Working well platform anti-falling method, device and equipment

Abstract

The invention discloses an anti-falling method, device and equipment for a working well platform, and relates to the technical field of engineering protection, and the method comprises the following steps: an anti-falling decision-making layer is selected according to decision-making time, and the anti-falling decision-making layer comprises a state decision-making layer, an image decision-making layer and a mixed decision-making layer; the state decision-making layer obtains a first anti-falling decision according to the state data; the image decision-making layer obtains a second anti-falling decision according to the falling credibility and the falling trend criticality, the falling credibility is obtained by processing state data to obtain a basic probability assignment and synthesizing the basic probability assignment based on an evidence synthesis algorithm, and the falling trend criticality is obtained by obtaining state data statistical characteristics and performing inspection processing; and the mixed decision-making layer obtains a third anti-falling decision by combining the predicted falling risk value, falling credibility and falling trend criticality. The method is used for solving the problem that an existing anti-falling technology of a working well platform is insufficient in adaptability.

Description

Technical Field

[0001] This invention relates to the field of engineering protection technology, and more specifically, to methods, devices and equipment for preventing falls from manhole platforms. Background Technology

[0002] In oilfield well repair, power cable laying and maintenance, and municipal pipeline maintenance, workers often need to enter or operate on high-rise well platforms. These environments are characterized by confined spaces and complex conditions, posing an extremely high risk of falls. A fall can easily result in serious personal injury or even death, causing incalculable loss of life and property.

[0003] Currently, fall prevention measures for manhole platforms mainly rely on physical protective devices, such as mesh structures installed inside the manhole to catch falling people or objects. Most existing devices are passive protection, meaning they only activate the interception or buffering mechanism after a fall has occurred. They cannot provide early warnings or interventions for pre-fall signs such as personnel instability or slipping, lacking proactive prevention capabilities. Furthermore, reliance on physical facilities presents reliability issues; fall protection nets and similar devices are susceptible to material aging and decreased impact resistance over time, potentially reducing their reliability. Some mechanical safety devices have complex structures, and in the harsh environment of manholes, key components may fail. They also lack the ability to perceive and intelligently analyze the real-time status of workers. Moreover, they cannot assess risk based on dynamic information such as personnel position, posture, and movement trends, thus failing to provide tiered and differentiated early warning and protection. Therefore, there is an urgent need for a fall prevention method capable of real-time perception of personnel status and intelligent assessment of fall risk to overcome the shortcomings of existing technologies and fundamentally improve the safety level of manhole operations.

[0004] For example, the invention patent announcement CN118628320B discloses a method and system for monitoring safety against falls during high-altitude operations. This includes acquiring physiological and behavioral data of high-altitude workers, usage status data of fall protection devices, and environmental data; determining human risk indicators, equipment risk indicators, and environmental risk indicators; judging whether any of these risk indicators exceeds a first preset threshold; if so, triggering an alarm and generating a first operation guide to push to the high-altitude workers; if not, determining a fall risk value based on the human risk indicators, equipment risk indicators, and environmental risk indicators, and triggering an alarm and generating a second operation guide to push to the high-altitude workers when the fall risk value exceeds a second preset threshold.

[0005] For example, the invention patent announcement CN120472642B discloses a high-altitude operation fall prevention visualization monitoring and early warning system and method, which includes the following steps: collecting workers' physiological state data to generate worker fatigue characteristics and attention distraction index; collecting behavioral posture data to analyze movement regularity; collecting work environment parameters and evaluating the work platform stability index; interactively correlating worker fatigue characteristics, attention distraction index, movement regularity, and work platform stability index to obtain worker-environment interaction state data; and extracting risk precursor features from the worker-environment interaction state data to form a multidimensional risk precursor feature set.

[0006] The above-disclosed technical solutions have at least the following technical problems: The current technology lacks sufficient depth of information integration, and is essentially still a passive response to current or existing risks, which may miss the best time for intervention due to computational delays.

