Artificial intelligence-based pon gateway fault prediction method, device and equipment

Abstract

The application provides an artificial intelligence-based PON gateway fault prediction method, device and equipment, and relates to the technical field of artificial intelligence.In the application, first, a target device temperature matrix is constructed based on temperature data formed by a plurality of gateway hardware devices at a plurality of time points;second, a semantic mining unit is used to perform a plurality of semantic mining operations on the target device temperature matrix to form a plurality of device temperature semantic vectors corresponding to the target device temperature matrix;then, a semantic fusion unit is used to fuse the plurality of device temperature semantic vectors to form a fused device temperature semantic vector corresponding to the target device temperature matrix;finally, a fault prediction unit is used to predict and output a gateway fault prediction result corresponding to the target PON gateway based on the fused device temperature semantic vector.Based on the above, the problem of not timely handling of PON gateway hardware faults in the prior art can be improved.

Description

Artificial Intelligence-Based PON Gateway Fault Prediction Method, Device, and Equipment Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, and device for predicting PON gateway faults based on artificial intelligence. Background Technology

[0002] PON (Passive Optical Network) gateways are key devices for implementing fiber optic access solutions such as Fiber to the Home (FTTH) or Fiber to the Building (FTTB). As network access equipment, hardware failures in PON gateways can affect network stability and performance. Currently, maintenance of PON gateway hardware is typically performed only after a hardware failure has occurred. This leads to delays in addressing PON gateway hardware failures. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a PON gateway fault prediction method, apparatus and equipment based on artificial intelligence, so as to improve the problem of untimely handling of PON gateway hardware faults in the prior art.

[0004] To achieve the above objectives, this application adopts the following technical solution:

[0005] An artificial intelligence-based PON gateway fault prediction method includes:

[0006] Based on the temperature data generated by multiple gateway hardware devices in the target PON gateway at multiple time points, a target device temperature matrix is ​​constructed. In the target device temperature matrix, the temperature data in the same row belongs to the temperature data of one gateway hardware device at multiple time points, and the temperature data in the same column belongs to the temperature data of the multiple gateway hardware devices at one time point.

[0007] Using the semantic mining unit in the target gateway fault prediction model, the target device temperature matrix is ​​subjected to semantic mining operations in various ways to form multiple device temperature semantic vectors corresponding to the target device temperature matrix. The target gateway fault prediction model is a trained neural network and also includes a semantic fusion unit and a fault prediction unit. The multiple device temperature semantic vectors are used to represent multiple semantic information in the target device temperature matrix.

[0008] Using the semantic fusion unit, the multiple device temperature semantic vectors are fused to form the fused device temperature semantic vector corresponding to the target device temperature matrix;

[0009] Using the fault prediction unit, based on the temperature semantic vector of the fusion device, a gateway fault prediction result corresponding to the target PON gateway is predicted and output, wherein the gateway fault prediction result is used to characterize the probability of multiple gateway hardware devices in the target PON gateway failing.

[0010] In a preferred embodiment of this application, in the aforementioned AI-based PON gateway fault prediction method, the step of utilizing the semantic mining unit in the target gateway fault prediction model to perform various semantic mining operations on the target device temperature matrix to form multiple device temperature semantic vectors corresponding to the target device temperature matrix includes:

[0011] Using the normalization subunit included in the semantic mining unit of the target gateway fault prediction model, normalization operation is performed on each temperature data in the target device temperature matrix to form the device temperature normalization matrix corresponding to the target device temperature matrix. The semantic mining unit further includes a first semantic mining subunit, a second semantic mining subunit, and a third semantic mining subunit.

[0012] Using the first semantic mining subunit, a global semantic mining operation is performed on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix.

[0013] Using the second semantic mining subunit, a first local semantic mining operation is performed on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix.

[0014] Using the third semantic mining subunit, a second local semantic mining operation is performed on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix. The first local semantic mining operation and the second local semantic mining operation are different.

[0015] In a preferred embodiment of this application, in the aforementioned AI-based PON gateway fault prediction method, the step of using the first semantic mining subunit to perform global semantic mining on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix includes:

[0016] The device temperature normalization matrix is ​​loaded into the first semantic mining subunit;

[0017] Perform a convolution operation on the device temperature normalization matrix to form the device temperature convolution vector corresponding to the device temperature normalization matrix;

[0018] Perform a self-attention operation on the device temperature convolution vector to form a device temperature attention vector corresponding to the device temperature convolution vector;

[0019] The device temperature convolution vector and the device temperature attention vector are superimposed to form the device temperature semantic vector corresponding to the target device temperature matrix.

[0020] In a preferred embodiment of this application, in the aforementioned AI-based PON gateway fault prediction method, the step of using the second semantic mining subunit to perform a first local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix includes:

[0021] The device temperature normalization matrix is ​​loaded into the second semantic mining subunit;

[0022] For each row of temperature data in the device temperature normalization matrix, the difference between the temperature data in that row and the temperature data in the previous row is calculated to obtain the corresponding row temperature difference data, wherein the temperature data in the previous row of the first row of temperature data is 0.

[0023] Based on the row temperature difference data corresponding to each row of temperature data in the device temperature normalization matrix, a corresponding row temperature difference data matrix is ​​formed.

[0024] Perform a convolution operation on the row temperature difference data matrix to form the row temperature difference convolution vector corresponding to the row temperature difference data matrix;

[0025] Perform a self-attention operation on the row temperature difference convolution vector to form the row temperature difference attention vector corresponding to the row temperature difference convolution vector.

