Coal semi-automatic gelatinous layer index monitoring method and system

Abstract

The invention relates to the field of coal gelatinous layer index monitoring, in particular to a semi-automatic coal gelatinous layer index monitoring method and system. The method comprises the following steps: acquiring a data set; obtaining a pre-constructed index monitoring network; inputting samples in the data set into an index monitoring network in batches, and completing iterative training of the index monitoring network so as to realize semi-automatic gelatinous layer index monitoring of the coal; wherein any iteration training process comprises the steps of constructing a small sample feature extraction loss function, and supervising training of the index monitoring network by using the small sample feature extraction loss function. And the accuracy of monitoring the gelatinous layer index of the small sample coal is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of coal gelatinous layer index monitoring, and in particular to a semi-automatic coal gelatinous layer index monitoring method and system. Background Art

[0002] The colloidal layer index (GLI) is a key indicator for evaluating the quality of coking coal. It directly reflects the formation and solidification characteristics of colloids during coal pyrolysis and has important guiding significance for production practices in industries such as coal and metallurgy. Currently, the determination of the GLI is primarily based on a semi-automatic GLI instrument. This instrument records the temperature-displacement curve of the coal sample during heating and uses manual interpretation or algorithmic analysis to determine key parameters such as the maximum thickness and ultimate shrinkage of the GLI.

[0003] With the development of intelligent detection technology, neural network-based data analysis methods have gradually been applied to the automatic analysis of the colloidal layer index, significantly improving detection efficiency and standardization. However, existing neural networks have exposed obvious limitations when processing small-sample coal data. Due to the limited number of samples of small-sample coal types (such as scarce coal types and coal samples from special geological areas), traditional neural network architectures have difficulty effectively capturing the characteristic information contained in the data, which can easily lead to model underfitting or overfitting. In the context of colloidal layer index detection, the temperature-displacement curve characteristics of small-sample coal types are particularly complex and variable. Minor differences in curve morphology may correspond to different patterns of colloidal formation. Traditional networks are unable to accurately extract this unique information, ultimately resulting in large errors in the detection results of the colloidal layer index of small-sample coal types, seriously restricting the refined evaluation and rational utilization of coal resources. Therefore, it is urgent to develop a method for accurately detecting the colloidal layer index of small-sample coal types to overcome the technical bottleneck of traditional neural networks in processing small-sample data. Summary of the Invention

[0004] In order to solve the problem of how to improve the accuracy of colloid index monitoring of small sample coal types, the present invention provides a semi-automatic colloid index monitoring method and system for coal.

[0005] In a first aspect, the present invention provides a method for semi-automatic monitoring of the colloid index of coal, which adopts the following technical solution: A method for semi-automatic monitoring of coal colloid index comprises the following steps: A data set is obtained, where the data set includes a plurality of samples, where the samples can be divided into large samples and small samples, and each sample is a labeled index monitoring curve of a semi-automatic colloid layer of coal; Obtain a pre-built exponential monitoring network, which is a network with a small sample feature extraction structure added between the shared layer and the fully connected layer of the Conv1D network; The samples in the dataset are input into the index monitoring network in batches, and the iterative training of the index monitoring network is completed to realize the semi-automatic gelatinous layer index monitoring of coal. Among them, any iterative training process includes: constructing a small sample feature extraction loss function , the small sample input in the batch is recorded as the target small sample, the sample of the same type of coal as the target small sample is obtained and recorded as the reference sample, the common features of the output data of all reference samples in the shared layer are extracted and recorded as the common features, and the distinguishing features of the output data of the target small sample and the large sample input in the batch in the shared layer are extracted and recorded as the unique features, Indicates shared characteristics, Represents the output data of the target small sample in the small sample extraction structure, Indicates unique characteristics, Indicates the correlation of data; uses the small sample feature extraction loss function and the loss function of the Conv1D network to jointly supervise the training of the index monitoring network.