[0007] To address the above problems, this invention proposes a solution. Summary of the Invention

[0008] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method, apparatus, and equipment for preventing falls from manhole platforms. The method involves adaptively selecting a decision layer based on decision time, a state decision layer that reacts rapidly based on personnel status data, an image decision layer that makes accurate judgments by extracting data features, and a hybrid decision layer that introduces a time-series prediction model for forward-looking predictions. This addresses the problem of insufficient adaptability of existing manhole platform fall prevention technologies.

[0009] To achieve the above objectives, the present invention provides the following technical solution: A fall prevention method for a well platform includes the following steps: selecting a fall prevention decision layer based on the decision time, wherein the fall prevention decision layer includes a state decision layer, an image decision layer, and a hybrid decision layer; the state decision layer obtains a first fall prevention decision based on state data; the image decision layer obtains a second fall prevention decision based on fall confidence and fall trend urgency, wherein the fall confidence is obtained by processing state data to obtain a basic probability assignment, and the basic probability assignment is synthesized based on an evidence synthesis algorithm; the fall trend urgency is obtained by acquiring statistical features of state data and verifying them; and the hybrid decision layer obtains a third fall prevention decision by combining the predicted fall risk value, fall confidence, and fall trend urgency.

[0010] In a preferred embodiment, the step of selecting the fall protection decision layer based on the decision time specifically involves: the decision time including state decision time, image decision time, and mixed decision time; obtaining the state decision time; if the state decision time is lower than a preset state time threshold, selecting the state decision layer; otherwise, obtaining the image decision time and mixed decision time; if the mixed decision time is lower than a preset mixed time threshold, selecting the image decision layer; if the mixed decision time is higher than the mixed time threshold, selecting the mixed decision layer.

[0011] In a preferred embodiment, the method for obtaining the state decision time, image decision time, and hybrid decision time specifically comprises: collecting and preprocessing state data of personnel on the well platform from multi-source sensors, and calculating the state decision time using a preset state decision time formula; acquiring personnel image data of the well platform, and obtaining the image decision time based on the rate of change of the pixel area of ​​the workers in the personnel image data; and weightedly fusing the state decision time and the image decision time to obtain the hybrid decision time.

[0012] In a preferred embodiment, the state decision layer obtains a first fall prevention decision based on state data, specifically by: setting a state threshold for each type of state data based on historical accident data; and mapping the number of types of state data exceeding the state threshold to obtain the first fall prevention decision.

[0013] In a preferred embodiment, the fall confidence is obtained by processing state data to obtain a basic probability assignment, and synthesizing the basic probability assignment based on an evidence synthesis algorithm. Specifically, this involves: constructing a fuzzy neural network model; inputting instantaneous data of various types of state data into the fuzzy neural network, which includes a fuzzification layer and a fuzzy rule calculation layer; calculating the membership degree of each input relative to a preset fuzzy set through the fuzzification layer; obtaining the activation intensity of the fuzzy rule with the membership degree as an antecedent through the fuzzy rule calculation layer; normalizing the activation intensity to obtain the basic probability assignment; correcting the basic probability assignment through a pre-constructed machine learning model; and synthesizing the corrected basic probability assignment according to a preset state decision time formula to obtain the fall confidence.

[0014] In a preferred embodiment, the fall trend severity is obtained by acquiring and verifying the statistical characteristics of the state data, specifically: a preset sliding time window is used to extract the state data; the Mankendall algorithm is used to calculate the statistics and variance of the extracted state data; the statistics are standardized according to the variance to obtain a significance value; and the significance value is compared with a preset probability threshold to obtain the trend severity.

[0015] In a preferred embodiment, the image decision layer obtains a second fall prevention decision based on the fall confidence and the fall trend urgency, specifically by: inputting the fall confidence and the fall trend urgency into a pre-trained machine learning model, outputting the probability distribution of the second fall prevention decision; comparing the probability distribution of the second fall prevention decision with a preset probability threshold to obtain the final second fall prevention decision.

