High-altitude operation intelligent safety belt early warning method and system based on Internet of Things

Abstract

The invention relates to the technical field of Internet of Things perception, in particular to an intelligent safety belt early warning method and system for aloft work based on the Internet of Things, and the method comprises the following steps: collecting a real-time frequency through radio frequency oscillation, monitoring a change rate, and generating a transient oscillation time domain waveform; the method comprises the following steps: determining an oscillation stable moment by using a sliding window variance, calculating a damping convergence time length, setting a delay window, sending step excitation, collecting a dielectric frequency shift numerical sequence, extracting a surface layer and deep layer response difference, splicing the surface layer and deep layer response difference with the convergence time length to form a joint feature, mapping a material safety space, and triggering risk early warning when deviating from a safety cluster. Through joint description of radio frequency oscillation frequency change, damping convergence time sequence and dielectric frequency shift response, energy attenuation after mechanical contact is associated with material dielectric difference, physical attributes of a hanging point are reflected, a material discrimination basis is constructed in combination with response gradient, comprehensive determination of essential characteristics of a bearing material is turned, and the bearing material quality is improved. And the misjudgment that the surface can be connected but does not have the safe bearing capacity is effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) sensing technology, and in particular to an IoT-based intelligent safety belt early warning method and system for high-altitude operations. Background Technology

[0002] The field of IoT sensing technology involves the use of various types of sensors and sensor networks to collect data and monitor the status of objects in the physical world in real time. The core of this field is to obtain the identity, location information and motion status of objects through sensing devices such as radio frequency identification, infrared sensing, global positioning system and inertial measurement unit, and to connect the collected sensing data to the transmission network to realize the digital mapping and remote monitoring of physical entities.

[0003] Among them, the traditional intelligent safety belt early warning method for high-altitude operations refers to the technical matters of monitoring the standardization of safety belt wearing and the hanging status of hooks during the climbing process of high-altitude workers. Mechanical micro switches or spring switches are embedded in the waist buckle of the safety belt, and pressure sensors or limit switches are set on the inside of the safety rope hook. When the metal locking tongue is inserted into the buckle or the hook is fastened to the safety rope, the physical components squeeze the switch contacts to close the circuit and conduct the circuit. Alternatively, radio frequency identification electronic tags are sewn into the webbing of the safety belt and used in conjunction with reader devices fixed at the work site. The safety belt is locked and hooked by detecting the on / off level signal of the circuit or the response status of the radio frequency signal.

[0004] Existing high-altitude work safety belt warning methods use the connection of the buckle contact or radio frequency response as the criterion. The monitoring logic is limited to the connection level and lacks the ability to perceive the internal structure of the attachment point material, the load-bearing reliability, and the dynamic process of contact. When facing surface that can be fastened but the material is fragile, hollow, or non-load-bearing, it still outputs a safety result. At the same time, the mechanical contact is prone to vibration, wear, and temperature and humidity changes, and the radio frequency reading is affected by obstruction and attitude changes, resulting in the inconsistency between the status judgment and the actual risk, thus limiting the reliability of the warning. Summary of the Invention

[0005] To address the technical problems of existing high-altitude work safety belt warning methods that rely on buckle contact continuity or radio frequency response as criteria, with monitoring logic limited to connection status and lacking the ability to perceive the internal structure of the attachment point material, load-bearing reliability, and dynamic contact process, this invention provides an IoT-based intelligent safety belt warning method for high-altitude work.

[0006] To achieve the above objectives, this invention employs an IoT-based intelligent safety belt early warning method for high-altitude operations, comprising the following steps: S1: The real-time frequency of radio frequency oscillation is collected through the intelligent safety belt hook. The rate of change of the real-time frequency of radio frequency oscillation when in contact with the high-altitude work hook point is monitored. When the rate of change exceeds the preset mechanical trigger threshold, the hooking action is determined to have occurred, and a transient oscillation time-domain waveform is generated. S2: Call the transient oscillation time-domain waveform, determine the steady-state time of the waveform through the sliding window variance algorithm, calculate the time difference from the maximum offset peak to the steady-state time, and generate the transient attenuation time of the hanging point hardness. S3: Call the transient decay time of the hanging point hardness and calculate the sampling delay time window. At the end of the delay time window, send a stepped excitation signal to the seat belt hook, collect the oscillation frequency and subtract the no-load reference frequency to generate a dielectric frequency shift numerical sequence. S4: Call the dielectric frequency shift numerical sequence, extract the deep response value and the surface response value and calculate the differential gradient, and concatenate them with the transient decay time of the hanging point hardness to generate joint feature data of hanging point hardness dielectric. S5: Map the combined hardness and dielectric characteristics of the hanging point to a preset material safety classification feature space. When the hanging point deviates from the safety clustering area within the material safety classification feature space, generate a hanging point material risk warning instruction.

[0007] As a further embodiment of the present invention, the transient oscillation time-domain waveform includes discrete frequency sampling points, time-domain amplitude sequence, and waveform time axis data; the transient attenuation duration of the hanging point hardness includes hook contact start timestamp, steady-state determination time, and attenuation interval value; the dielectric frequency shift value sequence includes frequency difference quantization set, step response data packet, and time sequence index identifier; the joint feature data of hanging point hardness dielectric includes gradient feature vector, damping dimension coefficient, and fused attribute tensor; and the hanging point material risk warning instruction includes non-load-bearing material abnormality category code, risk level flag bit, and seat belt buckle locking protection code. After generating the risk warning instruction for the hanging point material, the intelligent safety belt is controlled to enter the protection response state. According to the risk level, the hook locking mechanism is driven to perform the refusal to lock action, and the risk warning instruction is uploaded to the high-altitude operation safety supervision platform through IoT communication.

[0008] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Monitor the RF oscillation circuit signal through the smart seat belt hook, set a fixed clock pulse interval as the sampling period, read the two frequencies at the beginning and end of the period, perform differential operation on the two frequencies, and calculate the ratio of the differential result to the clock pulse interval duration to obtain the RF frequency change slope value. S102: Call the RF frequency change slope value and compare it with the preset mechanical trigger threshold. If the slope value exceeds the mechanical trigger threshold, activate the acquisition command, continuously read the real-time RF oscillation frequency within the preset window period, store the frequency value into the buffer queue in time sequence, and establish the trigger state frequency data sequence. S103: Based on the triggered frequency data sequence, analyze the acquisition timestamp and frequency value of the elements in the sequence, construct a two-dimensional coordinate mapping relationship between time and frequency, connect the coordinate points in ascending order of timestamp, perform linear interpolation operation between adjacent coordinate points, and generate transient oscillation time-domain waveform.

