Environment-friendly public toilet equipment state monitoring method, device and equipment

Abstract

The invention provides an environment-friendly public toilet equipment state monitoring method, device and equipment, and relates to the technical field of equipment state monitoring. The method comprises the following steps: firstly, acquiring multi-dimensional state time sequence data of target environment-friendly public toilet equipment; secondly, performing cross-dimensional state time sequence semantic mining on the multi-dimensional state time sequence data to form multi-dimensional state time sequence features corresponding to the multi-dimensional state time sequence data; and then, network state prediction is carried out based on the multi-dimensional state time sequence characteristics, and a target network state prediction result is obtained. On the basis of the content, the problem that in the prior art, the reliability of equipment state monitoring is relatively low can be solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of device state monitoring, in particular to an environmental public toilet device state monitoring method, device and equipment. BACKGROUND

[0002] Intelligent environmental public toilets rely on various online devices to achieve intelligent functions. Device offline may cause data loss, service interruption and other problems. Traditional device management methods cannot monitor the online state of devices in real time and cannot timely discover and handle offline problems. With the development of Internet of Things technology, it is possible to build a remote offline state monitoring and alarm system, which will greatly improve the operation and maintenance efficiency and reliability of public toilets. The existing intelligent public toilet device management system mostly relies on simple timing heartbeat detection to monitor the offline state of devices, and lacks analysis and diagnosis of offline causes. However, this method can only be applied after network failure occurs, and cannot be maintained in a timely and effective manner. Therefore, in some conventional solutions, when the network level data (such as packet loss rate, etc.) is frequently in a poor state, it is determined that the environmental public toilet device may have network failure. This simple rule cannot reliably determine the actual state of the environmental public toilet device, and therefore has the problem of relatively low reliability. SUMMARY

[0003] Therefore, the purpose of the present application is to provide an environmental public toilet device state monitoring method, device and equipment to improve the problem of relatively low reliability of device state monitoring in the prior art.

[0004] To achieve the above-mentioned purpose, the present application adopts the following technical solutions: An environmental public toilet device state monitoring method comprises: Obtaining multi-dimensional state time series data of a target environmental public toilet device, wherein the multi-dimensional state time series data comprises network level time series data, device running time series data and device environment time series data, and the target environmental public toilet device is used to interact with a background device through a network; Performing cross-dimension state time series semantic mining on the multi-dimensional state time series data to form multi-dimensional state time series features corresponding to the multi-dimensional state time series data, wherein the multi-dimensional state time series features carry three-dimensional state time series semantic features corresponding to the network level time series data, the device running time series data and the device environment time series data; Performing network state prediction based on the multi-dimensional state time series features to obtain a target network state prediction result, wherein the target network state prediction result is used to reflect the probability of network failure of the target environmental public toilet device.

[0005] In the preferred selection of the present application, in the above-mentioned environmental public toilet equipment state monitoring method, the step of performing cross-dimension state time sequence semantic mining on the multi-dimensional state time sequence data to form multi-dimensional state time sequence characteristics corresponding to the multi-dimensional state time sequence data comprises: respectively performing semantic space mapping on the network layer time sequence data, the equipment running time sequence data and the equipment environment time sequence data to form network layer semantic features, equipment running semantic features and equipment environment semantic features; fusing the network layer semantic features and the equipment running semantic features to form network equipment fusion features, and performing significant feature screening on the network equipment fusion features based on the hidden features represented by the equipment environment semantic features to form multi-dimensional state time sequence features.

[0006] In the preferred selection of the present application, in the above-mentioned environmental public toilet equipment state monitoring method, the step of fusing the network layer semantic features and the equipment running semantic features to form network equipment fusion features, and performing significant feature screening on the network equipment fusion features based on the hidden features represented by the equipment environment semantic features to form multi-dimensional state time sequence features comprises: splicing the network layer semantic features and the equipment running semantic features to form network equipment fusion features; performing nonlinear mapping on the equipment environment semantic features to form equipment environment hidden features; performing significant feature screening on the network equipment fusion features based on the equipment environment hidden features to form multi-dimensional state time sequence features.

[0007] In the preferred selection of the present application, in the above-mentioned environmental public toilet equipment state monitoring method, the step of performing nonlinear mapping on the equipment environment semantic features to form equipment environment hidden features comprises: performing full connection processing on the equipment environment semantic features to form equipment environment full connection features; performing processing on the equipment environment full connection features based on a first nonlinear function to form equipment environment nonlinear features; performing full connection processing on the equipment environment nonlinear features to form full connection nonlinear features; performing processing on the full connection nonlinear features based on a second nonlinear function to form nonlinear mapping features; determining the equipment environment hidden features based on the nonlinear mapping features.