[0016] The fall prevention device for a well platform includes: a decision-making layer selection module for calculating the decision time and selecting the fall prevention decision layer based on the decision time; a first decision module for obtaining a first fall prevention decision based on status data; a second decision module for obtaining a second fall prevention decision based on fall confidence and fall trend urgency; and a third decision module for obtaining a third fall prevention decision by combining the predicted fall risk value, fall confidence, and fall trend urgency.

[0017] An electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the aforementioned method for preventing falls from a manhole platform.

[0018] The technical effects and advantages of the method, device and equipment for preventing falls from manhole platforms in this invention are as follows: This invention employs an adaptive decision-making time selection layer, a state-based decision-making layer that reacts rapidly based on personnel status data, an image-based decision-making layer that obtains fall credibility through fuzzy algorithms and evidence synthesis, and combines this with trend severity determined by trend testing of statistical data features. A machine learning model is then used for comprehensive judgment, and a hybrid decision-making layer introduces a time-series prediction model to obtain predicted risk values. These values ​​are then fused with the output of the image-based decision-making layer, significantly improving the accuracy and reliability of safety protection for manhole platforms and effectively solving the problem of insufficient adaptability in existing manhole platform fall prevention technologies. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the method for preventing falls from a well platform provided in an embodiment of the present invention.

[0020] Figure 2 A schematic diagram of the anti-fall device structure for a well platform provided in an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of the anti-fall device structure for a well platform provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1, Figure 1 The present invention provides a method for preventing falls from manhole platforms, comprising the following steps: S1, Select the fall protection decision layer according to the decision time, the fall protection decision layer includes a state decision layer, an image decision layer and a hybrid decision layer; S2, the state decision layer obtains the first fall protection decision based on the state data; S3, the image decision layer obtains a second fall prevention decision based on the fall confidence and the fall trend urgency. The fall confidence is obtained by processing the state data to obtain a basic probability assignment, and the basic probability assignment is synthesized based on the evidence synthesis algorithm. The fall trend urgency is obtained by acquiring the statistical features of the state data and verifying them. S4, the hybrid decision layer obtains the third fall prevention decision by combining the predicted fall risk value, fall confidence, and fall trend urgency.

[0024] This embodiment adaptively selects the decision layer based on decision time, the state decision layer reacts quickly based on personnel status data, the image decision layer obtains the fall credibility through fuzzy algorithm and evidence synthesis, combines the trend urgency obtained by trend testing of data statistical features, and uses a machine learning model for comprehensive judgment, the hybrid decision layer introduces a time series prediction model to obtain the predicted risk value, and fuses it with the output of the image decision layer, which significantly improves the accuracy and reliability of the safety protection of the well platform and effectively solves the problem of insufficient adaptability of existing well platform fall prevention technology.

[0025] S1, Select the fall protection decision layer according to the decision time. The fall protection decision layer includes a state decision layer, an image decision layer, and a hybrid decision layer.

[0026] In this embodiment, the step of selecting the fall protection decision layer based on the decision time specifically means: The decision time includes state decision time, image decision time, and hybrid decision time; Obtain the state decision time. If the state decision time is lower than the preset state time threshold, select the state decision layer; otherwise, obtain the image decision time and the mixed decision time. If the mixing decision time is lower than the preset mixing time threshold, select the image decision layer; if the mixing decision time is higher than the mixing time threshold, select the mixing decision layer.

[0027] In this embodiment, the method for obtaining the state decision time, image decision time, and hybrid decision time is specifically as follows: Collect and preprocess status data of personnel on the well platform from multiple sources, and calculate the status decision time by combining it with a preset status decision time formula; Acquire personnel image data of the well platform, and obtain the image decision time based on the rate of change of the pixel area of ​​the workers in the personnel image data; The state decision time and the image decision time are weighted and fused to obtain the hybrid decision time.

[0028] In this embodiment, the status data of the personnel on the well platform includes personnel location data, personnel movement status data, and personnel contact status data with the platform.

