Cascade operation state real-time monitoring and anti-collision risk intelligent early warning method and system

Abstract

The invention discloses a cascade operation state real-time monitoring and anti-collision risk intelligent early warning method and system, and the method comprises the steps: obtaining multi-sensor data, determining a key monitoring node, and extracting parameter changes to form an operation state trend sample; carrying out edge calculation processing on the trend sample, reversely analyzing and reconstructing component state parameters from the vibration spectrum signal, generating a step gap change curve, matching with a standard model to obtain a risk characteristic probability value, and generating a state traction chain according to the risk characteristic probability value; obtaining a displacement distribution coefficient and a vibration speed index from the state drag chain, predicting the collision probability distribution of the displacement distribution coefficient through cloud deep learning, determining an early warning level in combination with the vibration speed index, and generating early warning optimization configuration; emergency distribution is determined through linkage control analysis, differentiated response areas are identified, and linkage optimization factors are extracted from the differentiated response areas; finally, differential control parameters are generated and matched with the state traction chain to generate an early warning signal, and intelligent abnormal recognition and self-adaptive early warning of the step equipment are achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and in particular to a method and system for real-time monitoring of cascade operation status and intelligent early warning of collision risks. Background Technology

[0002] As a crucial vertical transportation facility in modern buildings and transportation hubs, escalator systems primarily rely on traditional methods of periodic manual inspections and simple sensor alarms for safety monitoring. Current monitoring technologies typically depend on maintenance personnel conducting on-site inspections of key parameters such as step clearance, component wear, and structural deformation at fixed intervals, combined with vibration and temperature sensors in the foundation for anomaly alarms. Some advanced systems are beginning to incorporate visual inspection and sound recognition technologies to monitor changes in step surface condition and operating noise.

[0003] However, existing technologies have revealed significant shortcomings when dealing with complex operating environments and varied failure modes: First, traditional periodic manual inspections cannot achieve continuous real-time monitoring, making it difficult to detect early signs of progressive failures in a timely manner, especially minute changes in step gaps and slow degradation of component performance; second, existing sensor systems mostly use single data sources and fixed threshold judgments, lacking the ability to fuse multi-dimensional information, and are prone to false alarms or missed detections when faced with complex factors such as environmental interference and equipment aging; finally, existing early warning mechanisms generally lack intelligent analysis and prediction capabilities, and cannot adaptively adjust according to the dynamic changes in equipment operating status, making it difficult to achieve the transformation from passive maintenance to proactive prevention, especially in high-traffic usage scenarios, where they cannot effectively predict and prevent sudden safety risks. Summary of the Invention

[0004] This invention provides a method and system for real-time monitoring of cascade operation status and intelligent early warning of collision risks. It aims to achieve intelligent perception and accurate prediction of the operation status of cascade equipment through multi-sensor data fusion and edge-cloud collaborative processing technology, build a differentiated response early warning system, and provide technical support for the safe operation and intelligent maintenance of cascade equipment.

[0005] The first aspect of this invention proposes a method for real-time monitoring of cascade operation status and intelligent early warning of collision risks, comprising the following steps: Acquire multi-sensor data for cascade operation, determine key monitoring nodes based on the multi-sensor data, and extract parameter changes of the key monitoring nodes to form an operational status trend sample; Edge computing processing is performed on the operating status trend sample to generate a step gap change curve. The step gap change curve is matched with the standard operating sample model to obtain the risk feature probability value. A status traction chain is generated based on the risk feature probability value. The displacement distribution coefficient and vibration velocity index are obtained from the state traction chain. The collision probability distribution of the displacement distribution coefficient is predicted by the cloud deep learning algorithm. The warning level is determined based on the vibration velocity index. The warning optimization configuration is generated by combining the collision probability distribution and the warning level. Based on the early warning optimization configuration, the emergency distribution is determined by linkage control analysis. Differentiated response areas are identified through the emergency distribution, and linkage optimization factors are extracted from the differentiated response areas. Differentiated control parameters are generated using the aforementioned linkage optimization factor, and these differentiated control parameters are matched with the state traction chain to generate an early warning signal.

[0006] The second aspect of this invention proposes a real-time monitoring system for cascade operation status and an intelligent early warning system for collision avoidance risks, comprising: The data acquisition module is used to acquire multi-sensor data of the cascade operation, determine key monitoring nodes based on the multi-sensor data, and extract parameter changes of the key monitoring nodes to form an operating status trend sample. The spectrum reconstruction module is used to perform edge computing processing on the operating status trend sample to generate a step gap change curve, match the step gap change curve with the standard operating sample model to obtain risk feature probability values, and generate a status traction chain based on the risk feature probability values. The attenuation compensation module is used to obtain the displacement distribution coefficient and vibration velocity index from the state traction chain, predict the collision probability distribution of the displacement distribution coefficient through a cloud-based deep learning algorithm, determine the warning level based on the vibration velocity index, and generate a warning optimization configuration by combining the collision probability distribution and the warning level. The linkage control module is used to perform linkage control analysis based on the early warning optimization configuration to determine the emergency distribution, identify differentiated response areas through the emergency distribution, and extract linkage optimization factors from the differentiated response areas. The early warning output module is used to generate differentiated control parameters using the linkage optimization factor, and to match the differentiated control parameters with the state traction chain to generate an early warning signal.

[0007] The beneficial effects of this invention are reflected in the following points: 1. By using multi-sensor data fusion and key monitoring node determination technology, dispersed sensor information such as vibration, current, and temperature is processed uniformly to form operating status trend samples, solving the problem of incomplete information from traditional single-sensor monitoring and improving the completeness of equipment status perception. 2. Edge computing is used to reverse analyze the vibration spectrum signal, reconstruct the mechanical component status parameters, and generate the cascade gap change curve, realizing real-time tracking of key operating parameters of the equipment and making up for the inability to continuously monitor periodic manual inspection. 3. An innovative state traction chain dynamic balance mechanism is constructed. Through the interaction of active traction and reverse drag of the normal operating state to the abnormal state, three adaptive adjustment modes are formed: forced traction, coordinated traction, and monitoring traction. This solves the technical bottleneck of passive response in traditional early warning systems and realizes active intervention and intelligent self-correction of equipment status. 4. A cloud-based deep learning prediction and differentiated response control mechanism is established. Through collision probability prediction and linkage optimization factor extraction, adaptive differentiated control parameters and early warning signals are generated, overcoming the limitations of traditional fixed threshold judgment. This realizes intelligent early warning and accurate response based on real-time equipment status, effectively improving the accuracy of cascade equipment fault prediction and the pertinence of early warning.

[0008] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0009] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0010] Unless otherwise specified or otherwise, the same reference numerals in different figures represent the same or similar technical features, and different reference numerals may be used to represent the same or similar technical features.

[0011] Figure 1 This is a flowchart illustrating a method for real-time monitoring of cascade operation status and intelligent early warning of collision risks according to the present invention.

[0012] Figure 2 This is a structural block diagram of a cascade operation status real-time monitoring and collision risk intelligent early warning system according to the present invention. Detailed Implementation

[0013] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0014] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0015] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0016] The technical solutions of the embodiments of this application are described below.

[0017] like Figure 1 As shown, this embodiment of the invention provides a method for real-time monitoring of cascade operation status and intelligent early warning of collision risks, including the following steps S110-S150: Step S110: Obtain multi-sensor data for cascade operation, determine key monitoring nodes based on multi-sensor data, and extract parameter changes of key monitoring nodes to form operating status trend samples.

[0018] Specifically, multi-sensor data is acquired for the ladder runway operation. Various types of sensors, including vibration sensors, temperature sensors, current sensors, speed sensors, and pressure sensors, are deployed at key locations within the ladder runway. Vibration sensors are installed under the ladder drive motor, guide rail supports, and ladder treads, with a sampling frequency of 1000Hz and a measurement range of 0-50g acceleration signals. Temperature sensors are located within the motor windings, bearing housings, and control cabinet, with a measurement accuracy of ±0.1°C and a data acquisition interval of 10 seconds. Current sensors are connected to the three-phase power lines of the main drive motor to measure real-time current changes during motor operation, with a sampling accuracy of 0.01A. Speed ​​sensors, in the form of photoelectric encoders, are installed on the drive wheel axle to monitor the ladder runway speed in real time, with a resolution of 0.01m / s. Pressure sensors are installed at the handrail tensioning device and the ladder chain tensioning position to monitor changes in mechanical tension. Data from all sensors is collected at a data acquisition station using a combination of wired and wireless methods. The acquisition station is equipped with a high-speed ADC converter and large-capacity storage devices. The raw data collected is time-synchronized to ensure consistency in the time base of data from different sensors. Multi-sensor data is then categorized and stored according to sensor type, installation location, and acquisition time, forming a structured multi-dimensional dataset.

