Elevator vibration reduction system based on active electromagnetic damping and elevator vibration reduction control method

Abstract

The invention provides an elevator vibration reduction system based on active electromagnetic damping and an elevator vibration reduction control method, and relates to the technical field of elevator vibration reduction, the elevator vibration reduction system comprises a vibration sensor, an electromagnetic damping vibration absorber and a control module; the vibration sensor is used for collecting vibration signals in each axial direction in the movement process of the elevator; the electromagnetic damping shock absorber is used for receiving the current control signal sent by the control module and adjusting the damping force provided for the elevator according to the current control signal; the control module is used for receiving a vibration signal sent by the vibration sensor, generating multi-dimensional time sequence data based on the vibration signal, arranging the multi-dimensional time sequence data into an input tensor according to a time window, inputting the input tensor into the composite neural network structure, and transmitting an output feature of the composite neural network structure as a state representation to the reinforcement learning model; and sending a current control signal to the electromagnetic damping shock absorber based on the output characteristics of the reinforcement learning model. According to the elevator vibration reduction system, the damping response can be flexibly adjusted according to the elevator running state, and elevator vibration is effectively restrained.

Description

Technical Field

[0001] This application relates to the field of elevator vibration reduction technology, and in particular to an elevator vibration reduction system and elevator vibration reduction control method based on active electromagnetic damping. Background Technology

[0002] With the acceleration of urbanization and the continuous emergence of high-rise buildings, traction elevators, as an indispensable vertical transportation tool in modern buildings, have received widespread attention for their operational safety and comfort. During long-term operation, elevator systems can experience vibrations due to various factors. These vibrations not only affect passenger experience but may also pose potential threats to the structural safety of the elevator.

[0003] Most current vibration reduction methods involve placing vibration damping pads under the main beam of the drive unit to control elevator vibration. However, these passive vibration reduction methods often suffer from problems such as response lag and limited vibration reduction effect when facing complex and variable vibration environments, making it difficult to effectively suppress vibrations of different frequencies and amplitudes. Summary of the Invention

[0004] In view of this, this application proposes an elevator vibration reduction system and elevator vibration reduction control method based on active electromagnetic damping.

[0005] In one aspect, this application provides an elevator vibration reduction system based on active electromagnetic damping, including: a vibration sensor, an electromagnetic damping vibration absorber, and a control module;

[0006] The vibration sensor is used to collect vibration signals along each axis during the elevator's movement;

[0007] The electromagnetic damping vibration damper is used to receive the current control signal sent by the control module and adjust the damping force provided to the elevator according to the current control signal.

[0008] The control module is used to receive vibration signals sent by the vibration sensor, generate multidimensional time series data based on the vibration signals, and organize them into an input tensor input composite neural network structure according to time windows. The output features of the composite neural network structure are used as state representations and passed to the reinforcement learning model. Based on the output features of the reinforcement learning model, the current control signal is sent to the electromagnetic damping vibration absorber. The composite neural network structure is used to extract short-term dynamic changes and long-term dependencies from the multidimensional time series data.

[0009] In one embodiment, the composite neural network structure includes: a GRU subnetwork, an LSTM subnetwork, and a pooling layer;

[0010] The GRU subnetwork is used to encode the input tensor and capture short-term dynamic changes; the LSTM subnetwork is used to receive the output features of the GRU subnetwork and further process the temporal context based on the received features to learn long-term dependencies; the pooling layer is used to receive the output features of the LSTM subnetwork, pool the output features at each time step, and obtain a fixed-length high-dimensional vector as the output feature of the composite neural network structure.

[0011] In one embodiment, the reinforcement learning model is a D3QN model based on the D3QN algorithm; the control module is further configured to receive new vibration signals during elevator operation, input the new vibration signals into the composite neural network structure, use the output features as the current state, determine the current damper current as the current action and the TD error of the current decision based on the current state, and solve for the reward parameters and predicted state parameters at the current time step; the state parameters, action parameters, reward parameters and predicted state parameters at the same time step are stored in the experience replay area in a four-tuple storage structure, and the data in the experience replay area is used as training samples to train the D3QN model, wherein the TD error is used to represent the difference between the expected long-term reward and the actual observed reward of the model under the current state-action pair.

[0012] In one embodiment, the reward function of the D3QN model is:

[0013]

[0014] in, Indicates the current vibration amplitude. This indicates the amplitude of the vibration at the previous moment. Indicates a time step. Indicates the standard deviation of the vibration signal. Indicates the mean value of the vibration signal. Represented as a positive minimal number. and This indicates an adjustable parameter.

