Villa elevator safety monitoring and early warning system and early warning method based on cloud platform

Abstract

The invention provides a villa elevator safety monitoring and early warning system and early warning method based on a cloud platform, and relates to the technical field of elevator monitoring and early warning. The villa elevator safety monitoring and early warning system comprises an elevator monitoring unit, a data processing and self-learning module, a cloud platform management module and a cloud platform management module; the data processing and self-learning module receives monitoring data and a cloud platform server, and the cloud platform server is used for remotely storing operation data of the elevator. According to the villa elevator safety monitoring and early warning system and early warning method based on the cloud platform, a deep learning model is introduced to carry out accurate fault prediction on elevator operation data, and an elevator operation strategy is optimized in combination with reinforcement learning. Through the meta-learning technology, the model can quickly adapt and make accurate prediction when facing a new elevator fault mode or an unknown environment, and the generative adversarial network enhances the diversity of training data, so that the model can show stronger robustness in different fault modes, and the accuracy of fault detection is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of elevator monitoring and early warning, in particular to a villa elevator safety monitoring and early warning system and method based on a cloud platform. BACKGROUND

[0002] As a convenient vertical transportation device, elevators have gradually become an indispensable facility in high-end residences. Due to the low frequency of use, complex operating environment and high value of villa elevators, once a failure or safety problem occurs, it will often cause serious consequences such as equipment damage, personal injury or property loss. Therefore, the safety monitoring and early warning of villa elevators are particularly important. At present, the existing elevator safety monitoring systems on the market are mostly applied to traditional commercial or high-rise residential elevators. Although these systems can provide basic operating state monitoring and fault detection, there are still some deficiencies in the existing technology under the special needs of villa elevators. First, most existing monitoring systems rely on fixed monitoring devices, which cannot be individually configured according to the specific use environment for small and private devices such as villa elevators. In addition, the existing technology has weak prediction and early warning capabilities for various potential hazards during elevator operation, mainly relying on manual inspection and regular maintenance, making it difficult to achieve real-time early warning and intelligent fault diagnosis. Traditional elevator monitoring systems often lack fine monitoring of key components of elevators such as elevator doors, carriages, drive systems, etc., making it difficult to discover small faults or abnormalities in components in a timely manner, thus failing to effectively warn of equipment problems at an early stage, increasing maintenance costs and use risks. Due to the lack of effective early warning means, users often only discover problems after the elevator fails, even missing the best maintenance opportunity, affecting the normal use and safety of the elevator.

[0003] How to implement a precise and real-time safety monitoring and early warning system in a villa elevator to timely discover and predict potential failures and risks is still a difficult problem to be solved in the existing technology. The existing technology has not fully solved the needs of villa elevator safety monitoring and early warning, and there is still a lot of room for improvement in actual application. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a villa elevator safety monitoring and early warning system and method based on a cloud platform, which solves the technical problem of how to improve the safety and efficiency of elevator operation through real-time monitoring, fault prediction and individualized maintenance recommendations.

[0005] To achieve the above purpose, the present application is implemented by the following technical scheme: a villa elevator safety monitoring and early warning system and method based on a cloud platform, comprising:

[0006] An elevator monitoring unit is used to collect real-time operation data of the elevator and generate monitoring data.

[0007] A data processing and self-learning module is used to receive the monitoring data.

[0008] A cloud platform server is used to remotely store the operation data of the elevator and generate elevator fault prediction information based on cloud big data analysis.

[0009] A user terminal is used to receive warning information sent by the cloud platform, including abnormal state of the elevator, fault type, maintenance suggestion and maintenance timing.

[0010] An intelligent environment perception module.

[0011] A cross-device remote collaborative management function is used to remotely view the elevator equipment state through the cloud platform, dispatch maintenance, and share information with other smart devices in the home, thereby realizing the linkage between the elevator and other devices.

[0012] Preferably, the elevator monitoring unit includes a sensor group for monitoring the state data of key components of the elevator, including elevator doors, carriages, drive systems, control systems, braking systems and electrical systems.