[0029] In this embodiment, the specific formula for calculating the state decision time is as follows:

[0030] In the formula, For state decision time, The distance between personnel and the edge of the platform, The speed at which personnel move toward the edge of the platform.

[0031] In this embodiment, the speed at which the person moves toward the edge of the platform is calculated by distance difference across consecutive frames.

[0032] It should be noted that this embodiment constructs a hierarchical decision-making mechanism by calculating the decision time in real time. When the state decision time is lower than the state time threshold, the fast-response state decision layer is immediately activated. When there is a certain buffer time, the image decision layer with higher accuracy or the hybrid decision layer with forward-looking capability is selected to be activated by comparing the hybrid decision time with the threshold. This ensures instantaneous response capability in extreme emergency situations and reduces false alarms through more complex analysis in non-emergency situations. It effectively balances the system response speed and judgment accuracy, and significantly improves the overall reliability of the well platform fall prevention system.

[0033] S2, the state decision layer obtains the first fall protection decision based on the state data.

[0034] In this embodiment, the state decision layer obtains a first fall protection decision based on the state data, specifically as follows: Set state thresholds for each type of state data based on historical accident data; The first fall prevention decision is obtained by mapping the number of types of state data that exceed the state threshold.

[0035] In this embodiment, the step of mapping the number of types of state data exceeding the state threshold to obtain the first fall protection decision specifically involves: If a type of data is abnormal, a low-level response is triggered, including logging and providing a minor notification.

[0036] If two types of data are abnormal, a medium-level response is triggered, including triggering an audible and visual alarm.

[0037] If any of the three types of data anomalies occur, an advanced response is triggered, including triggering interception devices and limiting the speed of surrounding equipment.

[0038] It should be noted that this embodiment constructs a progressive response strategy by setting independent thresholds for multiple types of state data and adopting an anomaly type counting triggering mechanism. This effectively avoids frequent false triggers caused by false alarms from a single sensor. At the same time, the redundancy of multi-source information ensures reliable identification of high-risk states. This not only ensures flexible handling of general abnormal states but also achieves rapid and accurate response to real fall risks, significantly improving the robustness and protective effectiveness of the system.

[0039] S3, the image decision layer obtains a second fall prevention decision based on the fall confidence and fall trend urgency. The fall confidence is obtained by processing the state data to obtain a basic probability assignment, and the basic probability assignment is synthesized based on the evidence synthesis algorithm. The fall trend urgency is obtained by acquiring the statistical features of the state data and verifying them.

[0040] In this embodiment, the fall confidence level is obtained by processing the state data to obtain a basic probability assignment, and then synthesizing the basic probability assignment based on the evidence synthesis algorithm, specifically as follows: A fuzzy neural network model is constructed by inputting instantaneous data of various types of state data into the fuzzy neural network, which includes a fuzzification layer and a fuzzy rule calculation layer. The membership degree of each input relative to a preset fuzzy set is calculated through a fuzzification layer; The activation strength of the fuzzy rule is obtained by using the fuzzy rule calculation layer with membership degree as the antecedent. The activation intensity is normalized to obtain the basic probability value; The base probability assignments are corrected using a pre-built machine learning model; The corrected basic probability values ​​are synthesized based on the preset state decision time formula to obtain the fall confidence level.

[0041] In this embodiment, the membership degree of each input relative to a preset fuzzy set is calculated through a fuzzification layer, specifically as follows: A fuzzy set is defined based on human body state data, and the membership function of the fuzzy set is defined; The membership vector of human state data is calculated based on the membership function.

[0042] In this embodiment, the membership function uses triangular membership.

[0043] In this embodiment, the activation intensity is specifically defined by the following formula:

[0044] In the formula, To activate intensity, , and Membership degrees for different types of state data.

[0045] It should be noted that fuzzy neural networks are a technology that combines fuzzy logic with artificial neural networks. By simulating human ability to process fuzzy information and learn to adapt, it effectively addresses the contradiction between system complexity and accuracy. It can automatically learn from data and generate fuzzy rules and membership functions, while processing both explicit and fuzzy information.

