Intelligent building data acquisition and analysis method and system based on Internet of Things

Abstract

The invention discloses a smart building data acquisition and analysis method and system based on the Internet of Things, and relates to the technical field of data analysis. The method comprises the steps that a vibration signal of a lift car and a sound signal of a traction machine of a target elevator in a preset time period are obtained; sequentially performing wavelet packet transformation, energy extraction and feature dimension reduction on the vibration signal and the sound signal to obtain a vibration feature vector and a sound feature vector; and the vibration feature vector and the sound feature vector serve as input of a pre-trained convolutional neural network model, and a fault diagnosis result of the target elevator is obtained. Through real-time monitoring and analysis of the vibration signals of the lift car and the sound signals of the traction machine during operation of the elevator, the abnormities during operation of the elevator can be captured, and the abnormities are often early signals of faults, so that intervention and timely maintenance before the faults deteriorate are facilitated, accidents are reduced, and the safety is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a method and system for collecting and analyzing data of a smart building based on the Internet of Things. Background Art

[0002] In modern urban construction, smart buildings are becoming a trend as an important solution to improve building management efficiency and living comfort. These buildings use IoT technology to achieve intelligent management of various equipment and provide more efficient services. In smart buildings, elevators, as key vertical transportation tools, are responsible for the daily travel of residents and visitors. However, the normal operation of elevators is crucial, and their failures will not only affect people's travel efficiency, but may also bring safety hazards. Therefore, fault monitoring of elevators is particularly necessary.

[0003] Patent CN118419721B discloses an elevator IoT fault alarm method and system, which analyzes the speed of the elevator guide wheel to deduce the possibility of elevator failure. At the same time, it rechecks the speed value and decibel value, and the feedback response is rapid, which improves the accuracy and reliability of alarm monitoring. However, the traditional detection method is to make judgments after the accident occurs, and it is impossible to predict potential faults before the fault occurs, resulting in situations where people are trapped or injured. The existing technology has the problem of untimely detection of potential faults and low safety. Summary of the invention

[0004] The purpose of the present invention is to solve the problem of untimely discovery of potential faults mentioned in the above background technology, and to propose a smart building data collection and analysis method and system based on the Internet of Things.

[0005] The first aspect of the present invention provides a method for collecting and analyzing smart building data based on the Internet of Things, the method comprising:

[0006] Acquiring the operation data of the target elevator in a preset time period; the operation data includes a vibration signal of the car and a sound signal of the traction machine;

[0007] Performing wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively, to obtain a first eigenvector and a second eigenvector;

[0008] Performing feature dimension reduction on the first eigenvector and the second eigenvector respectively to obtain a vibration eigenvector and a sound eigenvector;

[0009] The vibration feature vector and the sound feature vector are used as inputs of a pre-trained convolutional neural network model to obtain a fault diagnosis result of a target elevator.

[0010] Optionally, performing wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively, to obtain the first eigenvector and the second eigenvector comprises:

[0011] According to a preset wavelet function, the target signal is decomposed layer by layer until a preset number of layers is reached to obtain a plurality of sub-bands; the target signal is the vibration signal or the sound signal;

[0012] Calculate the energy eigenvalue of each subband separately:

[0013]

[0014] Among them, E i is the energy eigenvalue of the ith subband; x i (t) is the signal value of the i-th subband at the t-th sampling point; n is the number of sampling points;

[0015] The energy eigenvalues ​​of all sub-bands are combined to obtain a target eigenvector; the target eigenvector is the first eigenvector or the second eigenvector.

[0016] Optionally, performing feature dimension reduction on the first eigenvector and the second eigenvector respectively to obtain a vibration eigenvector and a sound eigenvector comprises:

[0017] According to the characteristic mean value vector and standard deviation vector of the vibration signal and the sound signal in the sample data, the first eigenvector and the second eigenvector are respectively standardized to obtain a third eigenvector and a fourth eigenvector;

[0018] Multiplying the third eigenvector by a principal component conversion matrix corresponding to the vibration signal to obtain a vibration eigenvector;

[0019] The fourth eigenvector is multiplied by the principal component conversion matrix corresponding to the sound signal to obtain a sound eigenvector.