[0019] Key monitoring nodes are identified based on multi-sensor data. Signal strength and sensitivity analysis are performed on the collected multi-sensor data to identify the measurement points most sensitive to the escalator's operating status. The standard deviation and coefficient of variation of each sensor signal are calculated; sensor locations with a standard deviation greater than twice the average are marked as high-sensitivity monitoring points. The correlation between different sensor data is analyzed, and the Pearson correlation coefficient is used to quantify the degree of association between sensors. Independent sensors with correlation coefficients below 0.3 are screened to avoid information redundancy at monitoring nodes. Principal component analysis is used to extract the main direction of change in sensor data; sensor combinations corresponding to principal components with a contribution rate exceeding 85% are identified as key monitoring nodes. For example, in the monitoring of an escalator in a shopping mall, the drive motor current sensor, the main guide wheel vibration sensor, and the step chain tension sensor constitute three key monitoring nodes; the data changes of these three nodes can reflect 95% of the equipment's operating status information. The spatial distribution of key monitoring nodes is calculated to ensure that nodes cover all important components of the escalator equipment. The key monitoring nodes are ranked by importance, and priority weights are assigned based on their contribution to fault detection. The identified key monitoring node information is recorded as a node list, which includes attributes such as node location, sensor type, data characteristics, and importance level.

[0020] Extracting parameter changes from key monitoring nodes forms operational status trend samples. Feature parameters reflecting changes in equipment operational status are extracted from time-series data collected from key monitoring nodes. A sliding window analysis is performed on the data from each monitoring node, with a window length of 30 minutes and a step size of 5 minutes. Statistical parameters such as the mean, maximum, minimum, and standard deviation of the data within the window are calculated. The temporal trend of parameter changes is analyzed, and the slope of parameter changes is calculated using linear regression. Parameter changes with an absolute slope value greater than 0.05 are marked as significant trends. Periodic features of parameter changes are extracted, and the main frequency components and periodic patterns in the data are identified using Fast Fourier Transform. Abrupt change features of parameter changes are calculated, and a change point detection algorithm is used to identify sudden jumps and abnormal change points in the parameter sequence. The extracted statistical parameters, trend parameters, periodic parameters, and abrupt change parameters are combined to form a multi-dimensional feature vector. For example, the feature vector of the motor current node includes five components: mean current, standard deviation current, slope of current change, amplitude of the dominant frequency, and number of abrupt changes. Multiple feature vectors are arranged in chronological order to form time-series samples, with each sample corresponding to the operational status characteristics of a time window. The operational status trend samples are standardized to map parameters of different dimensions to the same numerical range. The processed operational status trend samples are then stored in a sample database, containing complete information such as sample timestamps, feature vectors, and corresponding equipment operating condition labels.

[0021] Step S120: Perform edge computing processing on the running status trend sample to generate the step gap change curve, match the step gap change curve with the standard running sample model to obtain the risk feature probability value, and generate the status traction chain based on the risk feature probability value.

[0022] In some embodiments, the step of performing edge computing processing on the operating state trend sample to generate a step gap change curve includes: acquiring vibration spectrum signals from the operating state trend sample; reconstructing mechanical component state parameters by reverse analysis of the vibration spectrum signals through edge computing processing; evaluating the wear degree of the component based on the mechanical component state parameters to form a wear state mapping; and fitting the step gap change curve according to the wear state mapping.

[0023] Vibration spectrum signals were collected from operational status trend samples. The data components collected by vibration sensors were filtered from the operational status trend samples to extract vibration spectrum signals containing frequency domain information. The original vibration time-domain signals were processed using a window function, employing a Hanning window to reduce spectral leakage, with a window length set to 2048 sampling points. A Fast Fourier Transform was applied to convert the time-domain vibration signal into a frequency-domain spectrum, with a spectral resolution of 0.488 Hz and a frequency range covering 0-500 Hz. The amplitude distribution characteristics of the spectrum were analyzed to identify the frequency bandwidth and peak positions where vibration energy is concentrated. Characteristic parameters such as amplitude, phase, and power spectral density of the main frequency components were extracted. Background noise in the spectrum was filtered, using three times the average noise level as the threshold for valid signals. For the specific frequency response of the cascade equipment, vibration spectrum signals in the 20-200 Hz frequency band were collected, corresponding to the sensitive frequency range of changes in the cascade gaps. The collected vibration spectrum signals are grouped and stored according to frequency components. Low frequency components (20-50Hz) correspond to the overall structural vibration, mid frequency components (50-120Hz) correspond to the vibration of local components, and high frequency components (120-200Hz) correspond to friction and impact vibration.

[0024] The vibration spectrum signal is processed by edge computing to reverse-analyze and reconstruct the mechanical component state parameters. A pre-trained vibration-component state mapping table is loaded into the edge computing node, which contains the correspondence between different frequency components and specific mechanical component states. The mechanical component states causing the vibration are deduced from the vibration spectrum signal using a reverse-analysis algorithm. The vibration amplitude variation in the 20-30Hz frequency band is analyzed to reverse-derive the tilt angle and flatness parameters of the step pedal. The tension and wear degree of the main drive chain are analyzed from the spectral characteristics of the 50-80Hz frequency band. The surface roughness and clearance accuracy of the guide rail are reconstructed using vibration signals in the 100-150Hz frequency band. The specific values ​​of the component state parameters are determined: the tilt angle accuracy of the step pedal is 0.1 degrees, the chain tension accuracy is 10N, and the guide rail clearance accuracy is 0.1 mm. Through the fusion analysis of multi-band vibration signals, the radial and axial clearance parameters of the drive motor bearing are reconstructed. The reconstructed mechanical component state parameters are categorized and organized according to component type to form a complete parameter set containing position information, state values, and change trends.

[0025] Wear state mapping is generated based on the state parameters of mechanical components. Using the reconstructed state parameters, the current wear level of each component is determined through a wear analysis algorithm. For the step pedal, the wear depth is determined based on surface flatness parameters; a 0.5 mm increase in flatness deviation corresponds to a 0.2 mm increase in wear depth. For the main drive chain, chain elongation is analyzed by tension changes; a 100 N decrease in tension corresponds to a 2 mm chain elongation. The wear condition of the guide rail is analyzed; the surface wear is determined based on changes in clearance fit accuracy; a 0.5 mm increase in clearance corresponds to a 0.3 mm wear. The wear rate of each component is calculated, equal to the wear amount divided by the operating time, in millimeters per month. The wear level is classified into four grades: slight (0-20%), moderate (20-50%), severe (50-80%), and extremely severe (80-100%). A wear state mapping table is generated, recording the location coordinates, current wear amount, wear grade, and estimated remaining life of each component.

[0026] A step clearance variation curve is generated by fitting a wear state mapping. Using component wear information from the wear state mapping, the variation law of the step clearance is determined through the clearance-wear relationship model G=αW1+βW2+γW3, where G is the clearance variation, W1, W2, and W3 are the wear amounts of the pedal, chain, and guide rail, respectively, and α, β, and γ are the corresponding influence coefficients. A linear fitting method is used to correlate the quantitative relationship between component wear and clearance variation; 1 mm of wear on the step pedal corresponds to a 0.8 mm increase in clearance, and 1 mm of wear on the guide rail corresponds to a 1.2 mm increase in clearance. Considering the wear contribution of multiple components, the overall clearance variation is obtained through weighted summation. Time sampling points are set, and clearance values ​​are determined every hour to form discrete time-clearance data points. A cubic spline interpolation method is used to fit the discrete data points, generating a smooth and continuous step clearance variation curve. The shape characteristics of the fitted curve are analyzed to identify different stages of clearance variation, such as linear segments, acceleration segments, and steady segments.

