Embroidery machine thread breakage detection method and system

By constructing and deploying a disconnection detection model, combined with the detection of infrared sensor groups, the problem of insufficient accuracy and timeline detection of the embroidery machine is solved, efficient disconnection detection is achieved, power consumption is reduced, and detection reliability is improved Sex and traceability.

CN119083065AActive Publication Date: 2024-12-06FUZHOU HUICHUANG INTELLIGENT CONTROL TECHNOLOGY CO LTD
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202411246886.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-06
Publication Date
2024-12-06
Estimated Expiration
2044-09-06

AI Technical Summary

Technical Problem

The existing embroidery machine thread-break detection methods have problems such as low accuracy, insufficient timeliness and large power consumption, especially when mechanical jitter and long-term opening of infrared sensor groups.

Method used

By acquiring and preprocessing the historical data of the embroidery machine, a data set is constructed and a broken line detection model is created using the feature extraction module, feature fusion module and broken line detection module. The double verification is performed in combination with the detection of the infrared sensor group to achieve the breaking of the embroidery thread. Accurate detection of lines and improve model performance by optimizing hyperparameters and real-time logging.

Benefits of technology

It significantly improves the accuracy and timeline detection of the embroidery machine, reduces power consumption, and ensures the reliability and traceability of the disconnection detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119083065A_ABST
    Figure CN119083065A_ABST
Patent Text Reader

Abstract

The invention provides a method and a system for detecting broken threads of an embroidery machine in the technical field of embroidery machines. The method comprises the following steps: S1, acquiring a large amount of embroidery historical data to construct a data set; s2, creating a broken line detection model and setting a hyper-parameter group; s3, training the broken line detection model based on the data set and deploying the broken line detection model on an embroidery machine; s4, the embroidery machine obtains the embroidery operation file to control the corresponding embroidery needle to work; s5, starting an infrared sensor group at a corresponding needle position by the embroidery machine based on the currently working embroidery needle to carry out broken line detection, and obtaining a first detection result; collecting current embroidery pulse data, and inputting the embroidery pulse data into the broken thread detection model to obtain a second detection result; and S6, generating a disconnection detection report based on the first detection result and the second detection result, and giving an alarm based on the disconnection detection report. The method has the advantages that the precision and timeliness of embroidery machine thread breakage detection are greatly improved, and the power consumption is greatly reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of embroidery machines, and in particular to a method and system for detecting thread breakage in embroidery machines. Background Art

[0002] Embroidery machine, also known as computer embroidery machine, is the most advanced embroidery machine in contemporary times. It can make traditional hand embroidery be realized with high speed and high efficiency, and can also realize the requirements of "multi-level, multi-function, unity and perfection" that hand embroidery cannot achieve. It is an electromechanical product that embodies a variety of high-tech. As computer embroidery replaces hand embroidery, computer embroidery machine will also become the main machine type in the embroidery industry.

[0003] Embroidery threads are needed during embroidery machine embroidery. Due to the differences in embroidery techniques, there are many types of embroidery threads. Some types of embroidery threads will inevitably break during embroidery machine embroidery. If the breakage is not discovered in time and the thread is not replaced, the embroidery product will be unqualified. Therefore, there is a need to detect whether the embroidery thread is broken.

[0004] Traditionally, infrared sensors are used to detect broken embroidery threads. That is, an infrared transmitter and an infrared receiver are set on both sides of the embroidery thread. Under normal conditions, the infrared signal emitted by the infrared transmitter is blocked by the embroidery thread and cannot reach the infrared receiver. When the thread is broken, the infrared receiver receives the infrared signal. However, the traditional detection method has the following disadvantages:

[0005] 1. Since the embroidery machine will inevitably produce mechanical vibrations when working, this will cause the embroidery thread or the infrared sensor group to move, thereby affecting the accuracy of thread break detection; 2. An embroidery machine contains many embroidery needles, and the embroidery thread of each embroidery needle must be paired with a group of infrared sensor groups, and each infrared sensor group is in a long-term open state, and the power consumption remains high; 3. Detection can only be performed after the thread break occurs, which has already affected the production rhythm. In addition, due to the negligence of the staff, the broken thread may not be replaced in time, which seriously affects the embroidery efficiency.

[0006] Therefore, how to provide an embroidery machine thread breakage detection method and system to improve the accuracy and timeliness of embroidery machine thread breakage detection and reduce power consumption has become a technical problem that needs to be solved urgently. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a method and system for detecting thread breakage of an embroidery machine, so as to improve the accuracy and timeliness of the detection of thread breakage of the embroidery machine and reduce power consumption.

[0008] In a first aspect, the present invention provides a method for detecting thread breakage in an embroidery machine, comprising the following steps:

[0009] Step S1, obtaining a large amount of embroidery history data when the embroidery machine is working, and constructing a data set after preprocessing and marking each of the embroidery history data;

[0010] Step S2: creating a disconnection detection model based on the feature extraction module, the feature fusion module and the disconnection detection module, and setting a hyperparameter group of the disconnection detection model;

[0011] Step S3, training a thread break detection model based on the data set, and deploying the trained thread break detection model on an embroidery machine;

[0012] Step S4: the embroidery machine obtains an embroidery operation file, and controls the corresponding embroidery needle to work based on the embroidery operation file;

[0013] Step S5: the embroidery machine starts the infrared sensor group corresponding to the embroidery needle position to perform thread breakage detection based on the currently working embroidery needle, and obtains a first detection result; collects current embroidery pulse data, inputs the embroidery pulse data into the thread breakage detection model, and obtains a second detection result;

[0014] Step S6: the embroidery machine generates a thread breakage detection report based on the first detection result and the second detection result, and issues an alarm based on the thread breakage detection report;

[0015] Step S7, the embroidery machine records the embroidery log in real time, optimizes the thread breakage detection model based on the embroidery log, and stores the embroidery log;

[0016] The step S1 is specifically as follows:

[0017] Acquire a large amount of embroidery history data when the embroidery machine is working, wherein the embroidery history data is pulse data obtained when the infrared signal of the infrared sensor group is blocked by the grating wheel;

[0018] Each of the embroidery historical data is preprocessed including at least outlier removal and noise suppression, and each of the preprocessed embroidery historical data is marked with at least a thread break position and a position that is about to break. The thread break position and the position that is about to break are marked based on the selection of a target frame, and then a data set is constructed based on the marked embroidery historical data.

