A fault detection method, device and medium of a driving apparatus

Abstract

The present application relates to the technical field of industrial equipment automation detection and fault diagnosis, and particularly relates to a drive equipment fault detection method, equipment and medium. The method is cooperatively executed by a mobile robot and a cloud server. The method comprises: generating a global detection path based on mobile robot power, communication load and detection accuracy mapping; the mobile robot moves to each detection point in turn; an identification code is identified and uploaded to the cloud; the cloud dynamically generates a detection instruction set based on equipment historical data and issues the detection instruction set to the mobile robot; the mobile robot executes the instruction to collect equipment state data and uploads the equipment state data; the cloud compares the data with a normal threshold value, and instructs the mobile robot to re-detect or continue the task according to the result. The present application combines intelligent path planning and cloud dynamic decision-making, and realizes the efficiency, accuracy and self-adaptive ability of drive equipment fault detection in a complex environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial equipment automatic detection and fault diagnosis, in particular to a drive equipment fault detection method, device and medium. BACKGROUND

[0002] The tower crane equipment of a port terminal is the core equipment for cargo handling, and the health status of its drive system (such as motor, gear box, brake, etc.) is directly related to the operation efficiency and safety of the entire terminal. Therefore, it is crucial to regularly detect the drive equipment of the tower crane.

[0003] Currently, the main detection methods rely on manual inspection or fixed monitoring equipment. Manual inspection has problems such as low efficiency, high cost, and strong subjectivity, and in the harsh environment of the port with strong wind, heavy fog, and large waves, the safety risk of personnel is extremely high, and the detection work is often forced to be interrupted due to weather reasons. Although the fixed monitoring equipment can achieve continuous collection of some data, it has high installation and maintenance costs, limited coverage, and is difficult to flexibly respond to numerous and scattered detection points on the tower crane.

[0004] In recent years, attempts have been made to use mobile robots or unmanned aerial vehicles for equipment detection. However, when applied to the specific scenario of the port tower crane, these solutions still have obvious limitations. First, their path planning usually only considers basic spatial paths and obstacle avoidance, without fully considering the dynamic relationship between the robot's battery life and the quality and quantity of detection tasks, and the impact of unstable signal transmission in complex environments on the detection process, resulting in detection plans often being incomplete or detection quality being uneven. Second, the detection process is often pre-programmed, lacking dynamic adjustment based on the historical state of the equipment, and unable to achieve precise detection that is "tailored to the equipment". Finally, the analysis and decision of detection data are usually performed locally on the robot side, with limited computing power, making it difficult to perform deep data mining and global optimization, and also unable to form a real-time closed loop from detection to decision-making and execution.

[0005] Therefore, there is an urgent need in the art for a tower crane drive equipment detection scheme that can adapt to the harsh environment of the port and achieve efficient, accurate, and fully automatic detection. SUMMARY

[0006] To solve the above technical problems, the technical solution adopted by the present application is as follows:

[0007] According to the first aspect of the present application, a drive equipment fault detection method is provided, which is cooperatively executed by a mobile robot and a cloud server, and the method comprises the following steps:

[0008] S100, generating a global detection path based on the remaining power of the mobile robot, the expected data communication load, the mapping relationship between power and detection accuracy level, and the number of equipment to be detected.

[0009] S200, the mobile robot moves to each detection point in the global detection path in turn according to the global detection path.

[0010] S300, after the mobile robot arrives at the detection point, the mobile robot identifies the identification code of the detection point and uploads the identification information to the cloud server, and then receives a detection instruction set dynamically generated by the cloud server based on the device image information of the target driving device associated with the identification code.

[0011] S400, the mobile robot analyzes and executes the detection instruction set, controls the sensor carried by the mobile robot to collect data of the target driving device, and generates device state data.

[0012] S500, the mobile robot uploads the collected device state data to the cloud server; the cloud server compares the device state data with a preset normal operation parameter threshold range, and sends a subsequent action instruction to the mobile robot based on the comparison result.

[0013] According to the second aspect of the present application, an electronic device is provided, comprising a processor and a memory; the processor is used to execute the steps of the method of the first aspect of the present application by calling the program or instruction stored in the memory.

[0014] According to the third aspect of the present application, a computer readable storage medium is provided, the computer readable storage medium stores a program or instruction, the program or instruction makes the computer execute the steps of the method of the first aspect of the present application.