[0026] The row temperature difference convolution vector and the row temperature difference attention vector are superimposed to form the device temperature semantic vector corresponding to the target device temperature matrix.

[0027] In a preferred embodiment of this application, in the aforementioned AI-based PON gateway fault prediction method, the step of using the third semantic mining subunit to perform a second local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix includes:

[0028] The device temperature normalization matrix is ​​loaded into the third semantic mining subunit;

[0029] For each column of temperature data in the device temperature normalization matrix, the difference between the temperature data in that column and the temperature data in the previous column is calculated to obtain the corresponding column temperature difference data, wherein the temperature data in the previous column of the first column of temperature data is 0.

[0030] Based on the column temperature difference data corresponding to each column of temperature data in the equipment temperature normalization matrix, a corresponding column temperature difference data matrix is ​​formed.

[0031] Perform a convolution operation on the column temperature difference data matrix to form the column temperature difference convolution vector corresponding to the column temperature difference data matrix;

[0032] Perform a self-attention operation on the column temperature difference convolution vector to form the column temperature difference attention vector corresponding to the column temperature difference convolution vector;

[0033] The column temperature difference convolution vector and the column temperature difference attention vector are superimposed to form the device temperature semantic vector corresponding to the target device temperature matrix.

[0034] In a preferred embodiment of this application, in the aforementioned AI-based PON gateway fault prediction method, the step of fusing the multiple device temperature semantic vectors using the semantic fusion unit to form the fused device temperature semantic vector corresponding to the target device temperature matrix includes:

[0035] Using the first semantic fusion subunit in the semantic fusion unit, the second device temperature semantic vector among the multiple device temperature semantic vectors is fused into the first device temperature semantic vector to form the first fusion vector corresponding to the target device temperature matrix. The semantic fusion unit further includes a second semantic fusion subunit. The first device temperature semantic vector is used to represent the global semantic information in the target device temperature matrix, and the second device temperature semantic vector is used to represent the local semantic information in the target device temperature matrix.

[0036] Using the second semantic fusion subunit, the third device temperature semantic vector among the multiple device temperature semantic vectors is fused into the first device temperature semantic vector to form the second fusion vector corresponding to the target device temperature matrix. The third device temperature semantic vector is used to characterize the local semantic information in the target device temperature matrix, and the local semantic information is different from the local semantic information characterized by the second device temperature semantic vector.

[0037] The first type of device temperature semantic vector, the first fusion vector, and the second fusion vector are superimposed to form the fused device temperature semantic vector corresponding to the target device temperature matrix.

[0038] In a preferred embodiment of this application, in the aforementioned AI-based PON gateway fault prediction method, the step of using the first semantic fusion subunit in the semantic fusion unit to fuse the second device temperature semantic vector from the multiple device temperature semantic vectors into the first device temperature semantic vector to form the first fusion vector corresponding to the target device temperature matrix includes:

[0039] The first and second device temperature semantic vectors from the multiple device temperature semantic vectors are loaded into the first semantic fusion subunit of the semantic fusion unit;

[0040] Based on the second type of device temperature semantic vector, a cross-attention operation is performed on the first type of device temperature semantic vector to form the first fusion vector corresponding to the target device temperature matrix.

[0041] In a preferred embodiment of this application, the aforementioned AI-based PON gateway fault prediction method further includes:

[0042] A sample device temperature matrix is ​​determined, wherein, in the sample device temperature matrix, the temperature data in the same row belongs to the temperature data of a gateway hardware device at multiple time points, and the temperature data in the same column belongs to the temperature data of multiple gateway hardware devices at one time point.

[0043] Using the semantic mining unit in the candidate gateway fault prediction model, the sample device temperature matrix is ​​subjected to semantic mining operations in various ways to form multiple sample device temperature semantic vectors corresponding to the sample device temperature matrix. The candidate gateway fault prediction model belongs to a neural network and also includes a semantic fusion unit and a fault prediction unit. The multiple sample device temperature semantic vectors are used to represent multiple semantic information in the sample device temperature matrix.

[0044] Using the semantic fusion unit, the multiple sample device temperature semantic vectors are fused to form the sample fused device temperature semantic vector corresponding to the sample device temperature matrix;

[0045] Using the fault prediction unit, based on the temperature semantic vector of the sample fusion device, the corresponding sample gateway fault prediction result is predicted and output, wherein the sample gateway fault prediction result is used to characterize the probability of failure of the corresponding multiple gateway hardware devices.

[0046] Based on the error between the sample gateway fault prediction results and the fault probability labels corresponding to the sample device temperature matrix, the candidate gateway fault prediction model is updated to form the target gateway fault prediction model corresponding to the candidate gateway fault prediction model.

[0047] This application also provides an artificial intelligence-based PON gateway fault prediction device, comprising:

[0048] The temperature matrix construction module is used to construct a target device temperature matrix based on the temperature data generated by multiple gateway hardware devices in the target PON gateway at multiple time points. In the target device temperature matrix, the temperature data in the same row belongs to the temperature data of one gateway hardware device at multiple time points, and the temperature data in the same column belongs to the temperature data of the multiple gateway hardware devices at one time point.

[0049] The semantic mining module is used to perform semantic mining operations on the target device temperature matrix in various ways using the semantic mining unit in the target gateway fault prediction model, forming multiple device temperature semantic vectors corresponding to the target device temperature matrix. The target gateway fault prediction model is a trained neural network and also includes a semantic fusion unit and a fault prediction unit. The multiple device temperature semantic vectors are used to represent multiple semantic information in the target device temperature matrix.

