sequence of winding wire tension timestamps; a current-history winding wire tension local timing correlation feature differentiating sub-unit 1233, configured to use the winding wire tension local timing correlation feature vector corresponding to the current time point in the sequence of the winding wire tension local timing correlation feature vectors as the current node feature vector and use the winding wire tension local timing correlation feature vectors corresponding to other time points as the historical node feature vectors to obtain a sequence of the current winding wire tension local timing correlation feature vector and the historical winding wire tension local timing correlation feature vectors; a winding wire tension energy time decay factor calculation sub-unit 1234, configured to calculate an energy time decay factor of each historical winding wire tension local timing correlation feature vector in the sequence of the historical winding wire tension local timing correlation feature vectors relative to the current winding wire tension local timing correlation feature vector to obtain a sequence of winding wire tension energy time decay factors; a historical winding wire tension timing propagation sub-unit 1235, configured to perform timing propagation aggregation on the sequence of the historical winding wire tension local timing correlation feature vectors by using the sequence of the winding wire tension local timing energy significant descriptors as the forward adjustment factor and using the sequence of the winding wire tension energy time decay factors as the reverse adjustment factor to obtain a historical winding wire tension propagation aggregation representation feature vector; and a winding wire tension local timing information aggregation inference feature generation sub-unit 1236, configured to fuse the historical winding wire tension propagation aggregation representation feature vector and the current winding wire tension local timing correlation feature vector to obtain the winding wire tension local timing information aggregation inference feature vector.

[0027] More specifically, in the embodiments of the present application, the winding wire tension local timing energy significant descriptor calculation subunit is configured to: calculate the mean and standard deviation of the winding wire tension local timing correlation feature vectors respectively to obtain the winding wire tension local timing feature mean and the winding wire tension local timing feature standard deviation; calculate the position-wise difference between the winding wire tension local timing correlation feature vectors and the winding wire tension local timing feature mean to obtain the winding wire tension local timing difference vector; calculate the fourth power of each eigenvalue in the winding wire tension local timing difference vector to obtain the winding wire tension local timing modulation difference vector; calculate the expected value of the winding wire tension local timing modulation difference vector to obtain the winding wire tension local timing expected value; divide the winding wire tension local timing expected value by the fourth power of the winding wire tension local timing feature standard deviation to obtain the winding wire tension local timing energy significant descriptor.

[0028] More specifically, in the embodiments of the present application, the winding wire tension energy time decay factor calculation subunit is configured to: subtract the winding wire tension timestamps corresponding to the current winding wire tension local timing correlation feature vectors from the winding wire tension timestamps corresponding to the respective historical winding wire tension local timing correlation feature vectors and round down to obtain a sequence of historical winding wire tension time span values; calculate the square of the position-wise division of the sequence of historical winding wire tension time span values by the time decay inverse scaling parameter to obtain a sequence of historical winding wire tension time span squared modulation values, and then multiply the sequence of historical winding wire tension time span squared modulation values by the decay rate positive scaling parameter position-wise to obtain a sequence of winding wire tension timing span energy decay coefficients; use each winding wire tension timing span energy decay coefficient in the sequence of winding wire tension timing span energy decay coefficients as the exponent power, and calculate the natural exponential function value with the natural constant e as the base to obtain the sequence of the winding wire tension energy time decay factors.

[0029] More specifically, in the embodiments of the present application, the historical winding wire tension timing propagator subunit is configured to: divide the sequence of the winding wire tension local timing energy significant descriptors by the sequence of the winding wire tension energy time decay factors position-wise to obtain a sequence of winding wire tension adjustment factors, and calculate the position-wise weighted sum of the sequence of the winding wire tension adjustment factors and the sequence of the historical winding wire tension local timing correlation feature vectors to obtain the historical winding wire tension propagation aggregation representation feature vector.