[0015] The present application has at least the following beneficial effects:

[0016] 1. Global optimization and intelligent allocation of resources in the detection process are realized

[0017] By comprehensively considering the remaining power of the mobile robot, the communication load and the dynamic mapping relationship between the power and the detection accuracy for path planning, the defects of fixed path and rigid resource allocation in traditional inspection are overcome, and the endurance and overall efficiency of the detection task are significantly improved.

[0018] 2. The accuracy and intelligent level of fault detection are improved

[0019] By using the powerful data storage and analysis capability of the cloud server, the personalized detection instruction set is dynamically generated based on the complete historical archives of the device, so that each detection has clear pertinence, realizes the leap from "general inspection" to "accurate diagnosis", and greatly improves the fault recognition rate.

[0020] 3. A closed-loop adaptive system is built, and the adaptability to complex environment is enhanced

[0021] A complete closed loop is formed from "intelligent planning → accurate execution → data uploading → intelligent decision making → feedback control". Not only can the subsequent action be dynamically adjusted according to the detection result, but also the re-planning can be triggered autonomously when an emergency task occurs, communication is interrupted or the environment is abnormal, thereby ensuring the high robustness and completion rate of the task.

[0022] 4. Solving the safety and efficiency problems of traditional manual inspection and fixed automation solutions

[0023] The mobile robot replaces the manual entry into the harsh port environment, ensuring the safety of personnel; at the same time, the high degree of automation and intelligence overcomes the shortcomings of limited coverage and poor flexibility of fixed monitoring equipment, and provides reliable technical support for realizing the predictive maintenance of driving equipment.

[0024] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0026] Figure 1 A flowchart of a fault detection method of a driving equipment provided by the embodiments of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0029] It is to be understood that some of the example embodiments are described in terms of a process or method depicted as a flowchart. Although each step in a flowchart can be identified with a corresponding step in the process, many of the steps can be performed in parallel, concurrently or simultaneously with one another. In addition, the order of the steps can be re-arranged. A process can be terminated when its operations are completed, but can also terminate without regard to whether all exercises of the steps of the process have been completed or not. Processes can correspond in whole or in part to methods, functions, routines, sub-routines, subprograms, etc.

[0030] The embodiment of the present application provides a fault detection method of a driving device, which is cooperatively executed by a mobile robot and a cloud server, and is particularly used for realizing periodic and automatic fault detection of a driving device of a port tower crane.

[0031] The driving device in the embodiment of the present application refers to an electromechanical system providing core power for the port tower crane, and specifically includes but is not limited to a driving motor, a speed reducer, a brake and a transmission mechanism connected therewith.

[0032] The mobile robot, which can also be referred to as a mobile robot system in the present application, refers to a physical platform capable of autonomously moving in the wharf environment described in the present application and performing a detection task. The specific implementation can be a quadruped robot (usually referred to as a "robot dog"), but is not limited thereto. The hardware composition of the mobile robot system mainly includes:

[0033] Mobile platform: mechanical structure and driving system providing mobility, such as a quadruped walking mechanism, a wheeled or tracked chassis.

[0034] Sensor module: a set of various sensors integrated on the mobile platform for sensing the environment and the state of the equipment, at least including:

[0035] Visual sensor, used for collecting appearance images of the target equipment and identifying visual marks;

[0036] Infrared thermal imager, used for acquiring temperature distribution data of the surface of the equipment and detecting abnormal heating points;

[0037] Vibration sensor, used for monitoring mechanical vibration signals during operation of the driving equipment;

[0038] Identification code recognition device (such as a two-dimensional code scanner), used for quickly and accurately identifying and associating specific equipment at a detection point.

[0039] Data processing and communication unit: as the control core of the mobile robot, at least including a processor, a memory and a communication module. It is configured to perform the following core tasks:

[0040] Controlling the start and stop, working mode and parameters of the sensor module;

[0041] Processing and buffering the sensor data collected locally;

[0042] Through the communication module (such as a 5G / Wi-Fi module), a stable connection is established with a remote cloud server, and bidirectional transmission of task data, detection schemes and result data is performed.

[0043] The cloud server is a remote computing and control center of the method, and is at least configured with:

[0044] An equipment profile database is used to store and manage the complete health records of each drive equipment.

[0045] A path planning engine is used to perform global detection path calculation based on multiple constraint conditions.

[0046] An intelligent analysis engine is used to dynamically generate detection instruction sets according to historical data of the equipment, and to perform fault analysis and diagnosis on uploaded equipment state data.

[0047] A task scheduling and communication interface is used to issue instructions and paths to the mobile robot, and to receive data returned by the mobile robot.

