CN114580791B

CN114580791B - Method and device for identifying working state of bulking machine, computer equipment and storage medium

Info

Publication number: CN114580791B
Application number: CN202210460237.2A
Authority: CN
Inventors: 尹航; 严骅彬; 徐昊; 娄善平
Original assignee: Shenzhen Fengshang Wisdom Agriculture And Animal Husbandry Technology Co ltd
Current assignee: Shenzhen Fengshang Wisdom Agriculture And Animal Husbandry Technology Co ltd
Priority date: 2022-04-28
Filing date: 2022-04-28
Publication date: 2022-08-30
Anticipated expiration: 2042-04-28
Also published as: CN114580791A

Abstract

The application relates to a method and a device for identifying the working state of a bulking machine and computer equipment. The method comprises the following steps: acquiring corresponding initial dimension production parameter data of the bulking machine in the current time period; performing dimension conversion on each initial dimension production parameter data corresponding to the current time period to obtain target dimension production parameter data; loading a working state prediction model of the bulking machine, wherein the working state prediction model of the bulking machine is obtained by training through a neural network by using training data; and inputting the target dimension production parameter data into a working state prediction model of the bulking machine to predict the working state of the bulking machine, so as to obtain a corresponding working state result of the bulking machine in the current time period. The method can improve the working efficiency of the bulking machine.

Description

Method and device for identifying working state of bulking machine, computer equipment and storage medium

Technical Field

The present application relates to the field of artificial intelligence technologies, and in particular, to a method and an apparatus for identifying a working state of a bulking machine, a computer device, a storage medium, and a computer program product.

Background

Bulking machines are common equipment in the modern feed industry and are used for processing puffed foods, such as rice, corn, soybean, wheat and the like in daily life. The bulking machine comprises a feeder, a modulator, a host machine, a bulking cavity, a cutter motor and other sub-devices. In the production process of the bulking machine, along with the increase of the production capacity, the biggest problem is the waste of resources, such as that a worker does not use the most effective parameters to operate the machine or the machine is always idle and forgets to shut down, and the like. In order to avoid the resource waste, the working state of the bulking machine needs to be monitored in real time. The current bulking machine monitoring mode still needs to lean on manual monitoring, looks over and judges the operating condition of bulking machine through the manual work, however the manual work can not look over and judge the operating condition of bulking machine in real time, causes the extravagant problem of bulking machine resource easily.

Disclosure of Invention

Accordingly, it is necessary to provide a method, an apparatus, a computer device, a computer readable storage medium, and a computer program product for identifying a working state of a bulking machine, which can determine the working state in real time, avoid resource waste, and improve the working efficiency of the bulking machine.

In a first aspect, the application provides a method for identifying the working state of a bulking machine. The method comprises the following steps:

acquiring corresponding initial dimension production parameter data of the bulking machine in the current time period;

performing dimension conversion on each initial dimension production parameter data corresponding to the current time period to obtain target dimension production parameter data;

loading a working state prediction model of the bulking machine, wherein the working state prediction model of the bulking machine is obtained by training through a neural network by using training data;

and inputting the target dimension production parameter data into a working state prediction model of the bulking machine to predict the working state of the bulking machine, so as to obtain a corresponding working state result of the bulking machine in the current time period.

In a second aspect, the application further provides a device for identifying the working state of the bulking machine. The device comprises:

the acquisition module is used for acquiring the corresponding initial dimension production parameter data of the bulking machine in the current time period;

the conversion module is used for carrying out dimension conversion on each initial dimension production parameter data corresponding to the current time period to obtain target dimension production parameter data;

the loading module is used for loading a working state prediction model of the bulking machine, and the working state prediction model of the bulking machine is obtained by training through a neural network by using training data;

and the output module is used for inputting the target dimension production parameter data into the working state prediction model of the bulking machine to predict the working state of the bulking machine, so as to obtain a corresponding working state result of the bulking machine in the current time period.

In a third aspect, the application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the following steps when executing the computer program:

In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:

In a fifth aspect, the present application further provides a computer program product. The computer program product comprising a computer program which when executed by a processor performs the steps of:

According to the method, the device, the computer equipment, the storage medium and the computer program product for identifying the working state of the bulking machine, the dimension conversion is carried out on the production data of the bulking machine in the current time period to obtain the input data meeting a bulking machine working state prediction model, the working state of the bulking machine is predicted through the bulking machine working state prediction model to obtain the working state result of the bulking machine in the current time period, a manager can timely check the equipment according to the working state result of the bulking machine in the current time period to realize the real-time monitoring on the bulking machine, the normal production of the bulking machine is ensured, the resource loss is reduced, and the problem of bulking machine resource waste is avoided. For example, when the working production state of the bulking machine in the current time period is material spraying or machine blockage, managers can timely perform treatment such as equipment closing.