[0015] In one embodiment, the electromagnetic damping shock absorber includes: an electromagnetic coil, a permanent magnet column, and a housing;

[0016] The electromagnetic coil is located inside the housing, and the housing is provided with a channel for the displacement of the permanent magnet column;

[0017] The electromagnetic coil is used to adjust the current intensity in the electromagnetic coil according to the current control signal, so as to change the generated magnetic field intensity and adjust the displacement of the permanent magnet column relative to the electromagnetic coil.

[0018] The permanent magnet column is used to provide damping force to the elevator.

[0019] In one embodiment, the number of electromagnetic damping vibration dampers is four. The electromagnetic damping vibration dampers are located on the side of the drive host steel beam away from the drive host. The four electromagnetic damping vibration dampers support the edge portion of the drive host steel beam respectively and are arranged in a quadrilateral shape.

[0020] Secondly, this application also provides an elevator vibration reduction control method, which is executed based on the elevator vibration reduction system described in the first aspect, and the elevator vibration reduction control method includes:

[0021] Receive the vibration signal sent by the vibration sensor;

[0022] Multidimensional time series data is generated based on the vibration signal and organized into an input tensor input composite neural network structure according to the time window.

[0023] The output features of the composite neural network structure are passed as a state representation to the reinforcement learning model. Based on the output features of the reinforcement learning model, the current control signal is sent to the electromagnetic damping vibration damper. The composite neural network structure is used to extract short-term dynamic changes and long-term dependencies from multi-dimensional time series data.

[0024] In one embodiment, the reinforcement learning model is a D3QN model based on the D3QN algorithm; the elevator vibration reduction control method further includes:

[0025] During elevator operation, new vibration signals are received and input into the composite neural network structure.

[0026] The output features of the composite neural network structure are used as the current state. Based on the current state, the current damper current and the TD error of the current decision are determined as the current action. The reward parameters and predicted state parameters at the current time step are solved.

[0027] The state parameters, action parameters, reward parameters, and predicted state parameters at the same time step are stored in the experience replay area in a four-tuple storage structure. The data in the experience replay area is used as training samples to train the D3QN model.

[0028] Thirdly, this application also provides an electronic device, including a processor and a memory; the memory has a computer program stored thereon, wherein the computer program, when executed by the processor, implements the elevator vibration reduction control method as described in the second aspect.

[0029] Fourthly, this application also provides a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the elevator vibration reduction control method as described in the second aspect.

[0030] The elevator vibration reduction system based on active electromagnetic damping proposed in this application has the following advantages over related technologies:

[0031] 1. The elevator vibration reduction system of this application includes a vibration sensor, an electromagnetic damping vibration isolator, and a control module. The vibration sensor collects vibration signals during elevator movement in real time and transmits these signals to the control module. The control module generates multi-dimensional time-series data based on the vibration signals and organizes it into an input tensor input composite neural network structure according to time windows. The output features of the composite neural network structure are used as a state representation and transmitted to a reinforcement learning model. Based on the output features of the reinforcement learning model, a current control signal is sent to the electromagnetic damping vibration isolator. The electromagnetic damping vibration isolator can adjust the damping force provided to the elevator according to the current control signal, thereby achieving dynamic adjustment of the damping force provided to the elevator. This mechanism enables the electromagnetic damping vibration isolator to flexibly adjust its damping response according to the elevator's operating state, actively suppressing vibrations of different frequencies and amplitudes while balancing comfort and safety.

[0032] 2. Vibration signals exhibit not only short-term, sudden changes but also long-term, periodic variations. Therefore, a composite neural network structure can extract short-term dynamic changes and long-term dependencies from multi-dimensional time-series data, effectively improving the system's ability to represent complex vibration modes. The output of the composite neural network structure, as a state representation, is passed to the reinforcement learning model, enabling more stable and accurate assessment of motion value and optimization of current control strategies. During elevator operation, the reinforcement learning model can learn from the relationship between vibration signals and the current intensity applied by the electromagnetic damper, gradually optimizing the control strategy. Through a feedback mechanism, the reinforcement learning model continuously adjusts the current intensity to precisely regulate the damping coefficient, reducing vibration in real time and improving the vibration reduction effect. This allows the elevator vibration reduction system to adaptively adjust the control strategy according to the elevator's real-time operating state, achieving efficient vibration reduction control during elevator operation and improving elevator comfort and safety. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of the structure of an active electromagnetic damping elevator vibration reduction system in one embodiment of this application;

[0035] Figure 2This is a schematic diagram of the D3QN model in one embodiment of this application;

[0036] Figure 3 This is a schematic diagram of the structure of an electromagnetic damping vibration absorber in one embodiment of this application;

[0037] Figure 4 This is a schematic diagram of the application environment of an electromagnetic damping vibration reducer in one embodiment of this application;

[0038] Figure 5 This is a flowchart illustrating an elevator vibration reduction control method in one embodiment of this application.