[0009] As a further aspect of the present invention, the monitoring of radio frequency oscillation circuit signals via the intelligent seat belt hook is achieved by coupling the metal hook body to the radio frequency circuit as an inductive antenna; The preset mechanical trigger threshold is set based on the mechanical vibration spectrum characteristics generated by the return spring latch impacting the hook body at the moment of rebound closure. Within the silent monitoring window when the smart seat belt hook does not undergo mechanical triggering, sampling and slope calculation are continuously performed according to a fixed clock pulse interval to generate a static background noise slope set containing multiple data items. The static background noise slope set is traversed, and the absolute value of each item in the set is extracted. The value with the largest absolute value is selected by comparing the values ​​and determined as the peak value of the environmental noise fluctuation. The preset signal-to-noise ratio safety redundancy coefficient is read from the storage unit, and the peak value of the environmental noise fluctuation is multiplied by the signal-to-noise ratio safety redundancy coefficient. The result of the calculation is defined as the preset mechanical trigger threshold.

[0010] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Call the transient oscillation time-domain waveform, calculate the absolute deviation of the frequency amplitude of discrete sampling points relative to the center reference frequency, retrieve the maximum numerical index in the absolute deviation sequence, extract the sampling timestamp corresponding to the index position, and establish the time of the maximum offset peak. S202: Based on the transient oscillation time-domain waveform, a sliding window is set and translated point by point along the time axis. The statistical variance of the frequency amplitude within the window is calculated and compared with the preset stability error threshold. The turning point when the variance value first falls below the threshold and continues to be less than the threshold is identified, and the steady-state convergence time of the oscillation is obtained. S203: Call the steady-state convergence time and the peak value of the maximum offset of the oscillation, substitute the two time values ​​into the time difference formula, calculate the time span between them, quantify the transition period of the oscillation from the peak to the steady state, and generate the transient decay time of the hanging point hardness.

[0011] As a further aspect of the present invention, the preset stability error threshold is based on the initial oscillation interval of the transient oscillation time-domain waveform for a preset duration before the maximum offset peak time. The frequency amplitude of all discrete sampling points in the corresponding interval is extracted, the absolute deviation of the frequency amplitude of the discrete sampling points relative to the center reference frequency is calculated, statistical processing is performed on the absolute deviation in the initial oscillation interval, the arithmetic mean of the absolute deviation is obtained, the arithmetic mean is multiplied by a preset proportional coefficient and squared to generate a preset stability error threshold in a fixed numerical form.

[0012] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Call the transient decay time of the hanging point hardness, obtain the preset buffer time constant and perform linear superposition with the convergence time, accumulate the superposition result to the current time reference point, lock the absolute time coordinate corresponding to the end of the delay window, and establish the excitation trigger time node. S302: Real-time clock monitoring is performed for the excitation trigger time node. If the real-time clock reading coincides with the node value, a stepped excitation pulse is injected into the radio frequency front end of the smart seat belt hook to drive the circuit to oscillate and continuously quantize the frequency using a fixed sampling rate and store it into the buffer in sequence to obtain the excitation response frequency sampling set. S303: Based on the stimulated response frequency sampling set, call the pre-stored circuit no-load reference frequency, perform subtraction operations between the frequency values ​​in the set and the no-load reference frequency, arrange the frequency offsets obtained by the operation according to the acquisition time sequence, and generate a dielectric frequency shift value sequence.

[0013] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Based on the dielectric frequency shift value sequence, extract the first frequency offset value and the last frequency offset value according to the start address and end address of the sequence storage space, and combine and encapsulate the first frequency offset value and the last frequency offset value to establish a deep and surface dielectric response pair. S402: Based on the deep and surface dielectric response pairs, the first frequency offset value is extracted as the surface response value, the last frequency offset value is extracted as the deep response value, and the absolute value of the difference is calculated. A preset depth normalization factor is called to perform a multiplication weighting operation to obtain the dielectric depth differential gradient. S403: Call the dielectric depth differential gradient and the transient decay time of the hanging point hardness to construct a multi-dimensional feature vector container, map the two to mechanical and electrical property components in the vector container respectively, and perform vector splicing and fusion on the two-dimensional components to generate joint feature data of hanging point hardness and dielectric.

[0014] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Call the joint feature data of hardness and dielectric of the hanging point, load the preset material safety classification feature space, extract the centroid coordinates of the center of the safety clustering region, map the joint feature data into multi-dimensional vector nodes in the feature space, calculate the Euclidean distance between the node and the centroid coordinates, and obtain the feature space offset metric value. S502: Based on the feature space offset metric, compare it with the preset safe clustering boundary judgment threshold. If the feature space offset metric exceeds the coverage of the safe clustering boundary judgment threshold, when the value exceeds the limit, retrieve the feature dimension identifier with the largest offset contribution value and generate a material abnormal deviation index. S503: Call the material abnormal deviation index, load the risk alarm control font library according to the preset communication protocol standard, match the risk type code of the hanging point material in the risk alarm control font library, collect the intelligent safety belt device identifier and risk type code to perform message formatting and encapsulation, and generate the hanging point material risk warning instruction.

[0015] An IoT-based intelligent safety belt early warning system for high-altitude operations includes: The frequency vibration acquisition module collects the real-time frequency of radio frequency oscillation through the intelligent safety belt hook, monitors the rate of change of the real-time frequency of radio frequency oscillation when it comes into contact with the hanging point of high-altitude operation, and determines that a hooking action has occurred when it exceeds the preset mechanical trigger threshold, and generates transient oscillation time domain waveform; The damping determination module calls the transient oscillation time-domain waveform, determines the steady-state time of the waveform through the sliding window variance algorithm, calculates the time difference from the maximum offset peak to the steady-state time, and generates the transient attenuation time of the hanging point hardness. The dielectric sampling module calls the transient attenuation time of the hanging point hardness and calculates the sampling delay time window. When the delay time window ends, it sends a stepped excitation signal to the seat belt hook, collects the oscillation frequency and subtracts the no-load reference frequency, and generates a dielectric frequency shift numerical sequence. The feature fusion module calls the dielectric frequency shift numerical sequence, extracts the deep response value and the surface response value and calculates the differential gradient, and splices it with the transient attenuation time of the hanging point hardness to generate joint feature data of hanging point hardness dielectric. The risk warning module maps the combined hardness and dielectric characteristics of the hanging point to a preset material safety classification feature space. When the hanging point deviates from the safety clustering area within the material safety classification feature space, a risk warning command for the hanging point material is generated, and the hook locking mechanism is driven to perform protective actions according to the command.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by jointly characterizing the transient changes in radio frequency oscillation frequency, damping convergence timing, and dielectric frequency shift response, the energy attenuation characteristics after mechanical contact are correlated with the dielectric differences of materials. This reflects the true physical properties of the hanging point from both time and frequency dimensions. Combined with the deep and shallow response gradients, a material discrimination criterion is constructed, shifting risk identification from a single connection confirmation to a comprehensive judgment of the essential characteristics of the bearing material. This effectively avoids misjudgment problems in scenarios where the surface can be connected but does not have safe bearing capacity. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0020] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0021] Please see Figure 1 This invention provides an IoT-based intelligent safety belt early warning method for high-altitude operations, comprising the following steps: S1: The real-time frequency of radio frequency oscillation is collected through the intelligent safety belt hook. The rate of change of the real-time frequency of radio frequency oscillation when in contact with the high-altitude work hook point is monitored. When the rate of change exceeds the preset mechanical trigger threshold, the hooking action is determined to have occurred, and a transient oscillation time-domain waveform is generated. S2: Call the transient oscillation time-domain waveform, determine the steady-state time of the waveform through the sliding window variance algorithm, calculate the time difference from the maximum offset peak to the steady-state time, and generate the transient attenuation time of the hanging point hardness. S3: Call the transient decay time of the hanging point hardness and calculate the sampling delay time window. When the delay time window ends, send a stepped excitation signal to the seat belt hook, collect the oscillation frequency and subtract the no-load reference frequency to generate a dielectric frequency shift numerical sequence. S4: Call the dielectric frequency shift numerical sequence, extract the deep response value and the surface response value and calculate the differential gradient, and concatenate it with the transient decay time of the hanging point hardness to generate joint feature data of hanging point hardness and dielectric. S5: Map the combined hardness and dielectric characteristics of the hanging point to the preset material safety classification feature space. When it deviates from the safety clustering area within the material safety classification feature space, generate a hanging point material risk warning instruction.