[0030] In the embodiments of the present application, specifically, the winding wire tension local time series aggregation unit is configured to: pass the sequence of the winding wire tension local time series correlation feature vectors through an information forward propagation aggregation inference module based on node energy time series attenuation, and process it with the following forward propagation aggregation formula to obtain the winding wire tension local time series information aggregation inference feature vector; wherein, the forward propagation aggregation formula is: ; where is the sequence of the winding wire tension local time series correlation feature vectors, and are respectively the th and the th winding wire tension local time series correlation feature vectors in the sequence of the winding wire tension local time series correlation feature vectors, represents the number of feature vectors in the sequence of the winding wire tension local time series correlation feature vectors, is the eigenvalue at each position in the th winding wire tension local time series correlation feature vector in the sequence of the winding wire tension local time series correlation feature vectors, is the calculated expected value, and are respectively the mean and variance of the th winding wire tension local time series correlation feature vector, is the number of eigenvalues in each winding wire tension local time series correlation feature vector, is the energy significant descriptor of the th winding wire tension local time series correlation feature vector, is the energy significant descriptor of the th winding wire tension local time series correlation feature vector, and respectively represent the timestamps of the th and the th winding wire tension local time series correlation feature vectors, is the floor operation, is the time decay inverse scaling parameter, is the decay rate positive scaling parameter, and are weight hyperparameters, represents the value of the exponential function with the natural constant e as the base, is the winding wire tension local time series information aggregation inference feature vector.

[0031] In the embodiment of the present application, the control result generation unit 124 is configured to aggregate and infer feature vectors based on the local timing information of the winding wire tension to obtain a control result, where the control result is used to indicate whether to stop winding. Specifically, in the embodiment of the present application, the control result generation unit is configured to: pass the aggregated inference feature vector of the local timing information of the winding wire tension through a winding wire winding controller based on a classifier to obtain the control result, where the control result is used to indicate whether to stop winding. That is, the aggregated inference feature vector of the local timing information of the winding wire tension obtained by forward propagation aggregation inference using the sequence of the local timing correlation feature vectors of the winding wire tension is classified to automatically determine whether to stop winding. In this way, the tension data of the winding wire can be collected in real time, ensuring the accuracy of tension control and avoiding coil quality problems caused by uneven tension. At the same time, the deep learning model is used to analyze the timing characteristics of the tension data to automatically identify the optimal winding stop point, reducing the error of manual judgment and thus improving the automation level of stator winding control.

[0032] Particularly, considering that the sequences of the local timing correlation feature vectors of the winding wire tension respectively represent the one-dimensional local timing correlation features of the winding wire tension data in the local time domain, when performing information forward propagation aggregation inference based on the timing decay of node energy, the timing node energy distributions in each local time domain will have different aggregation inference weights based on information forward propagation, such that the aggregated inference feature vector of the local timing information of the winding wire tension will also have a diverse set expression distribution of aggregated inference features. Therefore, it is desired to improve the balance between the regression mapping accuracy and integrity when the aggregated inference feature vector of the local timing information of the winding wire tension is subjected to class regression through a winding wire winding controller based on a classifier, thereby improving the accuracy of the obtained control result.