[0020] Optionally, the principal component conversion matrix corresponding to the vibration signal is determined by sample data;

[0021] The specific process of determining the principal component conversion matrix corresponding to the vibration signal from the sample data is as follows:

[0022] Acquire sample data of the vibration signal, perform wavelet packet transformation and energy extraction, and obtain a sample matrix; each row of the sample matrix represents a sample, and each column represents a feature;

[0023] Standardizing the sample matrix to obtain a standardized matrix;

[0024] Calculating a covariance matrix of the standardized matrix;

[0025] Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors thereof;

[0026] Sort the eigenvalues ​​in descending order, select the eigenvectors corresponding to the first K eigenvalues, and form the principal component transformation matrix; where the K value satisfies:

[0027]

[0028] Among them, z is the sum of all eigenvalues; g is the preset selection threshold; l j is the jth eigenvalue.

[0029] Optionally, the convolutional neural network model includes a first convolution branch, a second convolution branch, a feature fusion layer, a fully connected layer and an output layer; the first convolution branch includes multiple convolution layers, and its activation function is a tanh function; the second convolution branch and the first convolution branch have the same structure and different weights, and their activation function is a ReLU function; the activation function of the output layer is a softmax function;

[0030] in:

[0031] The first convolution branch is used to process the vibration eigenvector to obtain a first eigentensor;

[0032] The second convolution branch is used to process the sound feature vector to obtain a second feature tensor;

[0033] The fusion layer is used to fuse the first feature tensor and the second feature tensor to obtain a fifth feature vector;

[0034] The fully connected layer is used to process the fifth eigenvector to obtain a sixth eigenvector;

[0035] The output layer is used to obtain a fault diagnosis result according to the sixth eigenvector.

[0036] The second aspect of the present invention provides a smart building data collection and analysis system based on the Internet of Things, the system comprising:

[0037] A data acquisition module, used to acquire the operation data of the target elevator in a preset time period; the operation data includes a vibration signal of the car and a sound signal of the traction machine;

[0038] A feature extraction module, used to perform wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively, to obtain a first feature vector and a second feature vector;

[0039] A feature dimension reduction module, used to perform feature dimension reduction on the first feature vector and the second feature vector respectively to obtain a vibration feature vector and a sound feature vector;

[0040] The fault diagnosis module is used to use the vibration feature vector and the sound feature vector as inputs of a pre-trained convolutional neural network model to obtain a fault diagnosis result of a target elevator.

[0041] Optionally, the feature extraction module includes:

[0042] A signal decomposition module, used for decomposing the target signal layer by layer according to a preset wavelet function until a preset number of layers is reached to obtain a plurality of sub-bands; the target signal is the vibration signal or the sound signal;

[0043] Energy calculation module, used to calculate the energy characteristic value of each sub-band separately:

[0044]

[0045] Among them, E i is the energy eigenvalue of the ith subband; x i (t) is the signal value of the i-th subband at the t-th sampling point; n is the number of sampling points;

[0046] The eigenvector determination module is used to combine the energy eigenvalues ​​of all sub-bands to obtain a target eigenvector; the target eigenvector is the first eigenvector or the second eigenvector.

[0047] Optionally, the feature dimension reduction module includes:

[0048] A standardization module, used to standardize the first eigenvector and the second eigenvector according to the characteristic mean value vector and the standard deviation vector of the vibration signal and the sound signal in the sample data, respectively, to obtain a third eigenvector and a fourth eigenvector;

[0049] The dimension reduction module is used to multiply the third eigenvector with the principal component conversion matrix corresponding to the vibration signal to obtain a vibration eigenvector; and multiply the fourth eigenvector with the principal component conversion matrix corresponding to the sound signal to obtain a sound eigenvector.

[0050] Optionally, the principal component conversion matrix corresponding to the vibration signal is determined by sample data;

[0051] The specific process of determining the principal component conversion matrix corresponding to the vibration signal from the sample data is as follows:

[0052] Acquire sample data of the vibration signal, perform wavelet packet transformation and energy extraction, and obtain a sample matrix; each row of the sample matrix represents a sample, and each column represents a feature;

[0053] Standardizing the sample matrix to obtain a standardized matrix;

[0054] Calculating a covariance matrix of the standardized matrix;

[0055] Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors thereof;

[0056] Sort the eigenvalues ​​in descending order, select the eigenvectors corresponding to the first K eigenvalues, and form the principal component transformation matrix; where the K value satisfies:

[0057]

[0058] Among them, z is the sum of all eigenvalues; g is the preset selection threshold; l j is the jth eigenvalue.

