Multi-mode self-calibration method for elevator weighing early warning

Abstract

The invention relates to the technical field of elevator safety control, and provides a multi-mode self-calibration method for elevator weighing early warning, which comprises the following steps: acquiring a zero-load height value when an elevator is in a no-load shutdown state and a full-load height value when the elevator is in a full-load state through a pull rope type sensor; acquiring a zero load height mean value of the plurality of zero load height values in a preset time period, and judging whether the car bottom rubber is aged or not according to the offset between the zero load height mean value and the zero load height values; and if the car bottom rubber is aged, updating the old zero-load height value by using the newest zero-load height mean value, and compensating and calibrating the full-load height value according to the offset to obtain the newest full-load height value. According to the invention, measurement error accumulation caused by long-term use can be effectively reduced, the long-term stability and adaptability of the system are improved, the maintenance cost is reduced, and the potential safety hazards of falling and tool slipping of operators during well operation are avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator safety control, and in particular to a multi-modal self-calibration method for elevator weighing early warning. Background Art

[0002] Traditional elevator weighing systems rely on mechanical pressure sensors at the bottom of the car for load detection. This requires frequent elevator stops and the deployment of personnel into the hoistway or confined spaces above the car roof for calibration and maintenance. These mechanically sensing solutions are susceptible to mechanical creep, temperature drift, and uneven wire rope tension. Limited by sensor accuracy, mounting location, and environmental interference, these solutions can suffer from error accumulation and zero-point drift, making it difficult to achieve long-term, stable, and high-precision weighing warnings.

[0003] Therefore, it is necessary to develop a multimodal self-calibration method for elevator weighing warning to achieve long-term stable and high-precision elevator weighing warning and reduce maintenance costs. Summary of the Invention

[0004] Based on this, in order to solve the problem that traditional elevator weighing systems are difficult to achieve long-term stable high-precision weighing warning, the present invention provides a multi-modal self-calibration method for elevator weighing warning, and its specific technical solution is as follows:

[0005] A multi-modal self-calibration method for elevator weighing warning includes the following steps:

[0006] A pull-rope sensor is installed between the car frame and the bottom of the car, and a zero-load height value when the elevator is in an unloaded and stopped state and a full-load height value when the elevator is in a fully loaded state are obtained by the pull-rope sensor;

[0007] Obtaining a zero-load height average of multiple zero-load height values ​​within a preset time period, and determining whether the car floor rubber has aged based on an offset between the zero-load height average and the zero-load height value;

[0008] If the car floor rubber is aged, the old zero-load height value is updated with the latest zero-load height average value, and the full-load height value is compensated and calibrated according to the offset to obtain the latest full-load height value.

[0009] Compared to traditional elevator weighing warning systems that rely on manual or periodic calibration, the multimodal self-calibration method for elevator weighing warnings employs a self-calibration mechanism. This mechanism determines whether the car floor rubber has aged based on the offset between the zero-load height mean and the zero-load height value. If the car floor rubber has aged, the old zero-load height value is updated with the latest zero-load height mean, and the full-load height value is compensated and calibrated based on the offset to obtain the latest full-load height value, compensating for errors caused by environmental factors. This mechanism effectively reduces the accumulation of measurement errors caused by long-term use and improves the long-term stability and adaptability of the system. It eliminates the need for frequent elevator stops and the deployment of personnel into the hoistway or confined spaces on the car roof for calibration and maintenance, reducing maintenance costs and avoiding the safety hazards of workers falling and tools slipping while working in the hoistway.

[0010] Preferably, the multimodal self-calibration method further includes the following steps:

[0011] Obtaining a real-time load height value of the elevator through the pull-rope sensor;

[0012] Acquire a real-time height difference according to the real-time load height value and the zero-load height value;

[0013] The real-time load of the car is obtained according to the real-time height difference.

[0014] Preferably, the multimodal self-calibration method further includes the following steps:

[0015] Determine whether the real-time load height value is less than the latest full-load height value. If so, determine that the car is overloaded and start the elevator sound and light alarm.

[0016] Preferably, the specific method for obtaining the real-time load of the car according to the real-time height difference includes the following steps:

[0017] Construct a polynomial model based on the real-time load and height difference of the car;

[0018] The polynomial model is fitted, and the real-time load of the car is obtained according to the fitted polynomial model.