[0027] In some embodiments, the step of matching the step gap variation curve with a standard operating sample model to obtain risk characteristic probability values ​​includes: identifying heterogeneous interference source signals from the step gap variation curve; separating the heterogeneous interference source signals to obtain interference components, wherein the interference components include electromagnetic interference components, mechanical interference components, and thermal interference components; constructing an environmental state mapping table using the interference components; and matching the environmental state mapping table with a standard operating sample model to obtain risk characteristic probability values.

[0028] Heterogeneous interference sources were identified from the gap variation curves. Signal analysis was performed on the generated gap variation curves to identify abnormal fluctuations that did not belong to normal wear changes. Wavelet transform was used to decompose the gap variation curves into signal components of different frequency scales, separating high-frequency noise and low-frequency trends. Local abnormal fluctuations in the curves were analyzed, and data points exceeding the normal variation range were detected using a sliding window statistical method. The second derivative of the gap variation was determined, and the locations of sudden changes in the curve slope were identified; these locations typically correspond to the influence of external interference. An anomaly detection threshold was set; when the gap variation rate exceeded 0.05 mm / min, it was considered abnormal interference. Periodic interference signals were identified through spectral analysis: the 50 Hz harmonics corresponded to electromagnetic interference, and the harmonics of the mechanical rotation frequency corresponded to mechanical interference. The duration and intensity characteristics of the interference signals were analyzed; short-duration, high-intensity interference usually originated from electromagnetic sources, while long-duration, low-intensity interference usually originated from thermal expansion.

[0029] The heterogeneous interference source signals are separated to obtain the interference components. Independent component analysis (ICA) separates the mixed interference signals into independent components. Electromagnetic interference (EMI) components are identified through amplitude analysis of 50Hz and its harmonic frequencies; the intensity of EMI is closely related to the power of surrounding electrical equipment. EMI components are extracted from the gap variation curve; these components exhibit high-frequency periodic fluctuations with an amplitude range of 0.1-0.5 mm. Mechanical interference components correspond to the vibration effects of the equipment's mechanical movement and are identified through harmonic analysis of the ladder runway frequency. When extracting mechanical interference components, the focus is on low-frequency vibration components of 1-10Hz, as these components have the most significant mechanical coupling with the ladder gap. Thermal interference components reflect the impact of temperature changes on the ladder gap and are extracted through correlation analysis between gap changes and temperature changes. The relative contribution of each interference component is determined: EMI accounts for 30% of the total interference, mechanical interference accounts for 50%, and thermal interference accounts for 20%. Amplitude normalization is performed on the separated interference components to map the influence of different types of interference to a unified evaluation standard.

[0030] For example, constructing an environment state mapping table using the interference components includes: detecting data inconsistency features between the interference components; converting the data inconsistency features into verification confidence enhancement parameters; using the verification confidence enhancement parameters to perform data consistency correction to generate corrected interference components; and establishing an environment state mapping table based on the corrected interference components.

[0031] The system detects data inconsistencies among interference components. It analyzes the temporal synchronicity of each interference component, detecting time delay differences through cross-correlation analysis. It analyzes the amplitude correlation of interference components; under normal circumstances, electromagnetic interference and mechanical interference should have a weak correlation, while thermal interference has a low correlation with other components. It identifies abnormally strong correlations; when the correlation coefficient between electromagnetic and thermal components exceeds 0.7, it is considered data inconsistency. It detects abnormal causal relationships between components, using Granger causality tests to analyze whether the direction of influence between components conforms to physical laws. It analyzes the statistical consistency of component data, comparing the mean, variance, and distribution shape of each component to ensure they are within reasonable ranges. It identifies time-domain inconsistencies; when one component experiences a sudden change while other components remain stable, it is marked as time-domain inconsistency. It records frequency-domain inconsistencies, analyzing whether there are abnormal overlaps or missing spectral components in each component.

[0032] Data inconsistency features are transformed into credibility enhancement parameters. Based on the type and severity of the detected inconsistencies, corresponding credibility impact coefficients are determined. The impact coefficient for time synchronization inconsistencies is 0.8, meaning time delay reduces data credibility by 20%. The impact coefficient for anomalous amplitude correlation is 0.7, meaning abnormally strong correlations reduce credibility by 30%. The impact coefficient for anomalous causal relationships is 0.6, meaning causal relationships violating physical laws reduce credibility by 40%. The impact coefficient for anomalous statistical consistency is 0.9, meaning minor deviations in statistical features only reduce credibility by 10%. A comprehensive impact coefficient is derived, and the overall credibility coefficient is obtained by multiplying the multiple individual impact coefficients. Credibility coefficients are then converted into enhancement parameters, which are equal to 1 minus the credibility coefficient; a larger value indicates the need for more correction processing. Grading thresholds for enhancement parameters are set: 0-0.2 for slight enhancement, 0.2-0.5 for moderate enhancement, 0.5-0.8 for strong enhancement, and above 0.8 for critical enhancement.

[0033] Data consistency correction is performed using confidence enhancement parameters to generate corrected interference components. Based on the obtained confidence enhancement parameters, an appropriate data correction algorithm is selected to correct the consistency of the interference components. For slight inconsistencies (0-0.2), a simple linear filtering method is used to eliminate random fluctuations in the data, with the filter window length set to 5 data points. For moderate inconsistencies (0.2-0.5), a cubic spline interpolation algorithm is used to fill in missing data and anomalous jumps, with the interpolation node spacing controlled within half of the original sampling interval. For severe inconsistencies (0.5-0.8), wavelet denoising technology is used to separate signal and noise components, preserving the main signal features while removing inconsistent noise. For severe inconsistencies (above 0.8), a signal reconstruction algorithm is used to completely replace the anomalous data segments, with the reconstructed data generated based on the statistical characteristics of adjacent normal data segments. For time synchronization inconsistencies, a time alignment algorithm is used to eliminate time delays between components, and the optimal time offset is determined through a cross-correlation function. For amplitude correlation anomalies, an amplitude normalization method is used to unify the numerical range of each component and eliminate the influence of dimensional differences. When correcting causal anomalies, the influence weights between components are readjusted according to physical constraints to ensure that the causal relationships conform to the physical laws of equipment operation. The corrected electromagnetic, mechanical, and thermal interference components are reorganized into a complete dataset to ensure the inherent consistency and logical rationality of the data.

[0034] An environmental state mapping table is constructed based on the corrected interference components. The electromagnetic environment level is redefined using the corrected electromagnetic interference components. These corrected components eliminate the influence of measurement errors and coupling interference, and the threshold for determining the electromagnetic environment level is adjusted accordingly to three levels: mild electromagnetic environment, moderate electromagnetic environment, and strong electromagnetic environment. The standard for mechanical environment complexity is adjusted based on the corrected mechanical interference components: single-frequency components correspond to simple mechanical environments, and multi-frequency components correspond to complex mechanical environments. The thermal environment stability index is redefined using the corrected thermal interference components: thermal environments with small amplitude and rate of change are stable thermal environments, while those with drastic changes are unstable thermal environments. A new version of the three-dimensional environmental state mapping table is constructed, with the three dimensions corresponding to the corrected electromagnetic environment, mechanical environment, and thermal environment parameters, respectively. A classification system for different environment combinations is established in the new mapping table: strong electromagnetic-complex mechanical-unstable thermal environment constitutes a high-complexity environment type, while mild electromagnetic-simple mechanical-stable thermal environment constitutes a low-complexity environment type. The environmental state mapping table is stored according to typical environment combinations, with each combination including the corresponding environmental complexity level and environmental characteristic description.

[0035] Risk characteristic probability values ​​are obtained by matching an environmental state mapping table with a standard operating sample model. The obtained environmental state mapping table is input into the standard operating sample model for environmental adaptability matching analysis. The standard operating sample model includes equipment performance benchmarks and risk threshold parameters under different environmental conditions. The corresponding environmental type in the model is queried through the environmental state mapping table to determine the equipment operating benchmark under the current environmental conditions. The degree of deviation between the actual operating state and the environmental benchmark is analyzed; a larger deviation indicates a higher level of risk. The degree of deviation is converted into risk characteristic probability values ​​using a probability density function: a deviation within 1 standard deviation corresponds to a probability value of 0.1-0.3, a deviation within 2 standard deviations corresponds to a probability value of 0.3-0.6, and a deviation exceeding 2 standard deviations corresponds to a probability value of 0.6-1.0. The influence weights of different interference components on risk are considered: electromagnetic interference has a weight of 0.3, mechanical interference has a weight of 0.5, and thermal interference has a weight of 0.2. A weighted average is used to obtain a comprehensive risk characteristic probability value, which comprehensively reflects the operational risk level of the equipment under the current environmental conditions.

