An instrument processing production line management system based on artificial intelligence

Abstract

The present invention discloses an instrument processing production line management system based on artificial intelligence, which belongs to the technical field of production line management. By collecting surface images of processed instruments on the production line and corresponding operating data of the production line, and using an artificial intelligence model to analyze the surface images of the processed instruments and the corresponding operating data of the production line, not only the production quality can be monitored, but also the operation of the production line can be monitored, and faults in the production line can be discovered in time, thereby comprehensively managing the working conditions of the instrument processing production line, solving the problems of low production efficiency, excessive manual intervention and unstable product quality in the prior art.

Description

Technical Field

[0001] The invention belongs to the technical field of production line management, and in particular relates to an instrument processing production line management system based on artificial intelligence. Background Art

[0002] With the development of science and technology, production automation has been widely used in various fields. Especially in the field of instrument processing, production automation can not only improve production efficiency and reduce production costs, but also ensure product quality. For example, in the field of optical instrument processing, it is necessary not only to ensure that the processing is completed in an orderly manner, but also to ensure that the product is free of defects. Otherwise, it may not only lead to product quality problems, but more seriously, it may lead to product function problems. The same is true in the process of other instrument processing. When there are defects or production failures in the production process, the yield rate may decrease. However, the existing production line management system still has some problems in actual application, such as low production efficiency, excessive manual intervention, and unstable product quality. Summary of the invention

[0003] The present invention provides an instrument processing production line management system based on artificial intelligence, which is used to solve the problems of low production efficiency, excessive manual intervention and unstable product quality in the prior art.

[0004] An instrument processing production line management system based on artificial intelligence, comprising: a data acquisition module, a data transmission module, a data analysis module, a production management module and a user interface module;

[0005] The data acquisition module is used to collect surface images of processed instruments on the production line and corresponding operation data of the production line to obtain data to be analyzed;

[0006] The data transmission module is used to encrypt the data to be analyzed by using a data encryption algorithm, and then transmit the encrypted data to be analyzed to the data analysis module;

[0007] The data analysis module is used to decrypt the encrypted data to be analyzed, and then use artificial intelligence to analyze the data to be analyzed to determine the quality analysis results and the operation analysis results;

[0008] The production management module is used to manage the instrument processing production line according to the quality analysis results and the operation analysis results to obtain management results;

[0009] The user interface module is used to display historical data, current data to be analyzed, quality analysis results, operation analysis results and management results to the user.

[0010] Furthermore, the data acquisition module includes an image data acquisition unit and an operation data acquisition unit;

[0011] The image data acquisition unit is used to acquire surface images of instruments on the production line;

[0012] The operation data collection unit is used to collect the operation data corresponding to the production line;

[0013] Among them, the surface image and the operating data corresponding to the running motor together constitute the data to be analyzed, and the operating data includes the electric spindle temperature, electric spindle current, electric spindle voltage, electric spindle vibration, electric spindle speed, transmission chain lubrication status, servo controller current, overtravel parameters and zero point signal that cannot be found.

[0014] Further, after the data to be analyzed is encrypted by a data encryption algorithm, the encrypted data to be analyzed is transmitted to a data analysis module, including:

[0015] Encrypting the data to be analyzed by using an asymmetric encryption algorithm to obtain encrypted data to be analyzed;

[0016] The encrypted data to be analyzed is transmitted to the data analysis module through the local area network.

[0017] Further, the data analysis module includes a first data analysis unit and a second data analysis unit;

[0018] The first data analysis unit is used to use the first artificial intelligence model to identify the surface image in the data to be analyzed and determine the quality analysis result;

[0019] The second data analysis unit is used to use a second artificial intelligence model to identify the operating data in the data to be analyzed and determine the operating analysis result.

[0020] Furthermore, the data analysis module also includes an artificial intelligence model deployment unit;

[0021] The artificial intelligence model deployment unit is used to deploy the first artificial intelligence model and the second artificial intelligence model so that the first data analysis unit and the second data analysis unit schedule artificial intelligence to perform data recognition.