[0039] Explanation of reference numerals in the attached figures:

[0040] 10-Elevator vibration reduction system based on active electromagnetic damping, 11-Vibration sensor, 12-Electromagnetic damping vibration absorber, 121-Electromagnetic coil, 122-Permanent magnet column, 123-Shell, 13-Control module, 21-Drive host, 22-Drive host steel beam. Detailed Implementation

[0041] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0042] In some embodiments, such as Figure 1 As shown, the elevator vibration reduction system 10 based on active electromagnetic damping provided in this application includes: a vibration sensor 11, an electromagnetic damping vibration absorber 12, and a control module 13.

[0043] Vibration sensor 11 is used to collect vibration signals in each axis during elevator movement.

[0044] The electromagnetic damping shock absorber 12 is used to receive the current control signal sent by the control module 13 and adjust the damping force provided to the elevator according to the current control signal.

[0045] The control module 13 is used to receive the vibration signal sent by the vibration sensor 11, generate multi-dimensional time series data based on the vibration signal, and organize it into an input tensor input composite neural network structure according to the time window. The output features of the composite neural network structure are used as state representations and transmitted to the reinforcement learning model. Based on the output features of the reinforcement learning model, a current control signal is sent to the electromagnetic damping vibration absorber 12. The composite neural network structure is used to extract short-term dynamic changes and long-term dependencies from the multi-dimensional time series data.

[0046] It is understandable that the core principle of the electromagnetic damping vibration damper 12 is based on the principle of electromagnetic induction. During operation, the vibration sensor 11 monitors the elevator's vibration signals in real time and transmits these signals to the control module 13. Based on the elevator's operating conditions, the control module 13 sends a current control signal to adjust the damping force provided by the electromagnetic damping vibration damper 12. Through this current regulation mechanism, the electromagnetic damping vibration damper 12 can dynamically adjust the damping force according to real-time vibration conditions. For example, when the elevator experiences significant vibration during startup, acceleration, or braking, the damper automatically adjusts to a larger damping force to effectively absorb vibration energy; when the elevator is running at a stable speed, the damping force is appropriately reduced to ensure comfortable and efficient operation. Because it can precisely adjust according to the actual operating conditions of the elevator, the electromagnetic damping vibration damper 12 significantly improves the elevator's stability and safety, avoiding mechanical damage or decreased comfort caused by excessive vibration.

[0047] In applications, vibration signals exhibit not only short-term sudden changes but also long-term periodic variations. Therefore, a composite neural network structure is needed to extract meaningful dynamic features from multidimensional time-series data reflecting elevator vibration. This involves extracting short-term dynamic changes and long-term dependencies from the multidimensional time-series data. The features extracted by the composite neural network structure can accurately reflect the current vibration state, capture short-term abrupt changes and long-term trends, and predict future vibration behavior, effectively improving the system's ability to represent complex vibration modes. The features output by the composite neural network structure are closely related to the control strategy. Passing the features output by the composite neural network structure as a state representation to the reinforcement learning model can influence the selection of the current control signal, achieving more stable and accurate motion value assessment and current control strategy optimization, thereby optimizing the elevator vibration reduction process.

[0048] During elevator operation, the reinforcement learning model can learn from the relationship between vibration signals and the current intensity applied by the electromagnetic damping damper 12, and gradually optimize the control strategy. Through a feedback mechanism, the reinforcement learning model continuously adjusts the current intensity to precisely regulate the damping coefficient, reducing vibration in real time and improving the damping effect. This allows the elevator vibration reduction system to adaptively adjust the control strategy according to the elevator's real-time operating status, achieving efficient vibration reduction control during elevator operation and improving elevator comfort and safety.