[0019] Preferably, the polynomial model is expressed as L=a·(ΔH) 4 +b·(ΔH) 3 +c·(ΔH) 2 +d·ΔH+e;

[0020] Wherein, L represents the real-time load of the car, ΔH represents the real-time height difference, and a, b, c, d, and e represent the coefficient of the quartic term, the coefficient of the cubic term, the coefficient of the quadratic term, the coefficient of the linear term, and the preset constant, respectively.

[0021] Preferably, according to formula H' m =H m -α·(H0-H a ) Compensate and calibrate the full load height value;

[0022] Among them, H0-H a Indicates the offset, H0, H a Represent the zero load height value and the mean zero load height, H m Indicates the full load height value before compensation calibration, H' m It represents the full load height value after compensation calibration, and α represents the compensation coefficient.

[0023] Preferably, the multimodal self-calibration method further includes the following steps:

[0024] Acquire real-time image data of the interior of the elevator car, and determine the number of passengers and whether an electric vehicle has entered the elevator car based on the real-time image data;

[0025] If the number of people taking the elevator is greater than a preset threshold or an electric car enters the elevator car, the elevator sound and light alarm is activated.

[0026] Preferably, the body of the pull-wire sensor is fixedly mounted on a central rigid support point at the bottom of the car, the pull-wire axis of the pull-wire sensor is arranged in a vertical direction and the pulling direction of the pull wire is consistent with the sinking direction of the car.

[0027] Preferably, real-time image data is obtained by installing a network camera at a diagonal position on the top of the car away from the car door, and the real-time image data is used to identify the passengers based on the YOLO target detection algorithm to obtain the number of passengers.

[0028] Preferably, when it is detected that the number of passengers is zero and the data fed back by the pull-rope sensor remains stable for a preset time value, it is determined that the elevator is in an unloaded parking state. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but rather the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0030] Figure 1 This is a schematic diagram of the overall process of a multi-modal self-calibration method for elevator weighing warning in one embodiment of the present invention;

[0031] Figure 2 This is a flow diagram of a multi-modal self-calibration method for elevator weighing warning in another embodiment of the present invention. Figure 1 ;

[0032] Figure 3 1 is a flow chart of a specific method for obtaining the real-time load of a car in one embodiment of the present invention;

[0033] Figure 4 This is a flow diagram of a multi-modal self-calibration method for elevator weighing warning in another embodiment of the present invention. Figure 2 ;

[0034] Figure 5 Schematic diagram of the overall working process of the system in one embodiment of the present invention;

[0035] Figure 6 1 is a diagram showing the relationship between the car load and the raw data of the rope sensor according to an embodiment of the present invention;

[0036] Figure 7 1 is a diagram showing the relationship between the car load and the relative height in one embodiment of the present invention;

[0037] Figure 8 This is a schematic diagram of the debugging software interface in one embodiment of the present invention;

[0038] Figure 9 1 is a schematic diagram of the recognition effect of the visual recognition module in one embodiment of the present invention;

[0039] Figure 10 1. It is a schematic diagram of the format of sending and returning codes for reading registers of a drawstring sensor encoder according to one embodiment of the present invention;

[0040] Figure 11 This is a schematic diagram of the hardware deployment of each module in one embodiment of the present invention;

[0041] Figure 12 1 is a flowchart of the self-calibration function program in one embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with its embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0043] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly attached to the other element or there may be an intermediate element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or there may be an intermediate element. The terms "vertical," "horizontal," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only implementation methods.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used herein in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0045] The "first" and "second" in the present invention do not represent specific quantities and orders, but are only used to distinguish names.

[0046] Before describing the embodiments of the present invention, the prior art is briefly introduced.

[0047] Traditional elevator weighing technology has many pain points and shortcomings, mainly manifested in the following aspects:

[0048] 1. The weighing accuracy is not enough. Due to interference from environmental factors, the sensor has poor long-term operating stability, making it difficult to accurately monitor the elevator load.

[0049] 2. Traditional sensors are susceptible to aging of the car floor rubber and require frequent manual calibration, resulting in high maintenance costs and low efficiency.

[0050] 3. There is a general lack of effective real-time passenger monitoring methods, resulting in safety hazards such as overloading and failure to promptly report the incident to the police;

[0051] 4. To ensure safety, maintenance personnel often set the overload alarm threshold too low, reducing the load capacity of the elevator and resulting in a waste of resources and energy.