[0022] Further, deploying the first artificial intelligence model includes:

[0023] Constructing a defect feature initial extraction unit, a multi-level feature fusion feature extraction unit, and a feature recognition classification unit that are connected in sequence to obtain a first artificial intelligence model, and after optimizing the first artificial intelligence model using an intelligent optimization algorithm, deploying a second artificial intelligence model in a computer device;

[0024] The defect feature initial extraction unit is set to an improved VGG16, and the improved VGG16 removes the fully connected layer and the output layer;

[0025] The multi-level feature fusion feature extraction unit includes a first multi-scale attention structure, a first maximum pooling layer and a second multi-scale attention structure connected in sequence;

[0026] The first multi-scale attention structure has the same structure as the second multi-scale attention structure, both of which include: a multi-scale feature fusion substructure, an attention mechanism substructure, and a multi-level feature fusion substructure;

[0027] The multi-scale feature fusion substructure includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, a sixth convolutional layer, a second maximum pooling layer and a concatenate layer. The first convolutional layer, the second convolutional layer, the third convolutional layer and the second maximum pooling layer all take the feature map output by the multi-level feature fusion feature extraction unit as input, the output of the second convolutional layer is used as the input of the fourth convolutional layer, the output of the third convolutional layer is used as the input of the fifth convolutional layer, the output of the second maximum pooling layer is used as the input of the sixth convolutional layer, and the concatenate layer superimposes the features output by the first convolutional layer, the fourth convolutional layer, the fifth convolutional layer and the sixth convolutional layer and outputs them;

[0028] The attention mechanism substructure includes a channel attention mechanism, a spatial attention mechanism, a first data multiplication layer, and a second data multiplication layer; the feature maps output by the multi-scale feature fusion substructure enter the channel attention mechanism, the first data multiplication layer, and the second data multiplication layer respectively, the output of the channel attention mechanism also enters the first data multiplication layer, and the output of the first data multiplication layer enters the spatial attention mechanism and the second data multiplication layer respectively, and the second data multiplication layer outputs different feature maps after element-wise multiplication;

[0029] The multi-level feature fusion substructure includes an Add layer, and the Add layer is used to connect the feature map output by the multi-level feature fusion feature extraction unit, the feature map output by the multi-scale feature fusion substructure, and the feature map output by the second data multiplication layer along the feature channel and then output them to obtain a multi-level feature fusion map;

[0030] The feature recognition classification unit includes a first global average pooling layer, a fully connected layer and a softmax layer connected in sequence.

[0031] Further, the first artificial intelligence model is used to identify the surface image in the data to be analyzed to determine the quality analysis result, including:

[0032] The defect feature initial extraction unit in the first artificial intelligence model is used to extract defect features from the surface image in the data to be analyzed to obtain a defect feature map;

[0033] Using a first multi-scale attention structure in a multi-level feature fusion feature extraction unit to extract multi-scale feature information from the defect feature map, to obtain a first multi-scale feature map;

[0034] Using the first maximum pooling layer in the multi-level feature fusion feature extraction unit to remove redundant information in the first multi-scale feature map and reduce the amount of parameters to obtain an intermediate feature map;

[0035] The second multi-scale attention structure in the multi-level feature fusion feature extraction unit is used to extract multi-scale feature information from the intermediate feature map to obtain a second multi-scale feature map;

[0036] A feature recognition and classification unit is used to recognize and classify the second multi-scale feature map to obtain a quality analysis result.

[0037] Further, deploying the second artificial intelligence model includes:

[0038] Constructing a sequentially connected input layer, hidden layer, and output layer to obtain a second artificial intelligence model;

[0039] After using an intelligent optimization algorithm to enable the second artificial intelligence model to learn the data relationship between the operating data and the fault type, the second artificial intelligence model is deployed in the computer device.

[0040] Further, the second artificial intelligence model is used to identify the operation data in the data to be analyzed and determine the operation analysis results, including:

[0041] The operating data in the data to be analyzed is used to construct input vector data, and the input vector data is used as the input of the second artificial intelligence model to obtain the predicted probability of each fault type and normal operation type output by the second artificial intelligence model, and the type with the largest predicted probability is used as the operation analysis result.

[0042] Furthermore, according to the quality analysis results and the operation analysis results, the instrument processing production line is managed to obtain management results, including:

[0043] When the quality analysis result is abnormal, a mechanical arm is used to mark the corresponding instrument and generate quality abnormality information;

[0044] When the operation analysis result is abnormal, check whether there is a corresponding parameter adjustment plan stored in the database. If so, use the corresponding parameter adjustment plan to adjust the operation parameters of the production line, and generate operation abnormality information according to the operation analysis result and the adjustment plan of the operation parameters. Otherwise, directly generate operation abnormality information according to the operation analysis result;

[0045] The quality abnormality information and the operation abnormality information are taken together as management results.

