Dispensing needle life prediction method and device

Abstract

The application discloses a dispensing needle life prediction method, and relates to the technical field of material processing. The method comprises the following steps: acquiring glue path data of historical materials and corresponding processing condition data of dispensing needles; performing model training based on the glue path data, the processing condition data and a long short-term memory network to obtain a life prediction model; collecting glue path data and real-time processing condition data in real time, and inputting the collected real-time glue path data and real-time processing condition data into the life prediction model to predict the remaining life of the dispensing needle. The method can improve the production line efficiency and save manpower and cost. The application also discloses a dispensing needle life prediction device, which also has the above technical effects.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material processing, in particular to a dispensing needle life prediction method; and also relates to a dispensing needle life prediction device. BACKGROUND

[0002] Dispensing is an important process in the material processing process. The dispensing needle used in the dispensing process directly affects the processing quality. At present, the dispensing needle is replaced in two ways: one is to replace the dispensing needle regularly, and the other is to replace the dispensing needle when the pressure value exceeds the standard during processing, causing the machine to alarm, and then manually diagnosing whether the dispensing needle is blocked, and replacing the dispensing needle when the dispensing needle is blocked.

[0003] However, diagnosing the dispensing needle blockage after processing a defective product and causing the machine to alarm will cause waste of materials, and the process of processing materials with a blocked dispensing needle to obtain a defective product will affect the overall production efficiency of the production line, causing waste of production time. In addition, manually diagnosing whether the dispensing needle is blocked will cause part of the manpower to be wasted in the dispensing needle detection process.

[0004] Therefore, how to solve the above technical defects has become a technical problem to be solved by those skilled in the art. SUMMARY

[0005] The purpose of the present application is to provide a dispensing needle life prediction method that can improve production line efficiency and save manpower and cost. Another purpose of the present application is to provide a dispensing needle life prediction device that also has the above technical effects.

[0006] To solve the above technical problems, the present application provides a dispensing needle life prediction method, comprising:

[0007] obtaining glue path data of historical materials and processing condition data of corresponding dispensing needles;

[0008] based on the glue path data, the processing condition data and a long short-term memory network, model training is performed to obtain a life prediction model;

[0009] real-time collection of glue path data and real-time processing condition data, and input of the collected real-time glue path data and real-time processing condition data into the life prediction model to predict the remaining life of the dispensing needle.

[0010] Optionally, based on the glue path data, the processing condition data and the long short-term memory network, model training is performed to obtain the life prediction model, comprising:

[0011] based on the glue path data, the processing condition data and the long short-term memory network, an attention mechanism is adopted for model training to obtain the life prediction model.

[0012] Optionally, the step of obtaining the life prediction model according to the weight further comprises:

[0013] According to the weight, the weight is normalized by using a softmax function to obtain an attention weight;

[0014] According to the attention weight, the life prediction model is obtained.

[0015] Optionally, the step of obtaining the life prediction model by using the attention mechanism for model training comprises:

[0016] The glue path data and the processing condition data are applied to the long short-term memory network by using the attention mechanism, weights of at least part of feature data in the glue path data and the processing condition data are obtained, and the glue path data and the processing condition data of the last dispensing of the dispensing needle are applied to the glue path data and the processing condition data of the next dispensing, so as to perform the model training and obtain the life prediction model according to the weights.

[0017] Optionally, the step of obtaining the life prediction model by using the glue path data, the processing condition data and the long short-term memory network for model training comprises:

[0018] The glue path data and the processing condition data are used as sample data, the sample data are randomly divided into several parts, the sample data are used for multiple times of model training, and the life prediction model is obtained; wherein, each time, a first number of the sample data in the several parts of the sample data are randomly used as a training set for model training, and a second number of the sample data are used as a test set for model testing.

[0019] Optionally, the step of obtaining the life prediction model by using the glue path data, the processing condition data and the long short-term memory network for model training comprises:

[0020] Based on the glue path data, the processing condition data and the long short-term memory network, an optimal parameter is found by using a stochastic gradient descent algorithm, and the life prediction model is obtained based on the optimal parameter.

[0021] Optionally, the attention weight α = softmax(W T M), wherein W represents encoding of input feature data, and M represents a mapping matrix of a hidden state of the long short-term memory network.

[0022] The prediction value s = hα T of the life prediction model, wherein h represents an output of the hidden state of the long short-term memory network, and T represents a matrix transpose.