[0052] 5. Load detection often relies on mechanical pressure sensors at the bottom of the car, requiring frequent stops and dispatching personnel into the hoistway or confined space on the car roof for calibration and maintenance. Manual calibration takes 2-3 hours per time, increasing the average annual maintenance cost and creating risks of falls and tool slips during hoistway operations.

[0053] like Figure 1 As shown, in order to solve the problem of difficulty in achieving long-term stable high-precision weighing warning in traditional elevator weighing systems, a multi-modal self-calibration method for elevator weighing warning in one embodiment of the present invention includes the following steps:

[0054] S1, installing a pull-rope sensor between the car frame and the bottom of the car, and obtaining the zero-load height value of the elevator when it is in an unloaded parking state and the full-load height value when it is in a fully loaded state through the pull-rope sensor.

[0055] Preferably, for the no-load stop state of the elevator, the acquisition method includes: obtaining real-time image data through a network camera installed at a diagonal position on the top of the car away from the car door, and identifying the passengers in the real-time image data based on the YOLO target detection algorithm to obtain the number of passengers.

[0056] Specifically, if Figure 11 As shown, this embodiment uses a development board as the elevator system controller. The network camera is fixed at a 25-degree or 30-degree pitch angle, covering the entire interior of the elevator car. It avoids facing the car lighting to prevent direct exposure to strong light and image overexposure. The corresponding data and signal cables are routed through reserved holes in the car roof and connected to the embedded development board and the public power supply on the car roof, respectively. The USB 3.0 port on the development board is connected to the drawstring sensor through a signal conversion module (specifically, an RS485-to-USB conversion module: a CH340 chip, bidirectional half-duplex), enabling communication between the drawstring sensor data and the calculation and control module.

[0057] The drawstring sensor is fixedly mounted on a central rigid support point at the bottom of the car. Its drawstring axis is vertically aligned, and its pull aligns with the car's downward movement. The drawstring sensor utilizes a high-precision BRT27 displacement sensor with a range of 0-650mm, a linear accuracy of ±0.05%, IP67 protection, a baud rate of 9600-115200bps, and support for the Modbus-RTU protocol. It monitors the vertical displacement of the car bottom in real time due to load changes.

[0058] The development board uses a LubanCat2 RK3568 card computer (quad-core Cortex-A55, 4GB LPDDR4), equipped with a Linux operating system and loaded with code for cable-pull sensor communication and relay control. It can receive and calculate data from cable-pull sensors and webcams and control relays. The relay, which uses an optocoupler isolation relay module, is used to control the overload alarm signal in the car-top control cabinet.

[0059] When the number of passengers detected is zero and the data fed back by the pull-wire sensor remains stable for a preset time, the elevator is determined to be in an unloaded and stopped state. For example, during normal operation of the elevator, when the number of passengers detected is N=0 and the height data fed back by the pull-wire sensor remains stable for 10 minutes, the elevator is considered to be in an unloaded and stopped state.

[0060] The YOLO object detection algorithm is deployed on the development board. The board's Gigabit Ethernet port communicates with the webcam's PoE cable to capture the elevator's internal video stream data, which is then used to obtain the real-time image data. A visual interface can be developed on the QT platform and the complete program burned into the development board, enabling local and remote operation. Users can access the QT interface via a mobile phone or tablet, viewing the elevator's operating status in real time and performing remote debugging and management.

[0061] The YOLO object detection algorithm, based on a deep convolutional neural network, is annotated and trained by collecting and analyzing a large number of datasets from similar application scenarios to optimize detection accuracy. After training on a computing server, it is applied to real-time inference of elevator internal video streams. The YOLO object detection algorithm is optimized for the computing resources on the development board and employs model quantization or pruning optimization strategies to improve inference speed and reduce computational costs.

[0062] The YOLO target detection algorithm model runs on the NPU of the development board and performs inference on the acquired video stream, identifies and counts people in the elevator, and transmits the detection results to the data processing module to determine whether the elevator is in an empty and stopped state.

[0063] S2, obtaining a zero-load height average of multiple zero-load height values ​​within a preset time period, and determining whether the car floor rubber has aged according to an offset between the zero-load height average and the zero-load height value.

[0064] S3, if the car floor rubber is aged, the old zero-load height value is updated with the latest zero-load height average value, and the full-load height value is compensated and calibrated according to the offset to obtain the latest full-load height value.