[0046] In this embodiment, the step of correcting the basic probability assignment using a pre-built machine learning model specifically involves: Collect data from fall events and simulated fall scenarios, and obtain true basic probability values ​​through expert annotation; The basic probability is modified by assigning values ​​through a pre-constructed neural network; The neural network is trained by using the generated fuzzy feature vectors and their corresponding true base probabilities as the training set.

[0047] In this embodiment, the formula for the fall confidence level is as follows:

[0048] In the formula, To improve the credibility of the fall, The conflict coefficient, This indicates that for all sets satisfying the condition that the intersection is exactly 1 The basic probabilities of propositions are first multiplied and then summed. A subset of the identification framework that constitutes all possible conclusions.

[0049] In this embodiment, the identification framework includes both falling and being safe.

[0050] In this embodiment, the conflict coefficient is specifically formulated as follows:

[0051] It should be noted that, The closer it is to 1, the greater the conflict between the pieces of evidence.

[0052] In this embodiment, the severity of the fall trend is obtained by acquiring statistical features of the state data and performing verification processing, specifically as follows: Set a preset sliding time window to capture status data; The Menkendall algorithm was used to calculate the statistics and variance of the extracted state data. The statistic is standardized based on the variance to obtain a significance value. The significance value is then compared with a preset probability threshold to obtain the trend severity.

[0053] In this embodiment, the specific formula for the statistic is as follows:

[0054] In the formula, For statistical purposes, It is a sign function; if the value inside the function is greater than 0, it is 1; if it is equal to 0, it is 0; and if it is less than 0, it is -1. and They are two adjacent data points in the state data. This represents the number of data points in the time window.

[0055] In this embodiment, the specific formula for the variance is:

[0056] In the formula, The variance of the statistic This represents the number of repeated values.

[0057] In this embodiment, the standardization of the statistic is specifically achieved using the following formula:

[0058] In the formula, This is the standardized statistic.

[0059] In this embodiment, the significance value is calculated using a standard normal distribution table or a statistical function.

[0060] In this embodiment, the image decision layer obtains a second fall prevention decision based on the fall confidence level and the fall trend urgency level, specifically as follows: The fall confidence and fall trend severity are input into a pre-trained machine learning model, which outputs the probability distribution of the second fall prevention decision. The probability distribution of the second fall protection decision is compared with the preset probability threshold to obtain the final second fall protection decision.

[0061] In this embodiment, the machine learning model is the LightGBM model.

[0062] It should be noted that LightGBM is a gradient boosting decision tree algorithm that discretizes continuous floating-point feature values ​​into multiple buckets to form a histogram. When finding the optimal split point, it is no longer necessary to traverse all the original data, but only to traverse the buckets, which greatly reduces the computational complexity. Each split involves all leaves of the same level, which makes it easy to control the depth of the tree and prevents overfitting.

[0063] It should be noted that this embodiment transforms multi-source sensor data into unified basic probability values ​​through a fuzzy neural network and resolves sensor conflict issues through evidence theory synthesis, significantly improving the reliability of fall risk perception. By introducing the Mankendall trend test to perform significance analysis on the temporal characteristics of the data, the system can capture the dynamic evolution of risk. Through a gradient boosting tree model, static risk credibility and dynamic trend urgency are efficiently integrated to generate probabilistic decision outputs, greatly reducing false alarms and missed alarms caused by misjudgment from a single information source or short-term data fluctuations. This forms an intelligent decision-making capability that combines anti-interference, foresight, and high credibility.

[0064] S4, the hybrid decision layer obtains the third fall prevention decision by combining the predicted fall risk value, fall confidence, and fall trend urgency.

[0065] In this embodiment, the hybrid decision layer obtains a third fall protection decision by combining the predicted fall risk value, fall confidence, and fall trend urgency, specifically as follows: Input the time series of state data into a pre-trained time series prediction model, and output state prediction data; Calculate the predicted risk value within the future time window based on the state prediction data; Based on the time-series dynamic preset time weight, the predicted risk values ​​are weighted by time weight, summed and averaged, and then fused with the probability distribution of the second fall prevention decision to obtain the fall risk value; The fall risk value is compared with the preset risk value range, and the corresponding third fall prevention decision is obtained by mapping the risk value range in which the fall risk value is located.