[0006] The present invention takes into account that the number of samples extracted from some coal types is small, which leads to the inability of traditional networks to obtain the feature extraction capabilities of small sample coal type information. Therefore, a small sample feature extraction structure is added to the traditional network to extract the unique information of small sample coal types, thereby better extracting the accuracy of the colloid layer index monitoring of small sample coal types; further, a small sample feature extraction loss function is set for the small sample structure, so that the small sample structure can obtain the ability to extract small sample features faster and better; further, unique features are introduced into the small sample feature extraction loss function to enable the small sample feature extraction structure to extract more unique information of small samples that is different from large samples, thereby making up for the inability of traditional networks to obtain the ability to extract unique information of small samples; further, common features are introduced into the small sample feature extraction loss function to enable the small sample feature extraction structure to extract common information between small samples, thereby preventing the small sample feature extraction structure from extracting some differentiated noise information, thereby improving the accuracy of feature extraction by the small sample feature extraction structure.

[0007] Preferably, the obtaining of a pre-built index monitoring network includes: The first preset number of network layers in the Conv1D network are used as shared layers, and the network layers after the shared layer to the fully connected layer in the Conv1D network are recorded as large sample feature extraction structures. A threshold structure and a small sample feature extraction structure are added after the shared layer. The small sample feature extraction structure is located after the threshold structure and is parallel to the large sample feature extraction structure, wherein the threshold structure is used to divert the data flow of small sample information to the small sample extraction structure, and divert the data flow of large sample information to the large sample feature extraction structure.

[0008] The present invention takes into account that the front layers of the network only have the ability to extract common shallow features, and therefore reserves a shared layer in the network to extract features for both large-eye samples and small samples, so that large samples can also participate in the training at the shared layer, effectively reducing the training scale of the unique network parameters of small samples, and preventing the problem of insufficient number of small samples to better complete the training of large-scale unique network parameters.

[0009] Preferably, the network layer in the small sample feature extraction structure is the same as the network layer in the large sample feature extraction structure.

[0010] Preferably, the method for realizing semi-automatic coal colloid index monitoring includes: The newly collected semi-automatic colloid layer index monitoring curve of the coal is input into the trained index monitoring network to obtain the newly collected semi-automatic colloid layer index of the coal.

[0011] Preferably, extracting the common features of the output data of all reference samples in the shared layer and recording them as the common features includes: The output data of each reference sample in the shared layer is obtained, and the mean of all output data of each reference sample in the shared layer is used as the comprehensive output data of each reference sample. The principal component analysis algorithm is used to reduce the dimension of the comprehensive output data of all reference samples, and the obtained reduced dimension data is used as the common feature.

[0012] The present invention eliminates the difference information between reference samples through the dimensionality reduction processing of the principal component analysis algorithm, and accurately extracts the common information between the reference samples.

[0013] Preferably, the method for obtaining the unique features includes: Obtain the output data of the large sample input in the batch in the shared layer, and use the average of the output data of each large sample in the shared layer as the comprehensive output data of each large sample; obtain the output data of the target small sample in the shared layer, and use the average of the output data of the target small sample in the shared layer as the comprehensive output data of the target small sample; calculate the absolute value of the difference between the corresponding positions of the comprehensive output data of the large sample and the comprehensive output data of the target small sample to obtain the difference data; set the position with a value greater than the preset difference threshold in the difference data to 1, and set the position with a value not greater than the preset difference threshold to 0 to obtain sparse data; perform an OR operation on the corresponding positions of the sparse data of all large samples to obtain comprehensive sparse data; perform a multiplication operation on the comprehensive output data of the target small sample and the comprehensive sparse data to obtain unique features.

[0014] The present invention excludes information with little difference from the large sample and more accurately extracts the unique information of the target small sample.

[0015] Preferably, the use of a small sample feature extraction loss function and a loss function inherent in the Conv1D network to jointly supervise the training of the index monitoring network includes: The first loss value is calculated using the small sample feature extraction loss function, the second loss value is calculated using the loss function of the Conv1D network, and the first loss value and the second loss value are used to perform reverse gradient update on the parameters in the exponential monitoring network.

[0016] Preferably, the index monitoring curve is a temperature-displacement curve measured using a semi-automatic colloid layer index measuring instrument.

[0017] Preferably, the samples are divided into large samples and small samples by comparing the number of samples of each type of coal with a preset value.

[0018] The present invention distinguishes large samples from small samples by comparing preset values ​​of the sample quantity. This method is relatively simple and has high implementation efficiency.

[0019] In a second aspect, the present invention provides a semi-automatic coal colloid index monitoring system, which adopts the following technical solution: A semi-automatic coal colloid index monitoring system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned semi-automatic coal colloid index monitoring method is implemented.