Drawings

Fig. 1 is a diagram illustrating an application environment of a method for recognizing an operating state of a puffing machine according to an embodiment;

fig. 2 is a schematic flow chart illustrating a method for identifying the operating state of the puffing machine according to an embodiment;

FIG. 3 is a schematic flow chart illustrating the training steps of the prediction model of the operating condition of the puffing machine in one embodiment;

FIG. 4 is a diagram illustrating data completion in one embodiment;

FIG. 5 is a schematic flow chart illustrating data preprocessing according to one embodiment;

fig. 6 is a schematic flow chart of the prediction of the operating state of the puffing machine in one embodiment;

FIG. 7 is a diagram of one-dimensional convolution in one embodiment;

FIG. 8 is a diagram of a relu activation function in one embodiment;

FIG. 9 is a diagram illustrating random elimination of a neuron node, according to one embodiment;

FIG. 10 is a schematic illustration of maximum pooling in one embodiment;

FIG. 11 is a schematic diagram of sample conversion in one embodiment;

FIG. 12 is a diagram illustrating the relationship of an input layer, hidden layer, and output layer in one embodiment;

FIG. 13 is a schematic diagram of the process for identifying the operating state of the puffing machine in one embodiment;

FIG. 14 is a schematic diagram of a network architecture in one embodiment;

fig. 15 is a block diagram showing the construction of the puffing machine operation state recognition device according to one embodiment;

FIG. 16 is a diagram showing an internal structure of a computer device in one embodiment;

fig. 17 is an internal structural view of a computer device in another embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The method for identifying the working state of the bulking machine, provided by the embodiment of the application, can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104, or may be located on the cloud or other network server. The server 104 acquires each initial dimension production parameter data corresponding to the current time period uploaded by the terminal 102; the server 104 performs dimension conversion on each initial dimension production parameter data corresponding to the current time period to obtain target dimension production parameter data; the server 104 loads a prediction model of the working state of the bulking machine, wherein the prediction model of the working state of the bulking machine is obtained by training through a neural network by using training data; the server 104 inputs the target dimension production parameter data into the working state prediction model of the bulking machine to predict the working state of the bulking machine, so as to obtain a working state result of the bulking machine corresponding to the current time period of the bulking machine, and the server 104 returns the working state result of the bulking machine to the terminal 102. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart car-mounted devices, and the like. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, and the like. The server 104 may be implemented as a stand-alone server or as a server cluster comprised of multiple servers.

In one embodiment, as shown in fig. 2, there is provided a method for identifying the operating state of a puffing machine, which is described by taking the method as an example applied to a server in fig. 1, and comprises the following steps:

and 202, acquiring the corresponding initial dimension production parameter data of the bulking machine in the current time period.

The initial dimension production parameter data refers to original production parameter data consisting of time dimensions and parameter dimensions of the bulking machine in the current data segment, and comprises data corresponding to each production parameter.

Specifically, after the bulking machine detects that the bulking machine is started, the server can acquire initial dimension production parameter data uploaded by each sensor inside the bulking machine in real time, and the server acquires corresponding initial dimension production parameter data within the current time period uploaded by the bulking machine in real time.

In one embodiment, the server may also obtain, from the data storage system, each initial dimension production parameter data corresponding to the bulking machine in the current time period; the server can also acquire the production parameter data of each initial dimension corresponding to the bulking machine in the current time period in the local storage space of the bulking machine, including the production parameter data of each corresponding initial dimension of each sub-device of the bulking machine in the current time period.

And 204, performing dimension conversion on each initial dimension production parameter data corresponding to the current time period to obtain target dimension production parameter data.

The dimension conversion refers to adjusting the dimension quantity of the initial dimension production parameter data. The target dimension production parameter data refers to production parameter data which is used for predicting a working state prediction model of the bulking machine after dimension conversion.

Specifically, when the server detects that the dimensionality of each initial dimensionality production parameter data corresponding to the bulking machine in the current time period does not meet the dimensionality predicted by the bulking machine working state prediction model, dimensionality conversion is carried out on each initial dimensionality production parameter data corresponding to the bulking machine in the current time period, the dimensionality of each initial dimensionality production parameter data is adjusted, and the server can traverse each initial dimensionality production parameter data in a sliding window mode according to the preset window size and the sliding step length to carry out dimensionality conversion to obtain target dimensionality production parameter data.

And step 206, loading a working state prediction model of the bulking machine, wherein the working state prediction model of the bulking machine is obtained by training through a neural network by using training data.

The working state prediction model of the bulking machine is an artificial intelligence model used for predicting the working state of the bulking machine, and can be an artificial intelligence model obtained by training data through a neural network.

Specifically, after the server obtains the target dimension production parameter data, the working state prediction model of the bulking machine is loaded from the model storage space. A trained prediction model of the working state of the bulking machine is pre-deployed in the model storage space.

And 208, inputting the target dimension production parameter data into a working state prediction model of the bulking machine to predict the working state of the bulking machine, so as to obtain a corresponding working state result of the bulking machine in the current time period.

The working state prediction refers to an operation process of obtaining a prediction result by a prediction model of the working state of the bulking machine based on target dimension production parameter data.

Specifically, the server inputs target dimension production parameter data into a loaded bulking machine working state prediction model, the bulking machine working state prediction model predicts the working state based on the target dimension production parameter data, and outputs a bulking machine working state result to obtain a corresponding bulking machine working state result of the bulking machine in the current time period. And then the server returns the working state result of the bulking machine to the bulking machine for displaying, so that a manager can check the working state of the bulking machine in the current time period in time and perform corresponding processing.

In one embodiment, as shown in fig. 3, the training of the prediction model of the operating state of the puffing machine comprises the following steps:

step 302, obtaining a historical production parameter data set corresponding to the bulking machine and a working state category corresponding to the historical production parameter data set.

The historical production parameter data set refers to a set of production parameter data uploaded by the bulking machine in a historical time period, and is the historical production parameter data set uploaded when the working state of the bulking machine is known. The working state category refers to predefined working state types of the bulking machine, including a normal working state and an abnormal working state.