[0066] In this embodiment, the time series prediction model adopts the Timer model.

[0067] It should be noted that Timer is a generative pre-trained large model for time series analysis. Its core training task is to predict the data at the next time point. By performing autoregressive pre-training on massive amounts of time series data, it learns to understand the inherent patterns and trends in the data, thereby gaining powerful generation and reasoning capabilities.

[0068] In this embodiment, the step of calculating the predicted risk value within the future time window based on the state prediction data specifically involves inputting the state prediction data into a pre-trained support vector regression model to obtain the predicted risk value within the future time window.

[0069] It should be noted that support vector regression sets a tolerance bias. As long as the deviation between the predicted value and the true value of a data point does not exceed the tolerance bias, the point is considered to have been perfectly fitted by the model. This makes support vector regression insensitive to small noises in the data and enhances the robustness of the model.

[0070] In this embodiment, the specific formula for the time dynamic weight is as follows:

[0071] In the formula, For the first The weight of each predicted risk value The preset time decay coefficient, For the first The time difference between a predicted risk value and the current moment This represents the data length.

[0072] In this embodiment, the step of weighting and summing the predicted risk values ​​according to time weights and then averaging them, and then fusing them with the probability distribution of the second fall prevention decision to obtain the fall risk value, specifically involves: The highest probability value in the probability distribution of the second fall prevention decision is selected and weighted by the preset weights of the corresponding decision types to obtain a probability weight value; The fall risk value is obtained by adding the probability weighted value and the predicted risk value.

[0073] It should be noted that this embodiment uses a time-series prediction model to perform autoregressive training on personnel status data to predict risk trends within future time windows. By introducing a time decay weight coefficient, the system focuses more on recent prediction results and reduces the interference of uncertain long-term predictions. The predicted risk value is quantified through a support vector regression model and weighted and fused with the probability distribution of the image decision layer to generate a comprehensive fall risk value. This significantly improves the system's accuracy in predicting progressive fall risks and forms an intelligent decision-making capability that combines foresight, anti-interference, and high precision.

[0074] Example 2, Figure 2 The present invention provides a fall prevention device for a well platform, comprising a decision-making selection module, a first decision module, a second decision module, and a third decision module, with connections between the modules: The decision-making layer selection module is used to calculate the decision time and select the fall protection decision layer based on the decision time. The first decision module is used to make the first fall protection decision based on the status data; The second decision module is used to make a second fall protection decision based on the fall credibility and the fall trend urgency. The third decision module is used to obtain a third fall prevention decision by combining the predicted fall risk value, fall confidence, and fall trend urgency.

[0075] Example 3: This example provides a computer electronic device, such as... Figure 3 As shown, the electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the aforementioned method for preventing falls from a well platform.

[0076] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0077] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0078] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0079] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0080] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0081] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for preventing falls from manhole platforms, characterized in that, Includes the following steps: The fall protection decision layer is selected based on the decision time, and the fall protection decision layer includes a state decision layer, an image decision layer, and a hybrid decision layer. The state decision layer makes the first fall protection decision based on the state data; The image decision layer obtains a second fall prevention decision based on the fall confidence and the fall trend urgency. The fall confidence is obtained by processing the state data to obtain a basic probability assignment, and the basic probability assignment is synthesized based on the evidence synthesis algorithm. The fall trend urgency is obtained by acquiring the statistical features of the state data and verifying them. The hybrid decision layer arrives at a third fall prevention decision by combining the predicted fall risk value, fall confidence, and fall trend severity.