[0046] The present invention provides an instrument processing production line management system based on artificial intelligence. By collecting surface images of processed instruments on the production line and the corresponding operating data of the production line, and using an artificial intelligence model to analyze the surface images of the processed instruments and the corresponding operating data of the production line, not only can the production quality be monitored, but also the operation of the production line can be monitored, and faults in the production line can be discovered in time, thereby comprehensively managing the working conditions of the instrument processing production line, solving the problems of low production efficiency, excessive manual intervention and unstable product quality in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0048] Figure 1 A schematic diagram of the structure of an instrument processing production line management system based on artificial intelligence provided by an embodiment of the present invention.

[0049] Figure 2 A schematic diagram of the structure of a data acquisition module provided in an embodiment of the present invention.

[0050] Figure 3 A schematic diagram of the structure of a data analysis module provided in an embodiment of the present invention.

[0051] Figure 4 A schematic diagram of the structure of a first artificial intelligence model provided in an embodiment of the present invention.

[0052] Figure 5 A schematic diagram of the structure of a second artificial intelligence model provided in an embodiment of the present invention.

[0053] Among them, 1-data acquisition module, 11-image data acquisition unit, 12-operation data acquisition unit, 2-data transmission module, 3-data analysis module, 31-first data analysis unit, 32-second data analysis unit, 33-artificial intelligence model deployment unit, 4-production management module, 5-user interface module.

[0054] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0055] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0056] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0057] like Figure 1 As shown, the embodiment of the present invention provides an instrument processing production line management system based on artificial intelligence, including: a data acquisition module 1, a data transmission module 2, a data analysis module 3, a production management module 4 and a user interface module 5;

[0058] The data acquisition module 1 is used to collect the surface images of the processed instruments on the production line and the corresponding operation data of the production line to obtain the data to be analyzed;

[0059] The data transmission module 2 is used to encrypt the data to be analyzed by using a data encryption algorithm, and then transmit the encrypted data to be analyzed to the data analysis module 3;

[0060] The data analysis module 3 is used to decrypt the encrypted data to be analyzed, and then use artificial intelligence to analyze the data to be analyzed to determine the quality analysis results and the operation analysis results;

[0061] Through quality analysis, the production quality of the production line can be guaranteed, and by analyzing the operating data, abnormal operation of the production line can be discovered in time, so that parameters can be automatically adjusted or staff can solve the abnormalities in time, thereby fully ensuring that the instruments processed by the instrument production line meet the factory requirements and can greatly improve the production efficiency of the instrument.

[0062] The production management module 4 is used to manage the instrument processing production line according to the quality analysis results and the operation analysis results to obtain management results;

[0063] The user interface module 5 is used to display historical data, current data to be analyzed, quality analysis results, operation analysis results and management results to the user.

[0064] The present invention provides an instrument processing production line management system based on artificial intelligence. By collecting surface images of processed instruments on the production line and the corresponding operating data of the production line, and using an artificial intelligence model to analyze the surface images of the processed instruments and the corresponding operating data of the production line, not only can the production quality be monitored, but also the operation of the production line can be monitored, and faults in the production line can be discovered in time, thereby comprehensively managing the working conditions of the instrument processing production line, solving the problems of low production efficiency, excessive manual intervention and unstable product quality in the prior art.

[0065] like Figure 2 As shown, the data acquisition module 1 includes an image data acquisition unit 11 and an operation data acquisition unit 12;

[0066] The image data acquisition unit 11 is used to acquire the surface image of the instrument on the production line;

[0067] The operation data collection unit 12 is used to collect the operation data corresponding to the production line;

[0068] Among them, the surface image and the operating data corresponding to the running motor together constitute the data to be analyzed, and the operating data includes the electric spindle temperature, electric spindle current, electric spindle voltage, electric spindle vibration, electric spindle speed, transmission chain lubrication status, servo controller current, overtravel parameters and zero point signal that cannot be found.

[0069] For example, X-axis electric cylinders are often used in automated instrument production lines, so the operating data of the X-axis electric cylinders can be collected as data to be analyzed. The fault types to be classified can be set to motor overload, limit switch failure, unbalanced rotating parts, servo controller failure, program failure, power failure, etc. At the same time, the operating status should also include the normal status, so as to determine whether the automated instrument production line is operating normally.

[0070] In the embodiment of the present invention, after the data to be analyzed is encrypted by a data encryption algorithm, the encrypted data to be analyzed is transmitted to the data analysis module 3, including:

[0071] Encrypting the data to be analyzed by using an asymmetric encryption algorithm to obtain encrypted data to be analyzed;

[0072] The encrypted data to be analyzed is transmitted to the data analysis module 3 using a local area network.