[0059] Optionally, the convolutional neural network model includes a first convolution branch, a second convolution branch, a feature fusion layer, a fully connected layer and an output layer; the first convolution branch includes multiple convolution layers, and its activation function is a tanh function; the second convolution branch and the first convolution branch have the same structure and different weights, and their activation function is a ReLU function; the activation function of the output layer is a softmax function;

[0060] in:

[0061] The first convolution branch is used to process the vibration eigenvector to obtain a first eigentensor;

[0062] The second convolution branch is used to process the sound feature vector to obtain a second feature tensor;

[0063] The fusion layer is used to fuse the first feature tensor and the second feature tensor to obtain a fifth feature vector;

[0064] The fully connected layer is used to process the fifth eigenvector to obtain a sixth eigenvector;

[0065] The output layer is used to obtain a fault diagnosis result according to the sixth eigenvector.

[0066] Beneficial effects of the present invention:

[0067] The present invention proposes a data collection and analysis method for a smart building based on the Internet of Things, the method comprising: obtaining the operating data of a target elevator in a preset time period; the operating data comprising a vibration signal of a car and a sound signal of a traction machine; performing wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively, to obtain a first eigenvector and a second eigenvector; performing feature dimension reduction on the first eigenvector and the second eigenvector respectively, to obtain a vibration eigenvector and a sound eigenvector; using the vibration eigenvector and the sound eigenvector as inputs of a pre-trained convolutional neural network model, to obtain a fault diagnosis result of the target elevator.

[0068] Through real-time monitoring and analysis of the elevator car vibration signal and traction machine sound signal during operation, it is possible to capture abnormalities in elevator operation. These abnormalities are often early signals of faults, which help to intervene before the fault worsens, conduct timely maintenance, reduce accidents, and improve safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The present invention will be further described below in conjunction with the accompanying drawings.

[0070] Figure 1 A flowchart of a method for collecting and analyzing smart building data based on the Internet of Things is provided for an embodiment of the present invention;

[0071] Figure 2 A decomposition schematic diagram is provided for an embodiment of the present invention;

[0072] Figure 3 An architecture diagram of a smart building data collection and analysis system based on the Internet of Things is provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] The embodiment of the present invention provides a method for collecting and analyzing data of a smart building based on the Internet of Things. Figure 1 , Figure 1 The present invention provides a flowchart of a method for collecting and analyzing data of a smart building based on the Internet of Things. The method comprises the following steps:

[0075] S101, obtaining the operation data of the target elevator in a preset time period.

[0076] S102, performing wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively, to obtain a first eigenvector and a second eigenvector.

[0077] S103, performing feature dimension reduction on the first eigenvector and the second eigenvector respectively to obtain a vibration eigenvector and a sound eigenvector.

[0078] S104, using the vibration feature vector and the sound feature vector as inputs of a pre-trained convolutional neural network model to obtain a fault diagnosis result of the target elevator.

[0079] The operation data includes the vibration signal of the car and the sound signal of the traction machine.

[0080] A smart building data collection and analysis method based on the Internet of Things provided by an embodiment of the present invention can capture abnormalities in elevator operation through real-time monitoring and analysis of elevator car vibration signals and traction machine sound signals during elevator operation. These abnormalities are often early signals of faults, which help to intervene before the fault worsens, conduct timely maintenance, reduce accidents, and improve safety.

[0081] In one implementation, a vibration sensor can be installed on the top of the elevator car and a sound sensor can be installed in the traction machine room. The collected data is sent to the server through wireless networking. The vibration signal of the elevator is a direct feedback generated when the mechanical system is running, which can reflect the mechanical structure state of the elevator during operation. For example, the uneven track of the elevator car, component wear, balance problems, etc. may all be reflected through vibration signals. The sound signal of the traction machine is usually closely related to the state of the elevator drive system. The sound signal can capture abnormal phenomena such as friction, vibration, and impact, especially abnormal sounds generated by components such as gears, bearings, or motors. Vibration signals and sound signals are the two main representations of the elevator mechanical system. By combining these two signals, a more comprehensive elevator operation status can be obtained, which can improve the comprehensiveness, sensitivity, and accuracy of fault diagnosis, thereby effectively realizing intelligent preventive maintenance of elevators.