2. The method for preventing falls from a well platform according to claim 1, characterized in that, The selection of the fall protection decision layer based on decision time is specifically as follows: The decision time includes state decision time, image decision time, and hybrid decision time; Obtain the state decision time. If the state decision time is lower than the preset state time threshold, select the state decision layer; otherwise, obtain the image decision time and the mixed decision time. If the mixing decision time is lower than the preset mixing time threshold, select the image decision layer; if the mixing decision time is higher than the mixing time threshold, select the mixing decision layer.

3. The method for preventing falls from a well platform according to claim 2, characterized in that, The method for obtaining the state decision time, image decision time, and hybrid decision time is as follows: Collect and preprocess status data of personnel on the well platform from multiple sources, and calculate the status decision time by combining it with a preset status decision time formula; Acquire personnel image data of the well platform, and obtain the image decision time based on the rate of change of the pixel area of ​​the workers in the personnel image data; The state decision time and the image decision time are weighted and fused to obtain the hybrid decision time.

4. The method for preventing falls from a well platform according to claim 3, characterized in that, The state decision layer obtains the first fall protection decision based on the state data, specifically as follows: Set state thresholds for each type of state data based on historical accident data; The first fall prevention decision is obtained by mapping the number of types of state data that exceed the state threshold.

5. The method for preventing falls from manhole platforms according to claim 4, characterized in that, The fall confidence level is obtained by processing state data to obtain a basic probability assignment, and then synthesizing the basic probability assignment based on an evidence synthesis algorithm, specifically as follows: A fuzzy neural network model is constructed by inputting instantaneous data of various types of state data into the fuzzy neural network, which includes a fuzzification layer and a fuzzy rule calculation layer. The membership degree of each input relative to a preset fuzzy set is calculated through a fuzzification layer; The activation strength of the fuzzy rule is obtained by using the fuzzy rule calculation layer with membership degree as the antecedent. The activation intensity is normalized to obtain the basic probability value; The base probability assignments are corrected using a pre-built machine learning model; The corrected basic probability values ​​are synthesized based on the preset state decision time formula to obtain the fall confidence level.

6. The method for preventing falls from a manhole platform according to claim 5, characterized in that, The severity of the fall trend is obtained by acquiring statistical characteristics of the state data and processing them, specifically as follows: Set a preset sliding time window to capture status data; The Menkendall algorithm was used to calculate the statistics and variance of the extracted state data. The statistic is standardized based on the variance to obtain a significance value. The significance value is then compared with a preset probability threshold to obtain the trend severity.

7. The method for preventing falls from a well platform according to claim 6, characterized in that, The image decision layer obtains a second fall prevention decision based on the fall credibility and the urgency of the fall trend, specifically: The fall confidence and fall trend severity are input into a pre-trained machine learning model, which outputs the probability distribution of the second fall prevention decision. The probability distribution of the second fall protection decision is compared with the preset probability threshold to obtain the final second fall protection decision.

8. The method for preventing falls from a well platform according to claim 7, characterized in that, The hybrid decision layer obtains a third fall protection decision by combining the predicted fall risk value, fall confidence, and fall trend urgency. Specifically: Input the time series of state data into a pre-trained time series prediction model, and output state prediction data; Calculate the predicted risk value within the future time window based on the state prediction data; Based on the time-series dynamic preset time weight, the predicted risk values ​​are weighted by time weight, summed and averaged, and then fused with the probability distribution of the second fall prevention decision to obtain the fall risk value; The fall risk value is compared with the preset risk value range, and the corresponding third fall prevention decision is obtained by mapping the risk value range in which the fall risk value is located.

9. An apparatus for using the method for preventing falls from a manhole platform as described in any one of claims 1-8, comprising: The decision-making layer selection module is used to calculate the decision time and select the fall protection decision layer based on the decision time. The first decision module is used to make the first fall protection decision based on the status data; The second decision module is used to make a second fall protection decision based on the fall credibility and the fall trend urgency. The third decision module is used to obtain a third fall prevention decision by combining the predicted fall risk value, fall confidence, and fall trend urgency.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the fall prevention method for a well platform as described in any one of claims 1 to 8.

Images