[0073] like Figure 3 As shown, the data analysis module 3 includes a first data analysis unit 31 and a second data analysis unit 32;

[0074] The first data analysis unit 31 is used to identify the surface image in the data to be analyzed by using the first artificial intelligence model to determine the quality analysis result;

[0075] The second data analysis unit 32 is used to use a second artificial intelligence model to identify the operation data in the data to be analyzed and determine the operation analysis result.

[0076] In the embodiment of the present invention, the data analysis module 3 further includes an artificial intelligence model deployment unit 33;

[0077] The artificial intelligence model deployment unit 33 is used to deploy the first artificial intelligence model and the second artificial intelligence model so that the first data analysis unit and the second data analysis unit schedule artificial intelligence to perform data recognition.

[0078] In an embodiment of the present invention, deploying the first artificial intelligence model includes:

[0079] A defect feature initial extraction unit, a multi-level feature fusion feature extraction unit and a feature recognition and classification unit are sequentially connected to obtain a first artificial intelligence model. After optimizing the first artificial intelligence model using an intelligent optimization algorithm, the second artificial intelligence model is deployed in a computer device.

[0080] The second artificial intelligence model is used to detect defects on the surface of instruments on the production line and mark or pick instruments with detected defects to achieve quality management. Common surface defects may include scratches, pits, inclusions, silver streaks, etc.

[0081] like Figure 4 As shown, the initial defect feature extraction unit is set to an improved VGG16 (Visual Geometry Group Network 16), and the improved VGG16 removes the fully connected layer and the output layer;

[0082] The embodiment of the present invention retains the convolution and pooling layers of VGG16, removes the fully connected layer, and adds multi-scale feature fusion and attention mechanism after the network to improve the defect feature extraction performance.

[0083] The multi-level feature fusion feature extraction unit includes a first multi-scale attention structure, a first maximum pooling layer and a second multi-scale attention structure connected in sequence;

[0084] The first multi-scale attention structure has the same structure as the second multi-scale attention structure, both of which include: a multi-scale feature fusion substructure, an attention mechanism substructure, and a multi-level feature fusion substructure;

[0085] The multi-scale feature fusion substructure includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, a sixth convolutional layer, a second maximum pooling layer and a Concatenate layer (i.e., a feature concatenation layer). The first convolutional layer, the second convolutional layer, the third convolutional layer and the second maximum pooling layer all take the feature map output by the multi-level feature fusion feature extraction unit as input, the output of the second convolutional layer is used as input of the fourth convolutional layer, the output of the third convolutional layer is used as input of the fifth convolutional layer, the output of the second maximum pooling layer is used as input of the sixth convolutional layer, and the Concatenate layer superimposes the features output by the first convolutional layer, the fourth convolutional layer, the fifth convolutional layer and the sixth convolutional layer and outputs them;

[0086] In the multi-scale feature fusion substructure, the first convolution layer uses 1×1 convolution to identify small targets; the second convolution layer first uses 1×1 convolution to reduce the number of channels and gather useful information, and the fourth convolution layer uses 3×3 convolution kernels to increase the receptive field; the third convolution layer first uses 1×1 convolution to reduce the number of channels and gather useful information, and the fifth convolution layer uses 5×5 convolution kernels to increase the receptive field to extract different features; the second maximum pooling layer is added before the 1×1 sixth convolution layer to reduce the dimension of the extracted feature information and increase the receptive field of the convolution kernel. After four branches, more abundant feature information of multiple scales can be obtained, and finally the features are superimposed and output.

[0087] The attention mechanism substructure includes a channel attention mechanism, a spatial attention mechanism, a first data multiplication layer, and a second data multiplication layer (i.e., an element-wise multiplication layer); the feature maps output by the multi-scale feature fusion substructure enter the channel attention mechanism, the first data multiplication layer, and the second data multiplication layer respectively, the output of the channel attention mechanism also enters the first data multiplication layer, and the output of the first data multiplication layer enters the spatial attention mechanism and the second data multiplication layer respectively, and the second data multiplication layer outputs different feature maps after element-wise multiplication;

[0088] Although a feature map fused with multiple scales is obtained through the multi-scale feature fusion substructure, some unimportant features among the acquired features may interfere with the detection accuracy. Therefore, the embodiment of the present invention extracts key features and suppresses secondary features through the attention mechanism substructure, thereby enhancing the expression of features and increasing the final defect recognition accuracy.