[0036] A state traction chain is generated based on the probability values ​​of risk characteristics. The state traction chain is a dynamic equilibrium mechanism for the interaction between equipment states. Normal operation has the ability to pull abnormal states back to a safe range, while abnormal states also exert a reverse drag on normal operation, creating a continuous tug-of-war. The obtained probability values ​​of risk characteristics are used as the driving signal for traction strength, constructing state traction modes of different intensities. For example, when an escalator in a shopping mall experiences an abnormal widening of the step gap, the normal vibration state of the equipment will be pulled back towards the standard range by the traction chain, while the abnormal gap will also pull the overall equipment state towards instability, eventually reaching a dynamic equilibrium. When a high-risk state (probability value 0.8-1.0) is detected, the chain activates a forced traction mode, forcefully pulling the abnormal state back to a safe range through emergency maintenance measures. A medium-risk state (probability value 0.6-0.8) triggers a coordinated traction mode, where normal operation and risk states form a balanced tug-of-war, gradually pulling the equipment back to a stable range through preventative maintenance. In low-risk states (probability value 0.4-0.6), a monitoring and traction mode is activated. In normal states, potential anomalies are continuously restrained to prevent the state from sliding towards the riskier direction. Different traction modes are constructed into a hierarchical state traction chain according to traction strength, with strong traction taking precedence over weak traction, forming a multi-level state interaction network.

[0037] Step S130: Obtain the displacement distribution coefficient and vibration velocity index from the state traction chain, predict the collision probability distribution of the displacement distribution coefficient using a cloud-based deep learning algorithm, determine the warning level based on the vibration velocity index, and generate an optimized warning configuration by combining the collision probability distribution and the warning level.

[0038] Displacement distribution coefficients and vibration velocity indices are obtained from the state traction chain. Using the state traction chain as a parametric traction source, the displacement distribution coefficient, reflecting the spatial displacement characteristics of the cascade equipment, is extracted through a chain traction mechanism. The displacement distribution coefficient plays a spatial state traction role in the traction chain, ensuring horizontal stability in the X-direction, vertical balance in the Y-direction, and symmetry before and after displacement in the Z-direction. When displacement deviates from the normal range in one direction, other directions are compensated by the traction chain, forming a dynamic traction balance in three-dimensional space. Statistical characteristic parameters of displacement distribution are extracted by analyzing the traction strength and direction of displacement parameters in the traction chain. The value of the displacement distribution coefficient reflects the stability of the traction balance; a larger value indicates a more severe traction imbalance and poorer equipment operational stability. Simultaneously, vibration velocity indices are extracted from the state traction chain. These indices reflect the mutual traction relationships between different frequency bands in the frequency domain. This index reflects the vibration intensity and severity of the moving parts of the equipment. Vibration velocity is obtained through time-domain integration; the vibration acceleration signal is integrated to obtain the time-domain sequence of vibration velocity. The main moving parts of the equipment are used as the velocity monitoring range, and velocity measurement and analysis are performed according to the component positions. The vibration velocity index is expressed by the formula V = √(1 / T∫v²(t)dt), where V is the RMS value of the vibration velocity, v(t) is the instantaneous vibration velocity, and T is the integration time window. Through time-series evolution analysis of the traction chain, the traction change trajectories of the displacement distribution coefficient and the vibration velocity index are obtained, forming a dynamic dataset of the mutual traction between parameters.

[0039] In some embodiments, predicting the collision probability distribution of the displacement distribution coefficient using a cloud-based deep learning algorithm includes: monitoring the performance degradation degree of each sensor in the displacement distribution coefficient; identifying the degradation mode of the performance degradation degree using a cloud-based deep learning algorithm; converting the degradation mode into system robustness enhancement parameters; and generating a collision probability distribution based on the system robustness enhancement parameters.

[0040] The performance degradation of each sensor in the displacement distribution coefficient is monitored. The performance status of each sensor constituting the displacement distribution coefficient is monitored, and performance degradation is identified by analyzing the characteristic changes in the sensor output signals. Sensor performance degradation is mainly manifested as decreased measurement accuracy, slower response speed, and increased signal noise. The degree of degradation is quantified by comparing the deviation of the current sensor output from the standard reference value: a deviation exceeding 5% is considered slight degradation, a deviation exceeding 15% is considered moderate degradation, and a deviation exceeding 30% is considered severe degradation. Changes in the signal-to-noise ratio (SNR) of the sensor signals are analyzed. A normal sensor's SNR should be maintained above 40dB; a SNR of 30-40dB indicates slight degradation, 20-30dB indicates moderate degradation, and below 20dB indicates severe degradation. Changes in sensor response time are monitored by measuring the delay time from excitation to response by applying a standard test signal. A normal response time is 10-50 milliseconds; a response time exceeding 100 milliseconds is considered performance degradation. The degree of degradation of each sensor over different time periods is statistically analyzed to form a time distribution chart of sensor performance degradation. Sort all sensors by the degree of attenuation and identify the critical sensors with the most severe attenuation.

[0041] The degradation patterns of sensor performance are identified using a cloud-based deep learning algorithm. Performance degradation data from each sensor is uploaded to a cloud-based deep learning platform, and pattern recognition algorithms are used to analyze the development patterns and typical modes of degradation. The deep learning algorithm employs a convolutional neural network structure, containing five convolutional layers and two fully connected layers, capable of automatically extracting feature patterns from the degradation data. Through training, six typical degradation patterns were identified, including linear degradation, exponential degradation, step degradation, fluctuating degradation, periodic degradation, and mixed degradation. Linear degradation is characterized by a steady decline in sensor performance at a constant rate. Exponential degradation is characterized by rapid performance deterioration with an accelerating rate of degradation over time. Step degradation is characterized by a sudden and significant drop in performance, typically caused by external shocks or device failure. Fluctuating degradation is characterized by repeated fluctuations in performance during the degradation process. Periodic degradation is characterized by performance decaying and recovering according to a fixed period. Mixed degradation involves a combination of various degradation characteristics. The cloud-based algorithm determines the dominant degradation pattern of the current sensor group by comparing the similarity between measured degradation data and typical patterns. Finally, the matching probability of each degradation pattern is output, with the pattern with the highest probability being the dominant pattern.

[0042] The decay patterns are transformed into system robustness enhancement parameters. Based on the identified decay pattern characteristics, the pattern information is converted into quantified system robustness enhancement parameters. Different decay patterns have different degrees of impact on system robustness, requiring corresponding enhancement parameters for compensation. The enhancement parameter for linear decay patterns is R = 1 + 0.2 × t, where t is the decay time; this parameter compensates for linear performance degradation. The enhancement parameter for exponential decay patterns uses an exponential form for compensation to prevent the impact of rapid performance deterioration on system stability. The enhancement parameter for step decay patterns is set to a fixed gain to offset sudden performance jumps. The enhancement parameter for fluctuating decay patterns uses adaptive filtering to smooth the impact of performance fluctuations. The enhancement parameter for periodic decay patterns uses phase compensation to offset periodic fluctuations. The enhancement parameter for hybrid decay patterns is generated through a weighted combination of multiple compensation methods. The numerical range of the enhancement parameters is 1.0-2.0; a larger value indicates a stronger need for robustness enhancement. Statistical analysis of the enhancement parameters of all sensors is performed to determine the overall robustness enhancement requirements of the system.

[0043] A collision probability distribution is generated based on system robustness enhancement parameters. Using these parameters, a probability generation algorithm is employed to derive the probability distribution of collision events. Higher enhancement parameter values ​​indicate severe sensor performance degradation, decreased system reliability, and a corresponding higher collision probability. A mapping relationship between enhancement parameters and collision probabilities is established: enhancement parameters of 1.0-1.2 correspond to low collision probabilities (0.1-0.2), 1.2-1.5 correspond to medium collision probabilities (0.2-0.5), and 1.5-2.0 correspond to high collision probabilities (0.5-0.8). Considering the contribution weights of different sensors to system safety, the degradation of critical sensors has a greater impact on collision probability. A weighted average method is used to synthesize the influence of all sensors, resulting in a baseline collision probability value for the entire system. Based on this baseline probability, considering uncertainties, a probability distribution function is used to describe the range of collision probability variation. The generated collision probability distribution includes statistical parameters such as the probability mean, standard deviation, and confidence interval. The collision probability distribution is organized according to a time series to reflect the trend of probability changes over time. The shape and characteristics of the probability distribution can be used to judge the development trend of system risk. The more dispersed the distribution, the higher the uncertainty of the risk.