[0019] Furthermore, in step S2, the feature extraction module adopts a separable convolution structure, including several feature blocks of different scales; each of the feature blocks includes a convolution operation unit, a skip connection unit, a pooling unit and a SE unit;

[0020] The feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse the multi-scale disconnection features output by the feature extraction module from bottom to top and from top to bottom;

[0021] The disconnection detection module is constructed based on a multi-layer convolutional neural network unit, a pooling unit, and a fully connected network unit, and is used to predict the size and position of the disconnection detection box;

[0022] The hyperparameter group includes at least a loss function, a learning rate, a batch size, an optimizer, a learning decay rate, and a discard rate;

[0023] The loss function adopts the cross entropy loss function.

[0024] Furthermore, the step S3 is specifically as follows:

[0025] The data set is divided into a training set, a test set, and a validation set based on a ratio of 6:2:2, and a disconnection detection model is trained using the training set based on a learning rate, a batch size, a learning decay rate, and a discard rate carried by the hyperparameter group to learn a pulse law until a loss value of a loss function carried by the hyperparameter group is less than a preset loss threshold, and each hyperparameter in the hyperparameter group is continuously optimized during the training process;

[0026] The trained disconnection detection model is tested by the test set to determine whether the detection accuracy is greater than a preset accuracy threshold. If not, the test fails, and the training set is expanded to continue training the disconnection detection model. If yes, the test passes, and:

[0027] The broken thread detection model that has passed the test is verified by using the verification set to determine whether the confidence of the broken thread detection model is greater than a preset confidence threshold. If not, the verification fails, and the training set is expanded to continue training the broken thread detection model. If so, the verification passes, the training of the broken thread detection model is terminated, and the verified broken thread detection model is deployed on the embroidery machine.

[0028] Furthermore, the step S4 is specifically as follows:

[0029] The embroidery machine obtains an embroidery operation file, parses the embroidery operation file to obtain an embroidery path carrying embroidery needle positions, and controls the corresponding embroidery needle to work based on the embroidery path;

[0030] The step S5 is specifically as follows:

[0031] The embroidery machine starts the infrared sensor group corresponding to the needle position based on the currently working embroidery needle, performs thread breakage detection based on whether the infrared receiver of the infrared sensor group receives the infrared signal of the infrared transmitter, and obtains a first detection result;

[0032] The current embroidery pulse data of the infrared sensor group of the embroidery machine is collected, and the embroidery pulse data is preprocessed including at least outlier removal and noise suppression, and the preprocessed embroidery pulse data is input into the thread breakage detection model to obtain a second detection result.

[0033] Furthermore, the step S6 is specifically as follows:

[0034] The embroidery machine analyzes the first detection result and the second detection result, and generates a thread breakage detection report when both the first detection result and the second detection result are thread breakage; generates a thread breakage detection report when the first detection result or the second detection result is thread breakage; generates a thread breakage detection report indicating that the embroidery thread needs to be replaced when the first detection result is that the thread is not broken and the second detection result is that the thread is about to be broken;

[0035] Displaying the disconnection detection report on a large screen, controlling the operation of an indicator light and a buzzer based on the disconnection detection report, and pushing the disconnection detection report to a pre-managed management terminal in real time to generate an alarm;

[0036] The step S7 is specifically as follows:

[0037] The embroidery machine records in real time an embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, optimizes the thread break detection model based on the embroidery log, and performs encrypted storage and distributed backup on the embroidery log.

[0038] In a second aspect, the present invention provides an embroidery machine thread breakage detection system, comprising the following modules:

[0039] A data set construction module is used to obtain a large amount of embroidery historical data when the embroidery machine is working, and construct a data set after preprocessing and marking each of the embroidery historical data;

[0040] A disconnection detection model creation module is used to create a disconnection detection model based on the feature extraction module, the feature fusion module and the disconnection detection module, and set a hyperparameter group of the disconnection detection model;

[0041] A thread break detection model training module, used for training the thread break detection model based on the data set, and deploying the trained thread break detection model on the embroidery machine;

[0042] The embroidery module is used for the embroidery machine to obtain an embroidery operation file, and control the corresponding embroidery needle to work based on the embroidery operation file;

[0043] The thread breakage detection module is used for the embroidery machine to start the infrared sensor group corresponding to the embroidery needle position to perform thread breakage detection based on the currently working embroidery needle, and obtain a first detection result; collect the current embroidery pulse data, input the embroidery pulse data into the thread breakage detection model, and obtain a second detection result;

[0044] A thread breakage detection report generating module, used for the embroidery machine to generate a thread breakage detection report based on the first detection result and the second detection result, and to issue an alarm based on the thread breakage detection report;

[0045] An embroidery log management module is used for the embroidery machine to record the embroidery log in real time, optimize the thread breakage detection model based on the embroidery log, and store the embroidery log;

[0046] The data set construction module is specifically used for:

[0047] Acquire a large amount of embroidery history data when the embroidery machine is working, wherein the embroidery history data is pulse data obtained when the infrared signal of the infrared sensor group is blocked by the grating wheel;

[0048] Each of the embroidery historical data is preprocessed including at least outlier removal and noise suppression, and each of the preprocessed embroidery historical data is marked with at least a thread break position and a position that is about to break. The thread break position and the position that is about to break are marked based on the selection of a target frame, and then a data set is constructed based on the marked embroidery historical data.

[0049] Furthermore, in the disconnection detection model creation module, the feature extraction module adopts a separable convolution structure, including several feature blocks of different scales; each of the feature blocks includes a convolution operation unit, a jump connection unit, a pooling unit and a SE unit;

[0050] The feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse the multi-scale disconnection features output by the feature extraction module from bottom to top and from top to bottom;

[0051] The disconnection detection module is constructed based on a multi-layer convolutional neural network unit, a pooling unit, and a fully connected network unit, and is used to predict the size and position of the disconnection detection box;

[0052] The hyperparameter group includes at least a loss function, a learning rate, a batch size, an optimizer, a learning decay rate, and a discard rate;

[0053] The loss function adopts the cross entropy loss function.

[0054] Furthermore, the disconnection detection model training module is specifically used for:

[0055] The data set is divided into a training set, a test set, and a validation set based on a ratio of 6:2:2, and a disconnection detection model is trained using the training set based on a learning rate, a batch size, a learning decay rate, and a discard rate carried by the hyperparameter group to learn a pulse law until a loss value of a loss function carried by the hyperparameter group is less than a preset loss threshold, and each hyperparameter in the hyperparameter group is continuously optimized during the training process;

[0056] The trained disconnection detection model is tested by the test set to determine whether the detection accuracy is greater than a preset accuracy threshold. If not, the test fails, and the training set is expanded to continue training the disconnection detection model. If yes, the test passes, and:

[0057] The broken thread detection model that has passed the test is verified by using the verification set to determine whether the confidence of the broken thread detection model is greater than a preset confidence threshold. If not, the verification fails, and the training set is expanded to continue training the broken thread detection model. If so, the verification passes, the training of the broken thread detection model is terminated, and the verified broken thread detection model is deployed on the embroidery machine.