Specifically, the bulking machine reads the historical production parameter data set of the bulking machine in a preset historical time period at regular time. The bulking machine can upload the historical production parameter data set to a database and can also store the historical production parameter data set locally. The timing reading of the server may be daily or weekly. The preset historical time period may be 7 days, 15 days, 30 days, etc. For example, the server reads the historical production parameter data set from the local, and loads the historical production parameter data set to the data space to be processed for data completion. The loaded historical production parameter data set is composed of a plurality of pieces of historical production parameter data, each piece of historical production parameter data is represented by date _ time, x1, x2, x3, … … and xm, wherein the date _ time represents the time for uploading data to the nearest second, m represents m production parameters in total of the bulking machine, xm represents the mth parameter, and the production parameters can be referred to as parameter characteristics. In the process of uploading the historical production parameter data by the bulking machine, the historical production parameter data can be uploaded only when the data changes, and the uploaded historical production parameter data can be lost at different time points, so that the server completes the data of the loaded historical production parameter data. And then the server acquires the working state category corresponding to the historical production parameter data set according to the historical production parameter data set.

In an embodiment, as shown in fig. 4, a data completion diagram is provided, in which missing data is updated to last uploaded data according to production parameters, for example, the first upload time is 2021-11-1223: 00: 00, the current data uploaded is 229.5. When the current data changes, the current data is uploaded again, and the uploading time is 2021-11-1223: 00: 02, the current data uploaded is 229.6. The upload time is 2021-11-1223: 00: 01, the uploading time is 2021-11-1223 in the data completion process because the current data is not changed and is lost: 00: the current data corresponding to 01 is filled with the last uploaded current data, namely 229.5.

And 304, performing data preprocessing based on the historical production parameter data set corresponding to the bulking machine and the working state category corresponding to the historical production parameter data set to obtain a working state label corresponding to each historical target parameter data in the historical target parameter data set and the historical target parameter data set.

The data preprocessing refers to a data processing process performed before the data is input into a working state prediction model of the bulking machine for training. The working state label is a label for identifying the working state of the bulking machine, and the working state label can be a discrete value, for example, 0 represents normal working, and 1 represents abnormal working.

Specifically, the server performs data preprocessing such as normalization, sample equalization processing, labeling, dimension conversion and the like on the historical production parameter data set after data completion and the working state category corresponding to the historical production parameter data set, so as to obtain a working state label corresponding to each historical target parameter data in the historical target parameter data set and the historical target parameter data set.

And step 306, dividing each historical target parameter data into a training data set and a testing data set, and performing feature selection from the training data set based on the data importance of the training data to obtain target training data.

The feature selection refers to a process of selecting a group of parameter features from various parameter features of the training data, and the selected group of parameter features is the group of parameter features with the highest identification accuracy of the working state of the bulking machine. The training data set refers to a data set used for training and updating model parameters. The test data set is a data set used to test the accuracy of the model.

Specifically, the server may divide each historical target parameter data in equal proportion, for example, 10: 1 or 30: 1, dividing to obtain a training data set and a test data set. And then the server selects the characteristics in each parameter characteristic of the training data set according to the importance of the parameter characteristics, and selects a group of parameter characteristics with the highest identification accuracy of the working state of the bulking machine. Because the production parameters of the bulking machine comprise various production parameters, if all the characteristic parameters are used for model training, the training speed is reduced, the training accuracy is not high, and when the highest training accuracy can be obtained by using a few characteristic parameters, the characteristic selection is needed. The server can select all the characteristic parameters of the bulking machine by using methods such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), information gain and the like, and select a group of characteristic parameters with the highest working state identification accuracy of the bulking machine to obtain target training data.

And 308, inputting the target training data into the initial bulking machine working state prediction model for classification training to obtain a training result, and performing loss calculation based on the training result and the working state label to obtain training loss information.

The initial expander working state prediction model refers to an expander working state prediction model with initialized model parameters. The training loss information is used for representing the error between the training result of the model and the working state label, and the smaller the error is, the higher the accuracy of model prediction is.

Specifically, the server inputs target training data into an initial bulking machine working state prediction model for classification training, wherein the classification training comprises parameter training corresponding to a convolutional layer, an active layer, a pooling layer and a full-connection layer, and the server classifies the target training data based on the parameter training results corresponding to the convolutional layer, the active layer, the pooling layer and the full-connection layer to obtain training results corresponding to the target training data, including a working state corresponding to the target training data; and then the server performs loss calculation according to the working state and the working state label in the training result to obtain training loss information. The loss predicted by the prediction model of the working state of the bulking machine can be calculated by using a loss function, for example, the following formula (1) can be used for calculation:

formula (1)

Wherein loss refers to training loss information,

in order to be the true value of the value,

for model prediction, n refers to the total number of training samples.

And 310, updating the working state prediction model of the initial bulking machine based on the training loss information to obtain an updated bulking machine working state prediction model.

Specifically, after the server obtains the training loss information, the server updates the initialization parameters in the initial bulking machine working state prediction model by using the optimizer based on the training loss information, so as to obtain an updated bulking machine working state prediction model after parameter updating. The optimizer may use an Adam (adaptive moment estimation) algorithm.

And step 312, taking the updated prediction model of the working state of the bulking machine as a prediction model of the working state of the initial bulking machine, returning target training data to be input into the prediction model of the working state of the initial bulking machine for classification training, and executing the step of obtaining a training result until a training completion condition is reached to obtain the prediction model of the working state of the bulking machine to be tested.

Specifically, the server inputs target training data into an updating puffing machine working state prediction model to perform model training and loss calculation again, parameter updating is performed again by using an optimizer according to an obtained training result, a training loss result and a working state label, then the server performs iteration on the puffing machine working state prediction model for multiple times by using the training data until the obtained training loss information reaches a preset threshold value, the parameter updating is finished, and the puffing machine working state prediction model to be tested is obtained.