[0022] The transient oscillation time-domain waveform includes discrete frequency sampling points, time-domain amplitude sequence, and waveform time axis data; the transient attenuation duration of the hanging point hardness includes the hook contact start timestamp, steady-state determination time, and attenuation interval value; the dielectric frequency shift numerical sequence includes the frequency difference quantization set, step response data packet, and time sequence index identifier; the joint feature data of hanging point hardness and dielectric includes gradient feature vector, damping dimension coefficient, and fused attribute tensor; the hanging point material risk warning instruction includes non-load-bearing material abnormality category code, risk level flag bit, and seat belt buckle locking protection code. After generating a risk warning command for the hanging point material, the intelligent safety belt is controlled to enter the protection response state. Based on the risk level, the hook locking mechanism is driven to perform a refusal to lock action, and the risk warning command is uploaded to the high-altitude operation safety supervision platform via IoT communication.

[0023] Please see Figure 2 The specific steps of S1 are as follows: S101: Monitor the RF oscillation circuit signal through the smart seat belt hook, set a fixed clock pulse interval as the sampling period, read the two frequencies at the beginning and end of the period, perform differential operation on the two frequencies, and calculate the ratio of the differential result to the clock pulse interval duration to obtain the RF frequency change slope value. The monitoring process begins by activating a high-precision analog-to-digital converter interface for the RF front-end, locking the signal output pin of the RF oscillation circuit with microsecond-level time resolution. This pin is connected to the high-speed acquisition channel of the main control unit. Notably, monitoring the RF oscillation circuit signal via the smart seatbelt hook is achieved by coupling a metal hook body to the RF circuit as an inductive antenna, thereby sensing changes in the external environment. A strict time reference is established, for example, using a frequency-divided signal from a crystal oscillator as the synchronization source. A fixed clock pulse interval is set as the sampling period, which must satisfy the Nyquist sampling theorem to avoid aliasing; for example, it is set to 500 microseconds. At the beginning of each sampling period, the first frequency count value is read from the register via direct memory access, and then the second frequency count value is read at the end of the sampling period. After the read operation is completed, a difference operation is immediately performed on these two frequency values: the second frequency value at the end of the period is subtracted from the first frequency value at the beginning of the period to obtain the frequency drift within that period. Then, this frequency drift is used as the dividend, and the fixed clock pulse interval is used as the divisor to perform a high-precision division operation. This calculation process aims to quantify the drastic change in frequency over time, and the resulting quotient is the RF frequency change slope value. This slope value is not simply a number, but a physical quantity reflecting the instantaneous impact of external mechanical stress on the inductor or capacitor components of the RF circuit. Specifically, it corresponds to the instantaneous impact generated when a worker operates a safety harness and the hook collides with a rigid attachment point, or the minute deformation stress caused to the circuit board by the sudden tightening of the webbing. For example, if the starting frequency is 13,560,000 Hz and the ending frequency is 13,560,200 Hz, and the sampling period is 500 microseconds (i.e., 0.0005 seconds), then the frequency difference is first calculated to be 200 Hz. Then, 200 Hz is divided by 0.0005 seconds, resulting in an RF frequency change slope value of 400,000 Hz per second. This calculation process is performed continuously in the floating-point unit inside the processor, ensuring that a corresponding slope monitoring value is output for each clock cycle. To eliminate quantization errors, the raw sampled data undergoes denoising preprocessing during actual computation. This includes removing outliers that exceed hardware physical limits, ensuring the validity of the data used in the differential calculation. The entire process runs as an interrupt service routine in the underlying driver, guaranteeing real-time response to changes in the RF signal and providing a precise basic data stream for subsequent trigger determination.

[0024] S102: Call the RF frequency change slope value and compare it with the preset mechanical trigger threshold. If the slope value exceeds the mechanical trigger threshold, activate the acquisition command, continuously read the real-time RF oscillation frequency within the preset window period, store the frequency value into the buffer queue in time sequence, and establish the trigger state frequency data sequence. The calculated RF frequency change slope value is then introduced into the threshold determination logic unit. Prior to this, the generation of the preset mechanical trigger threshold is completed within a separate silent monitoring window. This window is set in the first few seconds after the equipment starts, for example, 5 seconds, during which the equipment is in a static state, wearing safety harnesses but not experiencing a fall or violent pulling. This is the preparation phase where the worker is wearing a safety harness but standing on the platform without moving. During this period, frequency sampling and slope calculation are continuously performed according to the aforementioned 500-microsecond fixed clock pulse interval, generating a static background noise slope set containing tens of thousands of data points. Subsequently, each data item in this set is traversed, and its absolute value is extracted, eliminating the influence of positive and negative direction differences on the noise amplitude judgment. Using a bubble sort or quicksort algorithm, the value with the largest absolute value is selected from the set and determined as the peak value of the environmental noise fluctuation. For example, the maximum absolute value of the noise slope calculated during the silent period is 50,000 Hz per second. Next, the preset signal-to-noise ratio (SNR) safety redundancy coefficient in the storage unit is read. This coefficient is an empirical value derived from a large amount of experimental data and is used to prevent false triggering; for example, it is set to 3. The peak ambient noise fluctuation of 50,000 Hz per second is multiplied by the SNR safety redundancy coefficient of 3, resulting in 150,000 Hz per second. This result is defined as the preset mechanical trigger threshold. The preset mechanical trigger threshold is set based on the mechanical vibration spectrum characteristics generated by the return spring latch impacting the hook body at the moment of rebound closure, to ensure accurate capture of the actual hooking action. During the real-time monitoring phase, once the currently calculated RF frequency change slope value (400,000 Hz per second) exceeds this mechanical trigger threshold (150,000 Hz per second), it is immediately determined that a mechanical contact trigger action has occurred between the seat belt hook and the fixed object, and the acquisition command is activated. After the acquisition command is triggered, a preset high-speed acquisition window is opened, for example, lasting 200 milliseconds. During this period, simple slope calculations are not performed; instead, the real-time frequency values ​​of the radio frequency oscillation are continuously read and stored in a first-in-first-out (FIFO) buffer queue in strict chronological order of acquisition. This establishes a complete trigger-state frequency data sequence, which fully records the entire frequency fluctuation process from the moment of triggering. Table 1 shows the processing of background noise slope data and threshold generation process within the silent monitoring window.