[0050] The semantic fusion module is used to fuse the multiple device temperature semantic vectors using the semantic fusion unit to form the fused device temperature semantic vector corresponding to the target device temperature matrix.

[0051] The fault prediction module is used to predict and output the gateway fault prediction result corresponding to the target PON gateway based on the temperature semantic vector of the fusion device using the fault prediction unit. The gateway fault prediction result is used to characterize the probability of multiple gateway hardware devices in the target PON gateway failing.

[0052] Based on the above, this application also provides an electronic device, including:

[0053] Memory, used to store computer programs;

[0054] A processor connected to the memory is used to execute the computer program stored in the memory to implement the above-described AI-based PON gateway fault prediction method.

[0055] The AI-based PON gateway fault prediction method, apparatus, and device provided in this application first construct a target device temperature matrix based on temperature data generated by multiple gateway hardware devices at multiple time points. Second, a semantic mining unit performs various semantic mining operations on the target device temperature matrix to form multiple device temperature semantic vectors corresponding to the target device temperature matrix. Then, a semantic fusion unit fuses these multiple device temperature semantic vectors to form a fused device temperature semantic vector corresponding to the target device temperature matrix. Finally, a fault prediction unit predicts and outputs the gateway fault prediction result corresponding to the target PON gateway based on the fused device temperature semantic vector. Based on the above, since neural networks can be used to fully mine the semantic information in the temperature data, reliable fault prediction can be achieved. Therefore, after obtaining the prediction result, if the prediction result indicates a high probability of a fault, corresponding actions can be taken promptly, thereby improving the problem of untimely handling of PON gateway hardware faults in existing technologies. Attached Figure Description

[0056] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings.

[0057] Figure 1 is a structural block diagram of the electronic device provided in an embodiment of this application.

[0058] Figure 2 is a flowchart illustrating the AI-based PON gateway fault prediction method provided in this application embodiment.

[0059] Figure 3 is a schematic diagram of the target device temperature matrix provided in the embodiment of this application.

[0060] Figure 4 is a block diagram of an AI-based PON gateway fault prediction device provided in an embodiment of this application. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0062] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0063] As shown in Figure 1, this application embodiment provides an electronic device. The electronic device may include a memory, a processor, and an AI-based PON gateway fault prediction device.

[0064] Specifically, the memory and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, the memory and the processor can be electrically connected via one or more communication buses or signal lines. The AI-based PON gateway fault prediction device includes at least one software functional module stored in the memory in the form of software or firmware. The processor is used to execute executable computer programs stored in the memory, such as the software functional modules and computer programs included in the AI-based PON gateway fault prediction device, to implement the AI-based PON gateway fault prediction method provided in this application embodiment.

[0065] Optionally, the memory may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0066] Furthermore, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0067] It is understood that the structure shown in Figure 1 is only schematic. The electronic device may also include more or fewer components than shown in Figure 1, or have a different configuration than shown in Figure 1. For example, it may also include a communication unit for exchanging information with other devices.

[0068] Referring to Figure 2, this application embodiment also provides an AI-based PON gateway fault prediction method applicable to the aforementioned electronic device. The method steps defined in the process of the AI-based PON gateway fault prediction method can be implemented by the electronic device. The specific process shown in Figure 2 will be described in detail below.

[0069] Step S110: Based on the temperature data generated by multiple gateway hardware devices in the target PON gateway at multiple time points, construct the target device temperature matrix.

[0070] In this embodiment, the electronic device can construct a target device temperature matrix based on temperature data generated at multiple time points by multiple gateway hardware devices (such as processors, memory, optical modules, etc.) in the target PON gateway. In the target device temperature matrix, temperature data in the same row belongs to the temperature data of one gateway hardware device at multiple time points (the temperature data at multiple time points can be arranged according to the chronological order of the multiple time points, as shown in Figure 3), and temperature data in the same column belongs to the temperature data of the multiple gateway hardware devices at one time point (the order of the gateway hardware devices can be arbitrary, as long as it is consistent with the order of the sample device temperature matrix used during the training of the target gateway fault prediction model).

[0071] Step S120: Using the semantic mining unit in the target gateway fault prediction model, perform semantic mining operations on the target device temperature matrix in various ways to form multiple device temperature semantic vectors corresponding to the target device temperature matrix.

[0072] In this embodiment, the electronic device can utilize the semantic mining unit in the target gateway fault prediction model to perform various semantic mining operations on the target device temperature matrix, forming multiple device temperature semantic vectors corresponding to the target device temperature matrix. The target gateway fault prediction model is a trained neural network and also includes a semantic fusion unit and a fault prediction unit. The multiple device temperature semantic vectors are used to represent various semantic information (with different focuses) in the target device temperature matrix; that is, semantic information in the target device temperature matrix can be mined and represented by vectors.

[0073] Step S130: Using the semantic fusion unit, the multiple device temperature semantic vectors are fused to form the fused device temperature semantic vector corresponding to the target device temperature matrix.

[0074] In this embodiment, the electronic device can utilize the semantic fusion unit to fuse the various device temperature semantic vectors to form a fused device temperature semantic vector corresponding to the target device temperature matrix. This allows the fused device temperature semantic vector to acquire various semantic information from the target device temperature matrix, improving semantic representation capabilities.

[0075] Step S140: Using the fault prediction unit, based on the temperature semantic vector of the fusion device, predict and output the gateway fault prediction result corresponding to the target PON gateway.