[0065] Whenever the elevator is in the no-load stop state, the system's calculation control module will convert the rope sensor data H i (i=1,2,3……) Record and save, and calculate the average value of zero load height H once a week a , if the zero load height mean H a If the offset between the zero-load height value H0 and the zero-load height value H0 is large (e.g. the offset equivalent load exceeds 50kg), the calculation control module determines that the car bottom rubber is aging and the car full-load height needs to be recalibrated. The old zero-load height value is updated with the latest zero-load height average value and the zero-load height average H0 is used. a The difference between the zero load height value H0 and the full load height value H0 (ie the offset) is multiplied by the preset compensation coefficient α to compensate and calibrate the full load height value to form a new zero load height value H'0 and a full load height value H' m , that is, according to the formula H' m =H m-α·(H0-H a ) to compensate and calibrate the full load height value. Among them, H0-H a Indicates the offset, H m Indicates the full load height value before compensation calibration, H' m Indicates the full load height value after compensation calibration.

[0066] New zero load height value n represents the sampled rope sensor data H i Then, clear the recorded H i The new zero-load height value and full-load height value are used in daily judgment of the full load situation of the car, so as to realize automatic compensation for the deviation of the full and overload alarm values ​​of the elevator.

[0067] Compared to traditional elevator weighing warning systems that rely on manual or periodic calibration, the multimodal self-calibration method for elevator weighing warnings employs a self-calibration mechanism. This mechanism determines whether the car floor rubber has aged based on the offset between the zero-load height mean and the zero-load height value. If the car floor rubber has aged, the old zero-load height value is updated with the latest zero-load height mean, and the full-load height value is compensated and calibrated based on the offset to obtain the latest full-load height value, compensating for errors caused by environmental factors. This mechanism effectively reduces the accumulation of measurement errors caused by long-term use and improves the long-term stability and adaptability of the system. It eliminates the need for frequent elevator stops and the deployment of personnel into the hoistway or confined spaces on the car roof for calibration and maintenance, reducing maintenance costs and avoiding the safety hazards of workers falling and tools slipping while working in the hoistway.

[0068] As a preferred technical solution, Figure 2 As shown, the multimodal self-calibration method further includes the following steps:

[0069] S4, obtaining a real-time load height value of the elevator through the pull-rope sensor;

[0070] S5, obtaining a real-time height difference according to the real-time load height value and the zero-load height value;

[0071] S6, obtaining the real-time load of the car according to the real-time height difference.

[0072] like Figure 3 As shown, in step S6, the specific method for obtaining the real-time load of the car according to the real-time height difference includes the following steps:

[0073] S61, constructing a polynomial model according to the real-time load and real-time height difference of the car;

[0074] S62: Fit the polynomial model and obtain the real-time car load according to the fitted polynomial model. Here, the polynomial model can be fitted based on the least square method or the Newton iteration method.

[0075] Specifically, the polynomial model is expressed as L = a·(ΔH) 4 +b·(ΔH) 3 +c·(ΔH) 2 +d·ΔH+e; where L represents the real-time load of the car, ΔH represents the real-time height difference, and a, b, c, d, and e represent the coefficient of the quartic term, the coefficient of the cubic term, the coefficient of the quadratic term, the coefficient of the linear term, and the preset constant, respectively.

[0076] like Figure 10 As shown, the rope-drawing sensor has a built-in encoder and registers and complies with the RS485 communication protocol. After the host computer (such as a development board) completes the serial port configuration (baud rate 9600bps, 8 data bits, no parity, 1 stop bit), it sends standard MODBUS-RTU codes to the encoder to read and write to the corresponding registers, thus establishing serial communication with the rope-drawing sensor. The most important step is to read the distance between the rope end and the base (i.e., the encoder value). This data is stored in the register at address 0x0000-0x0001. If the host computer sends the hexadecimal code 01 03 00 00 00 02C40B, the encoder will return 01 03 04XX XX XX XX CC 40, where bits 4-7 are the hexadecimal encoder value.