[0033] Therefore, in a preferred example, passing the aggregated inference feature vector of the local timing information of the winding wire tension through a winding wire winding controller based on a classifier to obtain a control result includes: Determining the maximum eigenvalue of the aggregated inference of the local timing information of the winding wire tension of the aggregated inference feature vector of the local timing information of the winding wire tension 、and calculating the mean value of the aggregated inference of the local timing information of the winding wire tension of the feature set of the aggregated inference feature vector of the local timing information of the winding wire tension and the standard deviation of the aggregated inference of the local timing information of the winding wire tension ; Calculate the quotient of the aggregated inference mean of the local timing information of the winding wire tension divided by the maximum eigenvalue of the aggregated inference of the local timing information of the winding wire tension and the quotient of the standard deviation of the aggregated inference of the local timing information of the winding wire tension divided by the maximum eigenvalue of the aggregated inference of the local timing information of the winding wire tension to obtain the first predicted interval value of the aggregated inference of the local timing information of the winding wire tension and the second predicted interval value of the aggregated inference of the local timing information of the winding wire tension ; Calculate the reciprocal of each bit of the eigenvector of the aggregated inference of the local timing information of the winding wire tension, and after dot-multiplying with the first predicted interval value of the aggregated inference of the local timing information of the winding wire tension and the second predicted interval value of the aggregated inference of the local timing information of the winding wire tension, further perform dot-subtraction with the eigenvector of the aggregated inference of the local timing information of the winding wire tension and calculate the absolute value to obtain the probability fitting vector of the aggregated inference of the local timing information of the winding wire tension , where represents the eigenvector of the aggregated inference of the local timing information of the winding wire tension, represents dot-multiplication, represents dot-subtraction, represents the reciprocal of each bit of the eigenvector of the aggregated inference of the local timing information of the winding wire tension; Calculate the exponential function with the base of the natural constant and each eigenvalue of the probability fitting vector of the aggregated inference of the local timing information of the winding wire tension as the exponent, and perform dot-addition with the first predicted interval value of the aggregated inference of the local timing information of the winding wire tension and the second predicted interval value of the aggregated inference of the local timing information of the winding wire tension to obtain the predicted modeling vector of the aggregated inference of the local timing information of the winding wire tension , where represents dot-addition; Calculate the base-2 logarithm of the absolute value of each eigenvalue of the predicted modeling vector of the aggregated inference of the local timing information of the winding wire tension to obtain the optimized eigenvector of the aggregated inference of the local timing information of the winding wire tension ; Pass the optimized eigenvector of the aggregated inference of the local timing information of the winding wire tension through the winding wire coiling controller based on the classifier to obtain the control result.

[0034] That is, by performing non-parametric statistical normalization on the aggregated inference feature vector of the local time-series information of the winding wire tension, a prediction interval for the overall eigenvalue based on the aggregated inference feature vector of the local time-series information of the winding wire tension is constructed, and parametric continuous probability fitting is implemented for the causal inference verification of each component relative to the aggregated inference feature vector of the local time-series information of the winding wire tension, so as to establish the feasibility evaluation of the class-normalized feature distribution of the aggregated inference feature vector of the local time-series information of the winding wire tension to the multi-task optimization goal, to achieve the balanced executability of the input-output fidelity and information retention rate in the prediction modeling of the aggregated inference feature vector of the local time-series information of the winding wire tension, and to improve the accuracy of the control result obtained by the winding wire winding controller based on the classifier. In this way, the tension data of the winding wire can be collected in real time, ensuring the accuracy of tension control and avoiding coil quality problems caused by uneven tension. At the same time, the deep learning model is used to analyze the time-series characteristics of the tension data, so as to automatically identify the best winding stop point, reduce the error of human judgment, and thus improve the automation level of the stator winding control.

[0035] In summary, the full-automatic motor stator assembly control system 100 according to the embodiments of the present application is elucidated. It obtains the time series of the winding wire tension data collected by the tension sensor and uses the data analysis technology based on deep learning to perform local time-series feature analysis and association of the winding wire tension data, so as to automatically judge whether to stop winding according to the propagation aggregation inference feature of the winding wire tension data time-series feature information at each local time point in the entire time dimension. In this way, the tension data of the winding wire can be collected in real time, ensuring the accuracy of tension control and avoiding coil quality problems caused by uneven tension. At the same time, the deep learning model is used to analyze the time-series characteristics of the tension data, so as to automatically identify the best winding stop point, reduce the error of human judgment, and thus improve the automation level of the stator winding control.

[0036] As described above, the full-automatic motor stator assembly control system 100 according to the embodiments of the present application can be implemented in various terminal devices, such as a server for full-automatic motor stator assembly control. In one example, the full-automatic motor stator assembly control system 100 according to the embodiments of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the full-automatic motor stator assembly control system 100 can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the full-automatic motor stator assembly control system 100 can also be one of the many hardware modules of the terminal device.