[0008] In the preferred selection of the present application, in the above-mentioned environmental public toilet equipment state monitoring method, the step of performing significant feature screening on the network equipment fusion features based on the equipment environment hidden features to form multi-dimensional state time sequence features comprises: bitwise multiplication operation is performed on the device environment hidden features and the network device fusion features to form multi-dimensional state time sequence features.

[0009] In a preferred selection of the present application, in the above-mentioned environmental public toilet device state monitoring method, the step of respectively performing semantic space mapping on the network level time sequence data, the device running time sequence data and the device environment time sequence data to form network level semantic features, device running semantic features and device environment semantic features comprises: performing word embedding processing on the network level time sequence data to form a word embedding feature sequence; for each word embedding feature in the word embedding feature sequence, multiplying the current word embedding feature with a first shared weight matrix to obtain a first intermediate feature, multiplying an output feature corresponding to a previous word embedding feature of the word embedding feature with a second shared weight matrix to obtain a second intermediate feature, and adding the first intermediate feature, the second intermediate feature and a shared bias parameter to form an output feature of the current word embedding feature, wherein the output feature corresponding to the previous word embedding feature of the first word embedding feature is a zero feature; determining the network level semantic features based on an output feature corresponding to the last word embedding feature; respectively performing semantic space mapping on the device running time sequence data and the device environment time sequence data to form device running semantic features and device environment semantic features.

[0010] In a preferred selection of the present application, in the above-mentioned environmental public toilet device state monitoring method, the step of determining the network level semantic features based on an output feature corresponding to the last word embedding feature comprises: performing self-attention processing on the output feature corresponding to the last word embedding feature to form the network level semantic features.

[0011] In a preferred selection of the present application, in the above-mentioned environmental public toilet device state monitoring method, the environmental public toilet device state monitoring method further comprises: obtaining training multi-dimensional state time sequence data, wherein the multi-dimensional state time sequence data comprises training network level time sequence data, training device running time sequence data and training device environment time sequence data; using a semantic mining unit included in the network state prediction model to perform cross-dimension state time sequence semantic mining on the training multi-dimensional state time sequence data to form training multi-dimensional state time sequence features corresponding to the training multi-dimensional state time sequence data, wherein the training multi-dimensional state time sequence features carry three-dimension state time sequence semantic features corresponding to the training network level time sequence data, the training device running time sequence data and the training device environment time sequence data; The network state prediction model comprises a state prediction unit, and network state prediction is performed based on the training multi-dimensional state time sequence feature to obtain a training network state prediction result. Based on the error between the training network state prediction result and the network state label corresponding to the training multi-dimensional state time sequence data, the model parameters of the network state prediction model are updated to form an optimized network state prediction model.

[0012] The application also provides an environmental public toilet device state monitoring device, comprising: A state data acquisition module is configured to acquire multi-dimensional state time sequence data of a target environmental public toilet device, wherein the multi-dimensional state time sequence data comprises network layer time sequence data, device operation time sequence data and device environment time sequence data, and the target environmental public toilet device is configured to interact with a background device for data through a network. A semantic mining module is configured to perform cross-dimension state time sequence semantic mining on the multi-dimensional state time sequence data to form multi-dimensional state time sequence features corresponding to the multi-dimensional state time sequence data, wherein the multi-dimensional state time sequence features carry three-dimension state time sequence semantic features corresponding to the network layer time sequence data, the device operation time sequence data and the device environment time sequence data. A state prediction module is configured to perform network state prediction based on the multi-dimensional state time sequence features to obtain a target network state prediction result, wherein the target network state prediction result is used to reflect the probability of network failure of the target environmental public toilet device.

[0013] On the basis of the above, the application also provides an electronic device, comprising: A memory is configured to store a computer program. A processor connected with the memory is configured to execute the computer program stored in the memory to implement the environmental public toilet device state monitoring method.

[0014] The environmental public toilet equipment state monitoring method, device and equipment provided by the application first acquire multi-dimensional state time sequence data of a target environmental public toilet equipment; secondly, cross-dimension state time sequence semantic mining is performed on the multi-dimensional state time sequence data to form multi-dimensional state time sequence features corresponding to the multi-dimensional state time sequence data; then, network state prediction is performed based on the multi-dimensional state time sequence features to obtain a target network state prediction result. Based on the above, cross-dimension state time sequence semantic mining is performed on network layer time sequence data, equipment operation time sequence data and equipment environment time sequence data in three dimensions, so that semantic features in the equipment operation and equipment environment dimensions can assist in representing semantic information at the network layer. In a toilet, the humidity is relatively high, which affects the service life of hardware equipment, and the moisture in the air increases the propagation loss of wireless signals. Therefore, with the assistance of semantic features in the equipment environment dimension, the semantic representation accuracy of the formed multi-dimensional state time sequence features can be improved, the reliability of the determined target network state prediction result can be improved, and the problem of relatively low reliability of device state monitoring in the prior art can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the following preferred embodiments are specifically described below, and the accompanying drawings are described in detail as follows.