[0077] Write a program to control the host computer to send a command code to the encoder of the rope-pushing sensor every 0.5 seconds to read the encoder value. After waiting for the return code to appear complete 9 digits, intercept the 4th to 7th digits and convert them into decimal encoder values. Then calculate the rope end pull-out length according to the distance calculation formula. The formula is as follows:

[0078]

[0079] Where X' is the current encoder value, X is the preset reference position encoder value (set the position when the rope end is not pulled out as the reference position, then X = 0), C is the circumference of the grating encoder wheel (BRT27 rope-draw sensor C = 60), A is the single-turn resolution of the grating encoder (BRT27 rope-draw sensor A = 4096), and L' is the distance value of the current sensor rope end relative to the reference position (unit: mm).

[0080] After the pull-wire sensor and other hardware are correctly installed, the sensor encoder data when the car is unloaded is recorded as H0 (i.e., zero load height value), and the sensor encoder real-time data is recorded as H c , the real-time height difference ΔH=|H c-H0| Input the "load-relative height" polynomial model fitted during the data collection phase, use 500 as the guess value, and iteratively approximate the quartic polynomial to obtain the estimated car load.

[0081] The traditional elevator overload detection system is based on the mechanical deformation principle of the car bottom cushioning rubber: the elevator car bottom and the car frame are flexibly connected through the car bottom cushioning rubber. When the car load increases, the car bottom rubber undergoes vertical compression deformation, causing the distance between the car bottom and the car frame to decrease; by installing a micro switch on the car frame and calibrating the switch trigger position with standard weights, when the load reaches the threshold, the car bottom rubber deforms to cause the car bottom to contact the micro switch, disconnecting the normally closed contacts of the elevator main control overload signal, thereby realizing the elevator overload alarm function.

[0082] As elevators age, the car floor rubber ages. The same load causes greater deformation, triggering the microswitch before the car load reaches the overload threshold. This causes the elevator to sound an overload alarm prematurely, significantly reducing elevator efficiency. Therefore, elevator operation and maintenance companies must regularly calibrate and maintain elevators. The current maintenance solution requires loading the car with weights to fully load it and then recalibrating the position of the car floor microswitch. This calibration process is cumbersome, disrupts elevator operation, and consumes considerable manpower and resources.

[0083] This invention innovatively retains the principle of detecting floor rubber deformation, replacing the existing microswitch with a pull-rod sensor for distance monitoring. An embedded system continuously monitors sensor data during no-load and no-load shutdown conditions. Incorporating a multi-cycle mean comparison algorithm, it automatically identifies floor rubber degradation and dynamically compensates for and updates the full-load threshold. While maintaining the existing overload protection function, it also adds load estimation and automatic compensation and calibration for overload alarm offsets.

[0084] As a preferred technical solution, Figure 4 As shown, the multimodal self-calibration method further includes the following steps:

[0085] S7, judging whether the real-time load height value is less than the latest full-load height value, if so, judging that the car is overloaded, and activating the elevator sound and light alarm.

[0086] S8, obtaining real-time image data of the interior of the elevator car, and obtaining the number of passengers and determining whether an electric vehicle has entered the elevator car based on the real-time image data.

[0087] S9: If the number of passengers is greater than a preset threshold or an electric vehicle enters the elevator car, the elevator sound and light alarm is activated.

[0088] Specifically, this embodiment uses YOLOv8 target detection technology to monitor the situation inside the elevator car, including passenger counting and electric vehicle detection. First, video data of different scenes inside the elevator is collected and annotated. The YOLOv8 model is trained on a computing server, and the parameters are optimized to improve detection accuracy. The trained YOLOv8 model is converted to ONNX (Open Neural Network Exchange) format and optimized using RKNN (Rockchip Neural Network). It is then deployed on the RK3568 development board to improve inference efficiency.

[0089] To achieve efficient elevator car monitoring, Hikvision network cameras can be used to obtain RTSP video streams, and the QT front-end is used for data processing. OpenCV is used to decode the video frames, and then the RKNN API is called for target detection. The recognition results are displayed in real time on the QT interface.

[0090] To improve the system's adaptability and detection accuracy, the data collection phase included elevator car video samples taken at different times and under different lighting conditions. Data augmentation techniques were used to expand the dataset, and transfer learning was incorporated to optimize the model training process. After training, the YOLOv8 model was converted to the ONNX format and optimized using the RKNN toolchain, enabling faster inference speed and lower computational resource consumption when running on embedded devices.

[0091] By judging whether the elevator car is overloaded, whether the number of passengers is greater than the preset threshold, and whether an electric vehicle has entered the car, and taking corresponding sound and light alarm operations, the internal situation of the elevator car can be monitored in real time, improving evacuation efficiency and reducing safety hazards.