[0037] Alternatively, in another example, the fully automatic motor stator assembly control system 100 and the terminal device may also be separate devices, and the fully automatic motor stator assembly control system 100 can be connected to the terminal device through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0038] Figure 5 FIG. is a flowchart of a fully automatic motor stator assembly control method according to an embodiment of the present application. As Figure 5 shown, in the fully automatic motor stator assembly control method, it includes: S110, placing the motor stator to be assembled on an assembly table and fixing the motor stator to be assembled; S120, pasting insulating paper in the iron core slots of the motor stator to be assembled and winding winding wires into the iron core slots to obtain a stator winding coil; S130, performing a shaping process on the stator winding coil to obtain a shaped stator winding coil; S140, trimming, welding, and insulating the shaped stator winding coil to obtain a shaped motor stator; wherein, in step S120, pasting insulating paper in the iron core slots of the motor stator to be assembled and winding winding wires into the iron core slots to obtain a stator winding coil includes: obtaining a time series of winding wire tension data collected by a tension sensor; arranging the time series of the winding wire tension data in a time dimension as a winding wire tension time series input vector and then passing it through a time series encoder to obtain a sequence of winding wire tension local time series correlation feature vectors; passing the sequence of the winding wire tension local time series correlation feature vectors through an information forward propagation aggregation inference module based on node energy time series attenuation to obtain a winding wire tension local time series information aggregation inference feature vector; based on the winding wire tension local time series information aggregation inference feature vector, obtaining a control result, and the control result is used to indicate whether to stop winding.

[0039] Here, those skilled in the art can understand that the specific operations of each step in the above fully automatic motor stator assembly control method have been described in detail in the description of the above Figures 1 to 4 fully automatic motor stator assembly control system, and therefore, the repeated description thereof will be omitted.

[0040] In summary, the full-automatic motor stator assembly control method based on the embodiments of the present application is elucidated. It obtains the time series of the winding wire tension data collected by the tension sensor, and uses the data analysis technology based on deep learning to perform local time series feature analysis and correlation of the winding wire tension data, so as to automatically judge whether to stop winding according to the propagation aggregation inference features of the winding wire tension data time series feature information at each local time point in the entire time dimension. In this way, the tension data of the winding wire can be collected in real time, ensuring the accuracy of tension control and avoiding coil quality problems caused by uneven tension. At the same time, the deep learning model is used to perform time series feature analysis on the tension data, so as to automatically identify the best winding stop point, reduce the error of manual judgment, and thus improve the automation level of stator winding control.

[0041] In other embodiments provided by the present application, there is also provided a full-automatic motor stator assembly control method, which includes: performing a winding operation on the traditional coil enameled wire to obtain a wound coil; inserting an insulating paper into the stator core; embedding the wound coil into the stator core to obtain a stator coil with wire inserted; performing shaping, tying, measuring, wiring, and soldering operations on the stator coil with wire inserted to obtain a finished stator coil; performing quality inspection on the finished stator coil to screen out the stator coil products without errors; and performing dip painting and baking on the stator coil products without errors to obtain the final stator product.

[0042] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0043] The above description has been given for purposes of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions, and sub-combinations thereof.