[0048] As shown in Figure 1 The fault detection method of the drive equipment provided by the embodiment of the application comprises the following steps:

[0049] S100, generating a global detection path based on the remaining power of the mobile robot, the expected data communication load, the mapping relationship between the power and the detection accuracy level, and the number of equipment to be detected.

[0050] In the application, the global detection path, also referred to as a global detection path sequence, refers to an ordered set generated by a path planning algorithm, which specifies the optimal order and spatial route of the mobile robot visiting each detection point in turn.

[0051] Among them, the equipment to be detected is the drive equipment to be detected. The expected data communication load is estimated based on the data volume, data type of the detection data to be transmitted at each detection point, and the historical communication signal strength of the corresponding detection point.

[0052] In a preferred embodiment, the expected data communication load is estimated by the following formula: CL= (K type -D size ) / S strength , wherein CL represents the expected data communication load of a certain detection point, and the higher the value, the heavier the communication burden. K typeThe coefficient is adjusted for data type, and its value is set according to the urgency of data and transmission requirements; for example, for data with high real-time requirements (such as video streams), the coefficient is greater than 1.0; for non-real-time data (such as sensor readings), the coefficient is equal to or less than 1.0. size The estimated data volume (unit: MB) of the detection data to be uploaded at the detection point. strength The historical average communication signal strength (unit: dBm) of the detection point.

[0053] In another more preferred embodiment, the expected data communication load is estimated by the following formula: CL=α×(K type -D size ) + β × (1 / S strength ), α and β are weight coefficients, α + β = 1, used to adjust the relative importance of data volume factor and signal strength factor in total load calculation. For example, when emphasizing data transmission efficiency, the weight of α can be increased; when emphasizing connection reliability, the weight of β can be increased.

[0054] Further, S100 specifically comprises:

[0055] S101, constructing a port environment map containing the positions of all devices to be detected, obstacle areas and safe passages.

[0056] Specifically, this step aims to create a digital environment model that accurately reflects the physical space of the port and the task elements. Its construction process includes the following sub-steps:

[0057] S1011, environment data collection:

[0058] The port operation area is scanned by the laser radar, depth camera and inertial measurement unit carried by the mobile robot. Preferably, simultaneous localization and mapping technology can be used to make the mobile robot move in an unknown environment while gradually building a map and determining its own position. To obtain complete data, the mobile robot can first traverse the port along a pre-set exploration path.

[0059] S1012, map element extraction and labeling:

[0060] Based on the collected point cloud data and image data, key semantic information related to the inspection task is extracted from the original environment through point cloud segmentation, image recognition and other algorithms, and labeled in the environment map:

[0061] The position of the device to be detected: identify and label the accurate three-dimensional coordinates (such as X, Y, Z values in the world coordinate system) of each tower crane driving device, and associate it with its device identification code.

[0062] Obstacle region: Identify and label the boundary and occupied space of fixed obstacles (such as container stacks, fixed piles, buildings) and temporary obstacles (such as cargo stacking area, construction fence).

[0063] Safe passage: Identify and label the area that the mobile robot can pass through, such as the lane, the inspection dedicated passage, etc., and can classify the passage according to the flatness and risk level of the ground.

[0064] S1013, map fusion and optimization:

[0065] The real-time map constructed by the mobile robot can be fused with the pre-acquired port CAD drawing or high-precision satellite map to correct the cumulative error and provide a more macroscopic environmental context. The finally generated map is a three-dimensional grid map or point cloud map containing semantic information, which is used as the input of the path planning module.

[0066] Through the above steps, the constructed port environment map not only contains basic geometric spatial information, but also integrates task-oriented semantic information, which lays a reliable data foundation for subsequent generation of efficient and safe global detection paths.

[0067] S102, based on the port environment map, the remaining power of the mobile robot, the expected data communication load, and the mapping relationship between the power and the detection accuracy level are quantified as constraint conditions for path planning.

[0068] In the embodiment of the application, the constraint conditions can include:

[0069] (1) Power constraint: Establish the quantitative relationship between the remaining power E of the mobile robot and the maximum feasible action radius R, the expression is: R≤k×(E-E safe ), wherein k is the energy consumption coefficient, E safe is the minimum reserve power required to ensure the safe return of the mobile robot. The total length of any path must satisfy this action radius constraint.

[0070] (2) Communication load constraint: Define a communication load evaluation function for each detection point r: CL(r)=(K type -D size (r)) / S strength (r), CL(r) is the expected data communication load of the rth detection point, D size (r) represents the estimated data volume of the detection data to be uploaded at the rth detection point, S strength (r) is the historical average communication signal strength of the rth detection point, r takes a value from 1 to n, and n is the total number of detection points. When planning the path, the communication load of each detection point must not exceed the preset threshold.