[0092] like Figure 5 As shown, an embodiment of the present invention also provides the testing steps and principles of the development board, which are as follows:

[0093] like Figure 8 As shown in the figure, after completing all hardware installation and wiring, connect the development board and computer to the same local area network to access the development board's operating system and control operations. After opening the pre-burned debugger, the debugging software interface will pop up. Click the "Open Serial Port - Start Sending" button. If the height appears in the "Current Height" text box, the development board is successfully communicating with the sensor. Click the "Start Video Streaming" button and the camera monitoring screen appears, which indicates successful communication between the development board and the camera. Click the "Relay Action" button and if the elevator issues an overload alarm (the display panel displays "Overload" and emits a buzzer alarm), the wiring between the development board, relay, and elevator car top control board is correct.

[0094] Sensor module testing: In the debugging software interface, after clicking the "Open Serial Port - Start Sending" button, the "Current Height" text box will display the sensor's current distance value, and the "Sensor Connection Status" signal light will turn green. Then, after clicking the "Set Baseline" button, the car is assumed to be unloaded. The "Base Zero Load Height" text box will display the recorded base height value, and the "Current Load" text box will display the estimated car load (should be 0 kg). At this time, adding a 20 kg weight to the car will cause the "Current Height" and "Current Load" text boxes to change, indicating that the sensor is operating normally and its sensitivity meets the requirements for car load estimation.

[0095] Visual recognition module test: After clicking "Open Video Stream" in the debugging software interface, the camera monitoring screen is displayed and the "Camera Connection Status" signal light is green. When no one enters the car, the "Is it empty" signal light is green. When someone enters the car, the "Is it empty" signal light is red, and the "Persons" column in the upper left corner of the monitoring screen is not 0, such as Figure 9 If the above description is met, it means that the visual recognition module is functioning normally.

[0096] An embodiment of the present invention further provides steps for data collection and polynomial model construction, which are specifically as follows:

[0097] The first step is the car load estimation model

[0098] (1) Preparation for debugging software

[0099] Open the debugging software interface and click the "Open Serial Port - Start Sending" button in sequence, and wait until the value in the "Current Height" text box is stable.

[0100] (2) Data collection

[0101] ① Initial reference calibration: record the initial height value H0 in the no-load state.

[0102] ② Step loading: Add 20kg standard weights step by step into the car, wait for the height to stabilize after each step, and record the load corresponding to the height H c ,like Figure 6 shown.

[0103] ③ Data termination condition: When the cumulative load reaches 1100 kg (overload threshold 110%), the data collection stops.

[0104] ④ Curve generation: With H0 as the zero point, draw the load-relative height change curve (ΔH=|H c -H0|VS L), such as Figure 7 shown.

[0105] (3) Fitting model

[0106] A quartic polynomial model is fitted based on the weighted least squares method or Newton iteration method. The quartic polynomial model formula is as follows:

[0107] L=a·(ΔH) 4 +b·(ΔH) 3 +c·(ΔH) 2 +d·ΔH+e

[0108] Where L represents the load in the car (in kg), ΔH represents the height change (in mm), and the fitting weight for the 800-1000 kg segment is increased to 1.5 times to enhance fitting accuracy near the overload range. Additionally, model accuracy can be improved by collecting multiple sets of data for fitting.

[0109] (4) Initial benchmark calibration and real-time solution

[0110] ①Benchmark calibration: take the average of the empty height of each group of data as the empty state height value H0; take the average of the full load height (1000kg) of each group of data as the full load state height value H m (Right now Figure 5 in the Full_Load).

[0111] ② Real-time load estimation: The real-time height difference ΔH=|H c -H0| Input the polynomial model and use 500 as the guess value to iteratively approximate and solve the quartic polynomial to obtain the estimated value L of the car load.

[0112] Step 2: Visual Recognition Model

[0113] (1) Algorithm selection

[0114] The YOLO (You Only Look Once) object detection algorithm was chosen because it strikes a balance between real-time detection and accuracy, making it particularly well-suited for elevator environments that require fast response times. Furthermore, YOLO is easy to optimize and deploy on embedded platforms, which is crucial for resource-constrained devices.