Claims

1. A fully automatic motor stator assembly control system, characterized in that: include: A motor stator fixing module, used for placing the motor stator to be assembled on the assembly table and fixing the motor stator to be assembled; A stator winding coil module, used for pasting insulating paper in the core slots of the stator of the motor to be assembled, and winding the winding wire into the core slots to obtain the stator winding coil; A stator winding coil forming module, used for forming the stator winding coil to obtain a formed stator winding coil; A motor stator forming module, used for trimming, welding and insulating the formed stator winding coils to obtain a formed motor stator; Wherein, the stator winding coil module comprises: A winding wire tension data acquisition unit, used to acquire a time series of winding wire tension data acquired by a tension sensor; A winding wire tension local time series encoding unit, used for arranging the time series of the winding wire tension data into a winding wire tension time series input vector according to the time dimension and then passing it through a time series encoder to obtain a sequence of winding wire tension local time series associated feature vectors; A winding wire tension local time series aggregation unit, used for obtaining a winding wire tension local time series information aggregation reasoning feature vector by passing the sequence of the winding wire tension local time series associated feature vectors through an information forward propagation aggregation reasoning module based on node energy time series attenuation; A control result generating unit is used to aggregate and infer feature vectors based on the local time series information of the winding wire tension to obtain a control result, wherein the control result is used to indicate whether to stop winding.

2. The fully automatic motor stator assembly control system according to claim 1, characterized in that: The winding wire tension local time series encoding unit is used to: arrange the time series of the winding wire tension data into the winding wire tension time series input vector according to the time dimension, and then pass the winding wire tension time series input vector through a time series encoder based on a deep convolutional neural network to obtain a sequence of the winding wire tension local time series associated feature vectors.

3. The fully automatic motor stator assembly control system according to claim 2, characterized in that: The time series encoder based on deep convolutional neural network is a time series encoder based on 1D-CNN.

4. The fully automatic motor stator assembly control system according to claim 3, characterized in that: The winding wire tension local timing aggregation unit comprises: A winding wire tension local time series energy significant descriptor calculation subunit, used to calculate the energy significant descriptors of each winding wire tension local time series associated feature vector in the sequence of winding wire tension local time series associated feature vectors based on the mean and standard deviation of each winding wire tension local time series associated feature vector to obtain a sequence of winding wire tension local time series energy significant descriptors; A winding wire tension timestamp extraction subunit, used to extract the timestamp of each winding wire tension local time series associated feature vector in the sequence of the winding wire tension local time series associated feature vector to obtain a sequence of winding wire tension timestamps; A current historical winding wire tension local time series correlation feature partition subunit, used to use the winding wire tension local time series correlation feature vector corresponding to the current time point in the sequence of the winding wire tension local time series correlation feature vector as the current node feature vector and use the winding wire tension local time series correlation feature vectors corresponding to other time points as the historical node feature vector to obtain a sequence of the current winding wire tension local time series correlation feature vector and the historical winding wire tension local time series correlation feature vector; A winding wire tension energy time decay factor calculation subunit, used to calculate the energy time decay factor of each historical winding wire tension local time series associated characteristic vector in the sequence of historical winding wire tension local time series associated characteristic vectors relative to the current winding wire tension local time series associated characteristic vector to obtain a sequence of winding wire tension energy time decay factors; A historical winding wire tension time series propagation subunit, used to perform time series propagation aggregation on the sequence of the historical winding wire tension local time series associated feature vectors using the sequence of the winding wire tension local time series energy significant descriptors as positive adjustment factors and the sequence of the winding wire tension energy time decay factors as negative adjustment factors to obtain a historical winding wire tension propagation aggregation representation feature vector; The winding wire tension local time series information aggregation reasoning feature generation subunit is used to fuse the historical winding wire tension propagation aggregation representation feature vector and the current winding wire tension local time series association feature vector to obtain the winding wire tension local time series information aggregation reasoning feature vector.

5. The fully automatic motor stator assembly control system according to claim 4, characterized in that: The winding wire tension local time series energy significant descriptor calculation subunit is used for: Calculating the mean and standard deviation of the local time series associated characteristic vector of the winding wire tension respectively to obtain the mean of the local time series characteristic of the winding wire tension and the standard deviation of the local time series characteristic of the winding wire tension; Calculating the position difference between the winding wire tension local time series associated feature vector and the winding wire tension local time series feature mean to obtain a winding wire tension local time series difference vector; Calculating the fourth power of each eigenvalue in the winding wire tension local timing difference vector to obtain the winding wire tension local timing modulation difference vector; Calculating the expected value of the winding wire tension local timing modulation difference vector to obtain the winding wire tension local timing expected value; The expected value of the winding wire tension local time series is divided by the fourth power of the standard deviation of the winding wire tension local time series feature to obtain the winding wire tension local time series energy significant descriptor.