[0049] The real-time adjustment function of the electromagnetic damping vibration isolator 12 can also optimize the energy-saving effect of the elevator. When the elevator is under low load or in a relatively stable state, the reduction of damping force helps to reduce energy consumption, while when the elevator is under heavy load or running unstablely, the increase of damping force can effectively alleviate unnecessary energy waste. Therefore, the electromagnetic damping vibration isolator 12 not only improves the smoothness of elevator operation, but also provides support for the energy-saving management of the elevator.

[0050] The aforementioned elevator vibration reduction system includes a vibration sensor 11, an electromagnetic damping vibration isolator 12, and a control module 13. The vibration sensor 11 collects vibration signals during elevator movement in real time and transmits these signals to the control module 13. The control module 13 generates multi-dimensional time-series data based on the vibration signals and organizes it into an input tensor input composite neural network structure according to time windows. The output features of the composite neural network structure are then used as a state representation and passed to a reinforcement learning model. Based on the output features of the reinforcement learning model, a current control signal is sent to the electromagnetic damping vibration isolator 12. The electromagnetic damping vibration isolator 12 can adjust the damping force provided to the elevator according to the current control signal, thereby achieving dynamic adjustment of the damping force provided to the elevator. This mechanism enables the electromagnetic damping vibration isolator 12 to flexibly adjust its damping response according to the elevator's operating state, actively suppressing vibrations of different frequencies and amplitudes while balancing comfort and safety.

[0051] In some embodiments, the composite neural network structure includes: a GRU subnetwork, an LSTM subnetwork, and a pooling layer.

[0052] The GRU subnetwork is used to encode the input tensor and capture short-term dynamic changes; the LSTM subnetwork is used to receive the output features of the GRU subnetwork and further process the temporal context based on the received features to learn long-term dependencies; the pooling layer is used to receive the output features of the LSTM subnetwork, pool the output features at each time step, and obtain a fixed-length high-dimensional vector as the output feature of the composite neural network structure.

[0053] In this application, the vibration signal collected by the vibration sensor 11 is used as input and transmitted to a composite neural network structure. Meaningful dynamic features are extracted from the time-series data of elevator vibration. These meaningful dynamic features refer to the high-dimensional state representation automatically extracted from the elevator vibration time-series signal by the composite neural network. In this composite neural network, a GRU subnetwork is used to extract short-term dynamic changes, and an LSTM subnetwork is used to extract long-term dependencies. The output features are then transmitted to a reinforcement learning model, which can output the action-value function of the current state. The specific process is as follows:

[0054] First, the multidimensional time series data collected during elevator operation is organized into an input tensor according to time windows:

[0055]

[0056] Each of them It is an observation of the vibration state at a single frame.

[0057] The tensor is then input into the GRU (Gated Recurrent Unit) subnetwork. The GRU subnetwork first encodes the sequence data, focusing on capturing dynamic changes over a short period of time, such as local fluctuations and abrupt changes. The corresponding formula is as follows:

[0058]

[0059] This section can identify phenomena such as impacts, micro-vibrations, and sudden increases in speed.

[0060] The output of the GRU subnetwork serves as the input to the LSTM subnetwork. The LSTM subnetwork further processes the temporal context, learning long-term dependencies such as accumulated oscillation trends and slow drift, as shown in the following formula:

[0061]

[0062] Ultimately, this yields the ability to dynamically model the entire time window globally.

[0063] Finally, the pooling layer receives the output features from the LSTM subnetwork and pools the outputs from all time steps (e.g., average pooling or attention-weighted pooling) to obtain a high-dimensional vector of fixed length:

[0064]

[0065] The vector This refers to the meaningful dynamic features represented within the current time window.

[0066] It is understandable that vibration signals exhibit both short-term, sudden changes (such as rapid vibrations during elevator startup) and long-term, periodic changes (such as the continuous operation of elevators in high-rise buildings). The LSTM sub-network is responsible for extracting long-term dependencies in elevator vibration signals, such as the long-term operating trends and periodic vibration patterns of the elevator under different speed and load conditions; the GRU sub-network processes short-term dynamic changes in elevator vibration data, such as brief vibration variations, especially when the elevator accelerates, decelerates, or suddenly encounters external forces. The advantage of the LSTM-GRU composite neural network structure lies in its combination of the long short-term memory advantages of both, enabling it to capture both the long-term patterns of elevator vibration and respond promptly to short-term sudden events. This system introduces the LSTM-GRU composite neural network structure into the reinforcement learning process as a temporal modeling and feature extractor for elevator vibration states. This composite neural network structure effectively improves the system's ability to represent complex vibration patterns by integrating the long-term dependency capture capability of LSTM with the sensitivity of GRU to short-term dynamic changes. Compared with traditional state modeling methods, this composite network structure significantly enhances the adaptability and robustness of the reinforcement learning model to various elevator operation scenarios.