[0058] Furthermore, the embroidery module is specifically used for:

[0059] The embroidery machine obtains an embroidery operation file, parses the embroidery operation file to obtain an embroidery path carrying embroidery needle positions, and controls the corresponding embroidery needle to work based on the embroidery path;

[0060] The disconnection detection module is specifically used for:

[0061] The embroidery machine starts the infrared sensor group corresponding to the needle position based on the currently working embroidery needle, performs thread breakage detection based on whether the infrared receiver of the infrared sensor group receives the infrared signal of the infrared transmitter, and obtains a first detection result;

[0062] The current embroidery pulse data of the infrared sensor group of the embroidery machine is collected, and the embroidery pulse data is preprocessed including at least outlier removal and noise suppression, and the preprocessed embroidery pulse data is input into the thread breakage detection model to obtain a second detection result.

[0063] Furthermore, the disconnection detection report generating module is specifically used for:

[0064] The embroidery machine analyzes the first detection result and the second detection result, and generates a thread breakage detection report when both the first detection result and the second detection result are thread breakage; generates a thread breakage detection report when the first detection result or the second detection result is thread breakage; generates a thread breakage detection report indicating that the embroidery thread needs to be replaced when the first detection result is that the thread is not broken and the second detection result is that the thread is about to be broken;

[0065] Displaying the disconnection detection report on a large screen, controlling the operation of an indicator light and a buzzer based on the disconnection detection report, and pushing the disconnection detection report to a pre-managed management terminal in real time to generate an alarm;

[0066] The embroidery log management module is specifically used for:

[0067] The embroidery machine records in real time an embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, optimizes the thread break detection model based on the embroidery log, and performs encrypted storage and distributed backup on the embroidery log.

[0068] The advantages of the present invention are:

[0069] 1. By acquiring a large amount of embroidery historical data when the embroidery machine is working, each embroidery historical data is pre-processed and annotated to construct a data set; then, based on the feature extraction module, the feature fusion module and the thread break detection module, a thread break detection model is created, a hyperparameter group of the thread break detection model is set, the thread break detection model is trained based on the data set, and the trained thread break detection model is deployed on the embroidery machine; the embroidery machine obtains the embroidery operation file, controls the corresponding embroidery needle to work based on the embroidery operation file, starts the infrared sensor group of the corresponding embroidery needle position based on the currently working embroidery needle to perform thread break detection, and obtains a first detection result; collects the current embroidery pulse data, inputs the embroidery pulse data into the thread break detection model, obtains the second detection result, generates a thread break detection report based on the first detection result and the second detection result, issues an alarm based on the thread break detection report, and records the embroidery daily in real time. The thread break detection model is optimized based on the embroidery log, and the embroidery log is stored; that is, by combining the infrared detection of the infrared sensor group and the pulse regularity detection of the thread break detection model, the thread break of the embroidery thread can be double-checked and judged to ensure the accuracy of the thread break detection; based on the currently working embroidery needle, the infrared sensor group corresponding to the embroidery needle position is started to perform thread break detection, and the other infrared sensor groups are in a power-off state, so as to avoid all infrared sensor groups being in a long-term open state, and effectively reduce the power consumption of the thread break detection; by marking the position of the thread break that is about to break in the embroidery historical data, and then using the marked embroidery historical data to train the thread break detection model, the thread break detection model has the ability to predict whether the embroidery thread is about to break, and the embroidery thread that is about to break is replaced in time, which ultimately greatly improves the accuracy and timeliness of the embroidery machine's thread break detection and greatly reduces the power consumption.

[0070] 2. By preprocessing each embroidery historical data including at least outlier removal and noise suppression, the data quality is effectively improved, thereby greatly improving the training speed of the thread break detection model.

[0071] 3. By marking the pre-processed embroidery historical data at least including the thread break position and the position of the thread about to break, and then building a data set based on the marked embroidery historical data, the thread break detection model is trained using the data set. This enables the thread break detection model to not only detect the thread break position, but also predict whether the thread is about to break, greatly improving the reliability of the embroidery machine's thread break detection.

[0072] 4. The disconnection detection model is trained through the training set until the loss value of the loss function is less than the preset loss threshold. The hyperparameters in the hyperparameter group are continuously optimized during the training process. The detection accuracy of the trained disconnection detection model is tested through the test set. The confidence of the disconnection detection model that passes the test is verified through the verification set. That is, the disconnection detection model is continuously optimized during the training process, and relevant tests and verifications are performed, thereby greatly improving the detection accuracy of the disconnection detection model.

[0073] 5. By recording the embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report in real time, the thread break detection model is optimized based on the embroidery log, which further improves the detection accuracy of the thread break detection model. By encrypting and storing the embroidery log and performing distributed backup, the traceability is greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The present invention will be further described below in conjunction with embodiments with reference to the accompanying drawings.

[0075] Figure 1 The present invention is a flow chart of a method for detecting thread breakage in an embroidery machine.

[0076] Figure 2 The invention discloses a structural schematic diagram of a thread breakage detection system for an embroidery machine. DETAILED DESCRIPTION

[0077] The technical solution in the embodiment of the present application has the following overall idea: by combining the infrared detection of the infrared sensor group and the pulse regularity detection of the thread break detection model, the thread break of the embroidery thread can be double-checked and judged to ensure the accuracy of the thread break detection; based on the currently working embroidery needle, the infrared sensor group corresponding to the embroidery needle position is started to perform thread break detection, and the other infrared sensor groups are in a power-off state, so as to avoid all infrared sensor groups being in a long-term on state, and effectively reduce the power consumption of the thread break detection; by marking the position of the thread break that is about to break in the embroidery historical data, and then using the marked embroidery historical data to train the thread break detection model, the thread break detection model has the ability to predict whether the embroidery thread is about to break, and to replace the embroidery thread that is about to break in time, thereby improving the accuracy and timeliness of the embroidery machine's thread break detection and reducing power consumption.