[0020] By adopting the above technical solution, the above-mentioned coal semi-automatic colloid index monitoring method is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is made based on the memory and the processor for easy use.

[0021] The present invention has the following technical effects: The present invention takes into account that the number of samples extracted from some coal types is small, which makes the traditional network unable to obtain the feature extraction capability of small sample coal type information. Therefore, a small sample feature extraction structure is added to the traditional network to extract the unique information of small sample coal types, thereby improving the accuracy of the colloid layer index monitoring of small sample coal types. Furthermore, by setting a small sample feature extraction loss function for the small sample structure, the small sample structure can obtain the ability to extract small sample features faster and better; Furthermore, unique features are introduced into the small sample feature extraction loss function to enable the small sample feature extraction structure to extract more unique information of small samples that is different from large samples, thereby compensating for the inability of traditional networks to extract unique information of small samples; Furthermore, common features are introduced into the small sample feature extraction loss function to enable the small sample feature extraction structure to extract common information between small samples, thereby preventing the small sample feature extraction structure from extracting some differentiated noise information, thereby improving the accuracy of feature extraction by the small sample feature extraction structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of a method for semi-automatic coal colloid index monitoring according to an embodiment of the present invention; Figure 2 Schematic diagram of the network structure of the index monitoring network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The embodiment of the present invention discloses a method for semi-automatic monitoring of coal colloid index, referring to Figure 1 , including steps S1 to S3: S1: Obtain a data set, wherein the data set includes a plurality of samples, and the samples can be divided into large samples and small samples, and each sample is an index monitoring curve of a semi-automatic colloid layer of coal with a label.

[0024] Specifically, a semi-automatic gelatinous layer index measuring instrument is used to measure the temperature-displacement curves of various types of coal and record them as index monitoring curves. Based on experience, the index monitoring curves are analyzed to obtain the gelatinous layer index, which is used as a label for the corresponding index monitoring curve. The gelatinous layer index can be the maximum gelatinous layer thickness and final shrinkage, or other data, which is not limited in this embodiment.

[0025] The labeled exponential monitoring curve is taken as a sample, and all samples constitute the data set.

[0026] Obtain sample data for each type of coal in the data set, and record the samples of coal types with sample data less than the preset value as small samples; record the samples of coal types with sample numbers not less than the preset value as large samples.

[0027] S2: Obtain a pre-built exponential monitoring network, where the exponential monitoring network is a network with a small sample feature extraction structure added between the shared layer and the fully connected layer of the Conv1D network.

[0028] It should be noted that in order to perform glial layer index detection, a network for glial layer index detection needs to be constructed first.

[0029] It's important to note that traditional neural networks use a single network structure to extract features from all samples. However, some coal types have fewer samples, so less data from these types of coal participates in network training. This results in the network being more suited to extracting features from large samples, leading to poor performance on small samples. To improve the network's ability to extract features from small samples, we considered adding a network structure specifically designed to extract feature information from small samples to the traditional network.

[0030] Preferably, as an example, obtaining a pre-built index monitoring network includes: The first preset number of network layers in the Conv1D network are used as shared layers, and the network layers after the shared layer to the fully connected layer in the Conv1D network are recorded as large sample feature extraction structures. A threshold structure and a small sample feature extraction structure are added after the shared layer. The small sample feature extraction structure is located after the threshold structure and is parallel to the large sample feature extraction structure, wherein the threshold structure is used to divert the data flow of small sample information to the small sample extraction structure, and divert the data flow of large sample information to the large sample feature extraction structure.

[0031] The network layers of the small sample feature extraction structure are the same as those of the large sample feature extraction structure, except that the number of network layers of the small sample feature extraction structure may be smaller than that of the large sample feature extraction structure. Figure 2 Schematic diagram of the network structure of the index monitoring network. Figure 2 The part in the dotted rectangle box represents the shared layer, the diamond represents the threshold structure, the part in the solid rectangle box above represents the large sample feature extraction structure, the part in the implementation rectangle box below represents the small sample feature extraction structure, and the part in the cylinder represents the fully connected layer.

[0032] It is understandable that due to the small amount of small sample data, large-scale feature extraction structures trained with a small number of small samples are not accurate enough. Under normal circumstances, the features extracted by the early layers of the network are mostly shared shallow information, so the feature extraction structures of the early layers can be trained with all samples together. This can reduce the size of the feature extraction structure trained on small samples alone, thereby improving the ability to extract features.