[0016] Figure 1 The structural block diagram of the electronic device provided by the embodiment of the application is shown.

[0017] Figure 2 The flowchart of the environmental public toilet equipment state monitoring method provided by the embodiment of the application is shown.

[0018] Figure 3 The schematic diagram of semantic space mapping provided by the embodiment of the application is shown.

[0019] Figure 4 The block schematic diagram of the environmental public toilet equipment state monitoring device provided by the embodiment of the application is shown. DETAILED DESCRIPTION

[0020] In order to make the purposes, technical solutions and advantages of the embodiments of the application more clear, the technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, not all the embodiments. The components of the embodiments of the application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0021] 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.

[0022] like Figure 1 As shown in the figure, this application provides an electronic device. The electronic device may include a memory, a processor, and an environmentally friendly public toilet equipment status monitoring device.

[0023] In detail, 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 environmentally friendly public toilet equipment status monitoring 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 environmentally friendly public toilet equipment status monitoring device, to implement the environmentally friendly public toilet equipment status monitoring method provided in this application embodiment.

[0024] 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.

[0025] 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.

[0026] Understandable. Figure 1The illustrated structure is only schematic, and the electronic device can further include more or less components than those shown, or have a different configuration of components than those shown, such as, for example, a communication unit for information exchange with other devices. Figure 1 The illustrated structure is only schematic, and the electronic device can further include more or less components than those shown, or have a different configuration of components than those shown, such as, for example, a communication unit for information exchange with other devices. Figure 1 The illustrated structure is only schematic, and the electronic device can further include more or less components than those shown, or have a different configuration of components than those shown, such as, for example, a communication unit for information exchange with other devices.

[0027] In combination with Figure 2 , the embodiments of the present application further provide an environmental-friendly public toilet device state monitoring method applicable to the above-mentioned electronic device. The method steps defined by the process related to the environmental-friendly public toilet device state monitoring method can be implemented by the electronic device.

[0028] The specific process shown in Figure 2 will be described in detail below.

[0029] In step S110, multi-dimensional state time series data of a target environmental-friendly public toilet device is acquired.

[0030] In the embodiments of the present application, the electronic device can acquire multi-dimensional state time series data of a target environmental-friendly public toilet device. The multi-dimensional state time series data includes network level time series data, device running time series data and device environment time series data, and the target environmental-friendly public toilet device is used to interact with a background device through a network. Exemplarily, the network level time series data can be at least one of packet loss rate, signal-to-noise ratio (SNR) and network delay at each of a plurality of time instants. The device running time series data can be the usage of at least one of processor, memory and storage at each of a plurality of time instants. The device environment time series data can be at least one of temperature and humidity at each of a plurality of time instants.

[0031] In step S120, cross-dimension state time series semantic mining is performed on the multi-dimensional state time series data to form multi-dimensional state time series features corresponding to the multi-dimensional state time series data.

[0032] In the embodiments of the present application, after the multi-dimensional state time series data is acquired, the electronic device can perform cross-dimension state time series semantic mining on the multi-dimensional state time series data to form multi-dimensional state time series features corresponding to the multi-dimensional state time series data. The multi-dimensional state time series features carry three-dimension state time series semantic features corresponding to the network level time series data, the device running time series data and the device environment time series data. That is, not only the semantic features of the data itself, but also the semantic relationship between the data at different time instants, are mined in each dimension, and the semantic features of the three dimensions are fused to form multi-dimensional state time series features, so that the semantic richness and representation accuracy of the semantic features can be taken into account.

[0033] In step S130, network state prediction is performed based on the multi-dimensional state time sequence feature, and a target network state prediction result is obtained.

[0034] In the embodiments of the present application, after the multi-dimensional state time sequence feature is formed, the electronic device can perform network state prediction based on the multi-dimensional state time sequence feature to obtain a target network state prediction result. The target network state prediction result is used to reflect the probability of network failure of the target environmental public toilet device. For example, the multi-dimensional state time sequence feature can be fully connected to obtain a fully connected feature with a size of 1*1. Then, the fully connected feature can be mapped by an identity mapping (e.g., f(x)=x) or a linear mapping (i.e., mapped to 0-1), so as to obtain the probability of network failure, i.e., the target network state prediction result.