[0013] Preferably, the intelligent environment perception module includes humidity, temperature, vibration, noise and air quality sensors, which can automatically adjust the operation strategy of the elevator according to the changes in the environment around the elevator, improving the safety and energy efficiency of the elevator.

[0014] A villa elevator safety monitoring and early warning method based on a cloud platform includes the following steps:

[0015] S1. Use the elevator monitoring unit to collect real-time operation data and fault data of the elevator, including elevator doors, carriages, drive systems, control systems, braking systems and electrical systems, to generate raw monitoring data, and use the data preprocessing module to clean, denoise and normalize the data to ensure data quality and accuracy.

[0016] S2. Input the collected elevator operation data and fault data into a generative adversarial network model, generate simulated fault data similar to actual faults through a generator, and judge the generated data through a discriminator to optimize the adversarial network model training. This process aims to enhance the diversity of the data set, so that the deep learning model has stronger robustness and adaptability when processing different fault patterns.

[0017] S3. Train the deep learning model, which includes convolutional neural network and recurrent neural network, using the enhanced training data, and introduce meta-learning technology to enable the model to quickly adapt and make predictions when facing new failure modes or unknown environments, so as to improve the ability of failure prediction;

[0018] S4. The cloud platform server generates personalized elevator failure warning information and maintenance recommendations according to the failure prediction results of the deep learning model, the optimization strategies of the reinforcement learning, and the real-time data of the environment perception module, and sends the warning information to the user terminal in real time through the cloud platform to help the user take timely maintenance measures.

[0019] Preferably, the loss function of the discriminator can be represented by the following formula:

[0020]

[0021] where D(x) represents the discrimination output of the discriminator for real data x, G(z) represents the simulated failure data generated by the generator, p data (x) is the real data distribution, p z (z) is a random noise distribution, the simulated failure data further includes failure occurrence time, duration and speed mode of elevator operation, so as to enhance the diversity and adaptability of the data set, and the enhanced data set can enable the deep learning model to better perform failure prediction and pattern recognition.

[0022] Preferably, the learning process of the meta-learning model is optimized by the following formula:

[0023]

[0024] where θ is the parameter of the model, α is the learning rate, L(θ) is the loss function, and θ' is the model parameter updated by a small amount of data.

[0025] Preferably, the personalized elevator failure warning information includes failure type, failure occurrence time prediction, component damage prediction and recommended maintenance priority.

[0026] Preferably, the reinforcement learning algorithm optimizes the elevator operation strategy by using the Q-learning method, and the Q value update rule in the Q-learning is as follows:

[0027]

[0028] where Q(s t ,a t ) is the Q value of taking action a t in state s t , and rt+1 To get the reward in the next state s t+1 The reward obtained, gamma is the discount factor, alpha is the learning rate.

[0029] The application provides a cloud platform-based villa elevator safety monitoring and early warning system and method. It has the following beneficial effects:

[0030] The cloud platform-based villa elevator safety monitoring and early warning system and method introduces a deep learning model to accurately predict elevator operation data, and combines reinforcement learning to optimize elevator operation strategies. Through meta-learning technology, the model can quickly adapt and make accurate predictions when faced with new elevator failure modes or unknown environments. The generative adversarial network enhances the diversity of training data, enabling the model to exhibit stronger robustness under different failure modes, thereby improving the accuracy of fault detection, especially in rare or unknown failures.

[0031] This technical solution can generate personalized elevator fault warning information based on real-time monitoring data, deep learning fault prediction results, and reinforcement learning optimization strategies, including fault type, occurrence time, component damage prediction, and recommended maintenance priority. This personalized early warning system can help users take timely maintenance measures when potential elevator failures occur. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 A structure diagram for implementing the application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0034] As Figure 1 shown, the application provides a cloud platform-based villa elevator safety monitoring and early warning system, which includes an elevator monitoring unit for real-time collection of elevator operation data and generation of monitoring data. The elevator monitoring unit includes a sensor group for monitoring the state data of key components of the elevator, including elevator doors, carriages, drive systems, control systems, braking systems, and electrical systems.

[0035] A data processing and self-learning module receives monitoring data.