6. The fully automatic motor stator assembly control system according to claim 5, characterized in that: The winding wire tension energy time decay factor calculation subunit is used for: Subtracting the winding wire tension timestamp corresponding to the current winding wire tension local time series associated feature vector from the winding wire tension timestamp corresponding to each historical winding wire tension local time series associated feature vector and rounding down to obtain a sequence of historical winding wire tension time span values; After calculating the square of the positional division of the sequence of historical winding wire tension time span values ​​and the time attenuation inverse scaling parameter to obtain a sequence of historical winding wire tension time span square modulation values, the sequence of historical winding wire tension time span square modulation values ​​and the attenuation rate positive scaling parameter are positionally multiplied to obtain a sequence of winding wire tension time series span energy attenuation coefficients; Taking each winding wire tension time series span energy attenuation coefficient in the sequence of winding wire tension time series span energy attenuation coefficients as an exponential power, a natural exponential function value with a natural constant e as a base is calculated to obtain a sequence of winding wire tension energy time attenuation factors.

7. The fully automatic motor stator assembly control system according to claim 6, characterized in that: The historical winding wire tension timing propagation subunit is used to: divide the sequence of the winding wire tension local timing energy significant descriptors by the sequence of the winding wire tension energy time decay factors by position to obtain a sequence of winding wire tension adjustment factors, and calculate the position weighted sum of the sequence of the winding wire tension adjustment factors and the sequence of the historical winding wire tension local timing associated feature vectors to obtain the historical winding wire tension propagation aggregate representation feature vector.

8. The fully automatic motor stator assembly control system according to claim 7, characterized in that: The control result generating unit is used to aggregate the inference feature vector of the local time series information of the winding wire tension through a classifier-based winding wire winding controller to obtain the control result, and the control result is used to indicate whether to stop winding.

9. A fully automatic motor stator assembly control method, characterized in that: include: Placing the motor stator to be assembled on an assembly table, and fixing the motor stator to be assembled; Pasting insulating paper in the core slots of the stator of the motor to be assembled, and winding the winding wire into the core slots to obtain the stator winding coil; Performing a forming process on the stator winding coil to obtain a formed stator winding coil; The formed stator winding coils are trimmed, welded and insulated to obtain a formed motor stator; The method includes pasting insulating paper in the core slots of the stator of the motor to be assembled, and winding the winding wire into the core slots to obtain the stator winding coil, including: Acquire a time series of winding wire tension data collected by a tension sensor; Arranging the time series of the winding wire tension data according to the time dimension into a winding wire tension time series input vector and then passing it through a time series encoder to obtain a sequence of winding wire tension local time series associated feature vectors; The sequence of the winding wire tension local time series associated feature vectors is passed through an information forward propagation aggregation reasoning module based on node energy time series attenuation to obtain a winding wire tension local time series information aggregation reasoning feature vector; Based on the local time series information of the winding wire tension, a feature vector is aggregated and inferred to obtain a control result, and the control result is used to indicate whether to stop winding.

10. The fully automatic motor stator assembly control method according to claim 9, characterized in that: The time series of the winding wire tension data is arranged according to the time dimension as the winding wire tension timing input vector, and then passed through a timing encoder to obtain a sequence of winding wire tension local timing associated feature vectors, including: after the time series of the winding wire tension data is arranged according to the time dimension as the winding wire tension timing input vector, the winding wire tension timing input vector is passed through a timing encoder based on a deep convolutional neural network to obtain a sequence of winding wire tension local timing associated feature vectors.

Images