[0115] (2) Dataset creation

[0116] The dataset covers various elevator scenarios, including varying numbers of people, objects, and lighting conditions. A computer-based annotation platform was used to annotate the images, labeling the bounding boxes and categories of the objects. After annotation, the dataset was divided into a training set (70-80%), a validation set (10-15%), and a test set (10-15%) to ensure a balanced distribution of data across the dataset.

[0117] (3) Model training

[0118] The training process uses computing servers equipped with high-performance GPUs to handle the computational demands of deep learning models. Starting with a pretrained YOLO model, it is fine-tuned for the elevator scenario. Training is performed on the training set, while performance on the validation set is monitored, and hyperparameters (such as learning rate and batch size) are adjusted to prevent overfitting. Finally, the model is evaluated on the test set, using metrics such as mean average precision (mAP), precision, recall, and F1 score to ensure that it achieves the expected accuracy.

[0119] (3) Model deployment

[0120] Convert the trained YOLO model to a format suitable for the embedded platform. Test the converted model on a PC to ensure no significant performance degradation. Next, transfer the model to the embedded platform, install the necessary runtime libraries, and integrate it into the system software. Finally, conduct comprehensive testing on the embedded platform to verify real-time performance, accuracy, and robustness in various scenarios.

[0121] The following are the steps to implement the overall system functions:

[0122] 1. System startup and benchmark initialization

[0123] After completing module deployment, module testing, and data collection and no abnormalities are found, the intelligent weighing program can be started. The average value of the empty height of each group of data in the data collection phase is used as the initial empty state height value H0, and the average value of the full load height (1000kg) of each group of data is used as the initial full load state height value H m .

[0124] 2. Overload Judgment and Real-time Monitoring

[0125] When the elevator is operating normally, the pull-rope sensor module monitors the real-time height H of the car floor rubber in real time. c , and estimate the load size in the car based on the height difference. Real-time height H c Less than full load threshold height H m When the elevator is judged to be overloaded, the calculation control module controls the action of the optocoupler isolation relay through the GPIO pin and starts the elevator sound and light alarm.

[0126] The network camera monitors the situation inside the elevator in real time and counts the number of passengers. When the number of passengers N is greater than the rated number of passengers + 4 or an electric car enters the elevator, the elevator is judged to be abnormal. The calculation and control module controls the action of the optocoupler isolation relay through the GPIO pin and activates the elevator sound and light alarm.

[0127] 3. Self-calibration procedure

[0128] During the normal operation of the elevator, the control calculation module in the development board records the rope-type sensor data when the elevator is in the no-load state and regularly calculates the average value of the zero-load height. If the difference between the average value of the zero-load height and the initial zero-load height (i.e., the offset), Figure 5 If the equivalent load corresponding to (different_H) in the data is greater than or equal to 50 kg, or the offset is greater than the preset offset threshold, the car floor rubber is considered to be aged and the fully loaded car height needs to be recalibrated. The old zero-load height is then updated with the latest average zero-load height, and the difference between the two is used to compensate for the fully loaded height. The recorded rope-type sensor data is then cleared and recorded again. The new zero-load and fully loaded heights are used for daily car fully loaded determination, completing the automatic calibration process.

[0129] Here, yes Figure 5 Some parameters are explained in the following. avg=Sum / i+1 can be understood as difference_H=zeroLoad_H-avg, that is, the average value of zero load height H a The offset between the zero load height value H0, abs(EstimateLoad(difference_H,100)) can be understood as the equivalent load corresponding to the difference between the mean zero load height and the initial zero load height.

[0130] In summary, the present invention adopts a lightweight YOLO algorithm and deploys it to the development board, which can achieve accurate detection of elevator passengers. At the same time, a mathematical model of the rope-type sensor and the elevator load weight is established, and weighing calculations are performed based on the model to ensure the accuracy of the data. The rope-type sensor and the development board realize real-time data interaction through serial communication and are combined with an intelligent self-calibration mechanism. When the calibration conditions are met, the system automatically triggers the self-calibration mechanism, thereby improving long-term stability. In addition, the present invention develops a visual interface on the QT platform, which supports local and remote operations. Users can access the QT interface through mobile phones or tablets, and remote monitoring and control can be achieved. For elevator overload alarms, this weight is calculated based on the self-calibration height data. When it is detected that the full load threshold is exceeded, the development board drives the relay to control the elevator circuit to ensure the safe operation of the elevator.