[0036] The cloud platform server is used for remotely storing the operation data of the elevator and generating elevator fault prediction information based on cloud big data analysis.

[0037] The user terminal is used for receiving the early warning information sent by the cloud platform, and the information includes the abnormal state of the elevator, the fault type, the maintenance suggestion and the maintenance time. When the elevator appears abnormal, the system will display detailed fault information, including the fault type, the occurrence position and the influence range of the fault.

[0038] The fault type: the user can view the fault type through the graphical interface, such as elevator door not closed, drive system overload, brake failure, etc.

[0039] The maintenance suggestion: the system will give specific maintenance suggestions, including the recommended maintenance parts, the estimated maintenance time and the priority.

[0040] The maintenance time: the system will give the best time for the next maintenance based on the usage frequency and operation condition of the elevator.

[0041] The intelligent environment sensing module includes humidity, temperature, vibration, noise and air quality sensors, which can automatically adjust the operation strategy of the elevator according to the changes of the environment around the elevator, and improve the safety and energy efficiency of the elevator.

[0042] The cross-device remote collaborative management function can remotely view the elevator equipment state through the cloud platform, dispatch maintenance, and share information with other smart devices in the home, so as to realize the linkage between the elevator and other devices.

[0043] A villa elevator safety monitoring and early warning method based on a cloud platform, characterized by comprising the following steps:

[0044] S1. The elevator monitoring unit is used to collect real-time operation data and fault data of the elevator, including elevator door, car, drive system, control system, brake system and electrical system, to generate original monitoring data, and through the data preprocessing module, the data is cleaned, denoised and normalized to ensure data quality and accuracy.

[0045] For vibration signal data set, in order to eliminate the influence of noise on subsequent data analysis results, singular value decomposition method is used for denoising processing. Assuming that the collected vibration signal data containing noise is y=(y1,y2,…,y N Based on the theory of phase space reconstruction, an m×n order Hankel matrix can be constructed as follows:

[0046]

[0047] In the formula, A is an m x n matrix, N is the signal length, N = m + n - 1, and m ≥ n. Singular value decomposition of A can be expressed as:

[0048] A = UΣV T

[0049] In the formula, U and V T are m x m and n x n matrices, respectively, called the left and right singular matrices of A. Σ is an m x n diagonal matrix, and the elements λ on the main diagonal are called non-zero singular values of A, arranged in non-increasing order, i.e. λ1≥λ2≥…≥λ1.

[0050] Determine the effective rank of the singular value diagonal matrix Σ, i.e. the first p largest singular values, and then reconstruct the approximation matrix A p of A, which can be expressed as:

[0051]

[0052] In the formula, U p is the left singular vector corresponding to the first p larger singular values; V p is the right singular vector corresponding to the first p larger singular values; Σ p is the diagonal matrix corresponding to the first p larger singular values, which can be expressed as:

[0053]

[0054] The noise-free signal components in the obtained matrix A p are constructed into a matrix in the form of a Hankel matrix, which can be expressed as:

[0055]

[0056] In the formula, α = max(1, k - m + 1); β = min(n, k). The average of the elements on the anti-diagonal of the matrix A p is obtained to estimate the true signal, i.e. the denoised signal.

[0057] Elevator door sensor: used to detect the opening and closing state of the elevator door. Common sensor types include photoelectric sensors or magnetic switch sensors, used to determine whether the door is completely closed, preventing situations such as door jamming or failure to close.

[0058] Car sensor: used to monitor the running state of the car, including acceleration, speed, and position information, usually using acceleration sensors, position sensors, or photoelectric sensors.

[0059] Drive system sensor: monitors the state of the elevator drive system, including power, load, and speed, usually using current sensors, voltage sensors, and temperature sensors.

[0060] Control system sensors: used to detect the state of the control system, including the input and output signals of the elevator control panel. Common sensors include current sensors, temperature sensors, and voltage sensors.

[0061] Braking system sensors: detect whether the braking system is working properly, mainly through pressure sensors to judge the performance of the braking device.

[0062] Electrical system sensors: monitor the working condition of the electrical system, especially the power consumption and overload of the elevator. Current sensors and voltage sensors are commonly used.