[0082] In one implementation, feature dimensionality reduction is used to reduce computational complexity, making subsequent analysis and model training more efficient while retaining key information.

[0083] In one embodiment, step S102 includes:

[0084] Step 1: Decompose the target signal layer by layer according to a preset wavelet function until a preset number of layers is reached to obtain multiple sub-bands.

[0085] Step 2: Calculate the energy eigenvalue of each subband:

[0086]

[0087] Among them, E i is the energy eigenvalue of the ith subband; x i (t) is the signal value of the i-th subband at the t-th sampling point; n is the number of sampling points;

[0088] Step three, combining the energy eigenvalues ​​of all sub-bands to obtain a target eigenvector; the target eigenvector is the first eigenvector or the second eigenvector.

[0089] The target signal is a vibration signal or a sound signal.

[0090] In one implementation, the preset wavelet function may be Daubechies 4 (db4). The preset number of layers may be 9, that is, 512 subbands are obtained. Taking the three-layer decomposition as an example, see Figure 2 , Figure 2 A decomposition diagram provided by an embodiment of the present invention is in the form of a binary tree, and the last layer decomposes into 2 3 =8 subbands, numbered from left to right, AAA3 is the first subband, DDD3 is the eighth subband. S is the original signal, A1 is the low-frequency subband of S, D1 is the high-frequency subband of S; AA2 is the low-frequency subband of A1, DA2 is the high-frequency subband of A1; AD2 is the low-frequency subband of D1, DD2 is the high-frequency subband of D1; AAA3 is the low-frequency subband of AA2, DAA3 is the high-frequency subband of AA2; and so on, the left subtree root node of the parent node is the low-frequency subband of the parent node, and the right subtree root node of the parent node is the high-frequency subband of the parent node.

[0091] Wavelet packet transform is a multi-scale analysis method that can decompose signals into sub-bands of different frequency bands, covering all frequency bands from high frequency to low frequency. Since vibration signals and sound signals may carry different types of fault information at different frequencies, multi-scale decomposition can capture the characteristics of each frequency band in the signal, especially subtle, high-frequency abnormal signals, and enhance sensitivity to faults.

[0092] In one implementation, the sub-band energy eigenvalues ​​represent the signal strength in each frequency band. For vibration and sound signals of elevators, energy changes often reflect the presence of mechanical faults or anomalies. For example, abnormally high energy in certain frequency bands may indicate wear or failure of specific components (such as bearings and gears).

[0093] In one embodiment, step S103 includes:

[0094] Step 1: standardize the first eigenvector and the second eigenvector according to the characteristic mean value vector and the standard deviation vector of the vibration signal and the sound signal in the sample data, respectively, to obtain the third eigenvector and the fourth eigenvector.

[0095] Step 2: multiply the third eigenvector by the principal component conversion matrix corresponding to the vibration signal to obtain a vibration eigenvector; multiply the fourth eigenvector by the principal component conversion matrix corresponding to the sound signal to obtain a sound eigenvector.

[0096] In one implementation, principal component analysis (PCA) can be used to project the original high-dimensional feature vector into a lower-dimensional space, retaining most of the main information in the signal while removing irrelevant or redundant features. This reduces the dimensionality of model processing, reduces computational complexity, and speeds up model training and inference.

[0097] In one implementation, taking a vibration signal as an example, the standardized calculation method is:

[0098]

[0099] Among them, E′ i is the energy eigenvalue after normalization of the i-th element; E i is the i-th energy eigenvalue; μ i is the average value of the i-th energy eigenvalue in the vibration signal sample data; σ i is the standard deviation of the i-th energy eigenvalue in the vibration signal sample data.

[0100] In one embodiment, the principal component conversion matrix corresponding to the vibration signal and the sound signal is determined by the sample data;

[0101] The specific process of determining the principal component conversion matrix corresponding to the vibration signal from the sample data is as follows:

[0102] Step 1: Obtain sample data of the vibration signal, perform wavelet packet transformation and energy extraction, and obtain a sample matrix.

[0103] Step 2: Standardize the sample matrix to obtain a standardized matrix.

[0104] Step 3: Calculate the covariance matrix of the standardized matrix.

[0105] Step 4: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors.