[0078] Please refer to Figure 1 to Figure 2 As shown, a preferred embodiment of a method for detecting thread breakage in an embroidery machine according to the present invention comprises the following steps:

[0079] Step S1, obtaining a large amount of embroidery history data when the embroidery machine is working, and constructing a data set after preprocessing and marking each of the embroidery history data;

[0080] Step S2: creating a disconnection detection model based on the feature extraction module, the feature fusion module and the disconnection detection module, and setting a hyperparameter group of the disconnection detection model;

[0081] Step S3, training a thread break detection model based on the data set, and deploying the trained thread break detection model on an embroidery machine;

[0082] Step S4: the embroidery machine obtains an embroidery operation file, and controls the corresponding embroidery needle to work based on the embroidery operation file;

[0083] Step S5: the embroidery machine starts the infrared sensor group corresponding to the embroidery needle position to perform thread breakage detection based on the currently working embroidery needle, and obtains a first detection result; collects current embroidery pulse data, inputs the embroidery pulse data into the thread breakage detection model, and obtains a second detection result;

[0084] Step S6: the embroidery machine generates a thread breakage detection report based on the first detection result and the second detection result, and issues an alarm based on the thread breakage detection report;

[0085] Step S7, the embroidery machine records the embroidery log in real time, optimizes the thread breakage detection model based on the embroidery log, and stores the embroidery log;

[0086] The step S1 is specifically as follows:

[0087] Acquire a large amount of embroidery history data when the embroidery machine is working, wherein the embroidery history data is pulse data obtained when the infrared signal of the infrared sensor group is blocked by the grating wheel;

[0088] Each of the embroidery historical data is preprocessed including at least outlier removal and noise suppression, and each of the preprocessed embroidery historical data is marked with at least a thread break position and a position that is about to break. The thread break position and the position that is about to break are marked based on the selection of a target frame, and then a data set is constructed based on the marked embroidery historical data.

[0089] By preprocessing each embroidery history data including at least outlier removal and noise suppression, the data quality is effectively improved, thereby greatly improving the training speed of the thread break detection model.

[0090] By marking at least the thread break position and the position of the imminent thread break for each preprocessed embroidery historical data, and then constructing a data set based on the marked embroidery historical data, the thread break detection model is trained using the data set. This enables the thread break detection model to not only detect the thread break position, but also predict whether the thread is about to break, greatly improving the reliability of thread break detection in the embroidery machine.

[0091] In step S2, the feature extraction module adopts a separable convolution structure to extract the broken thread features of the embroidery thread, including several feature blocks of different scales; each of the feature blocks includes a convolution operation unit, a jump connection unit, a pooling unit and an SE unit (Squeeze-and-Excitat ion), and each unit is repeated N times;

[0092] The separable convolution structure modifies the traditional convolution operation into a two-layer convolution operation, decoupling the spatial and depth information, thereby effectively utilizing the model parameters. For example, a feature extraction module is formed by superimposing 9 feature blocks, the first 3 feature blocks are used for underlying feature mapping, and the last 6 feature blocks are used for disconnection feature extraction of different scales.

[0093] The feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse the multi-scale disconnection features output by the feature extraction module from bottom to top and from top to bottom to obtain fused features;

[0094] The feature fusion module is constructed based on the weighted bidirectional feature pyramid network, which can perform feature fusion in two directions (bottom-up and top-down), integrating high-level and low-level features, making the fused features more expressive and greatly improving the network construction efficiency.

[0095] The disconnection detection module is constructed based on a multi-layer convolutional neural network unit, a pooling unit, and a fully connected network unit, and is used to predict the size and position of the disconnection detection box according to the fusion features;

[0096] The hyperparameter group includes at least a loss function, a learning rate, a batch size, an optimizer, a learning decay rate, and a discard rate;

[0097] The loss function adopts the cross entropy loss function.

[0098] The step S3 is specifically as follows:

[0099] The data set is divided into a training set, a test set, and a validation set based on a ratio of 6:2:2, and a disconnection detection model is trained using the training set based on a learning rate, a batch size, a learning decay rate, and a discard rate carried by the hyperparameter group to learn a pulse law until a loss value of a loss function carried by the hyperparameter group is less than a preset loss threshold, and each hyperparameter in the hyperparameter group is continuously optimized during the training process;

[0100] The trained disconnection detection model is tested by the test set to determine whether the detection accuracy is greater than a preset accuracy threshold. If not, the test fails, and the training set is expanded to continue training the disconnection detection model. If yes, the test passes, and:

[0101] The broken thread detection model that has passed the test is verified by using the verification set to determine whether the confidence of the broken thread detection model is greater than a preset confidence threshold. If not, the verification fails, and the training set is expanded to continue training the broken thread detection model. If so, the verification passes, the training of the broken thread detection model is terminated, and the verified broken thread detection model is deployed on the embroidery machine.

[0102] The disconnection detection model is trained through the training set until the loss value of the loss function is less than the preset loss threshold. The hyperparameters in the hyperparameter group are continuously optimized during the training process. The detection accuracy of the trained disconnection detection model is tested through the test set, and the confidence of the disconnection detection model that passes the test is verified through the verification set. That is, the disconnection detection model is continuously optimized during the training process, and relevant tests and verifications are performed, thereby greatly improving the detection accuracy of the disconnection detection model.

[0103] In specific implementation, during the training process of the disconnection detection model, the disconnection detection frames of the disconnection detection model are merged and filtered using a soft connection non-maximum suppression algorithm, and a composite scaling algorithm (Neura l Architecture Search) is used to perform a network search operation on the feature extraction module, the feature fusion module and the disconnection detection module to determine the scaling coefficient of each dimension, and the computing power of each module is allocated using the scaling coefficient to select a network structure with the best efficiency and accuracy, thereby expanding the precision and accuracy of disconnection detection.

[0104] Since there are a lot of overlaps and invalid results in the predicted disconnection detection frames, the soft connection non-maximum suppression algorithm is used to filter the redundant disconnection detection frames according to the recognition probability value, merge some disconnection detection frames, and output the optimal disconnection detection frame.

[0105] The network search operation is performed through a composite scaling algorithm to determine the scaling coefficient of each dimension, that is, to solidify the network search space and path, effectively reducing the time complexity of the network search and thus improving the efficiency of disconnection detection.

[0106] The step S4 is specifically as follows:

[0107] The embroidery machine obtains an embroidery operation file, parses the embroidery operation file to obtain an embroidery path carrying embroidery needle positions, and controls the corresponding embroidery needle to work based on the embroidery path;

[0108] The step S5 is specifically as follows:

[0109] The embroidery machine starts the infrared sensor group corresponding to the needle position based on the currently working embroidery needle, performs thread breakage detection based on whether the infrared receiver of the infrared sensor group receives the infrared signal of the infrared transmitter, and obtains a first detection result;

[0110] The current embroidery pulse data of the infrared sensor group of the embroidery machine is collected, and the embroidery pulse data is preprocessed including at least outlier removal and noise suppression, and the preprocessed embroidery pulse data is input into the thread breakage detection model to obtain a second detection result.