[0044] The warning level is determined based on the vibration velocity index. Using the acquired vibration velocity index, the current warning level of the equipment is determined according to preset level judgment criteria. The warning levels are divided into four levels: Normal (0-5 mm / s), Attention (5-10 mm / s), Warning (10-20 mm / s), and Emergency (above 20 mm / s). When the vibration velocity index is within the Normal range, it indicates that the vibration intensity of the moving parts of the equipment is low and the operating status is stable. Attention indicates that the vibration velocity is beginning to increase, the dynamic response of the moving parts of the equipment is enhanced, requiring increased monitoring but not yet reaching an abnormal level. Warning indicates a significant increase in vibration velocity, indicating that the equipment has motion overload or mechanical loosening, posing a potential risk of failure, and requiring inspection and maintenance. Emergency indicates that the vibration velocity has reached a dangerous level, and the moving parts of the equipment may experience serious mechanical failure, requiring immediate shutdown and repair. The warning level is adjusted according to the changing trend of the vibration velocity index; when the index rises continuously, the warning level is upgraded in advance, and when the index falls, the warning level is downgraded later. For example, if the vibration velocity index of a subway escalator increases from 4.2 mm / s to 6.8 mm / s, the warning level is raised from normal to concern, triggering enhanced monitoring measures. The determined warning level is then associated with and stored along with the corresponding timestamp, forming a time-series record of the warning level.

[0045] In some embodiments, generating an optimized warning configuration by combining the collision probability distribution and the warning level includes: identifying regions with missing monitoring data in the collision probability distribution; converting the regions with missing monitoring data into an interpolation algorithm training domain; performing data completion processing using the interpolation algorithm training domain to obtain a complete dataset; and generating an optimized warning configuration by fusing the complete dataset with the warning level.

[0046] Identify missing regions in the collision probability distribution. Perform an integrity check on the generated collision probability distribution to identify missing, anomaly, and blank areas in the data sequence. Data loss is typically caused by sensor malfunction, communication interruption, or data transmission errors. Detect the location and extent of missing data through time series continuity analysis; data loss is defined as occurring when the time interval between adjacent data points exceeds twice the normal sampling interval. Analyze the distribution pattern of missing data to identify different types, such as random loss, continuous loss, and periodic loss. Random loss manifests as isolated missing points scattered throughout the time series, continuous loss manifests as data gaps over long periods, and periodic loss manifests as data loss occurring at fixed intervals. Statistically count the number and proportion of each type of loss; when missing data exceeds 10% of the total data volume, specialized completion processing is required. Mark the boundary locations and duration of missing regions to determine the working range for subsequent interpolation processing. Analyze the characteristics of the data surrounding the missing regions to identify the trends and statistical features of data changes before and after the loss.

[0047] The missing data regions are transformed into training domains for interpolation algorithms. Based on the characteristics of the identified missing data regions, corresponding training domains for each type of missing data are designed. The training domain includes the valid data surrounding the missing region and the corresponding interpolation target region. For random missing points, the training domain is set to a range of 10 data points before and after the missing point, utilizing the local features of neighboring data for interpolation. For continuous missing segments, the training domain is expanded to 20 data points before and after the missing segment, utilizing the features of a larger range for long-segment interpolation. For periodic missing data, the training domain is set to a data segment containing the complete cycle, utilizing the periodic pattern to reconstruct the missing data. Statistical features of the data are extracted from the training domain, including parameters such as mean, variance, autocorrelation function, and spectral characteristics. The most suitable interpolation algorithm type is determined through feature analysis: linear interpolation is used for linearly varying data, spline interpolation for nonlinearly varying data, and harmonic interpolation for periodic data. Corresponding training domains for different missing regions are assigned to form a complete interpolation processing scheme. The setting of the training domain considers the boundary effects of the data and the interpolation accuracy requirements to ensure the reliability and accuracy of the interpolation results.

[0048] A complete dataset is obtained by performing data completion processing on the training domain using an interpolation algorithm. Based on the designed interpolation algorithm training domain, corresponding numerical interpolation methods are used to complete the missing data. For linear interpolation, a straight line is fitted using two valid data points before and after the missing point, and the completed data is obtained through a linear function. For cubic spline interpolation, a cubic polynomial function is constructed using multiple data points before and after the missing segment to ensure the continuity and smoothness of the interpolation curve at the boundary. For harmonic interpolation, the frequency components of the training domain data are extracted through Fourier transform, and the missing data is reconstructed using the main frequency components. The physical constraints of the data are considered during interpolation processing to ensure that the completed data is within a reasonable numerical range. The interpolation results are quality checked, and the reliability of the interpolation is judged by consistency analysis with neighboring data. When there is a significant inconsistency between the interpolated data and the surrounding data, the interpolation parameters are adjusted or the interpolation method is changed. The completed data of all missing regions are integrated into the original data sequence to form a complete dataset with continuous time. Statistical analysis is performed on the complete dataset to confirm that the statistical characteristics of the completed data are consistent with the original data.

[0049] An optimized early warning configuration is generated by fusing a complete dataset with early warning levels. Using the complete collision probability distribution dataset after data completion, combined with the early warning levels determined by vibration velocity indicators, an optimal early warning configuration scheme is generated through a fusion algorithm. The fusion processing employs a multi-criteria decision-making method, simultaneously considering the magnitude and trend of collision probabilities, as well as the constraints of early warning levels. When the complete dataset shows a continuously rising collision probability and the early warning level is warning, a progressive early warning configuration is generated, gradually increasing the early warning intensity and response speed. When the collision probability suddenly jumps and the early warning level becomes emergency, an emergency early warning configuration is generated, immediately initiating the highest level of emergency response measures. The optimized early warning configuration includes specific details such as early warning trigger thresholds, response time requirements, information delivery targets, and processing procedures. The early warning threshold is dynamically adjusted based on the statistical characteristics of the complete dataset; when data fluctuations are large, the threshold is lowered to increase sensitivity, and when data is stable, the threshold is raised to reduce false alarms. Early warning levels and collision probabilities are weighted, with a weight coefficient of 0.7 for high-level warnings and 0.3 for high-probability risks. A comprehensive early warning intensity index is derived through weighted fusion, which serves as the primary basis for selecting early warning configurations. The generated early warning optimization configurations are categorized and stored according to different risk scenarios, and each configuration includes detailed execution steps and parameter settings.

[0050] Step S140: Based on the early warning optimization configuration, perform linkage control analysis to determine the emergency distribution, identify differentiated response areas through the emergency distribution, and extract linkage optimization factors from the differentiated response areas.

[0051] Specifically, emergency distribution is determined through linkage control analysis based on the early warning optimization configuration. Using the early warning optimization configuration as input data, linkage control analysis determines the response distribution pattern of each component in an emergency state. The early warning optimization configuration includes response strategies and execution parameters corresponding to different risk levels, which need to be converted into specific component control commands. The linkage control analysis adopts a multi-level response architecture: the first layer is the immediate response of core safety components, the second layer is the coordinated response of auxiliary monitoring components, and the third layer is the delayed response of peripheral protection components. For high-risk early warning configurations, the response time of core components is set to within 0.5 seconds, the response time of auxiliary components is 1-2 seconds, and the response time of peripheral components is 3-5 seconds. The spatial configuration of emergency distribution is divided according to the physical structure of the equipment: 40% of the response resources are allocated to the drive system area, 35% to the safety protection system area, and 25% to the monitoring and sensing system area. The linkage sequence of each component is determined through a response priority matrix, with the safety braking system having the highest priority, followed by the emergency stop switch, and the status indication system having the lowest priority. The emergency response time configuration adopts a phased response mode: 0-2 seconds is the emergency braking phase, 2-10 seconds is the status stabilization phase, and 10-30 seconds is the safety confirmation phase. The response intensity of the linkage control is matched with the warning level: emergency level warning triggers 100% response intensity, warning level triggers 70% response intensity, and attention level triggers 40% response intensity.