[0023] Optionally, the processing condition data comprises a material waiting time of the dispensing needle and an initial installation time of the dispensing needle.

[0024] Optionally, the method further comprises:

[0025] According to the remaining life of the dispensing needle, a replacement time of the dispensing needle is recommended.

[0026] Optionally, the method further comprises:

[0027] The obtained dispensing path data of different sources and the processing condition data are data integrated and data cleaned.

[0028] To solve the above technical problems, the application also provides a dispensing needle life prediction device, comprising:

[0029] a memory for storing a computer program;

[0030] a processor for executing the computer program to realize the steps of the dispensing needle life prediction method according to any one of the above.

[0031] The dispensing needle life prediction method provided by the application comprises: obtaining dispensing path data of historical materials and processing condition data of corresponding dispensing needles; performing model training based on the dispensing path data, the processing condition data and a long short-term memory network to obtain a life prediction model; collecting real-time dispensing path data and real-time processing condition data in real time, inputting the collected real-time dispensing path data and real-time processing condition data into the life prediction model, and predicting the remaining life of the dispensing needle.

[0032] It can be seen that the dispensing needle life prediction method provided by the application is based on the dispensing path data of the materials, the processing condition data of the dispensing needles and the long short-term memory network to train the life prediction model. The life prediction model can be used to predict the remaining life of the dispensing needle before the next material processing according to the dispensing path data generated after the current material processing and the processing condition data of the dispensing needle, that is, to predict how long the dispensing needle will be blocked, and then to replace the dispensing needle according to the predicted remaining life of the dispensing needle. This can not only avoid the cost and time waste caused by the dispensing needle blockage during the material processing, but also improve the production line efficiency, and no longer need manual detection of the dispensing needle blockage, which can save manpower allocation.

[0033] The dispensing needle life prediction device provided by the application also has the above technical effects. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the prior art and the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0035] Figure 1 A flowchart of a dispensing needle life prediction method provided by an embodiment of the present application is shown in

[0036] Figure 2 A schematic diagram of an LSTM network architecture provided by an embodiment of the present application is shown in

[0037] Figure 3 A schematic diagram of an Attention-LSTM network provided by an embodiment of the present application is shown in

[0038] Figure 4 A schematic diagram of a dispensing needle life prediction device provided by an embodiment of the present application is shown in DETAILED DESCRIPTION

[0039] The core of the present application is to provide a dispensing needle life prediction method, which can improve the efficiency of the production line and save manpower and cost. Another core of the present application is to provide a dispensing needle life prediction device, which also has the above technical effects.

[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0041] Reference is made to Figure 1 , Figure 1 A flowchart of a dispensing needle life prediction method provided by an embodiment of the present application is shown in Figure 1 The method comprises the following steps.

[0042] S101: Obtain the glue path data of historical materials and the processing condition data of corresponding dispensing needles.

[0043] The glue path data of the materials that have completed dispensing and the processing condition data of the dispensing needles can reflect or affect the clogging of the dispensing needles. Therefore, the glue path data of historical materials and the processing condition data of dispensing needles are obtained in this embodiment, so as to train the model and finally predict the remaining life of the dispensing needles.

[0044] The glue path data includes glue edge distance, glue width, etc. The glue dispenser applies pressure to the glue dispensing needle during the production process, so that the glue dispensing needle extrudes glue to dispense glue on the material. After dispensing glue on the material, the dispensing condition of the material is converted into glue path data by taking a photo. For each piece of material that has completed dispensing, the glue path data thereof is obtained.

[0045] In addition, the processing condition data can include the material waiting time of the glue dispensing needle before dispensing glue on each piece of material and the initial installation time of the glue dispensing needle, i.e., the installation start time of the glue dispensing needle. The material waiting time refers to the time interval between completing dispensing on the previous piece of material and starting dispensing on the current piece of material.

[0046] In some embodiments, the method further comprises:

[0047] The obtained glue path data and processing condition data of different sources are data integrated and data cleaned.

[0048] The glue path data and processing condition data of different sources refer to the glue path data and processing condition data placed in different databases. For example, the glue path data is placed in database A, the material waiting time is placed in database B, and the installation start time of the glue dispensing needle is placed in database C. For the glue path data and processing condition data of different sources, the embodiment is data integrated and further data cleaned to select data in the effective time section. The effective time section can be the time section from the installation to the disassembly of the glue dispensing needle.