[0063] S2. Input the collected elevator operation data and fault data into the generative adversarial network model. The generator generates simulated fault data similar to the actual fault, and the discriminator judges the generated data. The adversarial network model training is optimized. This process aims to enhance the diversity of the data set, so that the deep learning model has stronger robustness and adaptability when dealing with different fault patterns. The loss function of the discriminator can be represented by the following formula:

[0064]

[0065] where D(x) represents the discriminator's judgment output for the real data x, G(z) represents the simulated fault data generated by the generator, p data (x) is the real data distribution, p z (z) is the random noise distribution. The simulated fault data also includes the fault occurrence time, duration, and speed pattern of elevator operation to enhance the diversity and adaptability of the data set, and the enhanced data set can better enable the deep learning model to perform fault prediction and pattern recognition;

[0066] S3. Use the enhanced training data to train the deep learning model, which includes convolutional neural networks and recurrent neural networks. Introduce meta-learning technology to enable the model to quickly adapt and make predictions when facing new fault patterns or unknown environments. The meta-learning model can quickly learn new tasks with a small amount of data, improving the ability of fault prediction. The learning process of the meta-learning model is optimized by the following formula:

[0067]

[0068] where θ is the model parameter, α is the learning rate, L(θ) is the loss function, θ ′ is the model parameter updated by a small amount of data. Meta-learning optimizes the initialization parameters of the model so that the model can quickly adapt to new task data with a small amount of data.

[0069] Assuming a deep learning model based on convolutional neural networks (CNNs) and recurrent neural networks (RNNs) is being trained to predict elevator failure patterns. The following dataset is used to train the model:

[0070] Elevator monitoring data: Each elevator failure task contains the following features:

[0071] Elevator door status (open / close)

[0072] Car position (in meters)

[0073] Running speed (in m / s)

[0074] Current consumption (in A)

[0075] Temperature (in °C)

[0076] Brake status (normal / failure)

[0077] Assuming the dataset has 5000 elevator monitoring records, each containing the above-mentioned features and corresponding failure labels (such as "electrical failure," "mechanical failure," etc.).

[0078] 2. Training of the deep learning model

[0079] Convolutional Neural Network:

[0080] Input data: Assume that the elevator door status, car position, temperature, and current consumption data are processed into a two-dimensional array (e.g., a matrix of size (5000,4)(5000,4)(5000,4), where 4 is the feature dimension).

[0081] CNN architecture: Use a 3-layer convolutional layer followed by a pooling layer and ReLU activation function, and finally pass through a fully connected layer for failure prediction.

[0082] Training data: Use 8000 elevator operation records (part of which comes from known failure patterns and another part comes from unseen environments) to train the convolutional neural network. The convolutional layer extracts spatial features of each part of the elevator, and finally predicts the elevator failure type.

[0083] Recurrent Neural Network:

[0084] Input data: Elevator running speed, temperature changes, and brake status as time series data input. Assume that each task data point is a time step, and the continuous state data of the elevator operation is input as time series.

[0085] RNN architecture: Use a long short-term memory network to process time series data and capture time-dependent dynamic features before and after elevator failure.

[0086] S4. The cloud platform server generates personalized elevator fault warning information and maintenance recommendations based on the fault prediction results of the deep learning model, the optimization strategies of reinforcement learning, and the real-time data of the environment perception module, and sends the warning information to the user terminal in real time through the cloud platform to help users take timely maintenance measures. The personalized elevator fault warning information includes fault type, fault occurrence time prediction, component damage prediction, and recommended maintenance priority. The reinforcement learning algorithm optimizes the elevator operation strategy using the Q-learning method. The Q-value update rule in Q-learning is as follows:

[0087]

[0088] where Q(s t ,a t ) is the Q-value of taking action a t in state s t , r t+1 is the reward obtained in the next state s t+1 , γ is the discount factor, and α is the learning rate.

[0089] The elevator motor temperature is T = 85℃, and the current I = 12A. The model predicts that the motor may overheat based on historical data.

[0090] The fault type is "electrical fault: motor overheating".