[0033] To facilitate understanding, the following explains the data flow of the index monitoring network: Samples are input into the index monitoring network and first processed by the shared layer. After passing through the shared layer, the data flows into the threshold structure. The threshold structure determines whether the sample is a large or small sample by identifying the number of samples of the coal category to which the sample belongs. If the sample is determined to be a large sample, the data flows into the large sample feature extraction structure; if the sample is determined to be a small sample, the data flows into the small sample extraction structure. After being processed by the corresponding large sample feature extraction structure or small sample extraction structure, the data flows into the fully connected layer, and the output is obtained after further processing by the fully connected layer.

[0034] S3: Input the samples in the data set into the index monitoring network in batches to complete the iterative training of the index monitoring network to realize semi-automatic coal colloid index monitoring.

[0035] S30: Input the samples in the data set into the exponential monitoring network in batches to complete the iterative training of the exponential monitoring network.

[0036] It should be noted that due to the small number of small samples, traditional neural networks are unable to extract the unique feature information of small samples, resulting in poor monitoring of the colloidal layer index of small sample coals. In order to enable the index monitoring network to learn the unique feature information of small samples, a loss function must be set for the small sample feature extraction structure. This loss function supervises the small sample extraction structure to extract the unique information of small samples.

[0037] Among them, any iterative training process includes: constructing a small sample feature extraction loss function; using the small sample feature extraction loss function and the loss function of the Conv1D network to jointly supervise the training of the index monitoring network.

[0038] S300: Construct a small sample feature extraction loss function.

[0039] Preferably, as an example, constructing a small sample feature extraction loss function includes:

[0040] Among them, the small sample input in the batch is recorded as the target small sample, the sample of the same type of coal as the target small sample is obtained and recorded as the reference sample, the common features of the output data of all reference samples in the shared layer are extracted and recorded as the common features, and the distinguishing features of the output data of the target small sample and the large sample input in the batch in the shared layer are extracted and recorded as the unique features. Indicates shared characteristics, Represents the output data of the target small sample in the small sample extraction structure, Indicates unique characteristics, represents the correlation of data; S represents the small sample feature extraction loss function.

[0041] It is understandable that the output data of the target small sample in the small sample extraction structure It reflects the target small sample feature information extracted by the small sample extraction structure, and the unique features It reflects the unique characteristic information of the target small sample that is different from the large sample. Reflects the correlation between the characteristic information and unique information of the target small sample extracted by the small sample extraction structure. The larger the value, the better the small sample extraction structure can extract the unique information of the target small sample. It reflects the common characteristic information of samples of the same type of coal as the target small sample. The correlation between the characteristic information of the target small sample extracted by the small sample extraction structure and the common information. Since the distinguishing information between the target small sample and the large sample may also be noise information, etc., the extraction of this information will interfere with the detection results. Therefore, by analyzing the correlation between the characteristic information of the target small sample extracted by the small sample extraction structure and the common information, the small sample extraction structure can extract more unique common information of the small sample rather than some noise information.

[0042] The above embodiments involve common features, unique features, the correlation between common features and the output data of target small samples in the small sample extraction structure, and the correlation between unique features and the output data of target small samples in the small sample extraction structure. The following describes the method for determining the correlation between common features, unique features, the correlation between common features and the output data of target small samples in the small sample extraction structure, and the correlation between unique features and the output data of target small samples in the small sample extraction structure.

[0043] First, the method of obtaining common features is introduced.

[0044] Preferably, as an example, the common feature acquisition method includes: The output data of each reference sample in the shared layer is obtained, and the mean of all output data of each reference sample in the shared layer is used as the comprehensive output data of each reference sample. The principal component analysis algorithm is used to reduce the dimension of the comprehensive output data of all reference samples, and the obtained reduced dimension data is used as the common feature.

[0045] It can be understood that after the comprehensive output data is subjected to dimensionality reduction processing by the principal component analysis algorithm, the common correlation information in the comprehensive output data will be retained, and some specific information will be removed, thereby extracting the common feature information.

[0046] Then the method of obtaining unique features is introduced.