[0106] Step 5: Sort the eigenvalues ​​in descending order, select the eigenvectors corresponding to the first K eigenvalues, and form the principal component transformation matrix; where the K value satisfies:

[0107]

[0108] Among them, z is the sum of all eigenvalues; g is the preset selection threshold; l j is the jth eigenvalue.

[0109] Among them, each row of the sample matrix represents a sample, and each column represents a feature.

[0110] In one implementation, the eigenvalues ​​are sorted in descending order, and the K value is determined based on a preset selection threshold g and the sum of all eigenvalues ​​z, that is, the eigenvectors corresponding to the first K most important eigenvalues ​​are selected to form a principal component conversion matrix. This achieves data dimensionality reduction while retaining the main information in the original data as much as possible. The selection threshold g can be 0.85 or 0.9.

[0111] In one implementation, the principal component transformation matrix is ​​calculated using sample data to ensure that the selected features are closely related to the actual data. This data-driven approach can maximize the extraction of information related to fault diagnosis and avoid bias caused by subjective judgment.

[0112] In one implementation, the K value in the principal component analysis process of the sound signal can be obtained through the same steps. By comparing the two K values ​​of the vibration signal and the sound signal, the larger K value is selected for feature vector selection. The consistency of the dimensions of the vibration feature vector and the sound feature vector is ensured.

[0113] In one embodiment, the convolutional neural network model includes a first convolution branch, a second convolution branch, a feature fusion layer, a fully connected layer and an output layer; the first convolution branch includes multiple convolution layers, and its activation function is a tanh function; the second convolution branch has the same structure as the first convolution branch but different weights, and its activation function is a ReLU function; the activation function of the output layer is a softmax function;

[0114] in:

[0115] The first convolution branch is used to process the vibration eigenvector to obtain the first eigentensor;

[0116] The second convolution branch is used to process the sound feature vector to obtain the second feature tensor;

[0117] A feature fusion layer, used to fuse the first feature tensor and the second feature tensor to obtain a fifth feature vector;

[0118] A fully connected layer is used to process the fifth eigenvector to obtain the sixth eigenvector;

[0119] The output layer is used to obtain the fault diagnosis result according to the sixth eigenvector.

[0120] The feature fusion layer includes a pooling layer, which is used to downsample the two-dimensional tensor to obtain a one-dimensional vector. The two one-dimensional vectors can be concatenated to form the fifth feature vector.

[0121] In one implementation, the model can extract and fuse the feature expressions of two different types of sensor data by processing vibration feature vectors and sound feature vectors in parallel. This multi-source information fusion helps to capture the equipment status more comprehensively and improve the accuracy and reliability of fault diagnosis.

[0122] In one implementation, the fault diagnosis result includes the probability of occurrence of each type of fault and the probability of normality, which helps decision makers understand and judge the fault situation more intuitively.

[0123] The embodiment of the present invention provides a method for collecting and analyzing data of a smart building based on the Internet of Things. Figure 3 , Figure 3 The present invention provides an architecture diagram of a smart building data collection and analysis system based on the Internet of Things. The system includes:

[0124] The data acquisition module is used to acquire the operation data of the target elevator in a preset time period.

[0125] The feature extraction module is used to perform wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively to obtain a first feature vector and a second feature vector.

[0126] The feature dimension reduction module is used to perform feature dimension reduction on the first feature vector and the second feature vector respectively to obtain a vibration feature vector and a sound feature vector.

[0127] The fault diagnosis module is used to use the vibration feature vector and the sound feature vector as inputs of a pre-trained convolutional neural network model to obtain a fault diagnosis result of a target elevator.

[0128] The operation data includes the vibration signal of the car and the sound signal of the traction machine.

[0129] A smart building data collection and analysis method based on the Internet of Things provided by an embodiment of the present invention can capture abnormalities in elevator operation through real-time monitoring and analysis of elevator car vibration signals and traction machine sound signals during elevator operation. These abnormalities are often early signals of faults, which help to intervene before the fault worsens, conduct timely maintenance, reduce accidents, and improve safety.

[0130] In one embodiment, the feature extraction module includes:

[0131] The signal decomposition module is used to decompose the target signal layer by layer according to a preset wavelet function until a preset number of layers is reached to obtain multiple sub-bands.