[0089] The multi-level feature fusion substructure includes an Add layer (i.e., a feature addition layer), which is used to connect the feature map output by the multi-level feature fusion feature extraction unit, the feature map output by the multi-scale feature fusion substructure, and the feature map output by the second data multiplication layer along the feature channel and then output them to obtain a multi-level feature fusion map;

[0090] Connecting the first multi-scale attention structure with the second multi-scale attention structure through a 3×3 first maximum pooling layer can not only remove redundant information in the features extracted by the first multi-scale attention structure and reduce the number of parameters, but also prevent overfitting and improve the generalization ability of the model.

[0091] The feature recognition classification unit includes a first global average pooling layer, a fully connected layer and a softmax layer (ie, a normalization layer) which are connected in sequence.

[0092] In an embodiment of the present invention, the first artificial intelligence model is used to identify the surface image in the data to be analyzed and determine the quality analysis result, including:

[0093] The defect feature initial extraction unit in the first artificial intelligence model is used to extract defect features from the surface image in the data to be analyzed to obtain a defect feature map;

[0094] Using a first multi-scale attention structure in a multi-level feature fusion feature extraction unit to extract multi-scale feature information from the defect feature map, to obtain a first multi-scale feature map;

[0095] Using the first maximum pooling layer in the multi-level feature fusion feature extraction unit to remove redundant information in the first multi-scale feature map and reduce the amount of parameters to obtain an intermediate feature map;

[0096] The second multi-scale attention structure in the multi-level feature fusion feature extraction unit is used to extract multi-scale feature information from the intermediate feature map to obtain a second multi-scale feature map;

[0097] A feature recognition and classification unit is used to recognize and classify the second multi-scale feature map to obtain a quality analysis result.

[0098] In an embodiment of the present invention, deploying the second artificial intelligence model includes:

[0099] Construct the input layer, hidden layer and output layer connected in sequence, and get the following Figure 5 The second artificial intelligence model shown;

[0100] After using an intelligent optimization algorithm to enable the second artificial intelligence model to learn the data relationship between the operating data and the fault type, the second artificial intelligence model is deployed in the computer device.

[0101] Since the embodiment of the present invention collects less operating data, the operating data can be constructed into a vector, and then the second artificial intelligence model is used to perform data recognition to determine the fault type. For example, the second artificial intelligence model can be constructed using a classic BP neural network to complete the fault recognition process.

[0102] In the prior art, only classification loss is often considered, while the performance index of the artificial intelligence model is not paid attention to, which leads to the situation that the artificial intelligence model trained by classification loss has missed detection, which is fatal for the surface defects of the detection instrument and will reduce the factory yield. Therefore, the embodiment of the present invention provides a loss function combined with the performance index of the artificial intelligence model, which can be:

[0103]

[0104]

[0105]

[0106] Where L represents the loss function, Represents a constant term, which can be set to 3; Indicated in t The cross entropy loss during the training process is Indicates in front t The average value of the cross entropy loss during the training process, Indicates in front t The variance of the cross entropy loss during the training process, Indicated in t The intermediate variables in the training process, Indicates in front t The average value of the intermediate variables during the training process, Indicates in front t The variance of the intermediate variables during the training process; represents the preset weight, Indicates the accuracy of the artificial intelligence model, Represents the recall rate of the artificial intelligence model, represents the number of input samples, Indicated in t During the training process m The actual value (i.e. label value) corresponding to each sample, Indicated in t During the training process m The predicted value corresponding to the sample.

[0107] The loss function provided by the embodiment of the present invention combines cross entropy loss, accuracy and recall rate together, which can effectively balance the optimization process of the artificial intelligence model. While avoiding missed recognition as much as possible, it can reduce the workload of staff and enable the artificial intelligence model to run in the most optimized direction.

[0108] It is worth noting that the above loss function is a preferred embodiment of the present invention. Those skilled in the art may also use other loss functions (such as root mean square loss) to evaluate the training process of the artificial intelligence model.

[0109] In the prior art, a particle swarm algorithm is often used as an artificial intelligence algorithm to train an artificial intelligence model to obtain the optimal parameters of the artificial intelligence model. However, the particle swarm algorithm has the disadvantage of being easily trapped in a local optimum. Therefore, an embodiment of the present invention provides an improved particle swarm algorithm, which may include:

[0110] Initialize a particle swarm; wherein each particle in the particle swarm is a hyperparameter to be optimized; the particle includes the hyperparameter of the artificial intelligence model to be trained; the artificial intelligence model to be trained is the first artificial intelligence model or the second artificial intelligence model, and the hyperparameter may be the connection weights and thresholds between layers.