[0131] Experimental test results demonstrate that this method achieves a weighing accuracy of 98.7% and an overload warning response time of less than 50ms, significantly outperforming traditional single-sensor solutions. Furthermore, the multimodal data fusion algorithm ensures efficient computing power, enabling high-precision, real-time load monitoring even during peak hours and in complex load distribution environments. This method possesses significant innovation, practical value, and market potential, and can be widely applied in the field of elevator safety control, providing intelligent safety assurance for elevator manufacturers, property management, and maintenance companies, thereby creating greater economic and social benefits.

[0132] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0133] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A multi-modal self-calibration method for elevator weighing warning, characterized in that: The multimodal self-calibration method comprises the following steps: A pull-rope sensor is installed between the car frame and the bottom of the car, and a zero-load height value when the elevator is in an unloaded and stopped state and a full-load height value when the elevator is in a fully loaded state are obtained by the pull-rope sensor; Obtaining a zero-load height average of multiple zero-load height values ​​within a preset time period, and determining whether the car floor rubber has aged based on an offset between the zero-load height average and the zero-load height value; If the car floor rubber is aged, the old zero-load height value is updated with the latest zero-load height average value, and the full-load height value is compensated and calibrated according to the offset to obtain the latest full-load height value.

2. A multi-modal self-calibration method for elevator weighing warning according to claim 1, characterized in that: The multimodal self-calibration method further comprises the following steps: Obtaining a real-time load height value of the elevator through the pull-rope sensor; Acquire a real-time height difference according to the real-time load height value and the zero-load height value; The real-time load of the car is obtained according to the real-time height difference.

3. A multi-modal self-calibration method for elevator weighing warning according to claim 2, characterized in that: The multimodal self-calibration method further comprises the following steps: Determine whether the real-time load height value is less than the latest full-load height value. If so, determine that the car is overloaded and start the elevator sound and light alarm.

4. A multi-modal self-calibration method for elevator weighing warning according to claim 3, characterized in that: The specific method for obtaining the real-time load of the car according to the real-time height difference includes the following steps: Construct a polynomial model based on the real-time load and height difference of the car; The polynomial model is fitted, and the real-time load of the car is obtained according to the fitted polynomial model.

5. A multi-modal self-calibration method for elevator weighing warning according to claim 4, characterized in that: The polynomial model is expressed as L = a·(ΔH) 4 +b·(ΔH) 3 +c·(ΔH) 2 +d·ΔH+e; Wherein, L represents the real-time load of the car, ΔH represents the real-time height difference, and a, b, c, d, and e represent the coefficient of the quartic term, the coefficient of the cubic term, the coefficient of the quadratic term, the coefficient of the linear term, and the preset constant, respectively.

6. A multi-modal self-calibration method for elevator weighing warning according to claim 5, characterized in that: According to the formula H' m =H m -α·(H0-H a ) Compensate and calibrate the full load height value; Among them, H0-H a Indicates the offset, H0, H a Represent the zero load height value and the mean zero load height, H m Indicates the full load height value before compensation calibration, H' m It represents the full load height value after compensation calibration, and α represents the compensation coefficient.

7. A multi-modal self-calibration method for elevator weighing warning according to claim 1, characterized in that: The body of the pull-wire sensor is fixedly mounted on a central rigid support point at the bottom of the car. The pull-wire axis of the pull-wire sensor is arranged in a vertical direction and the pulling direction of the pull wire is consistent with the sinking direction of the car.

8. A multi-modal self-calibration method for elevator weighing warning according to claim 1, characterized in that: The multimodal self-calibration method further comprises the following steps: Acquire real-time image data of the interior of the elevator car, and determine the number of passengers and whether an electric vehicle has entered the elevator car based on the real-time image data; If the number of people taking the elevator is greater than a preset threshold or an electric car enters the elevator car, the elevator sound and light alarm is activated.

9. A multi-modal self-calibration method for elevator weighing warning according to claim 8, characterized in that: Real-time image data is obtained by installing a network camera at a diagonal position on the top of the car away from the car door, and the real-time image data is used to identify the elevator passengers based on the YOLO target detection algorithm to obtain the number of passengers.

10. A multi-modal self-calibration method for elevator weighing warning according to claim 9, characterized in that: When it is detected that the number of passengers is zero and the data fed back by the pull-rope sensor remains stable for a preset time value, it is determined that the elevator is in an unloaded parking state.

Images