[0049] For example, the glue dispensing needle is installed at 8:00 and disassembled at 12:00, and the effective time section is 8:00-12:00, and the data between 8:00-12:00 is selected.

[0050] In addition, the glue path data and processing condition data can also be normalized. The normalization method can refer to the prior art, which will not be described here.

[0051] S102: Based on the glue path data, the processing condition data, and the long short-term memory network, a model is trained to obtain a life prediction model;

[0052] Based on the obtained glue path data and processing condition data, the glue path data and processing condition data are used as sample data for model training to obtain a life prediction model for predicting the remaining life of the glue dispensing needle, i.e., how long the glue dispensing needle will be blocked.

[0053] The glue needle blockage is caused by the glue accumulation in the needle due to dispensing of multiple pieces of material. The dispensing result of the next piece of material is related to the dispensing result of the previous piece of material. In order to better meet the above characteristics of the dispensing process, the embodiment adopts a long short-term memory network for model training. The network architecture of the long short-term memory network (LSTM) can learn the relevance between each piece of data, and obtain a life prediction model with high precision.

[0054] The network architecture of the long short-term memory network can refer to Figure 2 . Figure 2 x t represents the input of the data in the current state; h t-1 represents the output of the previous hidden state; h t represents the output of the current LSTM hidden state; c t-1 represents the output of the previous cell state; c t represents the output of the current LSTM cell state; f t represents the forget gate, and a value of 0-1 is input to c t-1 , which determines what information the cell state discards; i t represents the input gate, which determines whether the current input is added to the LSTM; o represents the newly created candidate cell state of the current time; o t represents the output gate, which takes a value of 0-1 and determines the output of the hidden state; σ represents the sigmoid function, and tanh represents the activation function.

[0055] The relationship between the above parameters is:

[0056] f t = σ (W f · h t-1 , x t + b f ) ;

[0057] i t = σ (W i · [h t-1 , x t + b i ) ;

[0058]

[0059]

[0060] o t = σ(W o · [h t-1 , x t ] + b o );

[0061] h t = o t * tanh(c t );

[0062] wherein W f , W i , W c , W o represent the weights of each gate in the LSTM module, which are learned by the neural network; b f , b i , b c , b o represent the offsets of each gate.

[0063] Further, in some embodiments, the model training based on the glue road data, the processing condition data and the long short-term memory network obtains the life prediction model, which comprises:

[0064] Based on the glue road data, the processing condition data and the long short-term memory network, the attention mechanism is adopted for model training to obtain the life prediction model.

[0065] wherein the model training based on the attention mechanism obtains the life prediction model, which comprises:

[0066] The attention mechanism is adopted to apply the glue road data and the processing condition data to the long short-term memory network to obtain the weights of at least part of the feature data in the glue road data and the processing condition data, and to apply the glue road data and the processing condition data of the last dispensing of the dispensing needle to the glue road data and the processing condition data of the next dispensing, so as to perform the model training and obtain the life prediction model according to the weights.

[0067] The attention mechanism is a technology in the neural network that simulates cognitive attention. By adopting the attention mechanism, the weights of some features in the neural network input data can be enhanced, while the weights of other features can be weakened, so as to focus the attention of the neural network on the most important part of the data, thereby improving the overall accuracy of the life prediction model.

[0068] For example, assuming that the life prediction model is y = ax1+ bx2+ cx3, where x1, x2, x3 are input data, and y is the true value, the life prediction model will learn how to adjust a, b, c, so that y approaches ax1+ bx2+ cx3. In this application, x1, x2, x3 are the glue path data, the waiting time, and the interval between the current time and the time when the dispensing needle head is installed, and y is the interval between the current time and the time when the dispensing needle head is disassembled. The attention mechanism can enhance the weights a, b, c of x1, x2, x3. The attention mechanism is to enhance the weight of some features in the neural network input data through query vectors (query), key vectors (key), and value vectors (value), while weakening the weight of other features. It generally includes the following three steps: the first step is to calculate the similarity between the query and each key to obtain the weight. Common similarity functions include dot product, concatenation, and perception, etc. The second step is to use a softmax function to normalize these weights to convert them into attention. The third step is to weight and sum the weights and the corresponding values to obtain the final attention, i.e. the attention weight. Finally, according to the attention weight, the related parameters in the life prediction model can be obtained, i.e. the life prediction model.