[0025] Table 1: Background Noise Sampling and Threshold Generation Table Sampling sequence number Sampling time (seconds) Frequency slope value (Hertz / second) Absolute value of slope (Hertz / second) Remark 1 0.0005 -12000 12000 Initial noise ... ... ... ... ... 4520 2.2600 48500 48500 transient disturbance 4521 2.2605 -50000 50000 Noise peak point ... ... ... ... ... 10000 5.0000 15000 15000 End of window As shown in Table 1, Table 1 lists some of the sampled data within a 5-second silent window. By traversing the data in the "Absolute value of slope" column of Table 1, the maximum value of 50,000 Hz per second was selected. Combined with the coefficient 3, the threshold was finally determined to be 150,000 Hz per second.

[0026] S103: Based on the triggered frequency data sequence, the acquisition timestamp and frequency value of the elements in the sequence are analyzed, a two-dimensional coordinate mapping relationship between time and frequency is constructed, the coordinate points are connected in ascending order of timestamp, and linear interpolation is performed between adjacent coordinate points to generate transient oscillation time-domain waveforms. Based on the triggered frequency data sequence stored in the cache queue, the data processing flow first performs a parsing operation, extracting the sampling timestamp and corresponding frequency value for each element in the sequence. A two-dimensional coordinate system data structure is constructed in memory, mapping the timestamp to the horizontal axis value and the frequency value to the vertical axis value. At this point, the data points are discretely distributed in the coordinate system. To recreate the actual physical fluctuation process, adjacent coordinate points are connected sequentially in ascending order of timestamps. For each pair of adjacent coordinate points, a linear interpolation operation is performed to fill the gaps in the sampling interval. Specifically, assuming the current coordinate point is (time 1, frequency 1) and the next coordinate point is (time 2, frequency 2), for any time t between time 1 and time 2, its corresponding frequency f is calculated through a linear equation: f equals frequency 1 plus (frequency 2 minus frequency 1) multiplied by (t minus time 1) divided by (time 2 minus time 1). For example, if time 10 milliseconds corresponds to a frequency of 13.6 MHz and time 11 milliseconds corresponds to a frequency of 13.8 MHz, then at time 10.5 milliseconds, the frequency calculated using the above logic is 13.7 MHz. This high-density point-by-point calculation smoothly transitions the originally discrete sampling points, generating a continuous transient oscillation time-domain waveform. This waveform not only contains information from the original sampling points but also completes the details of the signal change trend through interpolation, providing a high-resolution time-domain signal model for subsequent feature extraction. This process can accurately reproduce the frequency oscillation trajectory of the seatbelt metal hook assembly at the moment of force application, caused by the mechanical deformation of the hook and changes in the contact surface impedance of the RF circuit.

[0027] Please see Figure 3 The specific steps of S2 are as follows: S201: Call the transient oscillation time-domain waveform, calculate the absolute deviation of the frequency amplitude of discrete sampling points relative to the center reference frequency, retrieve the maximum numerical index in the absolute deviation sequence, extract the sampling timestamp corresponding to the index position, and establish the time of the maximum offset peak. The generated transient oscillation time-domain waveform is invoked. The processing first determines a center reference frequency, which is the nominal operating frequency of the circuit in a no-load state with the seatbelt unloaded, for example, 13.56 MHz. Next, each discrete sampling point on the waveform is traversed, and the difference between its frequency amplitude and the center reference frequency is calculated. The absolute value of this difference is taken to obtain an absolute deviation sequence. This sequence is retrieved, and the index of the element with the largest value is found using a comparison algorithm. The sampling timestamp corresponding to this index position is extracted, and this timestamp is marked as the peak time of the maximum offset. For example, in a waveform data segment with a center frequency of 13.56 MHz, the sampling point frequency at timestamp 15 milliseconds reaches 13.96 MHz, with an absolute deviation of 0.4 MHz; while the deviations at other time points are all less than this value. In this case, 0.4 MHz is identified as the maximum deviation, and 15 milliseconds is locked as the peak time of the maximum offset. Physically, this moment corresponds to the instant when the seatbelt locking mechanism experiences the greatest impact force or when the hook collides most violently with the anchor point, resulting in the greatest deformation of the radio frequency circuit. By establishing this moment, a key reference origin on the time axis is established for the subsequent analysis of the oscillation recovery process, giving the subsequent damping attenuation analysis a clear starting phase.

[0028] S202: Based on the transient oscillation time-domain waveform, a sliding window is set and translated point by point along the time axis. The statistical variance of the frequency amplitude within the window is calculated and compared with the preset stability error threshold. The turning point when the variance value first falls below the threshold and continues to be less than the threshold is identified, and the steady-state convergence time of the oscillation is obtained. Based on the transient oscillation time-domain waveform, a sliding window with a fixed width is defined, for example, covering 10 sampling points, and it is shifted point by point along the time axis starting from the moment of the maximum offset peak. At each position of the sliding window, the statistical variance of the frequency amplitude within the window is calculated to quantify the dispersion of frequency fluctuations within that time period. Simultaneously, a preset stability error threshold for determining steady state needs to be determined. This threshold is calculated based on an initial oscillation interval of the waveform before the moment of the maximum offset peak, for example, data from 5 milliseconds before the peak to the peak moment. The frequency amplitudes of all discrete sampling points within this interval are extracted, the absolute deviation relative to the center reference frequency is calculated, and the arithmetic mean of the absolute deviations is calculated. Assuming the average absolute deviation within the initial oscillation interval is 0.2 MHz, a preset scaling factor is read, for example, 0.05 (i.e., 5%), and 0.2 MHz is multiplied by 0.05 to generate 0.01 MHz as the linear error limit. This linear error limit is then squared to obtain the square of 0.0001 MHz, which is used as the preset stability error threshold in a fixed numerical form. Next, the variance calculated by the sliding window is continuously monitored to identify the turning point where the variance first falls below the threshold (i.e., less than the square of 0.0001 MHz) and remains below the threshold for several consecutive window positions (e.g., 5 consecutive windows). Once this condition is met, the time coordinate corresponding to this turning point is determined as the moment of steady-state convergence of the oscillation. This moment signifies that the violent oscillation caused by the mechanical shock has been attenuated to a negligible range of fluctuations by the damping effect at the connection between the seat belt webbing and the hook, and the hook has been securely attached to the anchor point.