And step 314, testing the working state prediction model of the bulking machine to be tested based on the test data set to obtain a test result.

Specifically, the server inputs the test data set into the prediction model of the working state of the puffing machine to be tested for testing, and the prediction model of the working state of the puffing machine to be tested outputs the test working state result corresponding to the test data set after the testing is finished, so as to obtain the test result corresponding to the test data set.

And step 316, obtaining a working state prediction model of the bulking machine when the test result meets the preset test condition.

Specifically, the server compares the working state result output by the working state prediction model of the bulking machine to be tested with the working state label corresponding to the test data set. And when the same proportion of the test result and the working state label corresponding to the test data set in the comparison result reaches a preset threshold value, the test result meets a preset test condition, and a working state prediction model of the bulking machine is obtained.

In the embodiment, the working state prediction model of the bulking machine is trained and tested by using the historical production parameter data set and the working state category corresponding to the historical production parameter data set of the bulking machine, so that the working state prediction model of the bulking machine can identify the working state of the production parameter data uploaded by the bulking machine in real time, the working state of the bulking machine can be monitored in real time, and the waste of resources of the bulking machine is avoided in time.

In an embodiment, as shown in fig. 5, in step 304, performing data preprocessing based on the historical production parameter data set corresponding to the bulking machine and the working state category corresponding to the historical production parameter data set, to obtain a working state label corresponding to each historical target parameter data in the historical target parameter data set, includes:

502, normalizing based on a historical production parameter data set to obtain a historical basic production parameter data set;

step 504, performing sample equalization processing on the historical basic production parameter data based on the working state type corresponding to the historical production parameter data set to obtain historical equalized production parameter data;

step 506, traversing historical balanced production parameter data with the same working state type by using a preset sliding window to obtain a historical target parameter data set;

and step 508, labeling the working state type corresponding to each historical target parameter data in the historical target parameter data set to obtain a working state label corresponding to each historical target parameter data.

The normalization refers to uniformly classifying the historical production parameter data with different value ranges into [0,1 ]. The sample equalization processing means that the number of samples corresponding to each working state is equalized, so that the number of samples corresponding to each working state is similar. Labeling is to determine a working state corresponding to the historical production parameter data set according to the working state category and generate a corresponding working state label.

Specifically, the server normalizes the historical production parameter data set after data completion to obtain a historical basic production parameter data set. The normalization formula is shown in formula (2):

formula (2)

Wherein,

a specific value is indicated for the characteristic,

the expression is taken to be the minimum value,

it is indicated that the maximum value is taken,

the normalized values are indicated.

The server marks the working state of the historical basic production parameter data set according to the working state category, for example, the marked data is represented as: y, date _ time, x1, x2, x3, … …, xm; y represents the working state of the bulking machine and is

discrete values

0,1,2 and the like, wherein the discrete value 0 can represent the normal working state, the discrete value 1 can represent the abnormal working state, and the discrete value 2 can represent the non-working state and the like.

And the server performs sample equalization processing on the marked historical basic production parameter data set to obtain historical equalized production parameter data. When the number of samples corresponding to a certain working state is excessive, randomly extracting a part of samples; when the number of samples corresponding to a certain working state is large, oversampling is carried out, namely, existing sample data is copied, so that the number of samples corresponding to the working state is basically equal to that of samples in other working states.

The server performs one-hot encoding calculation on the working state discrete value marked in the historical equilibrium production parameter data, that is, encodes the working state discrete value into a vector, for example, binary encoding can be performed, as shown in table 1, and the one-hot encoding calculation is performed on 0 to obtain 100. 1, calculating one-hot coding to obtain 010. 2 calculating the one-hot code to obtain 001.

TABLE 1

When the server detects that the coded historical balanced production parameter data are two-dimensional data, n in the two-dimensional data (n, m) represents the number of samples, m represents the number of characteristic parameters, dimension conversion is carried out on the historical balanced production parameter data by using sample windowing, the two-dimensional data are converted into three-dimensional data, and a historical target parameter data set is obtained.

In a specific embodiment, when some pieces of data in the historical equalized production parameter data are in the same working state, the pieces of data are used as one sample data, and a window with the size of win _ len and a sliding step size are set, so that the shape of one sample is (1, win _ len, m), when all the historical equalized production parameter data are traversed in the form of a sliding window, k samples are obtained, and the shape of each sample is (k, win _ len, m). And when at least 1 working state exists in a certain window in the process of traversing the sliding window, discarding the data of the window.

And the server performs labeling on the working state category corresponding to the coded historical target parameter data set, namely, generates a corresponding working state label, and obtains the working state label corresponding to each historical target parameter data in the historical target parameter data set.

In the embodiment, the historical production parameter data set and the working state category are subjected to data preprocessing to obtain the historical target parameter data and the corresponding working state label, so that the historical target parameter data can be directly used for training in the training process of the working state prediction model of the bulking machine, and the training efficiency of the working state prediction model of the bulking machine is improved.

In one embodiment, after obtaining the prediction model of the working state of the puffing machine when the test result meets the preset test condition, the method further includes:

writing the prediction parameters and the network structure corresponding to the prediction model of the working state of the bulking machine into a model information file, and storing the model information file into a model information database;

the prediction model of the working state of the loading bulking machine comprises the following steps:

and obtaining a model information file from the model information database, and establishing a working state prediction model of the bulking machine based on the model information file.

The prediction parameters refer to parameters corresponding to the trained prediction model of the working state of the bulking machine and are used for predicting the working state of production parameter data. The network structure refers to a network hierarchical structure of a trained prediction model of the working state of the bulking machine, and comprises a convolution activation layer, a pooling layer and a full-link layer. The model information file is a file for storing the network structure of the prediction model of the working state of the bulking machine and model prediction parameters, such as neuron weight, dropout (random inactivation) and other model prediction parameters. The model information database refers to a database for storing model information files.