[0111] The step S6 is specifically as follows:

[0112] The embroidery machine analyzes the first detection result and the second detection result, and generates a thread breakage detection report when both the first detection result and the second detection result are thread breakage; generates a thread breakage detection report when the first detection result or the second detection result is thread breakage; generates a thread breakage detection report indicating that the embroidery thread needs to be replaced when the first detection result is that the thread is not broken and the second detection result is that the thread is about to be broken;

[0113] Displaying the disconnection detection report on a large screen, controlling the operation of an indicator light and a buzzer based on the disconnection detection report, and pushing the disconnection detection report to a pre-managed management terminal in real time to generate an alarm;

[0114] The step S7 is specifically as follows:

[0115] The embroidery machine records in real time an embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, optimizes the thread break detection model based on the embroidery log, and performs encrypted storage and distributed backup on the embroidery log.

[0116] By recording in real time the embroidery log including at least the embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, the thread break detection model is optimized based on the embroidery log, which further improves the detection accuracy of the thread break detection model. By encrypting and storing the embroidery log and performing distributed backup, the traceability is greatly improved.

[0117] The encrypted storage and distributed backup of the embroidery log are specifically as follows:

[0118] A pair of public keys and private keys are created based on the RSA algorithm, an HMAC calculation is performed on the embroidery log to obtain an HMAC value, the embroidery log and the HMAC value are encrypted by the private key to obtain primary encrypted data, the primary encrypted data is mapped by a preset first mapping rule to obtain secondary encrypted data, the number 3 and the letter s in the secondary encrypted data are swapped to obtain the third-level encrypted data; the public key is mapped by a preset second mapping rule to obtain a primary key, the primary key is encrypted by the 3DES algorithm to obtain a secondary key, the third-level encrypted data and the primary and secondary keys are encrypted into an encrypted log by the XTEA algorithm, the encrypted log is stored in a file system, and the encrypted log is backed up in a distributed manner.

[0119] Since data encrypted by the private key can only be decrypted by the public key, the public key is encrypted at multiple levels, the embroidery log is also encrypted at multiple levels, and the HMAC value can also be used to verify the integrity of the embroidery log. The HMAC calculation has higher security than the ordinary hash calculation. If you don’t know the corresponding encryption algorithm or data transformation rule, you will not be able to crack the encrypted log. At least 9 security measures have been taken (RSA algorithm, public and private keys, HMAC calculation, first mapping rule, 3 and s swap, second mapping rule, 3DES algorithm, XTEA algorithm, distributed backup), which greatly improves the security of embroidery log storage.

[0120] A preferred embodiment of an embroidery machine thread breakage detection system of the present invention comprises the following modules:

[0121] A data set construction module is used to obtain a large amount of embroidery historical data when the embroidery machine is working, and construct a data set after preprocessing and marking each of the embroidery historical data;

[0122] A disconnection detection model creation module is used to create a disconnection detection model based on the feature extraction module, the feature fusion module and the disconnection detection module, and set a hyperparameter group of the disconnection detection model;

[0123] A thread break detection model training module, used for training the thread break detection model based on the data set, and deploying the trained thread break detection model on the embroidery machine;

[0124] The embroidery module is used for the embroidery machine to obtain an embroidery operation file, and control the corresponding embroidery needle to work based on the embroidery operation file;

[0125] The thread breakage detection module is used for the embroidery machine to start the infrared sensor group corresponding to the embroidery needle position to perform thread breakage detection based on the currently working embroidery needle, and obtain a first detection result; collect the current embroidery pulse data, input the embroidery pulse data into the thread breakage detection model, and obtain a second detection result;

[0126] A thread breakage detection report generating module, used for the embroidery machine to generate a thread breakage detection report based on the first detection result and the second detection result, and to issue an alarm based on the thread breakage detection report;

[0127] An embroidery log management module is used for the embroidery machine to record the embroidery log in real time, optimize the thread breakage detection model based on the embroidery log, and store the embroidery log;

[0128] The data set construction module is specifically used for:

[0129] Acquire a large amount of embroidery history data when the embroidery machine is working, wherein the embroidery history data is pulse data obtained when the infrared signal of the infrared sensor group is blocked by the grating wheel;

[0130] Each of the embroidery historical data is preprocessed including at least outlier removal and noise suppression, and each of the preprocessed embroidery historical data is marked with at least a thread break position and a position that is about to break. The thread break position and the position that is about to break are marked based on the selection of a target frame, and then a data set is constructed based on the marked embroidery historical data.

[0131] By preprocessing each embroidery history data including at least outlier removal and noise suppression, the data quality is effectively improved, thereby greatly improving the training speed of the thread break detection model.

[0132] By marking at least the thread break position and the position of the imminent thread break for each preprocessed embroidery historical data, and then constructing a data set based on the marked embroidery historical data, the thread break detection model is trained using the data set. This enables the thread break detection model to not only detect the thread break position, but also predict whether the thread is about to break, greatly improving the reliability of thread break detection in the embroidery machine.

[0133] In the thread break detection model creation module, the feature extraction module adopts a separable convolution structure to extract the thread break features of the embroidery thread, including several feature blocks of different scales; each of the feature blocks includes a convolution operation unit, a jump connection unit, a pooling unit and an SE unit (Squeeze-and-Excitation), and each unit is repeated N times;

[0134] The separable convolution structure modifies the traditional convolution operation into a two-layer convolution operation, decoupling the spatial and depth information, thereby effectively utilizing the model parameters. For example, a feature extraction module is formed by superimposing 9 feature blocks, the first 3 feature blocks are used for underlying feature mapping, and the last 6 feature blocks are used for disconnection feature extraction of different scales.

[0135] The feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse the multi-scale disconnection features output by the feature extraction module from bottom to top and from top to bottom to obtain fused features;

[0136] The feature fusion module is constructed based on the weighted bidirectional feature pyramid network, which can perform feature fusion in two directions (bottom-up and top-down), integrating high-level and low-level features, making the fused features more expressive and greatly improving the network construction efficiency.

[0137] The disconnection detection module is constructed based on a multi-layer convolutional neural network unit, a pooling unit, and a fully connected network unit, and is used to predict the size and position of the disconnection detection box according to the fusion features;

[0138] The hyperparameter group includes at least a loss function, a learning rate, a batch size, an optimizer, a learning decay rate, and a discard rate;

[0139] The loss function adopts the cross entropy loss function.