[0076] In this embodiment, the electronic device can utilize the fault prediction unit to predict and output a gateway fault prediction result corresponding to the target PON gateway based on the temperature semantic vector of the fused device. The gateway fault prediction result characterizes the probability of multiple gateway hardware devices in the target PON gateway failing. For example, the fault prediction unit may include a fully connected network and an output function (such as a sigmoid activation function). Thus, after performing a fully connected operation on the temperature semantic vector of the fused device through the fully connected network to obtain a fully connected vector, the output function can be used for processing to obtain the corresponding gateway fault prediction result.

[0077] Based on the above, since neural networks can be used to fully mine the semantic information in temperature data, reliable fault prediction can be achieved (abnormal temperature generally leads to hardware abnormalities, such as overheating which can easily cause device damage). Thus, after obtaining the prediction result, if the prediction result indicates that a fault is highly likely to occur, corresponding measures can be taken in a timely manner, thereby improving the problem of untimely handling of PON gateway hardware faults in the existing technology.

[0078] It should be noted that for step S120, the specific method of performing semantic mining operations on the target device temperature matrix is ​​not limited and can be selected according to actual needs.

[0079] For example, in an alternative implementation, in order to enable the mined multiple device temperature semantic vectors to have rich semantic information, the above step S120 may further include the following steps S121, S122, S123 and S124, the specific contents of which are as follows.

[0080] Step S121: Using the normalization sub-unit included in the semantic mining unit of the target gateway fault prediction model, normalize each temperature data in the target device temperature matrix to form the device temperature normalization matrix corresponding to the target device temperature matrix.

[0081] In this embodiment, the normalization subunit included in the semantic mining unit of the target gateway fault prediction model can be used to perform a normalization operation on each temperature data in the target device temperature matrix (this can be row normalization, i.e., the sum of temperature data in the same row is equal to 1, or global normalization, i.e., the sum of all temperature data is equal to 1), forming the device temperature normalization matrix corresponding to the target device temperature matrix. The semantic mining unit further includes a first semantic mining subunit, a second semantic mining subunit, and a third semantic mining subunit.

[0082] Step S122: Using the first semantic mining subunit, perform global semantic mining operation on the device temperature normalization matrix to form the device temperature semantic vector corresponding to the target device temperature matrix.

[0083] In this embodiment of the application, after obtaining the device temperature normalization matrix, the first semantic mining subunit can be used to perform global semantic mining operation on the device temperature normalization matrix to form the device temperature semantic vector corresponding to the target device temperature matrix. That is, the device temperature semantic vector is used to represent the global temperature semantic information.

[0084] Step S123: Using the second semantic mining subunit, perform a first local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix.

[0085] In this embodiment of the application, after obtaining the device temperature normalization matrix, the second semantic mining subunit can be used to perform a first local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix. That is, the device temperature semantic vector is used to represent local temperature semantic information.

[0086] Step S124: Using the third semantic mining subunit, perform a second local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix.

[0087] In this embodiment, after obtaining the device temperature normalization matrix, the third semantic mining subunit can be used to perform a second local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix. This device temperature semantic vector is used to represent local temperature semantic information. The first local semantic mining operation and the second local semantic mining operation are different. Thus, three types of device temperature semantic vectors can be obtained: one representing global semantic information and two representing local semantic information.

[0088] It is understood that the specific method of performing global semantic mining on the device temperature normalization matrix in step S122 above is not limited. For example, in an alternative implementation, in order to capture more abstract semantic features in the global semantic mining operation, step S122 above may further include the following:

[0089] First, the device temperature normalization matrix can be loaded into the first semantic mining subunit, so that subsequent processing can be performed in the first semantic mining subunit;

[0090] Secondly, a convolution operation can be performed on the device temperature normalization matrix to form a device temperature convolution vector corresponding to the device temperature normalization matrix. For example, the first semantic mining subunit may include a convolutional network, so that the device temperature normalization matrix can be convolved through the convolutional network.

[0091] Then, a self-attention operation can be performed on the device temperature convolution vector to form a device temperature attention vector corresponding to the device temperature convolution vector. For example, the first semantic mining subunit may also include a self-attention network, so that the self-attention operation can be performed on the device temperature convolution vector through the self-attention network.

[0092] Finally, the device temperature convolution vector and the device temperature attention vector can be superimposed to form the device temperature semantic vector corresponding to the target device temperature matrix. For example, in other embodiments, weighted superposition can also be performed, and the corresponding weight coefficients can be configured as needed or formed as parameters of the neural network during training.

[0093] It is understood that the specific method of performing the first local semantic mining operation on the device temperature normalization matrix in step S123 above is not limited. For example, in an alternative implementation, in order to capture more abstract semantic features and temperature difference features in the first local semantic mining operation, step S123 above may further include the following:

[0094] First, the device temperature normalization matrix can be loaded into the second semantic mining subunit, so that subsequent processing can be performed in the second semantic mining subunit;

[0095] Secondly, for each row of temperature data in the device temperature normalization matrix, the difference between the temperature data in that row and the temperature data in the previous row is calculated to obtain the corresponding row temperature difference data (as shown in Figure 3, the difference between temperature data 21 and temperature data 11, the difference between temperature data 22 and temperature data 12, the difference between temperature data 23 and temperature data 13, etc. are calculated respectively. In this way, the difference between the temperatures of different hardware devices can be characterized by the row temperature difference data). Among them, the temperature data in the previous row of the first row of temperature data is 0 (in this way, the row temperature difference data corresponding to the first row of temperature data is the first row of temperature data itself).