[0052] In some embodiments, identifying differentiated response regions through the emergency distribution includes: detecting system response delays in the emergency distribution; converting the system response delays into early warning window expansion parameters; using the early warning window expansion parameters to adaptively divide response regions to generate region configurations; and forming differentiated response regions based on the region configurations.

[0053] Detecting system response time delays in emergency distribution systems. This involves measuring and analyzing the actual response times of each component within a defined emergency distribution system to identify time delays during the response process. System response time delays primarily originate from factors such as signal transmission delays, processor processing delays, and actuator action delays. By deploying time measurement devices at key nodes in the emergency distribution system, the entire process from command issuance to action completion is accurately recorded. Signal transmission delays are obtained by measuring the propagation time of commands in the communication network; normal transmission delays should be controlled within 10 milliseconds. Processor processing delays are obtained by monitoring the command processing time of the controller; the processing time for complex commands can reach 50-100 milliseconds. Actuator action delays are obtained by measuring the time from receiving a command to completing an action in mechanical components; mechanical delays are typically 200-500 milliseconds. Various delays are categorized and statistically analyzed according to their sources to identify the main contributing factors. Statistical distributions of delay times are obtained through multiple measurements to analyze the stability and variation patterns of delays. Abnormal time delays are detected; a response time exceeding 150% of the normal range is considered an abnormal time delay.

[0054] The system response delay is converted into extended parameters for the early warning window. Based on the detected system response delay data, an extended parameter is generated using a conversion algorithm to adjust the early warning time window. The purpose of the extended parameters is to compensate for the impact of system response delay on the early warning effect, ensuring that early warning information can be delivered in time before a hazard occurs. The conversion algorithm uses a linear mapping method, with the extended parameter T_ext = α × T_delay + β, where T_ext is the extended parameter for the early warning window, T_delay is the system response delay, α is the proportional coefficient, and β is the base extended value. The value of the proportional coefficient α is determined according to the system's safety margin requirements; a larger α value is set for systems with higher safety requirements. The base extended value β is used to compensate for unpredictable random delays, typically 20% of the average response delay. Different conversion parameters are set for different types of response actions: α is set to 1.5 for braking actions, 1.2 for monitoring actions, and 1.0 for indicating actions. Considering the uncertainty of the response delay, a safety margin is added to the extended parameters, the size of which is determined based on the standard deviation of the delay. The converted extended parameters are grouped according to the response region, with each region corresponding to a specific extended parameter value.

[0055] An adaptive partitioning of response areas using early warning window extension parameters generates the area configuration. Utilizing the obtained early warning window extension parameters, the spatial distribution and functional allocation of emergency response areas are reconfigured through an adaptive partitioning algorithm. The principle of adaptive partitioning is to group components with similar extension parameters into the same response area, achieving synchronous response within each area. In practical applications of shopping mall escalators, when an abnormal widening of the step gap is detected, the system automatically regroups drive motors, main brakes, and safety sensors with similar response delays into the core control area, ensuring that these critical components can respond to abnormal situations in a coordinated manner. Simultaneously, safety devices requiring instantaneous response, such as emergency stop buttons and safety light curtains, are automatically merged into the fast response area, while auxiliary equipment such as status displays and ambient lighting are assigned to the delayed response area. The physical layout constraints of the equipment are fully considered during the area partitioning process; adjacent step sensors and guide rail monitoring devices are preferentially assigned to the same area, reducing the complexity of cross-area coordination. The load balancing mechanism of the area configuration ensures that no single area bears an excessively heavy response task, and the overall system response efficiency is optimized by dynamically adjusting the component allocation in each area. The regional configuration includes detailed boundary definitions, component ownership lists, response timing arrangements, and computing resource allocation schemes, forming a complete regional management framework.

[0056] Differentiated response zones are formed based on regional configuration. These differences are primarily reflected in key aspects such as response speed, control strategy, resource allocation, and priority settings. The fast response zone is specifically responsible for handling safety-critical functions, equipped with high-performance sensors and rapid actuators to ensure an instantaneous response upon detecting a danger signal. During peak hours in subway stations, when a surge in passenger flow causes abnormal escalator load, the fast response zone's braking system and safety protection devices can activate protection procedures in a very short time. The standard response zone handles daily monitoring and control tasks, using conventionally configured sensors and actuators. Upon receiving abnormal operation signals, it performs status checks and parameter adjustments according to established procedures. The delayed response zone primarily manages auxiliary functions and status indications, using shared resources and low-power devices. In emergencies, it can temporarily reduce priority or postpone processing. The control strategies of different response zones also differ significantly. The fast response zone employs a hard real-time control mechanism, bypassing software processing through direct hardware control; the standard response zone uses soft real-time control for logical decision-making; and the delayed response zone uses a non-real-time batch processing method. In terms of resource allocation, key areas receive more processing power, storage capacity and communication bandwidth support, and the priority setting mechanism ensures that high-priority areas can get services first when there is resource competition.

[0057] In some embodiments, extracting linkage optimization factors from the differentiated response region includes: collecting the intensity of complex environmental interference within the differentiated response region; converting the intensity of complex environmental interference into a system environmental adaptability evaluation index; generating adaptive test data based on the system environmental adaptability evaluation index; and extracting linkage optimization factors from the adaptive test data, wherein the linkage optimization factors include environmental adaptability adjustment parameters and vibration suppression adaptation parameters.

[0058] The intensity of complex environmental interference within the differentiated response area was collected. Complex environmental interference includes various types such as electromagnetic interference, mechanical vibration interference, temperature change interference, and humidity fluctuation interference. Electromagnetic interference was collected using a broadband electromagnetic field strength meter, covering a frequency range of 1MHz-1GHz, recording the peak and average values ​​of the electromagnetic field intensity in each frequency band. Mechanical vibration interference was measured using a triaxial accelerometer with a sampling frequency of 2000Hz, recording the vibration amplitude and spectral distribution in the X, Y, and Z directions. Temperature change interference was collected using a distributed temperature sensor network with a sensor spacing of 2 meters, recording the temperature gradient and rate of change. Humidity fluctuation interference was measured using a capacitive humidity sensor, recording the relative humidity value and its trend. The intensity of the collected interference data was quantified, converting the physical quantities of different types of interference into dimensionless intensity indices. The interference intensity was calculated using a normalization method, with the ratio of the measured value to the normal environmental baseline value used as the intensity index. The intensity index ranged from 0 to 10, with higher values ​​indicating stronger interference.

[0059] The intensity of interference in complex environments is transformed into a system environmental adaptability index. System environmental adaptability reflects the ability of equipment to maintain normal operation under complex environmental conditions. The transformation of the adaptability index adopts an inverse proportional relationship; the higher the intensity of environmental interference, the stronger the system adaptability requirement. A mapping function A=k / (1+I^n) is established between interference intensity and adaptability, where A is the adaptability index, I is the environmental interference intensity, k is the adaptability coefficient, and n is the nonlinear exponent. The value of the adaptability coefficient k is determined based on the design performance of the equipment; high-performance equipment has a larger k value, indicating stronger environmental adaptability potential. The nonlinear exponent n is used to adjust the sensitivity of adaptability to changes in interference intensity; a larger n value indicates greater sensitivity to strong interference. Different transformation parameters are set for different types of environmental interference: n=2.0 for electromagnetic interference, 1.5 for vibration interference, and 1.2 for temperature and humidity interference. A comprehensive environmental adaptability index is obtained by combining the effects of multiple interference types using a weighted average method. The weight allocation for each interference type is determined based on its impact on equipment performance: vibration interference has a weight of 0.4, electromagnetic interference has a weight of 0.3, and temperature and humidity interference has a weight of 0.3.

[0060] For example, obtaining adaptive test data based on the system environment adaptability evaluation index includes: identifying current scenario response strategy conflicts based on the system environment adaptability evaluation index; converting the current scenario response strategy conflicts into decision architecture reorganization trigger signals; using the decision architecture reorganization trigger signals to perform real-time decision architecture self-reorganization to generate a new decision architecture; and using the new decision architecture to perform adaptive capability testing to obtain adaptive test data.