[0140] The disconnection detection model training module is specifically used for:

[0141] The data set is divided into a training set, a test set, and a validation set based on a ratio of 6:2:2, and the disconnection detection model is trained using the training set based on the learning rate, batch size, learning decay rate, and dropout rate carried by the hyperparameter group to learn the pulse law until the loss value of the loss function carried by the hyperparameter group is less than a preset loss threshold, and each hyperparameter in the hyperparameter group is continuously optimized during the training process;

[0142] The trained disconnection detection model is tested by the test set to determine whether the detection accuracy is greater than a preset accuracy threshold. If not, the test fails, and the training set is expanded to continue training the disconnection detection model. If yes, the test passes, and:

[0143] The broken thread detection model that has passed the test is verified by using the verification set to determine whether the confidence of the broken thread detection model is greater than a preset confidence threshold. If not, the verification fails, and the training set is expanded to continue training the broken thread detection model. If so, the verification passes, the training of the broken thread detection model is terminated, and the verified broken thread detection model is deployed on the embroidery machine.

[0144] The disconnection detection model is trained through the training set until the loss value of the loss function is less than the preset loss threshold. The hyperparameters in the hyperparameter group are continuously optimized during the training process. The detection accuracy of the trained disconnection detection model is tested through the test set, and the confidence of the disconnection detection model that passes the test is verified through the verification set. That is, the disconnection detection model is continuously optimized during the training process, and relevant tests and verifications are performed, thereby greatly improving the detection accuracy of the disconnection detection model.

[0145] In specific implementation, during the training process of the disconnection detection model, the disconnection detection frames of the disconnection detection model are merged and filtered using a soft connection non-maximum suppression algorithm, and a composite scaling algorithm (Neura l Architecture Search) is used to perform a network search operation on the feature extraction module, the feature fusion module and the disconnection detection module to determine the scaling coefficient of each dimension, and the computing power of each module is allocated using the scaling coefficient to select a network structure with the best efficiency and accuracy, thereby expanding the precision and accuracy of disconnection detection.

[0146] Since there are a lot of overlaps and invalid results in the predicted disconnection detection frames, the soft connection non-maximum suppression algorithm is used to filter the redundant disconnection detection frames according to the recognition probability value, merge some disconnection detection frames, and output the optimal disconnection detection frame.

[0147] The network search operation is performed through a composite scaling algorithm to determine the scaling coefficient of each dimension, that is, to solidify the network search space and path, effectively reducing the time complexity of the network search and thus improving the efficiency of disconnection detection.

[0148] The embroidery module is specifically used for:

[0149] The embroidery machine obtains an embroidery operation file, parses the embroidery operation file to obtain an embroidery path carrying embroidery needle positions, and controls the corresponding embroidery needle to work based on the embroidery path;

[0150] The disconnection detection module is specifically used for:

[0151] The embroidery machine starts the infrared sensor group corresponding to the needle position based on the currently working embroidery needle, performs thread breakage detection based on whether the infrared receiver of the infrared sensor group receives the infrared signal of the infrared transmitter, and obtains a first detection result;

[0152] The current embroidery pulse data of the infrared sensor group of the embroidery machine is collected, and the embroidery pulse data is preprocessed including at least outlier removal and noise suppression, and the preprocessed embroidery pulse data is input into the thread breakage detection model to obtain a second detection result.

[0153] The disconnection detection report generation module is specifically used for:

[0154] The embroidery machine analyzes the first detection result and the second detection result, and generates a thread breakage detection report when both the first detection result and the second detection result are thread breakage; generates a thread breakage detection report when the first detection result or the second detection result is thread breakage; generates a thread breakage detection report indicating that the embroidery thread needs to be replaced when the first detection result is that the thread is not broken and the second detection result is that the thread is about to be broken;

[0155] Displaying the disconnection detection report on a large screen, controlling the operation of an indicator light and a buzzer based on the disconnection detection report, and pushing the disconnection detection report to a pre-managed management terminal in real time to generate an alarm;

[0156] The embroidery log management module is specifically used for:

[0157] The embroidery machine records in real time an embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, optimizes the thread break detection model based on the embroidery log, and performs encrypted storage and distributed backup on the embroidery log.

[0158] By recording in real time the embroidery log including at least the embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, the thread break detection model is optimized based on the embroidery log, which further improves the detection accuracy of the thread break detection model. By encrypting and storing the embroidery log and performing distributed backup, the traceability is greatly improved.

[0159] The encrypted storage and distributed backup of the embroidery log are specifically as follows:

[0160] A pair of public keys and private keys are created based on the RSA algorithm, an HMAC calculation is performed on the embroidery log to obtain an HMAC value, the embroidery log and the HMAC value are encrypted by the private key to obtain primary encrypted data, the primary encrypted data is mapped by a preset first mapping rule to obtain secondary encrypted data, the number 3 and the letter s in the secondary encrypted data are swapped to obtain the third-level encrypted data; the public key is mapped by a preset second mapping rule to obtain a primary key, the primary key is encrypted by the 3DES algorithm to obtain a secondary key, the third-level encrypted data and the primary and secondary keys are encrypted into an encrypted log by the XTEA algorithm, the encrypted log is stored in a file system, and the encrypted log is backed up in a distributed manner.

[0161] Since data encrypted by the private key can only be decrypted by the public key, the public key is encrypted at multiple levels, the embroidery log is also encrypted at multiple levels, and the HMAC value can also be used to verify the integrity of the embroidery log. The HMAC calculation has higher security than the ordinary hash calculation. If you don’t know the corresponding encryption algorithm or data transformation rule, you will not be able to crack the encrypted log. At least 9 security measures have been taken (RSA algorithm, public and private keys, HMAC calculation, first mapping rule, 3 and s swap, second mapping rule, 3DES algorithm, XTEA algorithm, distributed backup), which greatly improves the security of embroidery log storage.

[0162] In summary, the advantages of the present invention are:

[0163] 1. By acquiring a large amount of embroidery historical data when the embroidery machine is working, each embroidery historical data is pre-processed and annotated to construct a data set; then, based on the feature extraction module, the feature fusion module and the thread break detection module, a thread break detection model is created, a hyperparameter group of the thread break detection model is set, the thread break detection model is trained based on the data set, and the trained thread break detection model is deployed on the embroidery machine; the embroidery machine obtains the embroidery operation file, controls the corresponding embroidery needle to work based on the embroidery operation file, starts the infrared sensor group of the corresponding embroidery needle position based on the currently working embroidery needle to perform thread break detection, and obtains a first detection result; collects the current embroidery pulse data, inputs the embroidery pulse data into the thread break detection model, obtains the second detection result, generates a thread break detection report based on the first detection result and the second detection result, issues an alarm based on the thread break detection report, and records the embroidery daily in real time. The thread break detection model is optimized based on the embroidery log, and the embroidery log is stored; that is, by combining the infrared detection of the infrared sensor group and the pulse regularity detection of the thread break detection model, the thread break of the embroidery thread can be double-checked and judged to ensure the accuracy of the thread break detection; based on the currently working embroidery needle, the infrared sensor group corresponding to the embroidery needle position is started to perform thread break detection, and the other infrared sensor groups are in a power-off state, so as to avoid all infrared sensor groups being in a long-term open state, and effectively reduce the power consumption of the thread break detection; by marking the position of the thread break that is about to break in the embroidery historical data, and then using the marked embroidery historical data to train the thread break detection model, the thread break detection model has the ability to predict whether the embroidery thread is about to break, and the embroidery thread that is about to break is replaced in time, which ultimately greatly improves the accuracy and timeliness of the embroidery machine's thread break detection and greatly reduces the power consumption.