[0096] Then, based on the row temperature difference data corresponding to each row of temperature data in the equipment temperature normalization matrix, a corresponding row temperature difference data matrix is ​​formed, wherein the arrangement of the row temperature difference data can be consistent with the arrangement of the row temperature data in the equipment temperature normalization matrix.

[0097] Subsequently, a convolution operation can be performed on the row temperature difference data matrix to form the row temperature difference convolution vector corresponding to the row temperature difference data matrix; for example, the second semantic mining subunit may also include a convolutional network;

[0098] Furthermore, a self-attention operation can be performed on the row temperature difference convolution vector to form a row temperature difference attention vector corresponding to the row temperature difference convolution vector; for example, the second semantic mining subunit may also include a self-attention network.

[0099] Finally, the row temperature difference convolution vector and the row temperature difference attention vector can be superimposed to form the device temperature semantic vector corresponding to the target device temperature matrix.

[0100] It is understood that the specific method of performing the second local semantic mining operation on the device temperature normalization matrix in step S124 above is not limited. For example, in an alternative implementation, in order to capture more abstract semantic features and temperature change features in the second local semantic mining operation, step S124 above may further include the following:

[0101] First, the device temperature normalization matrix can be loaded into the third semantic mining subunit, so that subsequent processing can be performed in the third semantic mining subunit;

[0102] Secondly, for each column of temperature data in the device temperature normalization matrix, the difference between the temperature data in this column and the temperature data in the previous column is calculated to obtain the corresponding column temperature difference data (as shown in Figure 3, the difference between temperature data 12 and temperature data 11, the difference between temperature data 22 and temperature data 21, the difference between temperature data m2 and temperature data m1, etc. are calculated respectively. In this way, the difference between the temperature of the hardware device at different time points can be characterized by the column temperature difference data). Among them, the temperature data in the previous column of the first column of temperature data is 0 (in this way, the column temperature difference data corresponding to the first column of temperature data is the first column of temperature data itself).

[0103] Then, based on the column temperature difference data corresponding to each column of temperature data in the equipment temperature normalization matrix, a corresponding column temperature difference data matrix is ​​formed, wherein the arrangement of the column temperature difference data can be consistent with the arrangement of the column temperature data in the equipment temperature normalization matrix.

[0104] Subsequently, a convolution operation can be performed on the column temperature difference data matrix to form the column temperature difference convolution vector corresponding to the column temperature difference data matrix; for example, the third semantic mining subunit may also include a convolutional network.

[0105] Furthermore, a self-attention operation can be performed on the column temperature difference convolution vector to form a column temperature difference attention vector corresponding to the column temperature difference convolution vector; for example, the third semantic mining subunit may also include a self-attention network.

[0106] Finally, the column temperature difference convolution vector and the column temperature difference attention vector can be superimposed to form the device temperature semantic vector corresponding to the target device temperature matrix.

[0107] It should be noted that for step S130, the specific method of fusing the multiple device temperature semantic vectors is not limited and can be selected according to actual needs.

[0108] For example, in an alternative implementation, in order to fully and reliably fuse the various device temperature semantic vectors to obtain a fused device temperature semantic vector with high semantic representation accuracy, the above step S130 may further include the following steps S131, S132 and S133, the specific contents of each step are as follows.

[0109] Step S131: Using the first semantic fusion subunit in the semantic fusion unit, the second device temperature semantic vector among the multiple device temperature semantic vectors is fused into the first device temperature semantic vector to form the first fusion vector corresponding to the target device temperature matrix.

[0110] In this embodiment, the first semantic fusion subunit in the semantic fusion unit can be used to fuse a second type of device temperature semantic vector from the multiple device temperature semantic vectors into a first type of device temperature semantic vector, forming a first fusion vector corresponding to the target device temperature matrix. The semantic fusion unit further includes a second semantic fusion subunit. The first type of device temperature semantic vector is used to represent the global semantic information in the target device temperature matrix, and the second type of device temperature semantic vector is used to represent the local semantic information in the target device temperature matrix. That is, the mined first type of local semantic information can be fused into the global semantic information, such as fusing the aforementioned local semantic information representing the temperature differences between different hardware devices into the global semantic information to obtain the first fusion vector.

[0111] Step S132: Using the second semantic fusion subunit, the third device temperature semantic vector among the multiple device temperature semantic vectors is fused into the first device temperature semantic vector to form the second fusion vector corresponding to the target device temperature matrix.

[0112] In this embodiment, the second semantic fusion subunit can be used to fuse a third device temperature semantic vector from the multiple device temperature semantic vectors into the first device temperature semantic vector to form a second fusion vector corresponding to the target device temperature matrix. The third device temperature semantic vector is used to represent local semantic information in the target device temperature matrix, and this local semantic information is different from the local semantic information represented by the second device temperature semantic vector. That is, the mined second type of local semantic information can be fused into the global semantic information, such as fusing the aforementioned local semantic information representing the temperature difference of the hardware device at different time points into the global semantic information to obtain the second fusion vector. It is understood that the method of fusing the third device temperature semantic vector into the first device temperature semantic vector is the same as the method of fusing the second device temperature semantic vector into the first device temperature semantic vector; the specific fusion process can be referred to the explanation below.

[0113] Step S133: The first device temperature semantic vector, the first fusion vector, and the second fusion vector are superimposed to form the fused device temperature semantic vector corresponding to the target device temperature matrix.