[0067] In some embodiments, the reinforcement learning model is a D3QN model based on the D3QN algorithm; the control module 13 is also used to receive new vibration signals during elevator operation, input the new vibration signals into the composite neural network structure, take the output features as the current state, determine the current current of the electromagnetic damper 12 as the current action and the TD error of the current decision based on the current state, and solve the reward parameters and predicted state parameters at the current time step; the state parameters, action parameters, reward parameters and predicted state parameters at the same time step are stored in the experience replay area in the form of a quadruple storage structure, and the data in the experience replay area is used as training samples to train the D3QN model, wherein the TD error is used to represent the difference between the expected long-term reward and the actual observed reward of the model under the current state-action pair.

[0068] The reinforcement learning algorithm employs the D3QN architecture, which can distinguish between state value and action advantage, effectively suppressing Q-value overestimation. The Q-value measures the long-term reward obtained by taking a specific action in a given state. Specifically, the Q-value is an estimate of the future reward of the state-action pair, reflecting the total reward the agent expects to obtain in the long run after taking that action.

[0069] Then, combining the current state with historical experience, the system selects the action to apply the current intensity based on an ε-greedy strategy. (i.e., the aforementioned) The system then controls the electromagnetic damping vibration absorber 12 to apply the corresponding damping force. Subsequently, the elevator continues to operate, and the system receives new vibration feedback in real time. and status The TD error of the current decision is calculated, which measures the difference between the model's expected long-term reward and the actual observed reward in the current state-action pair. It is the difference between the target value and the current Q-value. The formula is as follows:

[0070]

[0071] in, The TD error represents the degree of error in the current decision. This represents the immediate reward at the current moment. This is a discount factor used to control the impact of future rewards; Indicates the next state Next, select the action that will bring the maximum Q value. , Indicates the current state Next, execute the action. The Q value.

[0072] All experience data (state, action, reward, next state) will be stored as samples in the Priority Experience Replay Pool (PER) for priority sampling during subsequent training.

[0073] The system employs a dual-network mechanism, utilizing the main network to select actions and the target network to evaluate Q-values, thereby improving learning stability. The reinforcement learning model continuously adjusts parameters in each iteration to reduce the error between Q-value predictions and actual returns, gradually optimizing the damping control strategy. As training progresses, the system adaptively establishes a mapping relationship between "vibration characteristics—current control—damping effect," automatically applying the optimal current under different operating conditions and adjusting the damping coefficient in real time, ultimately achieving intelligent suppression of elevator vibration and continuously optimizing the current adjustment strategy of the electromagnetic damping vibration absorber 12. Through real-time learning and adjustment, the system can automatically apply the optimal current according to different elevator operating conditions, intelligently controlling the elevator's vibration suppression effect and improving elevator comfort and safety.

[0074] The core of this method is to continuously optimize the relationship between state (vibration characteristics) and action (current control) through reinforcement learning, so as to achieve the best vibration reduction effect, making the elevator run more smoothly and the passenger experience more comfortable.

[0075] This process not only improves vibration reduction efficiency during elevator operation but also enhances its self-learning capabilities and environmental adaptability, making it widely applicable to various building elevators, freight elevators, and high-speed lifting platforms. See details for the specific structure. Figure 2 The input is the vibration characteristics of the current system. And the operation of adjusting the current of the electromagnetic damping shock absorber 12 The output is the elevator's vibration feedback reward. The vibration state of the elevator in the next moment And the action of adjusting the current according to this state.

[0076] In some embodiments, the reward function of the D3QN model is:

[0077]

[0078] in, Indicates the current vibration amplitude. This indicates the amplitude of the vibration at the previous moment. Indicates a time step. Indicates the standard deviation of the vibration signal. Indicates the mean value of the vibration signal. Represented as a positive minimal number. and This indicates an adjustable parameter.

[0079] The aforementioned reward function comprehensively considers the vibration improvement trend and overall operational stability during elevator operation. It employs a design combining a dynamic trend term based on vibration decay rate with an exponential stability score. This approach not only reflects in real-time whether the current control action effectively reduces vibration but also identifies whether the system operates smoothly and reliably. This function not only enhances the control strategy's ability to identify sudden vibrations and irregular fluctuations but also effectively guides the agent to avoid frequent and large-scale adjustments while achieving vibration reduction goals. This balances comfort, energy efficiency, and equipment protection, significantly improving the system's adaptability and control robustness.