[0091] Fault occurrence time prediction: Based on the current state of the elevator, combined with the environment and usage, the deep learning model predicts the occurrence time of the fault. For example, the motor fault may occur within the next 2 hours.

[0092] Model output prediction result: The fault occurrence time is 2 hours later, based on the overheating trend of the motor and the operation mode.

[0093] Component damage prediction: Through learning from historical fault patterns, the model can predict which components may be damaged in the future. For example, the brake system of the elevator may fail due to wear and tear due to frequent use.

[0094] The elevator brake system has been running for 3000 hours. According to historical data, the model predicts that the brake system will fail within the next 200 hours.

[0095] Maintenance priority: The model generates maintenance recommendations by comprehensively evaluating the fault risk of each component and the usage of the elevator, and gives the priority of maintenance. For example, the control system fault of the elevator is marked as high priority due to its greater impact on the overall elevator.

[0096] Experimental Example

[0097] 1. Simulating elevator fault data using a generative adversarial network (GAN). The elevator fault data and simulated data are input into the GAN model.

[0098] Real data: Sensor data from the elevator, including elevator door status, car position, fault type, etc. The following real data:

[0099] Elevator door status Car position Running speed Fault type

[0100]

[0101] These data are used as input to train the discriminator.

[0102] Random noise input: The generator receives random noise z from a normal distribution, which is used to generate simulated fault data. For example, suppose we generate the following noise samples from a standard normal distribution:

[0103] z1 = [0.5, -0.2, 0.8], z2 = [-1.0, 0.3, 0.1], z3 = [0.2, 0.7, -0.5]

[0104] 2. Generator and discriminator outputs

[0105] In the GAN, the goal of the generator is to generate simulated fault data that is as similar as possible to real data, while the discriminator distinguishes whether the input data is from real data or generated data.

[0106] Discriminator output for real data: Suppose the discriminator makes a judgment on the input real data (such as elevator door closed, car position 10 meters fault data), its output is D(x) = 0.95, indicating that the discriminator believes that the probability of this data being real data is 95%.

[0107] Discriminator output for generated data: The generator generates a simulated fault data according to the random noise z, and the discriminator judges the generated data, output D(G(z)) = 0.45, indicating that the discriminator believes that the probability of this data being real data is 45%. This is because the generated data cannot completely imitate the characteristics of real data, and the discriminator believes that it is more likely to be fake.

[0108] 3. Discriminator loss function

[0109] According to the training goal of GAN, the goal of the discriminator is to maximize the discrimination accuracy for real data and minimize the false judgment for generated data. The loss function of the discriminator is:

[0110]

[0111] where:

[0112] pdata (x) is the real data distribution, p z (z) is the random noise distribution.

[0113] D(x) represents the discriminator's output for real data, and D(G(z)) represents the discriminator's output for generated data.

[0114] Assuming the discriminator outputs D(x) = 0.95 for real data x and D(G(z)) = 0.45 for generated data G(z), we can plug these values into the loss function to calculate.

[0115] Loss for real data: For real data D(x) = 0.95, the loss is calculated as:

[0116] Loss for real data: For real data D(x) = 0.95, the loss is calculated as:

[0117] log(0.95) ≈ -0.022

[0118] Loss for generated data: For generated data D(G(z)) = 0.45, the loss is calculated as:

[0119] log(1 - 0.45) = log(0.55) ≈ -0.597

[0120] Therefore, the total loss for the discriminator is:

[0121] L D = -(-0.022) - (-0.597) = 0.022 + 0.597 = 0.619

[0122] 4. Generator's loss function

[0123] The generator's goal is to generate simulated data that is as realistic as possible, making it difficult for the discriminator to distinguish between real and generated data. The generator's loss function is:

[0124]

[0125] Substituting D(G(z)) = 0.45, the generator's loss is:

[0126] L G = -log(0.45) ≈ 0.798

[0127] The generator adjusts the parameters G(z) to increase the quality of the generated data, making it more difficult for the discriminator to recognize it as fake data.