[0114] In this embodiment, after fusing the first fused vector and the second fused vector respectively, the first device temperature semantic vector, the first fused vector, and the second fused vector can be superimposed to form the fused device temperature semantic vector corresponding to the target device temperature matrix. This not only yields a semantically richer vector but also avoids semantic distortion or loss caused by fusion.

[0115] It is understood that in step S131 above, the specific method of fusing the second device temperature semantic vector from the multiple device temperature semantic vectors into the first device temperature semantic vector is not limited. For example, in an alternative implementation, in order to capture more effective information during the fusion process, step S131 above may further include the following:

[0116] First, the first device temperature semantic vector and the second device temperature semantic vector from the multiple device temperature semantic vectors can be loaded into the first semantic fusion subunit of the semantic fusion unit, so that the fusion operation can be performed in the first semantic fusion subunit;

[0117] Secondly, based on the second type of device temperature semantic vector, a cross-attention operation can be performed on the first type of device temperature semantic vector to form a first fusion vector corresponding to the target device temperature matrix. For example, the first semantic fusion subunit has a cross-attention network. In this way, the corresponding cross-attention operation can be performed through the cross-attention network, so that semantic information related to the second type of device temperature semantic vector can be mined from the first type of device temperature semantic vector, which is used as the important semantic information mined, i.e., the first fusion vector.

[0118] Regarding the aforementioned AI-based PON gateway fault prediction method, it should be further noted that, to ensure the reliable execution of steps S120-S140, the AI-based PON gateway fault prediction method may further include a step of training to form the target gateway fault prediction model. For example, the specific content of this step can be described as follows:

[0119] First, the sample device temperature matrix can be determined. In the sample device temperature matrix, the temperature data in the same row belongs to the temperature data of a gateway hardware device at multiple time points, and the temperature data in the same column belongs to the temperature data of multiple gateway hardware devices at one time point. For details, please refer to the relevant explanation of the target device temperature matrix above.

[0120] Secondly, the semantic mining unit in the candidate gateway fault prediction model can be used to perform semantic mining operations on the sample device temperature matrix in various ways to form multiple sample device temperature semantic vectors corresponding to the sample device temperature matrix. The candidate gateway fault prediction model belongs to a neural network and also includes a semantic fusion unit and a fault prediction unit. The multiple sample device temperature semantic vectors are used to represent multiple semantic information in the sample device temperature matrix. For details, please refer to the relevant explanation of step S120 above.

[0121] Then, the semantic fusion unit can be used to fuse the multiple sample device temperature semantic vectors to form the sample fused device temperature semantic vector corresponding to the sample device temperature matrix; for details, please refer to the relevant explanation of step S130 above.

[0122] Subsequently, the fault prediction unit can be used to predict and output the corresponding sample gateway fault prediction result based on the temperature semantic vector of the sample fusion device. The sample gateway fault prediction result is used to characterize the probability of multiple corresponding gateway hardware devices failing. For details, please refer to the relevant explanation of step S140 above.

[0123] Finally, based on the error (such as cross-entropy error) between the sample gateway fault prediction results and the fault probability labels (1 or 0) corresponding to the sample device temperature matrix, the candidate gateway fault prediction model can be updated to form the target gateway fault prediction model corresponding to the candidate gateway fault prediction model (e.g., updating the parameters of the candidate gateway fault prediction model along the direction of reducing error until the error converges, thereby obtaining the target gateway fault prediction model). The formula for calculating the cross-entropy error is as follows:

[0124] L=-[y*log(p)+(1-y)*log(1-p)];

[0125] Where L is the error, y is the fault probability label, and p is the sample gateway fault prediction result.

[0126] Referring to Figure 4, this application embodiment also provides an AI-based PON gateway fault prediction device applicable to the aforementioned electronic devices. The AI-based PON gateway fault prediction device may include a temperature matrix construction module, a semantic mining module, a semantic fusion module, and a fault prediction module.

[0127] The temperature matrix construction module is used to construct a target device temperature matrix based on temperature data generated by multiple gateway hardware devices in the target PON gateway at multiple time points. In the target device temperature matrix, temperature data in the same row belongs to the temperature data of one gateway hardware device at multiple time points, and temperature data in the same column belongs to the temperature data of the multiple gateway hardware devices at one time point. In this embodiment, the temperature matrix construction module can be used to execute step S110 shown in Figure 2. For details regarding the temperature matrix construction module, please refer to the preceding description of step S110.

[0128] The semantic mining module is used to perform semantic mining operations on the target device temperature matrix in various ways using the semantic mining unit in the target gateway fault prediction model, forming multiple device temperature semantic vectors corresponding to the target device temperature matrix. The target gateway fault prediction model is a trained neural network and also includes a semantic fusion unit and a fault prediction unit. The multiple device temperature semantic vectors are used to represent various semantic information in the target device temperature matrix. In this embodiment, the semantic mining module can be used to execute step S120 shown in Figure 2. For details regarding the semantic mining module, please refer to the preceding description of step S120.

[0129] The semantic fusion module is used to fuse the multiple device temperature semantic vectors using the semantic fusion unit to form a fused device temperature semantic vector corresponding to the target device temperature matrix. In this embodiment, the semantic fusion module can be used to execute step S130 shown in Figure 2. For details regarding the semantic fusion module, please refer to the preceding description of step S130.

[0130] The fault prediction module is used to predict and output a gateway fault prediction result corresponding to the target PON gateway based on the temperature semantic vector of the fusion device, using the fault prediction unit. The gateway fault prediction result is used to characterize the probability of multiple gateway hardware devices in the target PON gateway failing. In this embodiment, the fault prediction module can be used to execute step S140 shown in Figure 2. For details regarding the fault prediction module, please refer to the preceding description of step S140.