[0035] Based on the above, since the cross-dimensional state time sequence semantic mining is performed on the network layer time sequence data, the device running time sequence data, and the device environment time sequence data, the semantic features of the device running and the device environment can assist in representing the semantic information of the network layer. In a toilet, the humidity is relatively high, which affects the service life of the hardware device, and the moisture in the air increases the propagation loss of the wireless signal. Therefore, with the assistance of the semantic features of the device environment, the semantic representation accuracy of the multi-dimensional state time sequence feature can be improved, the reliability of the target network state prediction result can be improved, and the problem of relatively low reliability of device state monitoring in the prior art can be solved.

[0036] It should be noted that the specific manner of performing cross-dimensional state time sequence semantic mining on the multi-dimensional state time sequence data is not limited, and can be selected according to requirements.

[0037] For example, in an alternative embodiment, the step S120 can include: First, the network layer time sequence data, the device running time sequence data, and the device environment time sequence data can be respectively mapped in a semantic space (in the process of mapping in a semantic space, the relationship between semantic features at different times needs to be captured), to form network layer semantic features, device running semantic features, and device environment semantic features. Second, the network layer semantic features, the device running semantic features, and the device environment semantic features can be spliced and processed by self-attention, so as to obtain the multi-dimensional state time sequence feature corresponding to the multi-dimensional state time sequence data.

[0038] For example, in another alternative implementation, in order to achieve reliable fusion of multidimensional semantic features and thereby further improve the semantic representation accuracy of the formed multidimensional state temporal features, the above step S120 may further include steps S121 and S122, the specific contents of which are as follows.

[0039] Step S121: Semantic space mapping is performed on the network layer time-series data, the device runtime time-series data, and the device environment time-series data respectively to form network layer semantic features, device runtime semantic features, and device environment semantic features.

[0040] In this embodiment, the network-level time-series data, the device runtime time-series data, and the device environment time-series data can be semantically mapped to form network-level semantic features, device runtime semantic features, and device environment semantic features, respectively. That is, the network-level time-series data can be mapped to a semantic space for representation, thereby forming corresponding network-level semantic features. Similarly, the device runtime time-series data can be mapped to a semantic space for representation, thereby forming corresponding device runtime semantic features. And the device environment time-series data can be mapped to a semantic space for representation, thereby forming corresponding device environment semantic features. It should be noted that the representation of each semantic feature can be a vector (or matrix), such as: [0.54, 0.76, 0.19, 0.96, 0.40, 0.13, 0.87, 0.60, 0.54, 0.25, 0.86, 0.35, 0.06, 0.11, 0.80, 0.26, 0.94, 0.67, 0.48, ..., 0.52].

[0041] Step S122: The network-level semantic features and the device operation semantic features are fused to form network device fusion features. The network device fusion features are then subjected to salient feature screening based on the latent features represented by the device environment semantic features to form multidimensional state temporal features.

[0042] In this embodiment, the network-level semantic features and the device operation semantic features can be fused to form network device fused features. Furthermore, based on the latent features represented by the device environment semantic features, salient feature filtering is performed on the network device fused features to form multidimensional state temporal features. It should be noted that since the semantic features of the device operation dimension and the semantic features of the network layer are more closely related, they can be fused first, and then fused again using the semantic features of the device environment dimension, which have a certain potential relationship.

[0043] It can be understood that the specific manner of performing semantic space mapping on the network layer timing data, the device runtime timing data and the device environment timing data respectively in step S121 is not limited, for example, in an alternative embodiment, in order to fully capture the semantic information about timing between data on the basis of mining the semantic information of the data itself, step S121 can further include steps S121a, S121b, S121c and S121d, and the specific contents of each step are as follows.

[0044] Step S121a, performing word embedding processing on the network layer timing data to form a word embedding feature sequence.

[0045] In the embodiment of the present application, the network layer timing data can be subjected to word embedding processing (such as through a word embedding model such as Word2Vec), to form a word embedding feature sequence. It should be noted that when each of the plurality of time points included in the network layer timing data corresponds to a plurality of data such as packet loss rate, signal-to-noise ratio and network delay, each type of data can be subjected to word embedding processing to obtain corresponding local word embedding features, and then the local word embedding features corresponding to each data can be subjected to mean value calculation to obtain the corresponding word embedding feature sequence. That is, in the word embedding feature sequence, each word embedding feature corresponds to a time point, and the word embedding features can be sorted according to the chronological relationship between the corresponding time points, such as from early to late.