[0029] S203: Call the steady-state convergence time of the oscillation and the peak time of the maximum offset, substitute the values ​​of the two times into the time difference formula, calculate the time span between the two, quantify the transition period of the oscillation from the peak to the steady state, and generate the transient decay time of the hanging point hardness. The system calls upon the steady-state convergence time and the peak offset time determined in the preceding steps to perform a time-difference calculation. Specifically, the value of the steady-state convergence time is subtracted from the value of the peak offset time. For example, if the peak offset time is 15 milliseconds and the steady-state convergence time is determined to be 155 milliseconds, then substituting 155 milliseconds and 15 milliseconds into the time difference formula yields a time span of 140 milliseconds. This value quantifies the transition period required for the oscillation to gradually recover from a violent fluctuation peak to a relatively calm steady state. This period directly reflects the mechanical damping characteristics of the system, namely the ability of the seatbelt hook material and the locking mechanism spring to dissipate impact energy. This calculation result is used to generate the transient decay time of the hook point hardness, serving as an important physical parameter for subsequent evaluation of the hardness and aging degree of the hook metal material. This parameter is obtained without relying on external sensors, entirely based on the time-domain response characteristics of radio frequency signals, enabling indirect measurement of the mechanical characteristics of key load-bearing components of the seatbelt.

[0030] Please see Figure 4 The specific steps of S3 are as follows: S301: Call the transient decay time of the hanging point hardness, obtain the preset buffer time constant and perform linear superposition with the convergence time, add the superposition result to the current timing reference point, lock the absolute time coordinate corresponding to the end of the delay window, and establish the excitation trigger time node. The calculated transient decay time of the hanging point hardness (e.g., 140 milliseconds) is used to first obtain the preset stabilization buffer factor. This factor is a coefficient greater than 1, used to add a certain amount of redundancy time to the convergence time to ensure that the circuit fully enters a stable state; for example, it is set to 1.5. The transient decay time of the hanging point hardness is linearly multiplied and superimposed with the stabilization buffer time constant, i.e., 140 milliseconds multiplied by 1.5, resulting in a delay time of 210 milliseconds. Subsequently, the current timing reference point is read and set as the peak time of the maximum offset (e.g., 15 milliseconds as mentioned above), and the calculated delay time is added to this reference point. That is, 15 milliseconds plus 210 milliseconds, the result is 225 milliseconds. This absolute time coordinate is locked and established as the excitation trigger time node. The determination of this node ensures that the subsequent active excitation operation is performed when the mechanical oscillation has completely disappeared and the electrical characteristics of the circuit are stable, avoiding interference from aftershocks on the dielectric response measurement.

[0031] S302: Real-time clock monitoring is performed for the excitation trigger time node. If the real-time clock reading coincides with the node value, a stepped excitation pulse is injected into the radio frequency front end of the smart seat belt hook to drive the circuit to oscillate and continuously quantize the frequency using a fixed sampling rate and store it into the buffer in sequence to obtain the excitation response frequency sampling set. For the calculated excitation trigger time node (e.g., 225 milliseconds), the system enters real-time clock monitoring mode. A high-precision timer inside the microprocessor continuously compares the current system time with the value at this node. Once the real-time clock reading perfectly coincides with 225 milliseconds, a stepped excitation pulse with a specific amplitude and pulse width is immediately injected into the RF front-end circuit integrated within the hook via a digital-to-analog converter or a dedicated pulse generator. This pulse, acting as a step signal, instantly disrupts the circuit's equilibrium state, driving the circuit to generate stimulated oscillation. Simultaneously, the oscillation frequency is continuously quantized at a fixed high sampling rate (e.g., 10 MHz) to capture the circuit's transient response to this step excitation. The quantized frequency values ​​are strictly stored in a high-speed cache according to the acquisition sequence until the oscillation decays back to the reference level, thus obtaining a complete sample set of stimulated response frequencies. This dataset records the purely electrical response process of the circuit under conditions of no mechanical deformation interference, influenced only by the dielectric properties of the hook material's internal and surface coatings.

[0032] S303: Based on the stimulated response frequency sampling set, call the pre-stored circuit no-load reference frequency, perform subtraction operation on the frequency values ​​in the set with the no-load reference frequency respectively, arrange the frequency offsets obtained by the operation according to the acquisition time sequence, and generate a dielectric frequency shift value sequence. Based on the stimulated response frequency sampling set, the circuit's no-load reference frequency, pre-stored in non-volatile memory, is first retrieved. This frequency was measured under standard conditions during factory calibration, for example, 13.5600 MHz. Next, each frequency value in the sampling set is iterated through, and a subtraction operation is performed on each value, subtracting the no-load reference frequency from the real-time frequency in the sampling set. For example, if the frequency in the sampling set at a certain moment is 13.5580 MHz, then 13.5580 minus 13.5600 yields a frequency offset of -0.0020 MHz (i.e., -2 kHz). The calculated frequency offsets are arranged according to the original acquisition timing sequence to generate a dielectric frequency shift value sequence. This sequence eliminates the influence of the environmental reference and purely reflects the pulling effect of the current seatbelt hook metal material and its insulating coating dielectric constant on the circuit frequency.

[0033] Table 2: Calculation Table of Stimulated Response and Frequency Shift Sampling sequence number Sampling time (milliseconds) Stimulated response frequency (megahertz) Unloaded reference frequency (megahertz) Dielectric frequency shift (kilohertz) 1 225.1 13.5580 13.5600 -2.0 2 225.2 13.5550 13.5600 -5.0 3 225.3 13.5520 13.5600 -8.0 ... ... ... ... ... 50 230.0 13.5590 13.5600 -1.0 Table 2 shows the correspondence between the stimulated response frequency sampling set and the dielectric frequency shift numerical sequence. It lists some of the sampled data after the excitation is triggered. The "dielectric frequency shift" in the last column is generated by subtracting the "stimulated response frequency" from the fixed "no-load reference frequency". The data form the dielectric frequency shift numerical sequence in sequence.