[0080] In some embodiments, such as Figure 3 As shown, the electromagnetic damping shock absorber 12 includes: an electromagnetic coil 121, a permanent magnet column 122, and a housing 123.

[0081] An electromagnetic coil 121 is located inside a housing 123, which has a channel for displacing the permanent magnet column 122. The electromagnetic coil 121 is used to adjust the current intensity within it according to a current control signal, thereby changing the generated magnetic field strength and adjusting the displacement of the permanent magnet column 122 relative to the electromagnetic coil 121. The permanent magnet column 122 is used to provide damping force to the elevator.

[0082] The permanent magnet column 122 is one of the core components of the electromagnetic damping vibration damper 12. Made of high-performance permanent magnetic material, it can stably generate a magnetic field without external power supply. The magnetic field strength and distribution of the permanent magnet column 122 are key factors affecting the magnitude and adjustment accuracy of the damping force. To ensure long-term stable operation, the permanent magnet column 122 uses high-temperature, highly antimagnetic materials that maintain their magnetism over a wide temperature range, adapting to the high-frequency operation requirements of the elevator system. The electromagnetic coil 121, another core component, generates a varying magnetic field by adjusting the current, thereby controlling the movement of the permanent magnet column 122. The magnitude of the coil current directly affects the magnetic field strength, thus precisely controlling the damping force. The coil is typically made of highly conductive material and wound to ensure good conductivity and current carrying capacity. When current passes through the electromagnetic coil 121, the generated magnetic field strength directly affects the displacement of the permanent magnet column 122, thereby adjusting the system's damping response. The displacement of the permanent magnet column 122 significantly affects the damping force because its specific position within the conductive structure determines the coupling range between the magnetic field and the conductor. When the permanent magnet column 122 extends deeper into the sensing area, the magnetic flux changes drastically, generating stronger eddy currents in the conductor, thus producing a greater reaction force, i.e., the damping force. Conversely, when the permanent magnet column 122 moves away from the sensing area, the magnetic flux change weakens, and the eddy currents and damping force decrease accordingly. By adjusting the current in the electromagnetic coil 121, the position of the permanent magnet column 122 can be precisely controlled, thereby achieving dynamic adjustment of the eddy current intensity and damping force. This mechanism enables the electromagnetic damping vibration damper 12 to flexibly adjust its damping response according to the elevator's operating status, balancing comfort and safety. The housing 123 is a crucial outer shell component of the entire electromagnetic damping vibration damper 12, primarily used to protect internal sensitive components and provide overall mechanical strength. The housing 123 of the electromagnetic damping shock absorber 12 is typically made of a wear-resistant metal alloy material. These materials have good corrosion resistance, friction resistance, and high strength, which can effectively prevent wear or corrosion during long-term elevator operation. The design of the housing 123 not only ensures the stability of the component but also has good heat dissipation performance to prevent malfunctions caused by overheating.

[0083] In the application, based on the elevator's operating status, the control module 13 sends a current control signal to adjust the current intensity in the electromagnetic coil 121, thereby changing the strength of the magnetic field and the movement amplitude of the permanent magnet column 122. Specifically, the change in current intensity determines the strength of the magnetic field, and thus the magnitude of the damping force. When the current increases, the magnetic field strength increases, the movement amplitude of the permanent magnet column 122 increases, providing a stronger damping force, and vice versa. The theoretical formula is as follows:

[0084]

[0085] in, Indicates electromagnetic damping force. Represents structural constants. Indicates magnetic field strength. This indicates relative velocity.

[0086] In some embodiments, such as Figure 4 As shown, there are four electromagnetic damping vibration dampers 12. The electromagnetic damping vibration dampers 12 are located on the side of the drive host steel beam 22 away from the drive host 21. The four electromagnetic damping vibration dampers 12 support the edge part of the drive host steel beam 22 respectively and are arranged in a quadrilateral.

[0087] It is understandable that the configuration of four electromagnetic damping vibration dampers 12 is chosen to precisely match the force distribution requirements of the drive host steel beam 22's edge, ensuring support stability. Furthermore, by arranging the four dampers to support the four edges of the drive host steel beam 22 in a quadrilateral pattern, a uniform support surface is formed, preventing deformation of the steel beam due to localized vibration stress concentration.