[0061] System environmental adaptability indicators are used to identify response strategy conflicts in the current scenario. Using these indicators, a strategy conflict detection algorithm identifies contradictions and conflicts between different response strategies in the current operating scenario. Response strategy conflicts typically manifest in various forms, such as resource allocation conflicts, timing conflicts, and priority conflicts. In the complex operating environment of office building escalators, multiple strategies are often activated simultaneously. For example, when the system simultaneously detects slight abnormal noise from the steps and passenger overload, the noise detection strategy requires reducing the operating speed for careful listening and analysis, while the overload protection strategy requires immediately stopping passenger transport. These two strategies exhibit significant contradictions in execution timing and control objectives. Another common conflict arises between vibration control and energy-saving operation. When the escalator experiences slight vibration, the vibration suppression strategy tends to adjust the drive frequency to reduce mechanical resonance, but the energy-saving operation strategy requires maintaining the current optimal efficiency frequency setting, creating a direct conflict in control parameters. Strategy compatibility analysis quantifies the coordination possibilities of different strategies by constructing a strategy matching matrix. When the environmental adaptability indicator is low, system resources become strained, and multiple monitoring and control strategies compete for limited processing resources and execution channels.

[0062] The current scenario response strategy conflict is transformed into a decision architecture reorganization trigger signal. The generation of this trigger signal is based on factors such as the severity, scope of impact, and duration of the conflict. Minor conflicts generate parameter adjustment trigger signals (intensity 1-3), triggering fine-tuning of local parameters. Moderate conflicts generate strategy switching trigger signals (intensity 4-6), triggering the replacement or reordering of some strategies. Severe conflicts generate architecture reorganization trigger signals (intensity 7-9), triggering the reconstruction of the entire decision architecture. The trigger signals are encoded using a multi-dimensional vector format, including signal type, intensity level, affected area, and time constraints. Signal type encoding distinguishes between parameter adjustment, strategy switching, and architecture reorganization, corresponding to encoding values ​​1, 2, and 3, respectively. Intensity level encoding uses a 9-level system, with higher levels indicating a more urgent reorganization need. The affected area encoding identifies the response area requiring reorganization, using a bitmap to represent a combination of multiple areas. The time constraint encoding indicates the time window requirement for the reorganization operation; urgent reorganization requires completion within 1 second, while regular reorganization allows 5-10 seconds of processing time.

[0063] A new decision architecture is generated through real-time self-reorganization using decision architecture reorganization trigger signals. Based on the generated reorganization trigger signals, the system's decision architecture is adjusted and reconstructed in real time through a self-reorganization algorithm. The decision architecture self-reorganization adopts a modular reconfiguration approach, decomposing the decision function into independently schedulable functional modules. Appropriate reorganization strategies are selected based on the type and strength of the trigger signals: parameter adjustment signals trigger parameter optimization reorganization, strategy switching signals trigger strategy replacement reorganization, and architecture reorganization signals trigger topology reorganization. Parameter optimization reorganization improves decision performance by adjusting the internal parameters of the decision modules, with the adjustment range controlled within ±20% of the original parameter values. Strategy replacement reorganization replaces conflicting strategies by activating backup decision strategies; the backup strategy library contains preset strategy schemes for different scenarios. Topology reorganization changes the information flow and control logic by reconnecting the decision modules, generating a decision architecture adapted to the new environment. The reorganization process employs a smooth switching method to avoid the impact of reorganization operations on system operation.

[0064] Adaptability testing was conducted using a new decision-making architecture to obtain adaptive test data. The newly restructured decision-making architecture was used to perform system performance tests through an adaptive capability testing program, obtaining test data reflecting the adaptive performance of the new architecture. Adaptability testing covered multiple key performance dimensions, including response speed, control accuracy, stability, and robustness. In actual testing of the shopping mall escalator, the response speed test simulated a sudden influx of passengers, observing the complete response time of the new decision-making architecture from detecting load changes to initiating corresponding control measures. The control accuracy test simulated the gradual widening of the step gap, quantifying the accuracy of the new architecture in perceiving subtle changes by comparing the system's predicted gap change trend with the actual measurement results. The stability test allowed the escalator to run continuously for several hours under the control of the new decision-making architecture, monitoring the fluctuation amplitude and change patterns of various operating parameters, paying particular attention to the consistency of system output during peak passenger flow periods. The robustness test examined the adaptability of the new architecture under different environmental disturbances, such as drastic temperature changes, increased electromagnetic interference, or intensified mechanical vibration, observing whether the system could still maintain normal monitoring and control functions. The test execution adopts a programmed automatic operation mode, which completes various performance tests in sequence according to the preset test process and automatically records the results data.

[0065] The system extracts linkage optimization factors from adaptive test data, including environmental adaptability adjustment parameters and vibration suppression adaptation parameters. Through in-depth analysis of the adaptive test data, optimization factors that significantly impact the system's linkage performance are identified and extracted. The extraction of environmental adaptability adjustment parameters focuses on the system's performance adjustment patterns under different environmental conditions such as temperature fluctuations, humidity changes, and electromagnetic interference. It was found that when the ambient temperature rises, the system needs to adjust the sensitivity threshold of the vibration sensor and the time window for data filtering accordingly; these adjustments are extracted as environmental adaptability adjustment parameters. By analyzing the correlation patterns between environmental changes and system performance, the applicable range and adjustment strategies for key adjustment parameters are determined. Environmental adaptability adjustment parameters mainly include core control elements such as sensor gain adjustment coefficients, signal filtering time constants, and anomaly detection response thresholds. The extraction of vibration suppression adaptation parameters focuses on the system's ability to suppress various mechanical vibration interferences and its adaptive adjustment mechanism. Vibration suppression adaptation parameters include technical elements such as filter configuration parameters for specific frequency bands, damping adjustment coefficients for vibration compensation, and frequency compensation parameters for dynamic response.

[0066] Step S150: Use the linkage optimization factor to generate differentiated control parameters, and match the differentiated control parameters with the state traction chain to generate an early warning signal.

[0067] Specifically, a linkage optimization factor is used to generate differentiated control parameters. This factor includes two main types: environmental adaptability adjustment parameters and vibration suppression adaptation parameters, which need to be converted into their corresponding control parameter forms. The conversion of environmental adaptability adjustment parameters focuses on key control elements such as gain coefficient, time constant, and response threshold. These parameters are allocated according to response regions. High-sensitivity parameters are assigned to the fast response region, with a gain coefficient set to 1.2-1.5 and a response threshold reduced by 20% to improve response speed. Standard sensitivity parameters are used in the standard response region, with the gain coefficient maintained at 1.0 and the response threshold kept at its original setting. Low-sensitivity parameters are used in the delayed response region, with a gain coefficient set to 0.8-0.9 and a response threshold increased by 30% to reduce false triggering. The conversion of vibration suppression adaptation parameters generates control elements such as filter parameters, damping coefficient, and frequency compensation parameters. For low-frequency vibration interference of 20-50Hz, corresponding low-pass filter parameters are generated, with a cutoff frequency set to 60Hz and a damping coefficient of 0.7. For mid-frequency vibration interference (50-120Hz), band-stop filter parameters are generated, with a center frequency of 85Hz and a bandwidth of 40Hz. For high-frequency vibration interference (120-200Hz), high-damping control parameters are generated, with the damping coefficient increased to 1.2. The generated differentiated control parameters are categorized and organized according to the controlled object, with different parameter sets corresponding to the drive system parameters, monitoring system parameters, and protection system parameters.

[0068] Differentiated control parameters are matched with the state traction chain to generate early warning signals. The state traction chain contains state relationships with different traction intensities and directions, which need to be matched and analyzed with the adjustment capabilities of the differentiated control parameters. The matching process adopts a multi-dimensional comparison method, considering factors such as traction intensity, response area, control type, and parameter adjustment range. When a forced traction mode appears in the state traction chain and the high-sensitivity parameters in the fast response area can provide sufficient traction force, a high-level early warning signal is generated. High-level early warning signals are characterized by high signal strength, extremely short response time requirements, and the highest information transmission priority. When the state traction chain exhibits a coordinated traction mode and the parameter configuration in the standard response area can maintain traction balance, a medium-level early warning signal is generated, with moderate signal strength, moderate response time, and medium information transmission priority. When the state traction chain is in a monitoring traction mode and the low-sensitivity parameters in the delayed response area are sufficient to maintain state restraint, a low-level early warning signal is generated, with low signal strength, lenient response time, and general information transmission priority. The generation of early warning signals adopts a real-time matching method, and the matching analysis is immediately re-performed whenever the adjustment capability of the differentiated control parameters or the traction relationship of the state traction chain changes. The warning signals contain complete information including signal type, intensity level, target object, response requirements, and traction adjustment suggestions. Signal types are divided into three main categories: equipment anomaly warnings, environmental risk warnings, and system fault warnings. Generated warning signals are sorted according to traction urgency and priority to ensure that warning signals requiring immediate traction intervention are processed and communicated first.