[0046] Step S121b, for each word embedding feature in the word embedding feature sequence, multiplying the current word embedding feature by a first shared weight matrix to obtain a first intermediate feature, multiplying the output feature corresponding to the previous word embedding feature of the word embedding feature by a second shared weight matrix to obtain a second intermediate feature, and adding the first intermediate feature, the second intermediate feature and a shared bias parameter to form the output feature of the current word embedding feature.

[0047] In the embodiments of the present application, after obtaining the word embedding feature sequence, for each word embedding feature in the word embedding feature sequence, the current word embedding feature is multiplied by the first shared weight matrix to obtain a first intermediate feature, and the output feature corresponding to the previous word embedding feature of the word embedding feature is multiplied by the second shared weight matrix to obtain a second intermediate feature, and the first intermediate feature, the second intermediate feature and the shared bias parameter are added to form the output feature of the current word embedding feature. Wherein, the output feature corresponding to the previous word embedding feature of the first word embedding feature is a zero feature, such as [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …, 0]. That is, the plurality of word embedding features can be fused in turn through a plurality of time steps, for example, in combination with Figure 3 At the first time step, the first word embedding feature and the zero feature can be fused, at the second time step, the second word embedding feature and the output feature of the first time step can be fused, and at the third time step, the third word embedding feature and the output feature of the second time step can be fused. Wherein, the fusion of the features can be realized by the first shared weight matrix, the second shared weight matrix and the shared bias parameter, and the sharing means that the weight matrix and the bias parameter of each time step are the same. In addition, it should be noted that in order to introduce nonlinearity, so that the nonlinear features in the semantic features can be captured, after the addition of the shared bias parameter, the result of the addition can also be subjected to nonlinear activation processing to form the output feature of the current word embedding feature. The specific fusion formula can be: H t =tanh(W h *H t-1 +W x *x t +b); Wherein, H t is the output feature of the current word embedding feature, tanh is a function of nonlinear activation processing, W h is the second shared weight matrix, H t-1 is the output feature corresponding to the previous word embedding feature, W x is the first shared weight matrix, x t is the current word embedding feature, and b is the shared bias parameter.

[0048] Step S121c, determining the network-level semantic feature based on the output feature corresponding to the last word embedding feature.

[0049] In the embodiments of the present application, after forming the output feature corresponding to the last word embedding feature, the network-level semantic feature can be determined based on the output feature corresponding to the last word embedding feature. For example, in an alternative implementation, the output feature corresponding to the last word embedding feature can be taken as the network-level semantic feature. For another example, in another alternative implementation, the output feature corresponding to the last word embedding feature can be subjected to self-attention processing (i.e., the output feature is subjected to linear transformation to generate three vectors: Q (query vector), K (key vector) and V (value vector); the dot product of Q and K is calculated, and then scaling and Softmax operation are performed to obtain an attention weight matrix A; the attention weight matrix A and the value vector V are weighted and summed to obtain a final output vector O, i.e., the network-level semantic feature), to form the network-level semantic feature.

[0050] In step S121d, the device runtime time series data and the device environment time series data are respectively subjected to semantic space mapping to form device runtime semantic features and device environment semantic features.

[0051] In the embodiments of the present application, the device runtime time series data and the device environment time series data can be respectively subjected to semantic space mapping to form device runtime semantic features and device environment semantic features. The specific processing manner can refer to the process of processing the network-level time series data.

[0052] It can be understood that the specific manner of forming the multi-dimensional state time series feature in the above step S122 is not limited. For example, in an alternative implementation, in order to enable important semantic features to be mined by fusing the semantic features of three dimensions, the above step S122 can further include steps S122a, S122b and S122c, and the specific contents of each step are as follows.

[0053] In step S122a, the network-level semantic feature and the device runtime semantic feature are spliced to form a network-device fusion feature.

[0054] In the embodiments of the present application, the network-level semantic feature and the device runtime semantic feature can be spliced to form a network-device fusion feature. The size of the network-level semantic feature and the size of the device runtime semantic feature can both be 1*n, and the size of the network-device fusion feature can be 1*m, where m=2n.

[0055] In step S122b, the device environment semantic feature is subjected to nonlinear mapping to form a device environment hidden feature.

[0056] In the embodiments of the present application, the device environment semantic features can be nonlinearly mapped to form device environment hidden features, i.e., to capture nonlinear information in the device environment semantic features.

[0057] In step S122c, based on the device environment hidden features, significant feature screening is performed on the network device fusion features to form multi-dimensional state time sequence features.