Specifically, the server stores the prediction parameters corresponding to the prediction model of the working state of the bulking machine as a model information file in a binary mode, and stores the model information file into a model information database. After the bulking machine is started, the server obtains the model information file from the model information database, and establishes a bulking machine working state prediction model according to the network structure and the model prediction parameters in the model information file.

In this embodiment, by generating the model information file and storing the model information file in the model information database, the working state prediction model of the bulking machine can be established by using the model information file in the model information database after the bulking machine is started, so that the use efficiency of the working state prediction model of the bulking machine is improved.

In one embodiment, as shown in fig. 6, the prediction model of the working state of the bulking machine comprises a first convolution layer, a first linear rectifying layer, a second convolution layer, a second linear rectifying layer, a pooling layer and a full-link layer;

inputting the target dimension production parameter data into a working state prediction model of the bulking machine to predict the working state of the bulking machine, and obtaining a corresponding working state result of the bulking machine in the current time period, wherein the working state result of the bulking machine comprises the following steps:

step 602, inputting target dimension production parameter data into a first convolution layer to perform a first convolution operation, so as to obtain a first convolution result;

step 604, inputting the first convolution result to a first linear rectification layer for performing a first linear rectification operation to obtain a first linear rectification result;

step 606, inputting the first linear rectification result to the second convolution layer for second convolution operation to obtain a second convolution result;

step 608, inputting the second convolution result to the second linear rectification layer for performing a second linear rectification operation to obtain a second linear rectification result;

step 610, inputting the second linear rectification result into a pooling layer to perform pooling operation to obtain a pooling result;

and step 612, inputting the pooling result into a full-link layer to perform full-link operation, so as to obtain a working state result of the bulking machine.

The first convolution layer, the first linear rectifying layer, the second convolution layer, the second linear rectifying layer, the pooling layer and the full-connection layer are hierarchical structures which are connected in sequence in a network structure corresponding to the prediction model of the working state of the bulking machine and have an operation function.

Specifically, the server inputs target dimension production parameter data into a first convolution layer in a working state prediction model of the bulking machine to perform first convolution operation, and a first convolution operation result is obtained. The server uses one-dimensional convolution, and performs inner product on data of different data windows in the target dimension production parameter data and convolution kernels to obtain a convolution result corresponding to the first convolution activation layer. For example, as shown in fig. 7, a one-dimensional convolution schematic diagram is provided, where the shape of a single target dimension production parameter data is (8, 6), the size of a convolution kernel is 5, the width of the convolution kernel is consistent with the width of the target dimension production parameter data, the shape of the convolution kernel is (5, 6), and data in each frame in the target dimension production parameter data is convolved with the convolution kernel to obtain corresponding data. The process is equivalent to that a window function traverses target dimension production parameter data, the window length is consistent with the convolution kernel size and is 5, the sliding step length is 1, the window slides by 1 unit step length each time, and the sliding is represented by an arrow in the figure to obtain 4 sliding windows. And performing convolution operation on the data of each window and a convolution kernel to obtain data output, wherein the shape of the output data is (4, 1), and a row of output data comprising 4 convolution values is represented. If n convolutions are performed by convolution operation and the sliding step is also 1, the shape of the output data is (4, n), which means that n output data comprising 4 convolution values. If m target dimension production parameter data exist, the shape of the output data is (m, 4, n), and m n pieces of output data comprising 4 convolution values are represented.

The server inputs the first convolution result into a first linear rectification layer to perform a first linear rectification operation, so as to obtain a first linear rectification result. The first linear rectifying operation includes an activation function, which may be a relu activation function (modified linear unit) with the formula f (x) = max (0, x), which means that the maximum value function is obtained, i.e. a value greater than or equal to 0 is retained, and all the remaining values less than 0 are directly rewritten to 0. For example, as shown in fig. 8, a schematic diagram of a relu activation function is provided, where data a in fig. 8 is subjected to a relu activation function operation to obtain data B in fig. 8, where B represents that a number smaller than 0 in a is rewritten to 0 and a number greater than or equal to 0 in a is reserved.

And the server inputs the first linear rectification result into the second convolution layer to carry out second convolution operation, so that a second convolution result is obtained. And then inputting the second convolution result into a second linear rectification layer to perform second linear rectification operation, so as to obtain a second linear rectification result. The operation process of the first volume operation and the second volume operation is the same; the first linear rectification operation and the second linear rectification operation have the same operation process.

The server can perform dropout operation on the convolution activation result by setting the value of the ratio rate to obtain the convolution activation result of the randomly deleted neuron node. The dropout operation is the weight for randomly deleting a neuron node by shutting down a portion of the network nodes, and the value of rate may be set between 0-1, representing the percentage of the neuron nodes deleted. For example, as shown in fig. 9, when the data shape output after convolution of certain target dimension production parameter data is (4, 3), and the setting ratio rate =0.5, 50% of the data output is randomly removed by dropout.

The server inputs the convolution activation result of the randomly deleted neuron node into the pooling layer to perform pooling operation, wherein the pooling operation can be maxporoling 1D (one-dimensional maximum pooling), namely, a point with the maximum value in the local acceptance domain is taken. And taking a group of data in the activation result according to the Pooling _ size, taking the maximum value in the group of data, taking the data of the obtained maximum value as a group of output, and synthesizing the output of the maximum value data of each group in the activation result to obtain a pooling result. For example, as shown in fig. 10, a schematic diagram of maximum pooling is provided, where the data shape is (4, 3) and pooling _ size) =2, and the pooled data shape is (2, 3), which is 1/2 of the original data.