[0088] To verify that the electromagnetic damping vibration damper 12 can support the drive host 21 and that the natural frequency of the electromagnetic damping vibration damper 12 will not resonate with the vibration of the elevator, the electromagnetic force generated by the electromagnetic damping vibration damper 12 and its natural frequency are calculated below:

[0089] The parameters of the electromagnetic damping vibration absorber 12 are as follows: permanent magnet column 122, diameter 12cm, area... Approximately 0.0177m 2 N52SH permanent magnet remanence It is 1.4T, and the number of coil turns is... The maximum current is 1500. 20A, magnetic levitation air gap The permanent magnet column is 5mm thick and 122mm high. The length is 20cm. The overall mass m is 17kg.

[0090] Calculate the maximum magnetic force provided by the electromagnetic damping shock absorber 12:

[0091]

[0092] Assuming the elevator's rated load capacity is 1000 kg, then the weight of a fully loaded elevator is approximately 2300 kg. Therefore, the full load weight is:

[0093]

[0094] The four electromagnetic damping shock absorbers 12 can provide a maximum force of 152.8KN, which is enough to support the weight of the entire elevator.

[0095] Natural frequency calculation:

[0096]

[0097]

[0098] The natural frequency of an elevator is approximately 1. ~10 Therefore, the device will not resonate with the elevator.

[0099] In summary, based on theoretical calculations, this device can not only support the weight of the elevator, but also will not resonate with the elevator.

[0100] Based on the same inventive concept, this application also provides an elevator vibration reduction control method, which is implemented based on the elevator vibration reduction system of any of the above schemes, such as... Figure 5 As shown, the elevator vibration reduction control method includes the following steps S501 to S503.

[0101] S501: Receives vibration signals sent by the vibration sensor.

[0102] S502: Generates multidimensional time series data based on vibration signals and organizes it into an input tensor input composite neural network structure according to time windows.

[0103] S503: The output features of the composite neural network structure are passed as state representations to the reinforcement learning model. Based on the output features of the reinforcement learning model, current control signals are sent to the electromagnetic damping vibration damper. The composite neural network structure is used to extract short-term dynamic changes and long-term dependencies from multi-dimensional time series data.

[0104] In some embodiments, the reinforcement learning model is a D3QN model based on the D3QN algorithm; the elevator vibration reduction control method further includes: receiving new vibration signals during elevator operation, inputting the new vibration signals into a composite neural network structure; using the output features of the composite neural network structure as the current state, determining the current damper current as the current action and the TD error of the current decision based on the current state, and solving for the reward parameters and predicted state parameters at the current time step; storing the state parameters, action parameters, reward parameters, and predicted state parameters at the same time step in an experience replay area in the form of a four-tuple storage structure, and using the data in the experience replay area as training samples to train the D3QN model.

[0105] It should be noted that the elevator vibration reduction control method provided in this application embodiment and the elevator vibration reduction control system provided in this application embodiment are based on the same inventive concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned elevator vibration reduction control system, and the repeated parts will not be described again.

[0106] In some embodiments, an electronic device provided in this application includes a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the elevator vibration reduction control method described above.

[0107] Specifically, the processor may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may be a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of this application.

[0108] Memory can be any medium capable of containing, storing, transmitting, propagating, or transmitting instructions. For example, memory can include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, instruments, or propagation media. Specific examples of memory include: magnetic storage devices such as magnetic tape or hard disk drives (HDDs); optical storage devices such as optical discs (CD-ROMs); and also random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0109] This application also provides a computer-readable medium storing a computer program thereon, which, when executed by a processor, implements the elevator vibration reduction control method described above. This computer-readable medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into that device / apparatus / system. The aforementioned computer-readable medium carries one or more programs, which, when executed, implement the method as described in the embodiments of this application.

[0110] According to embodiments of this application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wired, optical fiber, radio frequency signals, etc., or any suitable combination thereof.

[0111] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application. Therefore, the scope of this application should not be limited to the above embodiments. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An elevator vibration reduction system based on active electromagnetic damping, characterized in that, include: Vibration sensors, electromagnetic damping shock absorbers, and control modules; The vibration sensor is used to collect vibration signals along each axis during the elevator's movement; The electromagnetic damping vibration damper is used to receive the current control signal sent by the control module and adjust the damping force provided to the elevator according to the current control signal. The control module is used to receive vibration signals sent by the vibration sensor, generate multidimensional time series data based on the vibration signals, and organize them into an input tensor input composite neural network structure according to time windows. The output features of the composite neural network structure are used as state representations and passed to the reinforcement learning model. Based on the output features of the reinforcement learning model, the current control signal is sent to the electromagnetic damping vibration absorber. The composite neural network structure is used to extract short-term dynamic changes and long-term dependencies from the multidimensional time series data.