[0047] Preferably, as an example, the method for obtaining the unique features includes: Obtain the output data of the large sample input in the batch in the shared layer, and use the average of the output data of each large sample in the shared layer as the comprehensive output data of each large sample; obtain the output data of the target small sample in the shared layer, and use the average of the output data of the target small sample in the shared layer as the comprehensive output data of the target small sample; calculate the absolute value of the difference between the corresponding positions of the comprehensive output data of the large sample and the comprehensive output data of the target small sample to obtain the difference data; set the position with a value greater than the preset difference threshold in the difference data to 1, and set the position with a value not greater than the preset difference threshold to 0 to obtain sparse data; perform an OR operation on the corresponding positions of the sparse data of all large samples to obtain comprehensive sparse data; perform a multiplication operation on the comprehensive output data of the target small sample and the comprehensive sparse data to obtain unique features.

[0048] It can be understood that the data at each position in the comprehensive output data reflects different data characteristics. By screening out the data with large differences between the comprehensive output data of the target small sample and the comprehensive output data of the large sample, the data is used as the unique characteristics of the target small sample compared with the large sample.

[0049] Then, the calculation method of the correlation between the common features and the output data of the target small sample in the small sample extraction structure is introduced.

[0050] Preferably, as an example, a method for calculating the correlation between the common features and the output data of the target small sample in the small sample extraction structure includes: The mutual information between the common features and the output data of the target small sample in the small sample extraction structure is calculated as the correlation between the common features and the output data of the target small sample in the small sample extraction structure.

[0051] It can be understood that the mutual information can reflect the correlation between different data sequences. It can reflect both linear correlation and nonlinear correlation. Therefore, it can better reflect the correlation between common features and the output data of the target small sample in the small sample extraction structure.

[0052] Finally, a method for introducing the correlation between unique features and output data of target small samples in a small sample extraction structure is introduced.

[0053] Preferably, as an example, a method for calculating the correlation between the unique feature and the output data of the target small sample in the small sample extraction structure includes: The mutual information between the unique feature and the output data of the target small sample in the small sample extraction structure is calculated as the correlation between the unique feature and the output data of the target small sample in the small sample extraction structure.

[0054] S301: Use the small sample feature extraction loss function and the loss function of the Conv1D network to jointly supervise the training of the index monitoring network.

[0055] Preferably, as an example, the training of the index monitoring network is jointly supervised by using the small sample feature extraction loss function and the loss function of the Conv1D network itself, including: After the small samples of this batch are input into the index monitoring network, the loss value is calculated using the small sample feature extraction loss function and recorded as the first loss value; At the end of the exponential monitoring network, the output data is obtained. According to the output data and labels, the loss value is calculated using the loss function of the Conv1D network in the exponential monitoring network and recorded as the second loss value.

[0056] The first loss value and the second loss value are used to reversely update the network parameters in the exponential monitoring network through the gradient descent method.

[0057] It can be understood that since only small samples are extracted through the small sample structure, only small sample information is used to update the information of the small sample extraction structure, thereby preventing large sample information from overwhelmed by small sample information and causing the network to be unable to obtain the ability to extract small sample features.

[0058] It should be noted that the use of the gradient descent method to reversely update the network parameters in the network based on the loss value is an existing technology and will not be described in detail here.

[0059] S31: To realize semi-automatic monitoring of coal colloid index.

[0060] Preferably, as an example, to realize semi-automatic coal colloid index monitoring, the method includes: The newly collected semi-automatic colloid layer index monitoring curve of the coal is input into the trained index monitoring network to obtain the newly collected semi-automatic colloid layer index of the coal.

[0061] An embodiment of the present invention also discloses a semi-automatic coal colloid index monitoring system, including a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, a semi-automatic coal colloid index monitoring method according to the present invention is implemented.

[0062] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0063] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as resistive random access memory, dynamic random access memory, static random access memory, enhanced dynamic random access memory, high bandwidth memory, hybrid memory cube, etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium may be part of, accessible to, or connectable to the device.