[0058] In the embodiments of the present application, after the device environment hidden features and the network device fusion features are formed, based on the device environment hidden features, significant feature screening can be performed on the network device fusion features to form multi-dimensional state time sequence features, i.e., to screen important features. For example, to perform significant feature screening on the network device fusion features, the device environment features and the network device fusion features can be multiplied by each other to form multi-dimensional state time sequence features, so that important features in the network device fusion features can be focused on. In addition, it should be further explained that since the network device fusion features are actually formed by splicing the network level semantic features and the device running semantic features, after significant feature screening based on the device environment hidden features, i.e., after multiplication by each other, important features respectively possessed by the network level semantic features and the device running semantic features are actually captured, so that sufficient fusion between the two is not achieved. Therefore, in an alternative implementation, the result of multiplication by each other can be subjected to self-attention processing, so that important features of the two dimensions are cross-fused to capture associated semantic information, so that the representation accuracy of the obtained multi-dimensional state time sequence features can be further improved.

[0059] It can be understood that in the above step S122b, the specific manner of nonlinearly mapping the device environment semantic features is not limited, for example, in an alternative implementation, to effectively capture nonlinear high-order semantic features in the device environment semantic features, the above step S122b can further include the following contents: First, the device environment semantic features can be fully connected to form device environment fully connected features, for example, the device environment semantic features can be multiplied by a first fully connected weight matrix, and then the result of multiplication can be added to a first fully connected bias parameter to obtain device environment fully connected features. Second, the device environment fully connected features can be processed based on a first nonlinear function to form device environment nonlinear features, for example, the first nonlinear function can be a rectified linear unit (ReLU). Then, the device environment nonlinear features can be fully connected to form fully connected nonlinear features. For example, the device environment nonlinear features can be multiplied by a second fully connected weight matrix, and then the multiplication result can be added with a second fully connected bias parameter to obtain the fully connected nonlinear features. Further, the fully connected nonlinear features can be processed based on a second nonlinear function to form nonlinear mapping features. For example, the first nonlinear function can be sigmoid, which is used to map each feature parameter to the interval (0, 1). Finally, the nonlinear mapping features can be used to determine device environment latent features. For example, the nonlinear mapping features can be used as device environment latent features. Alternatively, the nonlinear mapping features can be further processed, for example, multiplied by a scaling factor, to obtain device environment latent features. The scaling factor can be 2, 3, etc.

[0060] For steps S120 and S130, the neural network model formed by training can be used to perform the corresponding steps to ensure reliable implementation of semantic mining and state prediction. The training process of the neural network model is as follows: First, training multi-dimensional state time series data can be obtained. The multi-dimensional state time series data includes training network level time series data, training device running time series data, and training device environment time series data. The multi-dimensional state time series data can be explained as described above. Second, the semantic mining unit included in the network state prediction model can be used to perform cross-dimension state time series semantic mining on the training multi-dimensional state time series data to form training multi-dimensional state time series features corresponding to the training multi-dimensional state time series data. The training multi-dimensional state time series features carry three-dimensional state time series semantic features corresponding to the training network level time series data, the training device running time series data, and the training device environment time series data. The first shared weight matrix, the second shared weight matrix, the shared bias parameter, and the parameters corresponding to the attention processing can be randomly generated at the initial stage. Then, the state prediction unit included in the network state prediction model can be used to perform network state prediction based on the training multi-dimensional state time series features to obtain training network state prediction results. The network state prediction can be explained as described above. Finally, the model parameters of the network state prediction model can be updated based on the error (such as mean square error, etc.) between the training network state prediction result and the network state label corresponding to the training multi-dimensional state time series data, and an optimized network state prediction model is formed. For example, the model parameters are adjusted in the direction of reducing the error until the error converges.

[0061] In combination Figure 4 The embodiments of the present application also provide an environmentally-friendly public toilet device state monitoring device applicable to the electronic device. The environmentally-friendly public toilet device state monitoring device can include a state data acquisition module, a semantic mining module, and a state prediction module.