2. The elevator vibration reduction system based on active electromagnetic damping as described in claim 1, characterized in that, The composite neural network structure includes: a GRU subnetwork, an LSTM subnetwork, and a pooling layer; The GRU subnetwork is used to encode the input tensor and capture short-term dynamic changes; the LSTM subnetwork is used to receive the output features of the GRU subnetwork and further process the temporal context based on the received features to learn long-term dependencies; the pooling layer is used to receive the output features of the LSTM subnetwork, pool the output features at each time step, and obtain a fixed-length high-dimensional vector as the output feature of the composite neural network structure.

3. The elevator vibration reduction system based on active electromagnetic damping as described in claim 1, characterized in that, The reinforcement learning model is a D3QN model based on the D3QN algorithm. The control module is also used to receive new vibration signals during elevator operation, input the new vibration signals into the composite neural network structure, use the output features as the current state, determine the current damper current as the current action and the TD error of the current decision based on the current state, and solve for the reward parameters and predicted state parameters at the current time step. The state parameters, action parameters, reward parameters and predicted state parameters at the same time step are stored in the experience replay area in the form of a four-tuple storage structure. The data in the experience replay area is used as training samples to train the D3QN model. The TD error is used to represent the difference between the expected long-term reward and the actual observed reward of the model under the current state-action pair.

4. The elevator vibration reduction system based on active electromagnetic damping as described in claim 3, characterized in that, The reward function of the D3QN model is: in, Indicates the current vibration amplitude. This indicates the amplitude of the vibration at the previous moment. Indicates a time step. Indicates the standard deviation of the vibration signal. Indicates the mean value of the vibration signal. Represented as a positive minimal number. and This indicates an adjustable parameter.

5. The elevator vibration reduction system based on active electromagnetic damping as described in claim 1, characterized in that, The electromagnetic damping vibration absorber includes: an electromagnetic coil, a permanent magnet column, and a housing; The electromagnetic coil is located inside the housing, and the housing is provided with a channel for the displacement of the permanent magnet column; The electromagnetic coil is used to adjust the current intensity in the electromagnetic coil according to the current control signal, so as to change the generated magnetic field intensity and adjust the displacement of the permanent magnet column relative to the electromagnetic coil. The permanent magnet column is used to provide damping force to the elevator.

6. The elevator vibration reduction system based on active electromagnetic damping as described in any one of claims 1 to 5, characterized in that, The number of electromagnetic damping vibration dampers is four. The electromagnetic damping vibration dampers are located on the side of the drive host steel beam away from the drive host. The four electromagnetic damping vibration dampers support the edge part of the drive host steel beam respectively and are arranged in a quadrilateral.

7. A method for controlling elevator vibration reduction, characterized in that, The elevator vibration reduction control method is performed based on the elevator vibration reduction system as described in any one of claims 1 to 6, and the elevator vibration reduction control method includes: Receive the vibration signal sent by the vibration sensor; Multidimensional time series data is generated based on the vibration signal and organized into an input tensor input composite neural network structure according to the time window. The output features of the composite neural network structure are passed as a state representation to the reinforcement learning model. Based on the output features of the reinforcement learning model, the current control signal is sent to the electromagnetic damping vibration damper. The composite neural network structure is used to extract short-term dynamic changes and long-term dependencies from multi-dimensional time series data.

8. The elevator vibration reduction control method as described in claim 7, characterized in that, The reinforcement learning model is a D3QN model based on the D3QN algorithm; the elevator vibration reduction control method further includes: During elevator operation, new vibration signals are received and input into the composite neural network structure. The output features of the composite neural network structure are used as the current state. Based on the current state, the current damper current and the TD error of the current decision are determined as the current action. The reward parameters and predicted state parameters at the current time step are solved. The state parameters, action parameters, reward parameters, and predicted state parameters at the same time step are stored in the experience replay area in a four-tuple storage structure. The data in the experience replay area is used as training samples to train the D3QN model.

9. An electronic device, characterized in that, It includes a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the elevator vibration reduction control method as described in claim 7 or 8.

10. A computer storage medium, characterized in that, It stores a computer program, wherein the computer program, when executed by a processor, implements the elevator vibration reduction control method as described in claim 7 or 8.

Images