[0069] It should be noted that using the softmax function to map the output of the neural network to the 0-1 interval can avoid the large difference between the results output each time during multiple iterations of the model training process. Thus, the accuracy of the model is improved.

[0070] Reference Figure 3 As shown, the long short-term memory network using the attention mechanism aims to maximize the accuracy of the important feature vector weight and the output. Figure 3 In the formula, x1 to x n represent the input of each LSTM, h1 to h n represent the output of the hidden state of each LSTM, and a represents the attention weight, and a = [a1, a2, …, a n ], ∑ i a i = 1.

[0071] In this embodiment, the attention weight a = softmax(W T M). Where M represents the mapping matrix of the hidden state, T represents the matrix transpose, W represents the word representation of the input (x), i.e. the encoding of the feature data, and W represents all the parameters to be learned, including W f , W i , W c , Wo , W h , W h , W h , W

[0072] In this embodiment, the prediction value s = hα T , h represents the output of the hidden state of the LSTM module, T represents the matrix transpose, and a represents the attention weight.

[0073] In this embodiment, the objective function is , s t , s t-1 , y t , y t-1 , y

[0074] In the process of training the model, the model parameters with the best model accuracy are found by minimizing the difference between the actual value and the predicted value. In this embodiment, the argmin function is used to find the best parameters of the neural network (Atten-LSTM) by minimizing the difference between the actual value and the predicted value, and then find a better model.

[0075] In some embodiments, the model training based on the rubber road data, the processing condition data and the long short-term memory network to obtain the life prediction model comprises:

[0076] The rubber road data and the processing condition data are used as sample data, and the sample data is randomly divided into several parts. The sample data is used for multiple model training to obtain the life prediction model. Each time, the first number of sample data in the sample data is randomly divided into several parts as a training set for model training, and the second number of sample data is used as a test set for model testing.

[0077] For example, the sample data is randomly divided into k parts, and k-1 parts are randomly selected each time as a training set for model training, and the remaining part is used as a test set for model testing. Repeating the above training and testing operations can effectively improve the accuracy of the life prediction model.

[0078] In some embodiments, the model training based on the rubber road data, the processing condition data and the long short-term memory network to obtain the life prediction model comprises:

[0079] Based on the glue road data, the processing condition data and the neural network algorithm, an optimal parameter is searched by using a stochastic gradient descend algorithm, and the life prediction model is obtained based on the optimal parameter.

[0080] In this embodiment, the stochastic gradient descend algorithm is used to search for the optimal parameter. The stochastic gradient descend algorithm is a first-order optimization method, and is one of the simplest methods for solving unconstrained optimization problems. For the specific principle of the stochastic gradient descend algorithm, reference can be made to the existing related description, which will not be described herein again. In this embodiment, the stochastic gradient descend algorithm is used to search for the optimal parameter, which is more convenient to implement and can improve the efficiency.

[0081] S103: Real-time glue road data and real-time processing condition data are collected, and the collected real-time glue road data and real-time processing condition data are input into the life prediction model to predict the remaining life of the dispensing needle.

[0082] After the optimal parameter is obtained, the collected real-time glue road data and real-time processing condition data are input into the life prediction model, and the remaining life of the dispensing needle, i.e., the remaining use time, is predicted before dispensing the next piece of material, so that the dispensing needle can be replaced according to the remaining life of the dispensing needle.

[0083] After the remaining life of the dispensing needle is predicted, the remaining life of the dispensing needle can be displayed for relevant personnel to check and manually determine the replacement time of the dispensing needle. Alternatively, after the remaining life of the dispensing needle is predicted, the system automatically determines and recommends the replacement time of the dispensing needle.

[0084] The real-time glue road data and real-time processing condition data can also be summarized and cleaned, and the glue road data and processing condition data can be normalized.

[0085] In order to reduce the labor consumption and improve the efficiency, in some embodiments, the method further comprises:

[0086] According to the remaining life of the dispensing needle, the replacement time of the dispensing needle is recommended.

[0087] In this embodiment, after the remaining life of the dispensing needle is predicted, the replacement time of the dispensing needle is recommended according to the remaining life of the dispensing needle and a preset replacement strategy. The replacement strategy can be set differently. For example, the replacement strategy can be to replace the dispensing needle when the remaining life of the dispensing needle is reached. For example, if the remaining life of the dispensing needle is 3 hours and the current time is 9 am, the dispensing needle will be replaced at 12 pm.