[0131] In this embodiment of the application, corresponding to the above-described AI-based PON gateway fault prediction method applied to the electronic device, a computer-readable storage medium is also provided, which stores a computer program that executes the various steps of the AI-based PON gateway fault prediction method when the computer program is run.

[0132] The steps executed by the aforementioned computer program during runtime will not be described in detail here, but can be found in the explanation of the AI-based PON gateway fault prediction method described above.

[0133] In summary, the AI-based PON gateway fault prediction method, apparatus, and device provided in this application first construct a target device temperature matrix based on temperature data generated by multiple gateway hardware devices at multiple time points. Second, a semantic mining unit performs various semantic mining operations on the target device temperature matrix to form multiple device temperature semantic vectors corresponding to the target device temperature matrix. Then, a semantic fusion unit fuses these multiple device temperature semantic vectors to form a fused device temperature semantic vector corresponding to the target device temperature matrix. Finally, a fault prediction unit predicts and outputs the gateway fault prediction result corresponding to the target PON gateway based on the fused device temperature semantic vector. Based on the above, since neural networks can be used to fully mine the semantic information in the temperature data, reliable fault prediction can be achieved. Therefore, after obtaining the prediction result, if the prediction result indicates a high probability of a fault, appropriate processing can be carried out in a timely manner, thereby improving the problem of untimely handling of PON gateway hardware faults in existing technologies.

[0134] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0135] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0136] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0137] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A PON gateway fault prediction method based on artificial intelligence, characterized in that, include: Based on temperature data generated by multiple gateway hardware devices in the target PON gateway at multiple time points, a target device temperature matrix is ​​constructed. In this matrix, temperature data in the same row belongs to the temperature data of one gateway hardware device at multiple time points, and temperature data in the same column belongs to the temperature data of multiple gateway hardware devices at a single time point. Using the semantic mining unit in the target gateway fault prediction model, various semantic mining operations are performed on the target device temperature matrix to form multiple device temperature semantic vectors corresponding to the target device temperature matrix. The target gateway fault prediction model is a trained neural network and includes a semantic fusion unit and a fault prediction unit. These multiple device temperature semantic vectors are used to represent various semantic information in the target device temperature matrix. The semantic fusion unit fuses these multiple device temperature semantic vectors to form a fused device temperature semantic vector corresponding to the target device temperature matrix. The fault prediction unit, based on the fused device temperature semantic vector, predicts and outputs a gateway fault prediction result for the target PON gateway. This gateway fault prediction result represents the probability of multiple gateway hardware devices in the target PON gateway experiencing a fault.

2. The PON gateway fault prediction method based on artificial intelligence according to claim 1, characterized in that, The step of using the semantic mining unit in the target gateway fault prediction model to perform semantic mining operations on the target device temperature matrix in various ways to form multiple device temperature semantic vectors corresponding to the target device temperature matrix includes: using the normalization subunit included in the semantic mining unit of the target gateway fault prediction model to normalize each temperature data in the target device temperature matrix to form a device temperature normalization matrix corresponding to the target device temperature matrix, wherein the semantic mining unit further includes a first semantic mining subunit, a second semantic mining subunit, and a third semantic mining subunit; using the first semantic mining subunit to perform a global semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix; using the second semantic mining subunit to perform a first local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix; using the third semantic mining subunit to perform a second local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix, wherein the first local semantic mining operation and the second local semantic mining operation are different.

3. The PON gateway fault prediction method based on artificial intelligence according to claim 2, characterized in that, The step of using the first semantic mining subunit to perform global semantic mining on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix includes: loading the device temperature normalization matrix into the first semantic mining subunit; performing a convolution operation on the device temperature normalization matrix to form a device temperature convolution vector corresponding to the device temperature normalization matrix; performing a self-attention operation on the device temperature convolution vector to form a device temperature attention vector corresponding to the device temperature convolution vector; and performing a superposition operation on the device temperature convolution vector and the device temperature attention vector to form a device temperature semantic vector corresponding to the target device temperature matrix.

4. The PON gateway fault prediction method based on artificial intelligence according to claim 2, characterized in that, The step of using the second semantic mining subunit to perform a first local semantic mining operation on the device temperature normalization matrix to form a device temperature semantic vector corresponding to the target device temperature matrix includes: loading the device temperature normalization matrix into the second semantic mining subunit; for each row of temperature data in the device temperature normalization matrix, calculating the difference between the temperature data in that row and the temperature data in the previous row to obtain the corresponding row temperature difference data, wherein the temperature data in the previous row of the first row of temperature data is 0; combining the row temperature difference data corresponding to each row of temperature data in the device temperature normalization matrix to form a corresponding row temperature difference data matrix; performing a convolution operation on the row temperature difference data matrix to form a row temperature difference convolution vector corresponding to the row temperature difference data matrix; performing a self-attention operation on the row temperature difference convolution vector to form a row temperature difference attention vector corresponding to the row temperature difference convolution vector; and performing a superposition operation on the row temperature difference convolution vector and the row temperature difference attention vector to form the device temperature semantic vector corresponding to the target device temperature matrix.