[0034] Please see Figure 5 The specific steps of S4 are as follows: S401: Based on the dielectric frequency shift numerical sequence, the first and last frequency offset values ​​are extracted according to the start and end addresses of the sequence storage space, and the first and last frequency offset values ​​are combined and encapsulated to establish a deep and surface dielectric response pair. Based on the generated dielectric frequency shift sequence, the first and last frequency offset values ​​are extracted according to the start and end physical addresses of the sequence in memory storage. The first value corresponds to the instantaneous response when the electromagnetic wave only penetrates the anti-rust coating or oxide layer on the hook surface, while the last value (or the mean value at the tail that tends to stabilize) corresponds to the response after the electromagnetic wave penetrates deep into the hook's metal alloy and establishes polarization equilibrium. These two values ​​are logically combined and encapsulated to establish a deep-layer and surface-layer dielectric response pair. For example, if the first value of the sequence is -2 kHz and the last value is -1 kHz, then the constructed data pair is [-2, -1]. This step, by extracting the transient and steady-state endpoint data, achieves the initial separation of the dielectric characteristics of the seat belt hook at different depths.

[0035] S402: Based on the deep and surface dielectric response pairs, the first frequency offset value is extracted as the surface response value, and the last frequency offset value is extracted as the deep response value. The absolute value of the difference is calculated, and the preset depth normalization factor is called to perform a multiplication weighting operation to obtain the dielectric depth differential gradient. Based on the constructed deep and surface dielectric response pairs, a deconstruction operation is first performed to explicitly extract the first frequency offset value (e.g., -2 kHz) as the surface response value and the last frequency offset value (e.g., -1 kHz) as the deep response value. Then, the absolute difference between the two is calculated, i.e., |-2 - (-1)| = 1 kHz. This difference reflects the gradient of dielectric properties as electromagnetic waves penetrate from the hook surface to the deep metal core. To unify the dimensions and enhance the characteristics, a preset depth normalization factor is invoked, which is set based on the penetration depth characteristics of the hook antenna radiation pattern, for example, 0.8. The absolute difference value of 1 kHz is multiplied and weighted with the depth normalization factor of 0.8, resulting in 0.8 kHz. This result is defined as the dielectric depth differential gradient, which can keenly reflect the aging differences from the outside in, such as surface hardening and internal corrosion, or surface coating peeling, in the seatbelt hook.

[0036] S403: Call the dielectric depth differential gradient and the transient decay time of the hanging point hardness to construct a multi-dimensional feature vector container, map the two to the mechanical and electrical property components in the vector container respectively, and perform vector splicing and fusion on the two-dimensional components to generate joint feature data of hanging point hardness and dielectric. The previously calculated dielectric depth differential gradient (e.g., 0.8 kHz) and the acquired hook point hardness transient decay time (e.g., 140 ms) are used. A multi-dimensional feature vector container is constructed to map these two physically distinct parameters into a unified vector space. The hook point hardness transient decay time is mapped as a mechanical property component, characterizing the elasticity and damping of the hook alloy and locking mechanism; the dielectric depth differential gradient is mapped as an electrical property component, characterizing the insulation aging degree of the hook material. Before performing vector concatenation, the data is standardized to eliminate dimensional differences; for example, 140 ms is standardized to 0.7 (relative to the standard value of 200 ms), and 0.8 kHz is standardized to 0.4 (relative to the standard value of 2 kHz). Finally, these two components are merged to generate a joint feature data of hook point hardness and dielectric properties, such as the vector [0.7, 0.4]. This feature data integrates mechanical feedback (such as latch fatigue) and electromagnetic dielectric feedback (such as hook corrosion), providing holistic data support for subsequent accurate material classification.

[0037] Please see Figure 6 The specific steps of S5 are as follows: S501: Call the joint feature data of hardness and dielectric of hanging point, load the preset material safety classification feature space, extract the centroid coordinates of the center of safety clustering region, map the joint feature data into multi-dimensional vector nodes in the feature space, calculate the Euclidean distance between the node and the centroid coordinates, and obtain the feature space offset metric. The joint feature data of hardness and dielectric properties of the anchor point (e.g., vector [0.7, 0.4]) is called, and a preset material safety classification feature space is loaded. This feature space is constructed based on test data of a large number of standard qualified seat belts and various aged and damaged seat belts, trained using clustering algorithms such as K-Means or DBSCAN. The centroid coordinates of the cluster regions marked as "safe" in this feature space are extracted, for example, the centroid coordinates are [0.8, 0.3]. Next, the current joint feature data is mapped to a multi-dimensional vector node in this feature space. The calculation is performed according to the Euclidean distance formula, that is, the square root of the sum of the squares of the difference between the dimensions of node [0.7, 0.4] and centroid [0.8, 0.3]. The specific calculation process is: (0.7-0.8)^2+(0.4-0.3)^2, that is, 0.01+0.01=0.02, and then the square root of 0.02 is taken to get approximately 0.1414. The value of 0.1414 is the feature space offset metric, which quantifies the similarity distance between the current detected seat belt hook material and the standard safety material.

[0038] Table 3: Spatial Distance Calculation Table for Qualitative Characteristics Sample number Mechanical components (normalized) Electrical components (normalized) Safety centroid coordinates Euclidean distance (offset metric) Current sample 0.7 0.4 [0.8，0.3] 0.1414 Reference Sample A 0.8 0.3 [0.8，0.3] 0.0000 Reference Sample B 0.2 0.9 [0.8，0.3] 0.8485 As shown in Table 3, the coordinates and offset measurements of samples of different materials in the feature space are displayed. The calculation result of 0.1414 for the "current sample" indicates that it is very close to the safety centroid, while the huge offset of the "reference sample B" suggests that the material properties of its seat belt hook have seriously deviated from the safety standard.

[0039] S502: Based on the feature space offset metric, compare it with the preset safe clustering boundary judgment threshold. If the feature space offset metric exceeds the range covered by the safe clustering boundary judgment threshold, when the value exceeds the limit, retrieve the feature dimension identifier with the largest offset contribution value and generate a material abnormal deviation index. Based on the calculated feature space offset metric (e.g., 0.1414), it is compared with a preset safety clustering boundary judgment threshold. This threshold defines the effective distribution radius of the safe material in the feature space, for example, set to 0.25. During the comparison process, if the offset metric 0.1414 is less than 0.25, the material is judged to be normal; if the offset metric exceeds 0.25 (e.g., 0.8485 for reference sample B), the material is judged to be abnormal. When the value exceeds the limit, attribution analysis is further performed to calculate the contribution ratio of multiple feature dimensions to the total distance. For example, for sample B, the mechanical component deviation is (0.2-0.8)^2=0.36, and the electrical component deviation is (0.9-0.3)^2=0.36. If the mechanical deviation contribution of a certain sample is 0.5 and the electrical deviation contribution is 0.1, the feature dimension with the largest contribution value is identified as "mechanical property", thereby generating a material abnormal deviation index, indicating that the risk is caused by "hardness / locking damping characteristics".