[0088] For example, the replacement strategy can be to replace the dispensing needle when a preset time length before the remaining life of the dispensing needle is reached. For example, if the remaining life of the dispensing needle is 3 hours, the current time is 9:00 am, and the preset time length is 30 minutes, the dispensing needle is replaced at 11:30 am.

[0089] In addition, different reminders of different emergency levels can be taken according to the size of the remaining life. For example, if the remaining life is short, an alarm can be used to remind, and if the remaining life is long, the reminder can be temporarily suspended.

[0090] In summary, the dispensing needle life prediction method provided by the present application is based on the glue path data of the material, the processing condition data of the dispensing needle, and the life prediction model trained by the neural network algorithm. The life prediction model can predict the remaining life of the dispensing needle according to the glue path data generated after the processing of the material and the processing condition data of the dispensing needle before the next material processing, that is, predict how long the dispensing needle will be blocked, and then facilitate the replacement of the dispensing needle according to the predicted remaining life of the dispensing needle. This not only avoids the cost and time waste caused by the dispensing needle blocking and still processing the material, improves the production line efficiency, but also no longer needs manual detection of the dispensing needle blockage, which can save manpower allocation.

[0091] The present application also provides a dispensing needle life prediction device, which can be referred to as the method described above. Please refer to Figure 4 , Figure 4 The dispensing needle life prediction device provided by the embodiment of the present application is shown in the schematic diagram, which includes a memory 1 and a processor 2. Figure 4

[0092] The memory 1 is used to store the computer program;

[0093] The processor 2 is used to execute the computer program to realize the following steps:

[0094] Obtain the glue path data of the historical material and the processing condition data of the corresponding dispensing needle;

[0095] Based on the glue path data, the processing condition data, and the long short-term memory network, the model is trained to obtain a life prediction model;

[0096] Real-time acquisition of glue path data and real-time processing condition data, and input of the collected real-time glue path data and real-time processing condition data into the life prediction model to predict the remaining life of the dispensing needle.

[0097] On the basis of the above embodiment, as a specific implementation, the processor 2 executes the computer program to specifically realize the following steps: ​

[0098] Based on the glue road data, the processing condition data and the long short-term memory network, an attention mechanism is adopted for model training to obtain the life prediction model.

[0099] On the basis of the above embodiment, as a specific implementation, the processor 2 executes a computer program to specifically implement the following steps:

[0100] The glue road data and the processing condition data are applied to the long short-term memory network by adopting an attention mechanism to obtain the weight of at least part of the feature data in the glue road data and the processing condition data, and the glue road data and the processing condition data of the last dispensing of the dispensing needle are applied to the glue road data and the processing condition data of the next dispensing to perform the model training and obtain the life prediction model according to the weight.

[0101] On the basis of the above embodiment, as a specific implementation, the processor 2 executes a computer program to specifically implement the following steps:

[0102] The glue road data and the processing condition data are taken as sample data, and the sample data are randomly divided into several parts. The sample data are used for multiple model training to obtain the life prediction model. In each time, a first number of the sample data in the several parts of the sample data are randomly taken as a training set for model training, and a second number of the sample data are taken as a test set for model testing.

[0103] On the basis of the above embodiment, as a specific implementation, the processor 2 executes a computer program to specifically implement the following steps:

[0104] Based on the glue road data, the processing condition data and the long short-term memory network, a random gradient descent algorithm is adopted to find the optimal parameters, and the life prediction model is obtained based on the optimal parameters.

[0105] On the basis of the above embodiment, as a specific implementation, the processing condition data includes the waiting time of the dispensing needle and the initial installation time of the dispensing needle.

[0106] On the basis of the above embodiment, as a specific implementation, the processor 2 executes a computer program to further implement the following steps:

[0107] According to the remaining life of the dispensing needle, the replacement time of the dispensing needle is recommended.

[0108] On the basis of the above embodiment, as a specific implementation, the processor 2 executes a computer program to further implement the following steps:

[0109] The obtained glue path data of different sources and the processing condition data are integrated and cleaned.