5. The PON gateway fault prediction method based on artificial intelligence according to claim 2, characterized in that, The step of using the third semantic mining subunit to perform a second local semantic mining operation on the equipment temperature normalization matrix to form the equipment temperature semantic vector corresponding to the target equipment temperature matrix includes: loading the equipment temperature normalization matrix into the third semantic mining subunit; calculating the difference between the temperature data in each column of the equipment temperature normalization matrix and the temperature data in the preceding column of the temperature data in that column to obtain the corresponding column temperature difference data, wherein the temperature data in the preceding column of the first column of the temperature data is 0; combining the column temperature difference data corresponding to each column of the temperature data in the equipment temperature normalization matrix to form a corresponding column temperature difference data matrix; performing a convolution operation on the column temperature difference data matrix to form a column temperature difference convolution vector corresponding to the column temperature difference data matrix; performing a self-attention operation on the column temperature difference convolution vector to form a column temperature difference attention vector corresponding to the column temperature difference convolution vector; and performing a superposition operation on the column temperature difference convolution vector and the column temperature difference attention vector to form the equipment temperature semantic vector corresponding to the target equipment temperature matrix.

6. The PON gateway fault prediction method based on artificial intelligence according to claim 1, characterized in that, The step of fusing the multiple device temperature semantic vectors using the semantic fusion unit to form a fused device temperature semantic vector corresponding to the target device temperature matrix includes: using a first semantic fusion subunit in the semantic fusion unit to fuse a second device temperature semantic vector from the multiple device temperature semantic vectors into a first device temperature semantic vector to form a first fused vector corresponding to the target device temperature matrix, wherein the semantic fusion unit further includes a second semantic fusion subunit, the first device temperature semantic vector being used to represent global semantic information in the target device temperature matrix, and the second device temperature semantic vector being used to represent local semantic information in the target device temperature matrix; using the second semantic fusion subunit to fuse a third device temperature semantic vector from the multiple device temperature semantic vectors into the first device temperature semantic vector to form a second fused vector corresponding to the target device temperature matrix, wherein the third device temperature semantic vector being used to represent local semantic information in the target device temperature matrix, and this local semantic information is different from the local semantic information represented by the second device temperature semantic vector; and performing a superposition operation on the first device temperature semantic vector, the first fused vector, and the second fused vector to form a fused device temperature semantic vector corresponding to the target device temperature matrix.

7. The PON gateway fault prediction method based on artificial intelligence according to claim 6, characterized in that, The step of fusing a second type of device temperature semantic vector from multiple device temperature semantic vectors into a first type of device temperature semantic vector using a first semantic fusion subunit in the semantic fusion unit to form a first fusion vector corresponding to the target device temperature matrix includes: loading the first type of device temperature semantic vector and the second type of device temperature semantic vector from the multiple device temperature semantic vectors into the first semantic fusion subunit in the semantic fusion unit; and performing a cross-attention operation on the first type of device temperature semantic vector based on the second type of device temperature semantic vector to form a first fusion vector corresponding to the target device temperature matrix.

8. The PON gateway fault prediction method based on artificial intelligence according to any one of claims 1-7, characterized in that, The AI-based PON gateway fault prediction method further includes: determining a sample device temperature matrix, wherein temperature data in the same row belongs to the temperature data of one gateway hardware device at multiple time points, and temperature data in the same column belongs to the temperature data of multiple gateway hardware devices at one time point; using the semantic mining unit in the candidate gateway fault prediction model, performing semantic mining operations on the sample device temperature matrix in various ways to form multiple sample device temperature semantic vectors corresponding to the sample device temperature matrix, wherein the candidate gateway fault prediction model belongs to a neural network and also includes a semantic fusion unit and a fault prediction unit, and the multiple sample device temperature semantic vectors are used to analyze the sample device temperature matrix. The device temperature matrix is ​​characterized by various semantic information. The semantic fusion unit fuses these various sample device temperature semantic vectors to form a sample fused device temperature semantic vector corresponding to the sample device temperature matrix. The fault prediction unit, based on the sample fused device temperature semantic vector, predicts and outputs a corresponding sample gateway fault prediction result, where the sample gateway fault prediction result represents the probability of failure for multiple corresponding gateway hardware devices. Based on the error between the sample gateway fault prediction result and the fault probability label corresponding to the sample device temperature matrix, the candidate gateway fault prediction model is updated to form a target gateway fault prediction model corresponding to the candidate gateway fault prediction model.

9. A PON gateway fault prediction device based on artificial intelligence, characterized in that, include: A temperature matrix construction module is used to construct a target device temperature matrix based on temperature data generated by multiple gateway hardware devices in the target PON gateway at multiple time points. In this matrix, temperature data in the same row belongs to the temperature data of one gateway hardware device at multiple time points, and temperature data in the same column belongs to the temperature data of the multiple gateway hardware devices at a single time point. A semantic mining module is used to perform various semantic mining operations on the target device temperature matrix using the semantic mining unit in the target gateway fault prediction model, forming multiple device temperature semantic vectors corresponding to the target device temperature matrix. The target gateway fault prediction model belongs to... The trained neural network also includes a semantic fusion unit and a fault prediction unit. The multiple device temperature semantic vectors are used to represent multiple semantic information in the target device temperature matrix. The semantic fusion module is used to fuse the multiple device temperature semantic vectors using the semantic fusion unit to form a fused device temperature semantic vector corresponding to the target device temperature matrix. The fault prediction module is used to use the fault prediction unit to predict and output a gateway fault prediction result corresponding to the target PON gateway based on the fused device temperature semantic vector. The gateway fault prediction result is used to represent the probability of multiple gateway hardware devices in the target PON gateway failing.

10. An electronic device, characterized in that, include: A memory for storing a computer program; a processor connected to the memory for executing the computer program stored in the memory to implement the method according to any one of claims 1-8.

Images