[0132] Energy calculation module, used to calculate the energy characteristic value of each sub-band separately:

[0133]

[0134] Among them, E i is the energy eigenvalue of the ith subband; x i (t) is the signal value of the i-th subband at the t-th sampling point; n is the number of sampling points.

[0135] The eigenvector determination module is used to combine the energy eigenvalues ​​of all sub-bands to obtain a target eigenvector.

[0136] Wherein, the target signal is the vibration signal or the sound signal; and the target feature vector is the first feature vector or the second feature vector.

[0137] In one embodiment, the feature dimension reduction module includes:

[0138] The standardization module is used to standardize the first eigenvector and the second eigenvector according to the characteristic mean value vector and the standard deviation vector of the vibration signal and the sound signal in the sample data, so as to obtain the third eigenvector and the fourth eigenvector.

[0139] The dimension reduction module is used to multiply the third eigenvector with the principal component conversion matrix corresponding to the vibration signal to obtain a vibration eigenvector; and multiply the fourth eigenvector with the principal component conversion matrix corresponding to the sound signal to obtain a sound eigenvector.

[0140] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for collecting and analyzing smart building data based on the Internet of Things, characterized in that: The method comprises: Acquiring the operation data of the target elevator in a preset time period; the operation data includes a vibration signal of the car and a sound signal of the traction machine; Performing wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively, to obtain a first eigenvector and a second eigenvector; Performing feature dimension reduction on the first eigenvector and the second eigenvector respectively to obtain a vibration eigenvector and a sound eigenvector; The vibration feature vector and the sound feature vector are used as inputs of a pre-trained convolutional neural network model to obtain a fault diagnosis result of a target elevator.

2. According to the method for collecting and analyzing data of a smart building based on the Internet of Things according to claim 1, it is characterized in that: The vibration signal and the sound signal are respectively subjected to wavelet packet transformation and energy extraction in sequence to obtain a first eigenvector and a second eigenvector, including: According to a preset wavelet function, the target signal is decomposed layer by layer until a preset number of layers is reached to obtain a plurality of sub-bands; the target signal is the vibration signal or the sound signal; Calculate the energy eigenvalue of each subband separately: Among them, E i is the energy eigenvalue of the ith subband; x i (t) is the signal value of the i-th subband at the t-th sampling point; n is the number of sampling points; The energy eigenvalues ​​of all sub-bands are combined to obtain a target eigenvector; the target eigenvector is the first eigenvector or the second eigenvector.

3. The method for collecting and analyzing smart building data based on the Internet of Things according to claim 2 is characterized in that: Performing feature dimension reduction on the first eigenvector and the second eigenvector respectively to obtain a vibration eigenvector and a sound eigenvector comprises: According to the characteristic mean value vector and standard deviation vector of the vibration signal and the sound signal in the sample data, the first eigenvector and the second eigenvector are respectively standardized to obtain a third eigenvector and a fourth eigenvector; Multiplying the third eigenvector by a principal component conversion matrix corresponding to the vibration signal to obtain a vibration eigenvector; The fourth eigenvector is multiplied by the principal component conversion matrix corresponding to the sound signal to obtain a sound eigenvector.

4. The method for collecting and analyzing smart building data based on the Internet of Things according to claim 3 is characterized in that: The principal component conversion matrix corresponding to the vibration signal is determined by the sample data; The specific process of determining the principal component conversion matrix corresponding to the vibration signal from the sample data is as follows: Acquire sample data of the vibration signal, perform wavelet packet transformation and energy extraction, and obtain a sample matrix; each row of the sample matrix represents a sample, and each column represents a feature; Standardizing the sample matrix to obtain a standardized matrix; Calculating a covariance matrix of the standardized matrix; Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors thereof; Sort the eigenvalues ​​in descending order, select the eigenvectors corresponding to the first K eigenvalues, and form the principal component transformation matrix; where the K value satisfies: Among them, z is the sum of all eigenvalues; g is the preset selection threshold; l j is the jth eigenvalue.