[0062] In detail, the state data acquisition module can be used to acquire multi-dimensional state time series data of a target environmentally-friendly public toilet device. The multi-dimensional state time series data includes network level time series data, device running time series data, and device environment time series data. The target environmentally-friendly public toilet device is used to interact with a background device through a network. In the embodiments of the present application, the state data acquisition module can be used to perform the steps S110 shown in the figure. The related content of the state data acquisition module can be referred to the description of the step S110 in the foregoing. Figure 2

[0063] In detail, the semantic mining module can be used to perform cross-dimension state time series semantic mining on the multi-dimensional state time series data, and form multi-dimensional state time series features corresponding to the multi-dimensional state time series data. The multi-dimensional state time series features carry three-dimensional state time series semantic features corresponding to the network level time series data, the device running time series data, and the device environment time series data. In the embodiments of the present application, the semantic mining module can be used to perform the step S120 shown in the figure. The related content of the semantic mining module can be referred to the description of the step S120 in the foregoing. Figure 2

[0064] In detail, the state prediction module can be used to perform network state prediction based on the multi-dimensional state time series features, and obtain a target network state prediction result. The target network state prediction result is used to reflect the probability of network failure of the target environmentally-friendly public toilet device. In the embodiments of the present application, the state prediction module can be used to perform the step S130 shown in the figure. The related content of the state prediction module can be referred to the description of the step S130 in the foregoing. Figure 2

[0065] ​​​In the embodiments of the present application, corresponding to the above-mentioned environmental-friendly public toilet device state monitoring method applied to the electronic device, a computer readable storage medium is also provided, and the computer readable storage medium stores a computer program. When the computer program runs, it executes each step of the environmental-friendly public toilet device state monitoring method.

[0066] Wherein, each step executed by the aforementioned computer program when running is not repeated here, and can be referred to the foregoing explanation and description of the environmental-friendly public toilet device state monitoring method.

[0067] In summary, the environmental-friendly public toilet device state monitoring method, device and equipment provided by the present application first acquire multi-dimensional state time series data of a target environmental-friendly public toilet device. Secondly, cross-dimension state time series semantic mining is performed on the multi-dimensional state time series data to form multi-dimensional state time series features corresponding to the multi-dimensional state time series data. Then, network state prediction is performed based on the multi-dimensional state time series features to obtain a target network state prediction result. Based on the above, since cross-dimension state time series semantic mining is performed on network layer time series data, device running time series data and device environment time series data in three dimensions, the semantic features of the device running and device environment dimensions can assist in representing the semantic information of the network layer. In a toilet, the humidity is relatively high, which can affect the service life of the hardware device, and the moisture in the air can increase the propagation loss of the wireless signal. Therefore, with the assistance of the semantic features of the device environment dimension, the semantic representation accuracy of the formed multi-dimensional state time series features can be improved, the reliability of the determined target network state prediction result can be improved, and the problem of relatively low reliability of device state monitoring in the prior art can be improved.

[0068] In several embodiments provided by the embodiments of the present application, it should be understood that the disclosed apparatus and method can also be implemented by other manners. The apparatus and method embodiments described above are only illustrative. For example, the flowchart and block diagram in the drawings show the possible implementation architecture, function and operation of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment or a part of code, which includes one or more executable instructions for implementing the specified logic function. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different order from that shown in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the function involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for implementing the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0069] In addition, each functional module in the embodiments of the present 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.

[0070] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, an electronic device, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes. It should be noted that in this document, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the processes, methods, articles or devices that include a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or device that includes the element.

[0071] The above descriptions are only the preferred embodiment of the present application, but not for limiting the present application. For those skilled in the art, the present application can have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An environmentally-friendly public lavatory device state monitoring method, characterized by, The method comprises the steps of: acquiring multi-dimensional state time series data of a target environmentally-friendly public toilet device, wherein the multi-dimensional state time series data comprises network-level time series data, device operation time series data and device environment time series data, and the target environmentally-friendly public toilet device is configured to interact with a background device through a network; performing cross-dimension state time series semantic mining on the multi-dimensional state time series data to form multi-dimensional state time series features corresponding to the multi-dimensional state time series data, wherein the multi-dimensional state time series features carry three-dimension state time series semantic features corresponding to the network-level time series data, the device operation time series data and the device environment time series data; performing network state prediction based on the multi-dimensional state time series features to obtain a target network state prediction result, wherein the target network state prediction result is used to reflect a probability of network failure of the target environmentally-friendly public toilet device.

2. The environmentally-friendly public toilet facility state monitoring method according to claim 1, characterized by, The step of performing cross-dimension state time series semantic mining on the multi-dimensional state time series data to form multi-dimensional state time series features corresponding to the multi-dimensional state time series data comprises the steps of: respectively performing semantic space mapping on the network-level time series data, the device operation time series data and the device environment time series data to form network-level semantic features, device operation semantic features and device environment semantic features; fusing the network-level semantic features and the device operation semantic features to form network-device fused features, and performing significant feature screening on the network-device fused features based on hidden features represented by the device environment semantic features to form multi-dimensional state time series features.