[0164] 2. By preprocessing each embroidery historical data including at least outlier removal and noise suppression, the data quality is effectively improved, thereby greatly improving the training speed of the thread break detection model.

[0165] 3. By marking the pre-processed embroidery historical data at least including the thread break position and the position of the thread about to break, and then building a data set based on the marked embroidery historical data, the thread break detection model is trained using the data set. This enables the thread break detection model to not only detect the thread break position, but also predict whether the thread is about to break, greatly improving the reliability of the embroidery machine's thread break detection.

[0166] 4. The disconnection detection model is trained through the training set until the loss value of the loss function is less than the preset loss threshold. The hyperparameters in the hyperparameter group are continuously optimized during the training process. The detection accuracy of the trained disconnection detection model is tested through the test set. The confidence of the disconnection detection model that passes the test is verified through the verification set. That is, the disconnection detection model is continuously optimized during the training process, and relevant tests and verifications are performed, thereby greatly improving the detection accuracy of the disconnection detection model.

[0167] 5. By recording the embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report in real time, the thread break detection model is optimized based on the embroidery log, which further improves the detection accuracy of the thread break detection model. By encrypting and storing the embroidery log and performing distributed backup, the traceability is greatly improved.

[0168] Although the specific implementation modes of the present invention are described above, those skilled in the art should understand that the specific implementation modes described are only illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for detecting thread breakage in an embroidery machine, characterized in that: The steps include: Step S1, obtaining a large amount of embroidery history data when the embroidery machine is working, and constructing a data set after preprocessing and marking each of the embroidery history data; Step S2: creating a disconnection detection model based on the feature extraction module, the feature fusion module and the disconnection detection module, and setting a hyperparameter group of the disconnection detection model; Step S3, training a thread break detection model based on the data set, and deploying the trained thread break detection model on an embroidery machine; Step S4: the embroidery machine obtains an embroidery operation file, and controls the corresponding embroidery needle to work based on the embroidery operation file; Step S5: the embroidery machine starts the infrared sensor group corresponding to the embroidery needle position to perform thread breakage detection based on the currently working embroidery needle, and obtains a first detection result; collects current embroidery pulse data, inputs the embroidery pulse data into the thread breakage detection model, and obtains a second detection result; Step S6: the embroidery machine generates a thread breakage detection report based on the first detection result and the second detection result, and issues an alarm based on the thread breakage detection report; Step S7, the embroidery machine records the embroidery log in real time, optimizes the thread breakage detection model based on the embroidery log, and stores the embroidery log; The step S1 is specifically as follows: Acquire a large amount of embroidery history data when the embroidery machine is working, wherein the embroidery history data is pulse data obtained when the infrared signal of the infrared sensor group is blocked by the grating wheel; Each of the embroidery historical data is preprocessed including at least outlier removal and noise suppression, and each of the preprocessed embroidery historical data is marked with at least a thread break position and a position that is about to break. The thread break position and the position that is about to break are marked based on the selection of a target frame, and then a data set is constructed based on the marked embroidery historical data.

2. The method for detecting thread breakage in an embroidery machine according to claim 1, wherein: In step S2, the feature extraction module adopts a separable convolution structure, including several feature blocks of different scales; each of the feature blocks includes a convolution operation unit, a skip connection unit, a pooling unit and a SE unit; The feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse the multi-scale disconnection features output by the feature extraction module from bottom to top and from top to bottom; The disconnection detection module is constructed based on a multi-layer convolutional neural network unit, a pooling unit, and a fully connected network unit, and is used to predict the size and position of the disconnection detection box; The hyperparameter group includes at least a loss function, a learning rate, a batch size, an optimizer, a learning decay rate, and a discard rate; The loss function adopts the cross entropy loss function.

3. The method for detecting thread breakage in an embroidery machine according to claim 1, wherein: The step S3 is specifically as follows: The data set is divided into a training set, a test set, and a validation set based on a ratio of 6:2:2, and the disconnection detection model is trained using the training set based on the learning rate, batch size, learning decay rate, and dropout rate carried by the hyperparameter group to learn the pulse law until the loss value of the loss function carried by the hyperparameter group is less than a preset loss threshold, and each hyperparameter in the hyperparameter group is continuously optimized during the training process; The trained disconnection detection model is tested by the test set to determine whether the detection accuracy is greater than a preset accuracy threshold. If not, the test fails, and the training set is expanded to continue training the disconnection detection model. If yes, the test passes, and: The broken thread detection model that has passed the test is verified by using the verification set to determine whether the confidence of the broken thread detection model is greater than a preset confidence threshold. If not, the verification fails, and the training set is expanded to continue training the broken thread detection model. If so, the verification passes, the training of the broken thread detection model is terminated, and the verified broken thread detection model is deployed on the embroidery machine.

4. The method for detecting thread breakage in an embroidery machine according to claim 1, wherein: The step S4 is specifically as follows: The embroidery machine obtains an embroidery operation file, parses the embroidery operation file to obtain an embroidery path carrying embroidery needle positions, and controls the corresponding embroidery needle to work based on the embroidery path; The step S5 is specifically as follows: The embroidery machine starts the infrared sensor group corresponding to the needle position based on the currently working embroidery needle, performs thread breakage detection based on whether the infrared receiver of the infrared sensor group receives the infrared signal of the infrared transmitter, and obtains a first detection result; The current embroidery pulse data of the infrared sensor group of the embroidery machine is collected, and the embroidery pulse data is preprocessed including at least outlier removal and noise suppression, and the preprocessed embroidery pulse data is input into the thread breakage detection model to obtain a second detection result.