Claims

1. A semi-automatic coal colloid index monitoring method, characterized in that: Including steps: A data set is obtained, where the data set includes a plurality of samples, where the samples can be divided into large samples and small samples, and each sample is a labeled index monitoring curve of a semi-automatic colloid layer of coal; Obtain a pre-built exponential monitoring network, which is a network with a small sample feature extraction structure added between the shared layer and the fully connected layer of the Conv1D network; The samples in the dataset are input into the index monitoring network in batches, and the iterative training of the index monitoring network is completed to realize the semi-automatic gelatinous layer index monitoring of coal. Among them, any iterative training process includes: constructing a small sample feature extraction loss function , the small sample input in the batch is recorded as the target small sample, the sample of the same type of coal as the target small sample is obtained and recorded as the reference sample, the common features of the output data of all reference samples in the shared layer are extracted and recorded as the common features, and the distinguishing features of the output data of the target small sample and the large sample input in the batch in the shared layer are extracted and recorded as the unique features, Indicates shared characteristics, Represents the output data of the target small sample in the small sample extraction structure, Indicates unique characteristics, Indicates the correlation of data; uses the small sample feature extraction loss function and the loss function of the Conv1D network to jointly supervise the training of the index monitoring network.

2. A method for monitoring coal semi-automatic colloid index according to claim 1, characterized in that: The obtaining of a pre-built index monitoring network includes: The first preset number of network layers in the Conv1D network are used as shared layers, and the network layers after the shared layer to the fully connected layer in the Conv1D network are recorded as large sample feature extraction structures. A threshold structure and a small sample feature extraction structure are added after the shared layer. The small sample feature extraction structure is located after the threshold structure and is parallel to the large sample feature extraction structure, wherein the threshold structure is used to divert the data flow of small sample information to the small sample extraction structure, and divert the data flow of large sample information to the large sample feature extraction structure.

3. A method for monitoring coal semi-automatic colloid index according to claim 2, characterized in that: The network layer in the small sample feature extraction structure is the same as the network layer in the large sample feature extraction structure.

4. A method for semi-automatic coal colloid index monitoring according to claim 1, characterized in that: The method for realizing semi-automatic coal colloid index monitoring includes: The newly collected semi-automatic colloid layer index monitoring curve of the coal is input into the trained index monitoring network to obtain the newly collected semi-automatic colloid layer index of the coal.

5. A method for semi-automatic coal colloid index monitoring according to claim 1, characterized in that: The common features of the output data of all reference samples in the shared layer are extracted and recorded as common features, including: The output data of each reference sample in the shared layer is obtained, and the mean of all output data of each reference sample in the shared layer is used as the comprehensive output data of each reference sample. The principal component analysis algorithm is used to reduce the dimension of the comprehensive output data of all reference samples, and the obtained reduced dimension data is used as the common feature.

6. A method for semi-automatic coal colloid index monitoring according to claim 1, characterized in that: The method for obtaining the unique features includes: Obtain the output data of the large sample input in the batch in the shared layer, and use the average of the output data of each large sample in the shared layer as the comprehensive output data of each large sample; obtain the output data of the target small sample in the shared layer, and use the average of the output data of the target small sample in the shared layer as the comprehensive output data of the target small sample; calculate the absolute value of the difference between the corresponding positions of the comprehensive output data of the large sample and the comprehensive output data of the target small sample to obtain the difference data; set the position with a value greater than the preset difference threshold in the difference data to 1, and set the position with a value not greater than the preset difference threshold to 0 to obtain sparse data; perform an OR operation on the corresponding positions of the sparse data of all large samples to obtain comprehensive sparse data; perform a multiplication operation on the comprehensive output data of the target small sample and the comprehensive sparse data to obtain unique features.

7. A method for semi-automatic coal colloid index monitoring according to claim 1, characterized in that: The training of the index monitoring network is jointly supervised by using the small sample feature extraction loss function and the loss function of the Conv1D network, including: The first loss value is calculated using the small sample feature extraction loss function, the second loss value is calculated using the loss function of the Conv1D network, and the first loss value and the second loss value are used to perform reverse gradient update on the parameters in the exponential monitoring network.

8. A method for semi-automatic coal colloid index monitoring according to claim 1, characterized in that: The index monitoring curve is a temperature-displacement curve measured using a semi-automatic colloid layer index measuring instrument.

9. A method for semi-automatic coal colloid index monitoring according to claim 1, characterized in that: The samples are divided into large samples and small samples by comparing the number of samples of each type of coal with the preset value.

10. A semi-automatic coal colloid index monitoring system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for monitoring the semi-automatic colloid layer index of coal according to any one of claims 1 to 9 is implemented.

Images