3. The method of claim 2, wherein the method further comprises: The step of fusing the network-level semantic features and the device operation semantic features to form network-device fused features, and performing significant feature screening on the network-device fused features based on hidden features represented by the device environment semantic features to form multi-dimensional state time series features comprises the steps of: splicing the network-level semantic features and the device operation semantic features to form network-device fused features; performing nonlinear mapping on the device environment semantic features to form device environment hidden features; performing significant feature screening on the network-device fused features based on the device environment hidden features to form multi-dimensional state time series features.

4. The environmentally-friendly public toilet facility state monitoring method according to claim 3, characterized by, The step of performing nonlinear mapping on the device environment semantic features to form device environment hidden features comprises the steps of: performing full connection processing on the device environment semantic features to form device environment full connection features; processing the device environment full connection features based on a first nonlinear function to form device environment nonlinear features; performing full connection processing on the device environment nonlinear features to form full connection nonlinear features; processing the full connection nonlinear features based on a second nonlinear function to form nonlinear mapping features; determining the device environment hidden features based on the nonlinear mapping features.

5. The method of claim 3, wherein the method further comprises: The step of performing significant feature screening on the network-device fused features based on the device environment hidden features to form multi-dimensional state time series features comprises the step of: performing bitwise multiplication operation on the device environment hidden features and the network-device fused features to form multi-dimensional state time series features.

6. The environmentally-friendly public toilet facility state monitoring method according to claim 2, characterized by, The step of respectively performing semantic space mapping on the network layer timing data, the device runtime timing data and the device environment timing data to form network layer semantic features, device runtime semantic features and device environment semantic features comprises: performing word embedding processing on the network layer timing data to form a word embedding feature sequence; for each word embedding feature in the word embedding feature sequence, multiplying the current word embedding feature by a first shared weight matrix to obtain a first intermediate feature, multiplying an output feature corresponding to a previous word embedding feature of the word embedding feature by a second shared weight matrix to obtain a second intermediate feature, and adding the first intermediate feature, the second intermediate feature and a shared bias parameter to form an output feature of the current word embedding feature, wherein the output feature corresponding to the previous word embedding feature of the first word embedding feature is a zero feature; determining the network layer semantic features based on the output feature corresponding to the last word embedding feature; respectively performing semantic space mapping on the device runtime timing data and the device environment timing data to form device runtime semantic features and device environment semantic features.

7. The environmentally friendly public toilet facility state monitoring method according to claim 6, characterized by, The step of determining the network layer semantic features based on the output feature corresponding to the last word embedding feature comprises: performing self-attention processing on the output feature corresponding to the last word embedding feature to form the network layer semantic features.

8. The environmentally-friendly public toilet facility state monitoring method according to any one of claims 1 to 7, characterized by, The environmental public toilet device state monitoring method further comprises: obtaining training multi-dimensional state timing data, wherein the multi-dimensional state timing data comprises training network layer timing data, training device runtime timing data and training device environment timing data; using a semantic mining unit included in a network state prediction model to perform cross-dimensional state timing semantic mining on the training multi-dimensional state timing data to form training multi-dimensional state timing features corresponding to the training multi-dimensional state timing data, wherein the training multi-dimensional state timing features carry three-dimensional state timing semantic features corresponding to the training network layer timing data, the training device runtime timing data and the training device environment timing data; using a state prediction unit included in the network state prediction model to perform network state prediction based on the training multi-dimensional state timing features to obtain training network state prediction results; updating model parameters of the network state prediction model based on errors between the training network state prediction results and network state labels corresponding to the training multi-dimensional state timing data to form an optimized network state prediction model.

9. A status monitoring device for environmentally friendly public toilet equipment, characterized in that, comprises: a state data acquisition module configured to acquire multi-dimensional state timing data of a target environmental public toilet device, wherein the multi-dimensional state timing data comprises network layer timing data, device runtime timing data and device environment timing data, and the target environmental public toilet device is configured to perform data interaction with a background device through a network; A semantic mining module is configured to perform cross-dimension state time sequence semantic mining on the multi-dimensional state time sequence data to form multi-dimensional state time sequence features corresponding to the multi-dimensional state time sequence data, wherein the multi-dimensional state time sequence features carry state time sequence semantic features of three dimensions corresponding to network layer time sequence data, device runtime time sequence data and device environment time sequence data. A state prediction module is configured to perform network state prediction based on the multi-dimensional state time sequence features to obtain a target network state prediction result, wherein the target network state prediction result is used to reflect a probability of network failure of the target environmental public toilet device.

10. An electronic device, comprising: The method comprises: a memory configured to store a computer program; a processor connected to the memory and configured to execute the computer program stored in the memory to implement the environmental public toilet device state monitoring method according to any one of claims 1-8.

Images