5. The method for collecting and analyzing smart building data based on the Internet of Things according to claim 1 is characterized in that: The convolutional neural network model includes a first convolution branch, a second convolution branch, a feature fusion layer, a fully connected layer and an output layer; the first convolution branch includes multiple convolution layers, and its activation function is a tanh function; the second convolution branch has the same structure as the first convolution branch but different weights, and its activation function is a ReLU function; the activation function of the output layer is a softmax function; in: The first convolution branch is used to process the vibration eigenvector to obtain a first eigentensor; The second convolution branch is used to process the sound feature vector to obtain a second feature tensor; The fusion layer is used to fuse the first feature tensor and the second feature tensor to obtain a fifth feature vector; The fully connected layer is used to process the fifth eigenvector to obtain a sixth eigenvector; The output layer is used to obtain a fault diagnosis result according to the sixth eigenvector.

6. A smart building data collection and analysis system based on the Internet of Things, characterized in that: The system comprises: A data acquisition module, used to acquire the operation data of the target elevator in a preset time period; the operation data includes a vibration signal of the car and a sound signal of the traction machine; A feature extraction module, used to perform wavelet packet transformation and energy extraction on the vibration signal and the sound signal respectively, to obtain a first feature vector and a second feature vector; A feature dimension reduction module, used to perform feature dimension reduction on the first feature vector and the second feature vector respectively to obtain a vibration feature vector and a sound feature vector; The fault diagnosis module is used to use the vibration feature vector and the sound feature vector as inputs of a pre-trained convolutional neural network model to obtain a fault diagnosis result of a target elevator.

7. The data collection and analysis system for smart buildings based on the Internet of Things according to claim 6 is characterized in that: The feature extraction module comprises: A signal decomposition module, used for decomposing the target signal layer by layer according to a preset wavelet function until a preset number of layers is reached to obtain a plurality of sub-bands; the target signal is the vibration signal or the sound signal; Energy calculation module, used to calculate the energy characteristic value of each sub-band separately: Among them, E i is the energy eigenvalue of the ith subband; x i (t) is the signal value of the i-th subband at the t-th sampling point; n is the number of sampling points; The eigenvector determination module is used to combine the energy eigenvalues ​​of all sub-bands to obtain a target eigenvector; the target eigenvector is the first eigenvector or the second eigenvector.

8. The data collection and analysis system for smart buildings based on the Internet of Things according to claim 7 is characterized in that: The feature dimension reduction module includes: A standardization module, used to standardize the first eigenvector and the second eigenvector according to the characteristic mean value vector and the standard deviation vector of the vibration signal and the sound signal in the sample data, respectively, to obtain a third eigenvector and a fourth eigenvector; The dimension reduction module is used to multiply the third eigenvector with the principal component conversion matrix corresponding to the vibration signal to obtain a vibration eigenvector; and multiply the fourth eigenvector with the principal component conversion matrix corresponding to the sound signal to obtain a sound eigenvector.

9. The data collection and analysis system for smart buildings based on the Internet of Things according to claim 8 is characterized in that: The principal component conversion matrix corresponding to the vibration signal is determined by the sample data; The specific process of determining the principal component conversion matrix corresponding to the vibration signal from the sample data is as follows: Acquire sample data of the vibration signal, perform wavelet packet transformation and energy extraction, and obtain a sample matrix; each row of the sample matrix represents a sample, and each column represents a feature; Standardizing the sample matrix to obtain a standardized matrix; Calculating a covariance matrix of the standardized matrix; Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors thereof; Sort the eigenvalues ​​in descending order, select the eigenvectors corresponding to the first K eigenvalues, and form the principal component transformation matrix; where the K value satisfies: Among them, z is the sum of all eigenvalues; g is the preset selection threshold; l j is the jth eigenvalue.

10. The smart building data collection and analysis system based on the Internet of Things according to claim 6 is characterized in that: The convolutional neural network model includes a first convolution branch, a second convolution branch, a feature fusion layer, a fully connected layer and an output layer; the first convolution branch includes multiple convolution layers, and its activation function is a tanh function; the second convolution branch has the same structure as the first convolution branch but different weights, and its activation function is a ReLU function; the activation function of the output layer is a softmax function; in: The first convolution branch is used to process the vibration eigenvector to obtain a first eigentensor; The second convolution branch is used to process the sound feature vector to obtain a second feature tensor; The fusion layer is used to fuse the first feature tensor and the second feature tensor to obtain a fifth feature vector; The fully connected layer is used to process the fifth eigenvector to obtain a sixth eigenvector; The output layer is used to obtain a fault diagnosis result according to the sixth eigenvector.

Images