[0111] Get the loss function value of each particle, divide half of the particles with the smallest loss function value into the local optimization particle swarm, and divide the other half of the particles into the global optimization particle swarm.

[0112] For the local search particle swarm, the particles are updated using adaptive inertia weights as follows:

[0113]

[0114]

[0115]

[0116] in, represents the adaptive inertia weight, Indicates t The first step of local search particle swarm in the training process i The speed of each particle is Indicates t +1 training session i The speed of each particle is represents the first acceleration coefficient, represents the second acceleration factor, represents the first random number between (0,1), represents the second random number between (0,1), Indicates iThe historical optimal value corresponding to the particle is represents the optimal particle; Indicates t During the training i Particles, Indicates updated ; represents the maximum inertia weight, represents the minimum inertia weight, T represents the preset maximum number of training times, i =1,2,…,I, where I represents the total number of particles in the local search particle swarm.

[0117] The embodiment of the present invention sets adaptive inertia weights to update particles, and can adjust the algorithm convergence speed in a nonlinear manner, thereby balancing the convergence effects of the early and late stages of the algorithm, so that the particle swarm algorithm has a certain global optimization capability in the early stage of the algorithm.

[0118] For the global optimization particle swarm, the momentum update method is used to update the particles as follows:

[0119]

[0120]

[0121] in, Indicates t The first step of the global optimization particle swarm in the training process j The speed of each particle is Indicates t +1 training session j The speed of each particle is represents the reduction factor, represents the third acceleration factor, represents the fourth acceleration factor, represents the third random number between (0,1), represents the fourth random number between (0,1), Indicates j The historical optimal value corresponding to the particle is represents the fifth random number between (0,1), e represents a natural constant, Indicates j The fitness value corresponding to each particle can be , Indicates j The loss function value corresponding to each particle.

[0122] By introducing a reduction factor to the global optimization particle swarm, we can have a stronger global optimization ability. At the same time, a larger disturbance is introduced to further improve the global optimization ability of the particle swarm. At the same time, the disturbance term is composed of poor particles and optimal particles. That is to say, a larger disturbance will be generated in the early and middle stages of the algorithm, while the disturbance will gradually decrease in the later stage of the algorithm, thereby ensuring the final convergence accuracy. Combined with the reduction factor, the problem that the existing particle swarm algorithm is prone to falling into local optimality can be effectively solved.

[0123] Before the number of training times reaches the maximum number of training times, return to the step of obtaining the local optimized particle swarm and the global optimized particle swarm, otherwise re-obtain the optimal particles, and use the optimal particles as the final hyperparameters of the artificial intelligence model to be trained to complete the training.

[0124] In an embodiment of the present invention, the second artificial intelligence model is used to identify the operation data in the data to be analyzed and determine the operation analysis result, including:

[0125] Using the operation data in the data to be analyzed to construct input vector data, and using the input vector data as the input of the second artificial intelligence model, obtaining the predicted probabilities of various fault types and normal operation types output by the second artificial intelligence model, and using the type with the largest predicted probability as the operation analysis result;

[0126] In an embodiment of the present invention, the instrument processing production line is managed according to the quality analysis result and the operation analysis result to obtain a management result, including:

[0127] When the quality analysis result is abnormal, a robot arm is used to mark (for example, spray or mark) the corresponding instrument, and quality abnormality information is generated at the same time;

[0128] When the operation analysis result is abnormal, check whether there is a corresponding parameter adjustment plan stored in the database. If so, use the corresponding parameter adjustment plan to adjust the operation parameters of the production line, and generate operation abnormality information according to the operation analysis result and the adjustment plan of the operation parameters. Otherwise, directly generate operation abnormality information according to the operation analysis result;

[0129] The quality abnormality information and the operation abnormality information are taken together as management results.

[0130] Optionally, when the production speed is known, the position of each instrument can be determined by the detection time and the speed of the production line, and defective instruments can be picked out by a robotic arm, thereby realizing automated management.

[0131] Some faults can be eliminated by adjusting parameters. Corresponding parameter adjustment plans can be preset for these faults to achieve automated management. For other faults that cannot be handled, the staff can be notified in time so that they can solve the faults in time.