[0110] The glue needle life prediction device provided in the present application is based on the glue path data of the material, the processing condition data of the glue needle, and a life prediction model trained by a neural network algorithm. The life prediction model can predict the remaining life of the glue needle before the next material processing according to the glue path data generated after the current material processing and the processing condition data of the glue needle, that is, how long the glue needle will be blocked, so as to facilitate the replacement of the glue needle according to the predicted remaining life of the glue needle. This not only can avoid the cost and time waste caused by the blockage of the glue needle during the material processing, and improve the production line efficiency, but also no longer needs manual detection of the blockage of the glue needle, and can save manpower allocation.

[0111] The embodiments in the description are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the device and computer readable storage medium disclosed in the embodiments, since they correspond to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0112] The skilled person can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of both. In order to clearly show the interchangeability of hardware and software, the components and steps of each example have been described in the above description. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0113] The steps of the method or algorithm described in combination with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0114] The above describes in detail the dispensing needle life prediction method and device provided by the present application. The principles and implementation manners of the present application are described by using specific examples, and the above description of the embodiments is only used to help understand the method of the present application and its core idea. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, the present application can be improved and modified in several ways, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A method for predicting the lifespan of a dispensing needle, characterized in that, include: Acquire historical material adhesive path data and corresponding dispensing needle processing condition data; adhesive path data includes adhesive edge distance and adhesive width; processing condition data includes the waiting time of the dispensing needle and the initial installation time of the dispensing needle; waiting time refers to the time interval between completing the dispensing of the previous piece of material and starting the dispensing of the current piece of material. Based on the adhesive path data, the processing condition data, and the long short-term memory network, a life prediction model is obtained through model training. Real-time glue path data and real-time processing condition data are collected and input into the life prediction model to predict the remaining life of the dispensing needle.

2. The method for predicting the lifespan of a dispensing needle according to claim 1, characterized in that, The lifetime prediction model is obtained by training the model based on the adhesive path data, the processing condition data, and the long short-term memory network. Based on the adhesive path data, the processing condition data, and the long short-term memory network, an attention mechanism is used to train the model, resulting in the lifetime prediction model.

3. The method for predicting the lifespan of a dispensing needle according to claim 2, characterized in that, The steps for training the lifespan prediction model using an attention mechanism include: An attention mechanism is used to apply the glue path data and the processing condition data to the long short-term memory network to obtain the weights of at least a portion of the feature data in the glue path data and the processing condition data; The glue path data and processing condition data of the previous dispensing on the dispensing needle are applied to the glue path data and processing condition data of the next dispensing to train the model and obtain the life prediction model according to the weights.

4. The method for predicting the lifespan of a dispensing needle according to claim 3, characterized in that, The step of obtaining the lifetime prediction model based on the weights further includes: Based on the weights, the weights are normalized using the softmax function and converted into attention weights; The lifetime prediction model is obtained based on the attention weights.

5. The method for predicting the lifespan of a dispensing needle according to claim 1, characterized in that, The lifespan prediction model obtained by training the model based on the adhesive path data, the processing condition data, and the long short-term memory network includes: Using the adhesive path data and the processing condition data as sample data, and randomly dividing the sample data into several parts, the model is trained multiple times using the sample data to obtain the life prediction model; wherein, each time, a first number of the sample data from the several parts of the sample data is randomly used as the training set for model training, and a second number of the sample data is used as the test set for model testing.

6. The method for predicting the lifespan of a dispensing needle according to claim 1, characterized in that, The lifespan prediction model obtained by training the model based on the adhesive path data, the processing condition data, and the long short-term memory network includes: Based on the adhesive path data, the processing condition data, and the neural network algorithm, the stochastic gradient descent algorithm is used to find the optimal parameters, and the life prediction model is obtained based on the optimal parameters.

7. The method for predicting the lifespan of a dispensing needle according to claim 1, characterized in that, Also includes: Based on the remaining lifespan of the dispensing needle, the recommended replacement time for the dispensing needle is as follows.

8. The method for predicting the lifespan of a dispensing needle according to claim 4, characterized in that, in, The attention weight α = softmax(W) T M), where W represents the encoding of the input feature data, M represents the mapping matrix of the hidden states of the Long Short-Term Memory network, and T represents the matrix transpose; The predicted value of the lifetime prediction model is s = hα T , where h represents the output of the hidden state of the Long Short-Term Memory network, and T represents the matrix transpose.

9. A device for predicting the lifespan of a dispensing needle, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the dispensing needle life prediction method as described in any one of claims 1 to 8 when executing the computer program.

Images