And the server alternately repeats convolution activation and pooling operation for multiple times in the prediction process of the working state prediction model of the bulking machine so as to obtain a final pooling result, and then performs full-connection operation. The server inputs the pooling result into a full connection layer to perform full connection operation, wherein the full connection layer comprises an input layer, a hidden layer and an output layer. And the server inputs the pooling result into an input layer in the full connection layer to obtain the pooling result after the operation of the input layer. The input layer is used for sample tiling of the data with the shape of (number of samples, x, y) to obtain the data with the shape of (number of samples, x y) after tiling. As shown in fig. 11, a schematic diagram of sample conversion is provided, which includes sample tiling data with a data shape of (1, 3, 3) to obtain data with a data shape of (1, 9).

The server inputs the pooling result after the operation of the input layer into the hidden layer, the weight value of each neuron in the hidden layer and each neuron in the input layer is operated, and the operation formula is shown as a formula (3):

formula (3)

Wherein,

represents the value of hidden layer neuron h,

representing the weights of the connections of input layer neuron i to hidden layer neuron h,

represents the value of input layer neuron i, and d represents the number of input layer neurons.

The server inputs the weight value of each neuron obtained by operation into the output layer, the weight value of each neuron in the output layer and each neuron in the hidden layer is operated, and the operation formula is shown as a formula (4):

formula (4)

Wherein,

representing output layer neurons jThe value of (a) is,

weights for the connections of hidden layer neurons h and output layer neurons j,

represents the value of hidden layer neuron h, and q represents the number of hidden layer neurons. As shown in fig. 12, a schematic diagram of the relationship between the input layer, hidden layer and output layer is provided, where yj represents the input of the jth output neuron and bh represents the input of the h hidden layer neuron.

The server processes the weight value of each neuron of the output layer by using an activation function softmax (normalized exponential function), and the softmax formula is shown as formula (5):

formula (5)

Wherein,

a value representing the jth neuron of the output layer, for a total of n neurons,

the j-th neuron is processed by softmax, after all neurons in an output layer are processed, the sum of the values is 1, the value with the largest numerical value is selected, and the production state of the bulking machine at the corresponding position is the predicted working state result of the bulking machine. For example, the puffing machine has three working states of starting, stabilizing production and stopping]Results obtained after softmax treatment were [0.1, 0.5, 0.4 ]]And the working state of the bulking machine is stable production.

In this embodiment, the target dimension production parameter data is predicted through the expander working state prediction model, and the expander working state result corresponding to the expander in the current time period is obtained, so that the manager can timely perform corresponding processing according to the expander working state result, and the resource waste of the expander is timely avoided.

In one embodiment, the bulking machine working state result comprises a normal working state and an abnormal working state; after inputting the target dimension production parameter data into the working state prediction model of the bulking machine to predict the working state of the bulking machine and obtaining the output working state result of the bulking machine, the method further comprises the following steps:

and when the working state result of the bulking machine is detected to be an abnormal working state, generating warning information and sending the warning information to the management terminal.

The warning information is a prompt message for indicating an abnormal operation state of the expansion machine. The abnormal working state comprises an idling state, an abnormal working state and the like of the bulking machine.

Specifically, the server detects the result of the working state of the bulking machine according to the result of the working state of the bulking machine output by the working state prediction model of the bulking machine, generates corresponding warning information when detecting that the result of the working state of the bulking machine is in an abnormal working state, and sends the warning information to the management terminal.

In this embodiment, when detecting that the bulking machine is in an abnormal working state, the warning information is generated and sent to the management terminal, so that the management personnel can check the production condition of the bulking machine in time and process the production condition, and the resource waste of the bulking machine is avoided in time.

In one embodiment, as shown in fig. 13, a schematic flow chart of the operation state recognition of the puffing machine is provided;

after detecting that the bulking machine is started, the bulking machine uploads production parameter data to a database or the local area in real time; the server loads a historical production parameter data set at regular time and performs data preprocessing on the historical production parameter data set, and the method comprises the following steps: completing data, normalizing, adding labels, performing single hot coding, performing sample equalization processing, windowing samples, and dividing a test data set and a training data set; and obtaining a training data set after data preprocessing and data division, and then carrying out one-time feature selection on the training data set to obtain target training data.

Then the server uses the target training data to carry out model training on the initial bulking machine working state prediction model, and the model training comprises the following steps: reading target training data, defining a network structure, training a model, and updating parameters by using an Adam optimizer to obtain a working state prediction model of the bulking machine; and the server generates a prediction parameter file according to the updated parameters and stores the prediction parameter file into a model parameter database.

As shown in fig. 14, a schematic diagram of a network structure is provided, which includes: and performing convolution operation twice, wherein the convolution operation comprises an activation function rule, and the convolution operation is followed by dropout random pruning operation with the pruning ratio of 0.5. Performing maximum pooling operation with a pooling size of 2 after dropout operation; and entering a full connection layer after the maximum pooling operation, wherein the full connection layer comprises an input layer, a hidden layer and an output layer, the hidden layer comprises an activation function rule, and the output layer comprises an activation function softmax.

After the server acquires the production parameter data of each initial dimension of the bulking machine in the current time period, the data preprocessing is carried out on the production parameter data of each initial dimension, and the method comprises the following steps: and normalizing and windowing the sample to obtain target dimension production parameter data.