[0132] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0133] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0134] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0135] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0136] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0137] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An instrument processing production line management system based on artificial intelligence, characterized in that: include: Data acquisition module, data transmission module, data analysis module, production management module and user interface module; The data acquisition module is used to collect surface images of processed instruments on the production line and corresponding operation data of the production line to obtain data to be analyzed; The data transmission module is used to encrypt the data to be analyzed by using a data encryption algorithm, and then transmit the encrypted data to be analyzed to the data analysis module; The data analysis module includes a first data analysis unit and a second data analysis unit; The first data analysis unit is used to use the first artificial intelligence model to identify the surface image in the data to be analyzed and determine the quality analysis result; The second data analysis unit is used to identify the operation data in the data to be analyzed by using the second artificial intelligence model to determine the operation analysis result; The training process of the first artificial intelligence model and the second artificial intelligence model includes: Initialize a particle swarm; wherein each particle in the particle swarm is a hyperparameter to be optimized; the particle includes a hyperparameter of the artificial intelligence model to be trained; the artificial intelligence model to be trained is the first artificial intelligence model or the second artificial intelligence model, and the hyperparameter is a connection weight and a threshold between each layer; Get the loss function value of each particle, divide half of the particles with the smallest loss function value into the local optimization particle group, and divide the other half of the particles into the global optimization particle group; For the local search particle swarm, the particles are updated using adaptive inertia weights as follows: in, represents the adaptive inertia weight, Indicates t The first step of local search particle swarm in the training process i The speed of each particle is Indicates t +1 training session i The speed of each particle is represents the first acceleration coefficient, represents the second acceleration factor, represents the first random number between (0,1), represents the second random number between (0,1), Indicates i The historical optimal value corresponding to the particle is represents the optimal particle; Indicates t During the training i Particles, Indicates updated ; represents the maximum inertia weight, represents the minimum inertia weight, T represents the preset maximum number of training times, i =1,2,…,I, where I represents the total number of particles in the local search particle swarm; For the global optimization particle swarm, the momentum update method is used to update the particles as follows: in, Indicates t The first step of the global optimization particle swarm in the training process j The speed of each particle is Indicates t +1 training session j The speed of each particle is represents the reduction factor, represents the third acceleration factor, represents the fourth acceleration factor, represents the third random number between (0,1), represents the fourth random number between (0,1), Indicates j The historical optimal value corresponding to the particle is represents the fifth random number between (0,1), e represents a natural constant, Indicates j The fitness value corresponding to a particle is , Indicates j The loss function value corresponding to each particle; Before the number of training times reaches the maximum number of training times, return to the step of obtaining the local optimized particle swarm and the global optimized particle swarm, otherwise re-obtain the optimal particles, and use the optimal particles as the final hyperparameters of the artificial intelligence model to be trained to complete the training; The data analysis module is used to decrypt the encrypted data to be analyzed, and then use artificial intelligence to analyze the data to be analyzed to determine the quality analysis results and the operation analysis results; The production management module is used to manage the instrument processing production line according to the quality analysis results and the operation analysis results to obtain management results; The user interface module is used to display historical data, current data to be analyzed, quality analysis results, operation analysis results and management results to the user.

2. The instrument processing production line management system based on artificial intelligence according to claim 1 is characterized in that: The data acquisition module includes an image data acquisition unit and an operation data acquisition unit; The image data acquisition unit is used to acquire surface images of instruments on the production line; The operation data collection unit is used to collect the operation data corresponding to the production line; Among them, the surface image and the operating data corresponding to the running motor together constitute the data to be analyzed, and the operating data includes the electric spindle temperature, electric spindle current, electric spindle voltage, electric spindle vibration, electric spindle speed, transmission chain lubrication status, servo controller current, overtravel parameters and zero point signal that cannot be found.

3. The instrument processing production line management system based on artificial intelligence according to claim 2 is characterized in that: After the data to be analyzed is encrypted by a data encryption algorithm, the encrypted data to be analyzed is transmitted to a data analysis module, including: Encrypting the data to be analyzed by using an asymmetric encryption algorithm to obtain encrypted data to be analyzed; The encrypted data to be analyzed is transmitted to the data analysis module through the local area network.

4. The instrument processing production line management system based on artificial intelligence according to claim 1 is characterized in that: The data analysis module also includes an artificial intelligence model deployment unit; The artificial intelligence model deployment unit is used to deploy the first artificial intelligence model and the second artificial intelligence model so that the first data analysis unit and the second data analysis unit schedule artificial intelligence to perform data recognition.