[0040] S503: Call the material abnormal deviation index, load the risk alarm control font library according to the preset communication protocol standard, match the hanging point material risk type code in the risk alarm control font library, collect the intelligent safety belt device identifier and risk type code to perform message formatting and encapsulation, and generate the hanging point material risk warning instruction; The system invokes the material anomaly deviation index and loads the risk alarm control font library stored in ROM according to a preset industrial IoT communication protocol standard (such as MQTT or CoAP). It searches the font library for the risk type code corresponding to the index; for example, "E01" represents "severe aging / metal fatigue risk of the safety belt hook." Simultaneously, it reads the unique device identifier (Device ID, such as "UID-X9527") of the safety belt stored in the device's non-volatile memory. It performs message formatting and encapsulation with the risk type code "E01" and the device identifier to construct a binary data packet conforming to the transmission standard. Finally, it sends this data packet as a material risk warning instruction to the back-end monitoring center via the wireless radio frequency transmission module, immediately controlling the smart safety belt to enter the protective response state. Based on the calculated risk level, it drives the hook locking mechanism to execute a refusal-locking action, preventing workers from forcibly using the equipment when it is in an unsafe condition. At the same time, the risk warning instruction is uploaded to the high-altitude operation safety supervision platform via IoT communication, achieving remote real-time monitoring and immediate intervention. The experimental results show that by using distance determination in the joint feature space, the accuracy of identifying early-aging safety belts is improved by about 18% compared to the traditional single threshold method, effectively preventing high-altitude operation accidents caused by material defects.

[0041] Please see Figure 7 A smart safety belt early warning system for high-altitude operations based on the Internet of Things, comprising: The frequency vibration acquisition module collects the real-time frequency of radio frequency oscillation through the intelligent safety belt hook, monitors the rate of change of the real-time frequency of radio frequency oscillation when it comes into contact with the high-altitude work hook point, and determines that a hooking action has occurred when the frequency exceeds the preset mechanical trigger threshold, generating a transient oscillation time-domain waveform. The damping determination module calls the transient oscillation time-domain waveform, determines the steady-state time of the waveform through the sliding window variance algorithm, calculates the time difference from the maximum offset peak to the steady-state time, and generates the transient attenuation time of the hanging point hardness. The dielectric sampling module calls the transient decay time of the hanging point hardness and calculates the sampling delay time window. When the delay time window ends, it sends a stepped excitation signal to the seat belt hook, collects the oscillation frequency and subtracts the no-load reference frequency, and generates a dielectric frequency shift numerical sequence. The feature fusion module calls the dielectric frequency shift numerical sequence, extracts the deep response value and the surface response value and calculates the differential gradient, and splices it with the transient decay time of the hanging point hardness to generate joint feature data of hanging point hardness and dielectric. The risk warning module maps the combined hardness and dielectric characteristics of the hanging point to a preset material safety classification feature space. When the hanging point deviates from the safety clustering area within the material safety classification feature space, a risk warning instruction for the hanging point material is generated, and the hook locking mechanism is driven to perform protective actions according to the instruction.

[0042] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.

Claims

1. A method for intelligent safety belt early warning in high-altitude operations based on the Internet of Things, characterized in that, Includes the following steps: S1: The real-time frequency of radio frequency oscillation is collected through the intelligent safety belt hook. The rate of change of the real-time frequency of radio frequency oscillation when in contact with the high-altitude work hook point is monitored. When the rate of change exceeds the preset mechanical trigger threshold, the hooking action is determined to have occurred, and a transient oscillation time-domain waveform is generated. S2: Call the transient oscillation time-domain waveform, determine the steady-state time of the waveform through the sliding window variance algorithm, calculate the time difference from the maximum offset peak to the steady-state time, and generate the transient attenuation time of the hanging point hardness. S3: Call the transient decay time of the hanging point hardness and calculate the sampling delay time window. At the end of the delay time window, send a stepped excitation signal to the seat belt hook, collect the oscillation frequency and subtract the no-load reference frequency to generate a dielectric frequency shift numerical sequence. S4: Call the dielectric frequency shift numerical sequence, extract the deep response value and the surface response value and calculate the differential gradient, and concatenate them with the transient decay time of the hanging point hardness to generate joint feature data of hanging point hardness dielectric. S5: Map the combined hardness and dielectric characteristics of the hanging point to a preset material safety classification feature space. When the hanging point deviates from the safety clustering area within the material safety classification feature space, generate a hanging point material risk warning instruction.

2. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 1, characterized in that, The transient oscillation time-domain waveform includes discrete frequency sampling points, time-domain amplitude sequence, and waveform time axis data; the transient attenuation duration of the hanging point hardness includes the hook contact start timestamp, steady-state determination time, and attenuation interval value; the dielectric frequency shift value sequence includes frequency difference quantization set, step response data packet, and time sequence index identifier; the joint feature data of hanging point hardness and dielectric includes gradient feature vector, damping dimension coefficient, and fused attribute tensor; the hanging point material risk warning instruction includes non-load-bearing material abnormality category code, risk level flag bit, and seat belt buckle locking protection code. After generating the risk warning instruction for the hanging point material, the intelligent safety belt is controlled to enter the protection response state. According to the risk level, the hook locking mechanism is driven to perform the refusal to lock action, and the risk warning instruction is uploaded to the high-altitude operation safety supervision platform through IoT communication.

3. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Monitor the RF oscillation circuit signal through the smart seat belt hook, set a fixed clock pulse interval as the sampling period, read the two frequencies at the beginning and end of the period, perform differential operation on the two frequencies, and calculate the ratio of the differential result to the clock pulse interval duration to obtain the RF frequency change slope value. S102: Call the RF frequency change slope value and compare it with the preset mechanical trigger threshold. If the slope value exceeds the mechanical trigger threshold, activate the acquisition command, continuously read the real-time RF oscillation frequency within the preset window period, store the frequency value into the buffer queue in time sequence, and establish the trigger state frequency data sequence. S103: Based on the triggered frequency data sequence, analyze the acquisition timestamp and frequency value of the elements in the sequence, construct a two-dimensional coordinate mapping relationship between time and frequency, connect the coordinate points in ascending order of timestamp, perform linear interpolation operation between adjacent coordinate points, and generate transient oscillation time-domain waveform.

4. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 3, characterized in that, The monitoring of radio frequency oscillation circuit signals via the smart seat belt hook is achieved by coupling the metal hook body to the radio frequency circuit as an inductive antenna. The preset mechanical trigger threshold is set based on the mechanical vibration spectrum characteristics generated by the return spring latch impacting the hook body at the moment of rebound closure. Within the silent monitoring window when the smart seat belt hook does not undergo mechanical triggering, sampling and slope calculation are continuously performed according to a fixed clock pulse interval to generate a static background noise slope set containing multiple data items. The static background noise slope set is traversed, and the absolute value of each item in the set is extracted. The value with the largest absolute value is selected by comparing the values ​​and determined as the peak value of the environmental noise fluctuation. The preset signal-to-noise ratio safety redundancy coefficient is read from the storage unit, and the peak value of the environmental noise fluctuation is multiplied by the signal-to-noise ratio safety redundancy coefficient. The result of the calculation is defined as the preset mechanical trigger threshold.

5. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Call the transient oscillation time-domain waveform, calculate the absolute deviation of the frequency amplitude of discrete sampling points relative to the center reference frequency, retrieve the maximum numerical index in the absolute deviation sequence, extract the sampling timestamp corresponding to the index position, and establish the time of the maximum offset peak. S202: Based on the transient oscillation time-domain waveform, a sliding window is set and translated point by point along the time axis. The statistical variance of the frequency amplitude within the window is calculated and compared with the preset stability error threshold. The turning point when the variance value first falls below the threshold and continues to be less than the threshold is identified, and the steady-state convergence time of the oscillation is obtained. S203: Call the steady-state convergence time and the peak value of the maximum offset of the oscillation, substitute the two time values ​​into the time difference formula, calculate the time span between them, quantify the transition period of the oscillation from the peak to the steady state, and generate the transient decay time of the hanging point hardness.

6. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 5, characterized in that, The preset stability error threshold is based on the initial oscillation interval of the transient oscillation time-domain waveform for a preset duration before the peak of the maximum offset. The frequency amplitude of all discrete sampling points in the corresponding interval is extracted, the absolute deviation of the frequency amplitude of the discrete sampling points relative to the center reference frequency is calculated, statistical processing is performed on the absolute deviation in the initial oscillation interval, the arithmetic mean of the absolute deviation is obtained, the arithmetic mean is multiplied by a preset proportional coefficient and squared to generate a preset stability error threshold in a fixed numerical form.

7. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 5, characterized in that, The specific steps for S3 are as follows: S301: Call the transient decay time of the hanging point hardness, obtain the preset buffer time constant and perform linear superposition with the convergence time, accumulate the superposition result to the current time reference point, lock the absolute time coordinate corresponding to the end of the delay window, and establish the excitation trigger time node. S302: Real-time clock monitoring is performed for the excitation trigger time node. If the real-time clock reading coincides with the node value, a stepped excitation pulse is injected into the radio frequency front end of the smart seat belt hook to drive the circuit to oscillate and continuously quantize the frequency using a fixed sampling rate and store it into the buffer in sequence to obtain the excitation response frequency sampling set. S303: Based on the stimulated response frequency sampling set, call the pre-stored circuit no-load reference frequency, perform subtraction operations between the frequency values ​​in the set and the no-load reference frequency, arrange the frequency offsets obtained by the operation according to the acquisition time sequence, and generate a dielectric frequency shift value sequence.

8. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 7, characterized in that, The specific steps of S4 are as follows: S401: Based on the dielectric frequency shift value sequence, extract the first frequency offset value and the last frequency offset value according to the start address and end address of the sequence storage space, and combine and encapsulate the first frequency offset value and the last frequency offset value to establish a deep and surface dielectric response pair. S402: Based on the deep and surface dielectric response pairs, the first frequency offset value is extracted as the surface response value, the last frequency offset value is extracted as the deep response value, and the absolute value of the difference is calculated. A preset depth normalization factor is called to perform a multiplication weighting operation to obtain the dielectric depth differential gradient. S403: Call the dielectric depth differential gradient and the transient decay time of the hanging point hardness to construct a multi-dimensional feature vector container, map the two to mechanical and electrical property components in the vector container respectively, and perform vector splicing and fusion on the two-dimensional components to generate joint feature data of hanging point hardness and dielectric.

9. The IoT-based intelligent safety belt early warning method for high-altitude operations according to claim 8, characterized in that, The specific steps of S5 are as follows: S501: Call the joint feature data of hardness and dielectric of the hanging point, load the preset material safety classification feature space, extract the centroid coordinates of the center of the safety clustering region, map the joint feature data into multi-dimensional vector nodes in the feature space, calculate the Euclidean distance between the node and the centroid coordinates, and obtain the feature space offset metric value. S502: Based on the feature space offset metric, compare it with the preset safe clustering boundary judgment threshold. If the feature space offset metric exceeds the coverage of the safe clustering boundary judgment threshold, when the value exceeds the limit, retrieve the feature dimension identifier with the largest offset contribution value and generate a material abnormal deviation index. S503: Call the material abnormal deviation index, load the risk alarm control font library according to the preset communication protocol standard, match the risk type code of the hanging point material in the risk alarm control font library, collect the intelligent safety belt device identifier and risk type code to perform message formatting and encapsulation, and generate the hanging point material risk warning instruction.

10. A smart safety belt early warning system for high-altitude operations based on the Internet of Things, characterized in that, The system is used to implement the IoT-based intelligent safety belt early warning method for high-altitude operations as described in any one of claims 1-9, the system comprising: The frequency vibration acquisition module collects the real-time frequency of radio frequency oscillation through the intelligent safety belt hook, monitors the rate of change of the real-time frequency of radio frequency oscillation when it comes into contact with the hanging point of high-altitude operation, and determines that a hooking action has occurred when it exceeds the preset mechanical trigger threshold, and generates transient oscillation time domain waveform; The damping determination module calls the transient oscillation time-domain waveform, determines the steady-state time of the waveform through the sliding window variance algorithm, calculates the time difference from the maximum offset peak to the steady-state time, and generates the transient attenuation time of the hanging point hardness. The dielectric sampling module calls the transient attenuation time of the hanging point hardness and calculates the sampling delay time window. When the delay time window ends, it sends a stepped excitation signal to the seat belt hook, collects the oscillation frequency and subtracts the no-load reference frequency, and generates a dielectric frequency shift numerical sequence. The feature fusion module calls the dielectric frequency shift numerical sequence, extracts the deep response value and the surface response value and calculates the differential gradient, and splices it with the transient attenuation time of the hanging point hardness to generate joint feature data of hanging point hardness dielectric. The risk warning module maps the combined hardness and dielectric characteristics of the hanging point to a preset material safety classification feature space. When the hanging point deviates from the safety clustering area within the material safety classification feature space, a risk warning instruction for the hanging point material is generated, and the hook locking mechanism is driven to perform protective actions according to the instruction.

Images