And then the server loads the working state prediction model of the bulking machine from the model parameter database, and carries out working state identification on the target dimension production parameter data through the working state prediction model of the bulking machine to obtain a corresponding working state result of the bulking machine in the current time period.

It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a part of the steps or stages in other steps.

Based on the same inventive concept, the embodiment of the present application further provides a device for recognizing the working state of the puffing machine, which is used for implementing the method for recognizing the working state of the puffing machine. The implementation scheme for solving the problem provided by the apparatus is similar to the implementation scheme described in the above method, so that the specific limitations in one or more embodiments of the bulking machine operating state identification apparatus provided below can refer to the limitations on the bulking machine operating state identification method in the above description, and are not described herein again.

In one embodiment, as shown in fig. 15, there is provided a bulking machine operation state identifying apparatus 1500 comprising: an obtaining module 1502, a converting module 1504, a loading module 1506, and an outputting module 1508, wherein:

and the acquisition module is used for acquiring the corresponding initial dimension production parameter data of the bulking machine in the current time period.

And the conversion module is used for carrying out dimension conversion on each initial dimension production parameter data corresponding to the current time period to obtain target dimension production parameter data.

And the loading module is used for loading the working state prediction model of the bulking machine, and the working state prediction model of the bulking machine is obtained by training through a neural network by using training data.

In one embodiment, the bulking machine operation state identifying apparatus 1500 further comprises:

the training unit is used for acquiring a historical production parameter data set corresponding to the bulking machine and a working state category corresponding to the historical production parameter data set; performing data preprocessing based on a historical production parameter data set corresponding to the bulking machine and a working state category corresponding to the historical production parameter data set to obtain a historical target parameter data set and a working state label corresponding to each historical target parameter data in the historical target parameter data set; dividing each historical target parameter data into a training data set and a testing data set, and performing feature selection from the training data set based on the data importance of the training data to obtain target training data; inputting target training data into an initial bulking machine working state prediction model for classification training to obtain a training result, and performing loss calculation based on the training result and a working state label to obtain training loss information; updating the working state prediction model of the initial bulking machine based on the training loss information to obtain an updated bulking machine working state prediction model; taking the updated prediction model of the working state of the bulking machine as a prediction model of the working state of the initial bulking machine, returning and inputting target training data into the prediction model of the working state of the initial bulking machine for classification training, and executing the step of obtaining a training result until a training completion condition is reached to obtain the prediction model of the working state of the bulking machine to be tested; testing the working state prediction model of the basic bulking machine to be tested based on the test data set to obtain a test result; and when the test result meets the preset test condition, obtaining a prediction model of the working state of the bulking machine.

the data preprocessing unit is used for carrying out normalization on the basis of the historical production parameter data set to obtain a historical basic production parameter data set; performing sample equalization processing on historical basic production parameter data based on the working state category corresponding to the historical production parameter data set to obtain historical equalized production parameter data; traversing historical balanced production parameter data with the same working state type by using a preset sliding window to obtain a historical target parameter data set; and labeling the working state category corresponding to each historical target parameter data in the historical target parameter data set to obtain a working state label corresponding to each historical target parameter data.

the parameter file generating unit is used for writing the prediction parameters and the network structure corresponding to the working state prediction model of the bulking machine into a model information file and storing the model information file into a model information database; the prediction model of the working state of the loading bulking machine comprises the following steps: and obtaining a model information file from the model information database, and establishing a prediction model of the working state of the bulking machine based on the model information file.

In one embodiment, the output module 1508 further includes:

the calculation unit is used for inputting the target dimension production parameter data into the first volume activation layer to obtain a first volume activation result; inputting the first convolution activation result into a second convolution activation layer to obtain a second convolution activation result; inputting the second convolution activation result into a pooling layer to perform pooling operation to obtain a pooling result; and inputting the pooling result into a full-connection layer to perform full-connection operation to obtain a working state result of the bulking machine.

and the warning unit is used for generating warning information and sending the warning information to the management terminal when detecting that the working state result of the bulking machine is an abnormal working state.

All or part of each module in the bulking machine working state identification device can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 16. The computer device includes a processor, a memory, an Input/Output interface (I/O for short), and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing historical production parameter data, initial dimension production parameter data and the like. The input/output interface of the computer device is used for exchanging information between the processor and an external device. The communication interface of the computer device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement a method for recognizing the operating state of a puffing machine.

In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 17. The computer apparatus includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input device. The processor, the memory and the input/output interface are connected by a system bus, and the communication interface, the display unit and the input device are connected by the input/output interface to the system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The input/output interface of the computer device is used for exchanging information between the processor and an external device. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a method for recognizing the operating state of a puffing machine. The display unit of the computer equipment is used for forming a visual and visible picture, and can be a display screen, a projection device or a virtual reality imaging device, the display screen can be a liquid crystal display screen or an electronic ink display screen, the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.

It will be appreciated by those skilled in the art that the configurations shown in fig. 16-17 are only block diagrams of some configurations relevant to the present disclosure, and do not constitute a limitation on the computing devices to which the present disclosure may be applied, and a particular computing device may include more or less components than those shown, or some components may be combined, or have a different arrangement of components.

In one embodiment, a computer device is further provided, which includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method embodiments when executing the computer program.

In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.

In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, carries out the steps in the method embodiments described above.