[0128] 5. Data augmentation and failure mode identification

[0129] In this way, the generator can generate diverse simulated fault data, including fault occurrence time, duration, speed pattern, and other characteristics. For example, the generated data may include:

[0130] Elevator door status Car position Running speed Fault type Close 12 m 0.8 m / s Electrical fault Open 7 m 0.6 m / s Mechanical fault Close 18 m 1.0 m / s Control system fault

[0131] The introduction of these simulated data helps the deep learning model to learn more types of fault patterns during training, improving its fault detection capability in the elevator system.

[0132] Although embodiments of the present application have been shown and described, it will be understood by those having ordinary skill in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cloud platform based safety monitoring and early warning system for villa elevators, characterized in that, The application relates to an elevator monitoring system, which comprises the following parts: An elevator monitoring unit for collecting real-time operation data of the elevator and generating monitoring data; A data processing and self-learning module for receiving the monitoring data; A cloud platform server for remotely storing the operation data of the elevator and generating elevator fault prediction information based on cloud big data analysis; A user terminal for receiving the early warning information sent by the cloud platform, which includes the abnormal state of the elevator, fault type, maintenance suggestion and maintenance timing; An intelligent environment perception module; A cross-device remote collaborative management function for remotely checking the state of the elevator equipment through the cloud platform, scheduling maintenance and sharing information with other smart devices in the family.

2. The cloud platform-based villa elevator safety monitoring and early warning system according to claim 1, characterized in that: The elevator monitoring unit comprises a sensor group for monitoring the state data of key components of the elevator, including the elevator door, the car, the drive system, the control system, the braking system and the electrical system.

3. The cloud platform-based villa elevator safety monitoring and early warning system according to claim 1, characterized in that: The intelligent environment perception module comprises humidity, temperature, vibration, noise and air quality sensors.

4. A cloud platform-based villa elevator safety monitoring and early warning method, characterized in that, The application comprises the following steps: S1. Collecting real-time operation data and fault data of the elevator by using the elevator monitoring unit, wherein the operation data includes the elevator door, the car, the drive system, the control system, the braking system and the electrical system, and then generating original monitoring data; S2. Inputting the collected elevator operation data and fault data into a generative adversarial network model, generating simulated fault data similar to the actual fault through a generator, and judging the generated data through a discriminator to optimize the adversarial network model training; S3. Training a deep learning model, including a convolutional neural network and a recurrent neural network, using the enhanced training data, and introducing a meta-learning technology to enable the model to quickly adapt and make predictions when facing new fault patterns or unknown environments; S4. The cloud platform server generates personalized elevator fault early warning information and maintenance suggestions according to the fault prediction results of the deep learning model, the optimization strategy of the reinforcement learning and the real-time data of the environment perception module, and sends the early warning information to the user terminal in real time through the cloud platform.

5. The cloud platform-based villa elevator safety monitoring and early warning method according to claim 4, characterized in that: The loss function of the discriminator can be represented by the following formula: where D(x) represents the discriminant output of the discriminator for real data x, G(z) represents the simulated fault data generated by the generator, p data (x) is a real data distribution, p z (z) is a random noise distribution, the simulated fault data further comprises a fault occurrence time, a duration time and a speed pattern of the elevator operation.

6. The cloud platform-based villa elevator safety monitoring and early warning method according to claim 4, characterized in that: The learning process of the meta-learning model is optimized by the following formula: Wherein, theta is the parameter of the model, alpha is the learning rate, L(theta) is the loss function, and theta' is the model parameter updated by a small amount of data.

7. The cloud platform-based villa elevator safety monitoring and early warning method according to claim 4, characterized in that: The personalized elevator fault early warning information includes fault type, fault occurrence time prediction, component damage prediction and recommended maintenance priority.

8. The cloud platform-based villa elevator safety monitoring and early warning method according to claim 4, characterized in that: The reinforcement learning algorithm optimizes the elevator operation strategy by using the Q-learning method, and the Q value updating rule in the Q-learning is as follows: where Q(s t ,a t ) is the Q-value of taking action a t in state s t , r t+1 is the reward obtained in the next state s t+1 , γ is the discount factor, and α is the learning rate.

Images