[0069] To implement the above-described method embodiments, a method for real-time monitoring of cascade operation status and intelligent early warning of collision risks is proposed to achieve the corresponding functions and technical effects. See also... Figure 2 , Figure 2 This diagram illustrates a structural block diagram of a ladder-type operational status real-time monitoring and collision risk intelligent early warning system 200 according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The ladder-type operational status real-time monitoring and collision risk intelligent early warning system 200 provided in this embodiment includes: The data acquisition module 201 is used to acquire multi-sensor data of the cascade operation, determine key monitoring nodes based on the multi-sensor data, and extract parameter changes of the key monitoring nodes to form an operating status trend sample. The spectrum reconstruction module 202 is used to perform edge computing processing on the operating status trend sample to generate a step gap change curve, match the step gap change curve with the standard operating sample model to obtain risk feature probability values, and generate a state traction chain based on the risk feature probability values. The attenuation compensation module 203 is used to obtain the displacement distribution coefficient and vibration velocity index from the state traction chain, predict the collision probability distribution of the displacement distribution coefficient through a cloud-based deep learning algorithm, determine the warning level based on the vibration velocity index, and generate a warning optimization configuration by combining the collision probability distribution and the warning level. The linkage control module 204 is used to perform linkage control analysis to determine the emergency distribution based on the early warning optimization configuration, identify differentiated response areas through the emergency distribution, and extract linkage optimization factors from the differentiated response areas. The early warning output module 205 is used to generate differentiated control parameters using the linkage optimization factor, and to match the differentiated control parameters with the state traction chain to generate an early warning signal.

[0070] The aforementioned cascade operation status real-time monitoring and collision risk intelligent early warning system 200 can implement the cascade operation status real-time monitoring and collision risk intelligent early warning method of the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can refer to the content of the above method embodiments, and will not be repeated in this embodiment.

[0071] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

[0072] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A method for real-time monitoring of cascade operation status and intelligent early warning of collision risk, characterized in that, include: Acquire multi-sensor data for cascade operation, determine key monitoring nodes based on the multi-sensor data, and extract parameter changes of the key monitoring nodes to form an operational status trend sample; Edge computing processing is performed on the operating status trend sample to generate a step gap change curve. The step gap change curve is matched with the standard operating sample model to obtain the risk feature probability value. A status traction chain is generated based on the risk feature probability value. The displacement distribution coefficient and vibration velocity index are obtained from the state traction chain. The collision probability distribution of the displacement distribution coefficient is predicted by the cloud deep learning algorithm. The warning level is determined based on the vibration velocity index. The warning optimization configuration is generated by combining the collision probability distribution and the warning level. Based on the early warning optimization configuration, the emergency distribution is determined by linkage control analysis. Differentiated response areas are identified through the emergency distribution, and linkage optimization factors are extracted from the differentiated response areas. Differentiated control parameters are generated using the aforementioned linkage optimization factor, and these differentiated control parameters are matched with the state traction chain to generate an early warning signal.

2. The method according to claim 1, characterized in that, The step of performing edge computing processing on the operating status trend sample to generate the step gap change curve includes: Vibration spectrum signals are collected from the operating status trend samples; The vibration spectrum signal is reverse-analyzed and reconstructed to obtain the state parameters of the mechanical component through edge computing processing; Based on the state parameters of the mechanical components, the wear degree of the components is assessed to form a wear state mapping; The step gap variation curve is generated by fitting the wear state mapping.

3. The method according to claim 1, characterized in that, The process of matching the risk characteristic probability value based on the step gap variation curve with the standard operating sample model includes: Identify heterogeneous interference source signals from the variation curve of the step gaps; The heterogeneous interference source signal is separated and processed to obtain interference components, which include electromagnetic interference components, mechanical interference components and thermal interference components. An environmental state mapping table is constructed using the aforementioned interference components; Risk feature probability values ​​are obtained by matching the environmental state mapping table with the standard operating sample model.

4. The method according to claim 1, characterized in that, The method of predicting the collision probability distribution of the displacement distribution coefficient using a cloud-based deep learning algorithm includes: Monitor the performance degradation of each sensor in the displacement distribution coefficient; The degradation pattern of the performance degradation is identified using a cloud-based deep learning algorithm; The decay mode is then converted into system robustness enhancement parameters; The collision probability distribution is generated based on the system robustness enhancement parameters.

5. The method according to claim 1, characterized in that, The step of generating an optimized warning configuration by combining the collision probability distribution and the warning level includes: Identify the regions with missing monitoring data in the collision probability distribution; The missing monitoring data area is transformed into the interpolation algorithm training domain; The interpolation algorithm is used to train the domain and perform data completion processing to obtain a complete dataset; An optimized early warning configuration is generated by fusing the complete dataset with the early warning level.

6. The method according to claim 1, characterized in that, The identification of differentiated response areas through the emergency distribution includes: Detect the system response time delay in the emergency distribution; The system response time delay is converted into an extended parameter for the early warning window; The aforementioned warning window extension parameters are used to adaptively divide the response area and generate a region configuration. Differentiated response regions are formed based on the regional configuration.

7. The method according to claim 1, characterized in that, The step of extracting linkage optimization factors from the differentiated response regions includes: The intensity of complex environmental interference within the differentiated response area is collected; The intensity of the complex environmental interference is transformed into an evaluation index of the system's environmental adaptability. Adaptability test data are obtained based on the system's environmental adaptability evaluation indicators; The linkage optimization factors are extracted from the adaptive test data. The linkage optimization factors include environmental adaptability adjustment parameters and vibration suppression adaptation parameters.

8. The method according to claim 3, characterized in that, The step of constructing an environmental state mapping table using the interference components includes: Detect the data inconsistency characteristics among the various interference components; The data inconsistency features are transformed into parameters to enhance the credibility of verification. The aforementioned verification credibility enhancement parameters are used to perform data consistency correction, generating corrected interference components. An environmental state mapping table is established based on the corrected interference components.

9. The method according to claim 7, characterized in that, The process of obtaining adaptive test data based on the system's environmental adaptability evaluation index includes: Identify current scenario response strategy conflicts based on the system's environmental adaptability evaluation index; The current scenario response strategy conflict is transformed into a decision architecture reorganization trigger signal; The decision architecture is reorganized in real time using the aforementioned decision architecture reorganization trigger signal to generate a new decision architecture. The new decision-making architecture is used to perform adaptive capability tests to obtain adaptive test data.

10. A real-time monitoring system for cascade operation status and an intelligent early warning system for collision avoidance risks, characterized in that, include: The data acquisition module is used to acquire multi-sensor data of the cascade operation, determine key monitoring nodes based on the multi-sensor data, and extract parameter changes of the key monitoring nodes to form an operating status trend sample. The spectrum reconstruction module is used to perform edge computing processing on the operating status trend sample to generate a step gap change curve, match the step gap change curve with the standard operating sample model to obtain risk feature probability values, and generate a status traction chain based on the risk feature probability values. The attenuation compensation module is used to obtain the displacement distribution coefficient and vibration velocity index from the state traction chain, predict the collision probability distribution of the displacement distribution coefficient through a cloud-based deep learning algorithm, determine the warning level based on the vibration velocity index, and generate a warning optimization configuration by combining the collision probability distribution and the warning level. The linkage control module is used to perform linkage control analysis based on the early warning optimization configuration to determine the emergency distribution, identify differentiated response areas through the emergency distribution, and extract linkage optimization factors from the differentiated response areas. The early warning output module is used to generate differentiated control parameters using the linkage optimization factor, and to match the differentiated control parameters with the state traction chain to generate an early warning signal.

Images