It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the relevant laws and regulations and standards of the relevant country and region.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), Magnetic Random Access Memory (MRAM), Ferroelectric Random Access Memory (FRAM), Phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for identifying the working state of a bulking machine is characterized by comprising the following steps:

performing dimension conversion on each initial dimension production parameter data corresponding to the current time period in a sliding window traversal mode by using a window with a preset size and a sliding step length to obtain target dimension production parameter data;

loading a prediction model of the working state of the bulking machine, wherein the prediction model of the working state of the bulking machine is obtained by training through a neural network by using training data, the training data is obtained by dividing each historical target parameter data, and each historical target parameter data is three-dimensional data obtained by performing dimension conversion on historical balanced production parameter data with the same working state type in a sliding window traversal mode by using a window with a preset size and a sliding step length; the historical balanced production parameter data is two-dimensional data obtained by normalizing based on a historical production parameter data set corresponding to the bulking machine to obtain a historical basic production parameter data set and performing sample balanced processing on the historical basic production parameter data set based on the working state type corresponding to the historical production parameter data set;

and inputting the target dimension production parameter data into the working state prediction model of the bulking machine to predict the working state of the bulking machine, so as to obtain a corresponding working state result of the bulking machine in the current time period.

2. The method of claim 1, wherein the training of the prediction model of the operating state of the puffing machine comprises the steps of:

acquiring a historical production parameter data set corresponding to the bulking machine and a working state category corresponding to the historical production parameter data set;

performing data preprocessing based on a historical production parameter data set corresponding to the bulking machine and a working state category corresponding to the historical production parameter data set to obtain a historical target parameter data set and a working state label corresponding to each historical target parameter data in the historical target parameter data set;

dividing each historical target parameter data into a training data set and a testing data set, and performing feature selection from the training data set based on the data importance of the training data to obtain target training data;

inputting the target training data into an initial bulking machine working state prediction model for classification training to obtain a training result, and performing loss calculation based on the training result and the working state label to obtain training loss information;

updating the working state prediction model of the initial bulking machine based on the training loss information to obtain an updated bulking machine working state prediction model;

taking the updated prediction model of the working state of the bulking machine as the prediction model of the working state of the initial bulking machine, returning and inputting the target training data into the prediction model of the working state of the initial bulking machine for classification training, and executing the step of obtaining a training result until the training completion condition is reached to obtain the prediction model of the working state of the bulking machine to be tested;

testing the working state prediction model of the bulking machine to be tested based on the test data set to obtain a test result;

and when the test result meets a preset test condition, obtaining the working state prediction model of the bulking machine.

3. The method according to claim 2, wherein the performing data preprocessing based on the historical production parameter dataset corresponding to the bulking machine and the working state category corresponding to the historical production parameter dataset to obtain a historical target parameter dataset and a working state label corresponding to each historical target parameter data in the historical target parameter dataset comprises:

and labeling based on the working state category corresponding to each historical target parameter data in the historical target parameter data set to obtain a working state label corresponding to each historical target parameter data.

4. The method according to claim 2, wherein after obtaining the expander work state prediction model when the test result satisfies the preset test condition, the method further comprises:

and acquiring the model information file from the model information database, and establishing the working state prediction model of the bulking machine based on the model information file.

5. The method of claim 1, wherein the bulking machine operating state prediction model comprises a first convolutional layer, a first linear rectifying layer, a second convolutional layer, a second linear rectifying layer, a pooling layer, and a fully-connected layer;

the inputting the target dimension production parameter data into the expander working state prediction model to predict the expander working state to obtain the corresponding expander working state result of the expander in the current time period includes:

inputting the target dimension production parameter data into the first convolution layer to perform first convolution operation to obtain a first convolution result;

inputting the first convolution result into the first linear rectification layer to perform a first linear rectification operation, so as to obtain a first linear rectification result;

inputting the first linear rectification result into the second convolution layer to carry out second convolution operation to obtain a second convolution result;

inputting the second convolution result into the second linear rectification layer to perform second linear rectification operation to obtain a second linear rectification result;

inputting the second linear rectification result into a pooling layer to perform pooling operation to obtain a pooling result;

and inputting the pooling result into a full-link layer to perform full-link operation, so as to obtain the working state result of the bulking machine.

6. The method of claim 1, wherein the bulking machine work state results comprise normal work states, abnormal work states;

after the target dimension production parameter data is input into the working state prediction model of the bulking machine to predict the working state of the bulking machine, and an output working state result of the bulking machine is obtained, the method further comprises the following steps:

and when the working state result of the bulking machine is detected to be an abnormal working state, generating warning information and sending the warning information to a management terminal.

7. A device for identifying the working state of a bulking machine, which is characterized in that the device comprises:

the conversion module is used for performing dimension conversion on each initial dimension production parameter data corresponding to the current time period in a sliding window traversal mode by using a window with a preset size and a sliding step length to obtain target dimension production parameter data;

the loading module is used for loading a prediction model of the working state of the bulking machine, the prediction model of the working state of the bulking machine is obtained by training through a neural network by using training data, the training data is obtained by dividing each historical target parameter data, and each historical target parameter data is three-dimensional data obtained by performing dimension conversion on historical balanced production parameter data with the same working state type in a sliding window traversal mode by using a window with a preset size and a sliding step length; the historical balanced production parameter data is two-dimensional data obtained by normalizing based on a historical production parameter data set corresponding to the bulking machine to obtain a historical basic production parameter data set and performing sample balanced processing on the historical basic production parameter data set based on the working state type corresponding to the historical production parameter data set;

and the output module is used for inputting the target dimension production parameter data into the working state prediction model of the bulking machine to predict the working state of the bulking machine so as to obtain a corresponding working state result of the bulking machine in the current time period.

8. The apparatus of claim 7, further comprising:

and the warning unit is used for generating warning information and sending the warning information to a management terminal when the working state result of the bulking machine is detected to be an abnormal working state.

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor realizes the steps of the method of any one of claims 1 to 6 when executing the computer program.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.