5. The method for detecting thread breakage in an embroidery machine according to claim 1, wherein: The step S6 is specifically as follows: The embroidery machine analyzes the first detection result and the second detection result, and generates a thread breakage detection report when both the first detection result and the second detection result are thread breakage; generates a thread breakage detection report when the first detection result or the second detection result is thread breakage; generates a thread breakage detection report indicating that the embroidery thread needs to be replaced when the first detection result is that the thread is not broken and the second detection result is that the thread is about to be broken; Displaying the disconnection detection report on a large screen, controlling the operation of an indicator light and a buzzer based on the disconnection detection report, and pushing the disconnection detection report to a pre-managed management terminal in real time to generate an alarm; The step S7 is specifically as follows: The embroidery machine records in real time an embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, optimizes the thread break detection model based on the embroidery log, and performs encrypted storage and distributed backup on the embroidery log.

6. An embroidery machine thread breakage detection system, characterized in that: Includes the following modules: A data set construction module is used to obtain a large amount of embroidery historical data when the embroidery machine is working, and construct a data set after preprocessing and marking each of the embroidery historical data; A disconnection detection model creation module, used to create a disconnection detection model based on the feature extraction module, the feature fusion module and the disconnection detection module, and set a hyperparameter group of the disconnection detection model; A thread break detection model training module, used for training the thread break detection model based on the data set, and deploying the trained thread break detection model on the embroidery machine; The embroidery module is used for the embroidery machine to obtain an embroidery operation file, and control the corresponding embroidery needle to work based on the embroidery operation file; The thread breakage detection module is used for the embroidery machine to start the infrared sensor group corresponding to the embroidery needle position to perform thread breakage detection based on the currently working embroidery needle, and obtain a first detection result; collect the current embroidery pulse data, input the embroidery pulse data into the thread breakage detection model, and obtain a second detection result; A thread breakage detection report generating module, used for the embroidery machine to generate a thread breakage detection report based on the first detection result and the second detection result, and to issue an alarm based on the thread breakage detection report; An embroidery log management module is used for the embroidery machine to record the embroidery log in real time, optimize the thread breakage detection model based on the embroidery log, and store the embroidery log; The data set construction module is specifically used for: Acquire a large amount of embroidery history data when the embroidery machine is working, wherein the embroidery history data is pulse data obtained when the infrared signal of the infrared sensor group is blocked by the grating wheel; Each of the embroidery historical data is preprocessed including at least outlier removal and noise suppression, and each of the preprocessed embroidery historical data is marked with at least a thread break position and a position that is about to break. The thread break position and the position that is about to break are marked based on the selection of a target frame, and then a data set is constructed based on the marked embroidery historical data.

7. The embroidery machine thread breakage detection system according to claim 6, characterized in that: In the disconnection detection model creation module, the feature extraction module adopts a separable convolution structure, including several feature blocks of different scales; each of the feature blocks includes a convolution operation unit, a jump connection unit, a pooling unit and a SE unit; The feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse the multi-scale disconnection features output by the feature extraction module from bottom to top and from top to bottom; The disconnection detection module is constructed based on a multi-layer convolutional neural network unit, a pooling unit, and a fully connected network unit, and is used to predict the size and position of the disconnection detection box; The hyperparameter group includes at least a loss function, a learning rate, a batch size, an optimizer, a learning decay rate, and a discard rate; The loss function adopts the cross entropy loss function.

8. The embroidery machine thread breakage detection system according to claim 6, characterized in that: The disconnection detection model training module is specifically used for: The data set is divided into a training set, a test set, and a validation set based on a ratio of 6:2:2, and the disconnection detection model is trained using the training set based on the learning rate, batch size, learning decay rate, and dropout rate carried by the hyperparameter group to learn the pulse law until the loss value of the loss function carried by the hyperparameter group is less than a preset loss threshold, and each hyperparameter in the hyperparameter group is continuously optimized during the training process; The trained disconnection detection model is tested by the test set to determine whether the detection accuracy is greater than a preset accuracy threshold. If not, the test fails, and the training set is expanded to continue training the disconnection detection model. If yes, the test passes, and: The broken thread detection model that has passed the test is verified by using the verification set to determine whether the confidence of the broken thread detection model is greater than a preset confidence threshold. If not, the verification fails, and the training set is expanded to continue training the broken thread detection model. If so, the verification passes, the training of the broken thread detection model is terminated, and the verified broken thread detection model is deployed on the embroidery machine.

9. The embroidery machine thread breakage detection system according to claim 6, characterized in that: The embroidery module is specifically used for: The embroidery machine obtains an embroidery operation file, parses the embroidery operation file to obtain an embroidery path carrying embroidery needle positions, and controls the corresponding embroidery needle to work based on the embroidery path; The disconnection detection module is specifically used for: The embroidery machine starts the infrared sensor group corresponding to the needle position based on the currently working embroidery needle, performs thread breakage detection based on whether the infrared receiver of the infrared sensor group receives the infrared signal of the infrared transmitter, and obtains a first detection result; The current embroidery pulse data of the infrared sensor group of the embroidery machine is collected, and the embroidery pulse data is preprocessed including at least outlier removal and noise suppression, and the preprocessed embroidery pulse data is input into the thread breakage detection model to obtain a second detection result.

10. The embroidery machine thread breakage detection system according to claim 6, characterized in that: The disconnection detection report generation module is specifically used for: The embroidery machine analyzes the first detection result and the second detection result, and generates a thread breakage detection report when both the first detection result and the second detection result are thread breakage; generates a thread breakage detection report when the first detection result or the second detection result is thread breakage; generates a thread breakage detection report indicating that the embroidery thread needs to be replaced when the first detection result is that the thread is not broken and the second detection result is that the thread is about to be broken; Displaying the disconnection detection report on a large screen, controlling the operation of an indicator light and a buzzer based on the disconnection detection report, and pushing the disconnection detection report to a pre-managed management terminal in real time to generate an alarm; The embroidery log management module is specifically used for: The embroidery machine records in real time an embroidery log including at least embroidery time, embroidery product model, embroidery product number, embroidery machine number and thread break detection report, optimizes the thread break detection model based on the embroidery log, and performs encrypted storage and distributed backup on the embroidery log.

Citation Information

Patent Citations

  • Dynamic-property-based early warning method for power transmission tower wire structural failure caused by icing

    CN103105257A

  • Device and method for detecting weft breakage and weft completion of circular knitting machine on basis of image technology

    CN104532452A

  • Yarn breakage early warning monitoring method, system and device and readable storage medium

    CN111519294A

  • Yarn broken end detection method, device and system of ring spinning frame and storage medium

    CN115690037A

  • Intelligent repairing method and system for empty needle and broken thread of sewing machine

    CN115787213A