5. The instrument processing production line management system based on artificial intelligence according to claim 4 is characterized in that: Deploying the first artificial intelligence model includes: Constructing a defect feature initial extraction unit, a multi-level feature fusion feature extraction unit, and a feature recognition classification unit that are connected in sequence to obtain a first artificial intelligence model, and after optimizing the first artificial intelligence model using an intelligent optimization algorithm, deploying a second artificial intelligence model in a computer device; The defect feature initial extraction unit is set to an improved VGG16, and the improved VGG16 removes the fully connected layer and the output layer; The multi-level feature fusion feature extraction unit includes a first multi-scale attention structure, a first maximum pooling layer and a second multi-scale attention structure connected in sequence; The first multi-scale attention structure has the same structure as the second multi-scale attention structure, both of which include: a multi-scale feature fusion substructure, an attention mechanism substructure, and a multi-level feature fusion substructure; The multi-scale feature fusion substructure includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, a sixth convolutional layer, a second maximum pooling layer and a concatenate layer. The first convolutional layer, the second convolutional layer, the third convolutional layer and the second maximum pooling layer all take the feature map output by the multi-level feature fusion feature extraction unit as input, the output of the second convolutional layer is used as the input of the fourth convolutional layer, the output of the third convolutional layer is used as the input of the fifth convolutional layer, the output of the second maximum pooling layer is used as the input of the sixth convolutional layer, and the concatenate layer superimposes the features output by the first convolutional layer, the fourth convolutional layer, the fifth convolutional layer and the sixth convolutional layer and outputs them; The attention mechanism substructure includes a channel attention mechanism, a spatial attention mechanism, a first data multiplication layer, and a second data multiplication layer; the feature maps output by the multi-scale feature fusion substructure enter the channel attention mechanism, the first data multiplication layer, and the second data multiplication layer respectively, the output of the channel attention mechanism also enters the first data multiplication layer, and the output of the first data multiplication layer enters the spatial attention mechanism and the second data multiplication layer respectively, and the second data multiplication layer outputs different feature maps after element-wise multiplication; The multi-level feature fusion substructure includes an Add layer, and the Add layer is used to connect the feature map output by the multi-level feature fusion feature extraction unit, the feature map output by the multi-scale feature fusion substructure, and the feature map output by the second data multiplication layer along the feature channel and then output them to obtain a multi-level feature fusion map; The feature recognition classification unit includes a first global average pooling layer, a fully connected layer and a softmax layer connected in sequence.

6. The instrument processing production line management system based on artificial intelligence according to claim 5 is characterized in that: The first artificial intelligence model is used to identify the surface image in the data to be analyzed and determine the quality analysis results, including: The defect feature initial extraction unit in the first artificial intelligence model is used to extract defect features from the surface image in the data to be analyzed to obtain a defect feature map; Using a first multi-scale attention structure in a multi-level feature fusion feature extraction unit to extract multi-scale feature information from the defect feature map, to obtain a first multi-scale feature map; Using the first maximum pooling layer in the multi-level feature fusion feature extraction unit to remove redundant information in the first multi-scale feature map and reduce the amount of parameters to obtain an intermediate feature map; The second multi-scale attention structure in the multi-level feature fusion feature extraction unit is used to extract multi-scale feature information from the intermediate feature map to obtain a second multi-scale feature map; A feature recognition and classification unit is used to recognize and classify the second multi-scale feature map to obtain a quality analysis result.

7. The instrument processing production line management system based on artificial intelligence according to claim 5 is characterized in that: Deploying the second artificial intelligence model includes: Constructing a sequentially connected input layer, hidden layer, and output layer to obtain a second artificial intelligence model; After using an intelligent optimization algorithm to enable the second artificial intelligence model to learn the data relationship between the operating data and the fault type, the second artificial intelligence model is deployed in the computer device.

8. The instrument processing production line management system based on artificial intelligence according to claim 7 is characterized in that: The second artificial intelligence model is used to identify the operation data in the data to be analyzed and determine the operation analysis results, including: The operating data in the data to be analyzed is used to construct input vector data, and the input vector data is used as the input of the second artificial intelligence model to obtain the predicted probability of each fault type and normal operation type output by the second artificial intelligence model, and the type with the largest predicted probability is used as the operation analysis result.

9. The instrument processing production line management system based on artificial intelligence according to claim 1 is characterized in that: According to the quality analysis results and operation analysis results, the instrument processing production line is managed to obtain management results, including: When the quality analysis result is abnormal, a mechanical arm is used to mark the corresponding instrument and generate quality abnormality information; When the operation analysis result is abnormal, check whether there is a corresponding parameter adjustment plan stored in the database. If so, use the corresponding parameter adjustment plan to adjust the operation parameters of the production line, and generate operation abnormality information according to the operation analysis result and the adjustment plan of the operation parameters. Otherwise, directly generate operation abnormality information according to the operation analysis result; The quality abnormality information and the operation abnormality information are taken together as management results.

Images