[0071] (3) Detection accuracy level adaptive constraint:

[0072] To realize the intelligent allocation of electric quantity resources, multiple detection accuracy levels are preset for the mobile robot. The mapping relationship between the electric quantity and the detection accuracy level is configured as a dynamic decision logic, the core of which is to allocate appropriate detection accuracy levels to each detection point according to the real-time remaining electric quantity, and each detection accuracy level is associated with different sensor working modes and data acquisition time lengths.

[0073] The mapping relationship is specifically implemented by the following strategies:

[0074] When the remaining electric quantity is higher than a first threshold (for example, 70% of the total electric quantity), a first accuracy level is allocated to the detection point. Under the first accuracy level, a first detection strategy is called, which is configured to: enable a specified combination of high-performance sensor types (such as synchronous working of a high-definition visible light camera and an infrared thermal imager), set a first group of high-precision sensor working parameters (such as high resolution and high sampling rate, for example, the visible light camera resolution is set to 1920x1080 pixels or above, and the sampling rate is not less than 30 frames / second; the infrared thermal imager temperature measurement accuracy is better than ±2°C), and adopt a first acquisition time length that guarantees data completeness (for example, data acquisition is performed on each detection surface of the device for not less than 10 seconds continuously).

[0075] When the remaining electric quantity is lower than the first threshold but higher than a second threshold (for example, 30% of the total electric quantity), a second accuracy level is allocated to the detection point. Under the second accuracy level, a second detection strategy is called, which is configured to: enable a subset of the sensor type combination (such as only enabling the high-definition visible light camera), set a second group of sensor working parameters with lower power consumption than the first group of sensor working parameters (such as standard resolution and reduced sampling rate, for example, the visible light camera resolution is set to 1280x720 pixels, and the sampling rate is reduced to 15 frames / second), and / or adopt a second acquisition time length shorter than the first acquisition time length (for example, the acquisition time length is shortened to 5 seconds) to achieve energy saving.

[0076] When the remaining electric quantity is lower than the second threshold, a conservative strategy is adopted to ensure that the mobile robot can safely return: a third accuracy level is allocated to the detection point or the detection point is directly excluded from the current path, and a path to return to the charging base is planned at the same time. The third accuracy level corresponds to a third detection strategy, which is configured to: only enable a single basic sensor (such as a low-power sensor, for example, only enable a global shutter camera with a resolution of 640x480 pixels) and / or adopt a shortest acquisition time length that can only guarantee basic data acquisition (for example, perform a single snapshot acquisition with a time length of not more than 2 seconds).

[0077] (4) Environmental constraint:

[0078] To further enhance the adaptability of the path in the real terminal environment, the current environment data is also introduced in S102, which at least includes wind speed and visibility. The path planning is based on the current environment data to dynamically correct the mobile energy consumption and feasible region, specifically including:

[0079] (a) Energy consumption correction: in response to the wind speed value exceeding a preset first wind speed threshold (e.g., 15 m / s), in the path calculation, for the path segment located in the geographic area where the wind speed value exceeds the first wind speed threshold, the budget estimate of the mobile energy consumption of the path segment is increased to quantify the additional energy consumption caused by wind resistance to the movement of the mobile robot.

[0080] In an illustrative embodiment, the energy consumption estimate of the corresponding regional path segment is corrected as E wind =E base ×(1+a×(v-v base )) where E wind represents the corrected energy consumption estimate of the mobile robot passing through a certain path segment under the current real-time wind speed value v (unit: m / s). E base represents the basic energy consumption estimate of the mobile robot passing through the same path segment under the reference wind speed value v base . This basic energy consumption estimate is usually calculated based on path length, terrain slope and roughness. The reference wind speed value is a preset reference wind speed used in energy consumption calibration under windless or light wind conditions. a represents the wind resistance influence coefficient (unit: m / s). This is a dimensionless coefficient fitted by experimental data, which is used to quantify the influence degree of unit wind speed increment on the energy consumption of the mobile robot. The coefficient is related to the size, weight and aerodynamic characteristics of the robot.

[0081] When v exceeds v base , (v-v base ) is positive, resulting in a corrected energy consumption estimate higher than the basic energy consumption value, so as to assign a higher energy consumption cost to the high wind speed area in the path planning, guiding the algorithm to preferentially select the path with smaller wind resistance.

[0082] (b) Area exclusion: in response to the visibility value being lower than a preset visibility threshold (e.g., 100 meters), in the path planning, the geographic area located in the visibility value lower than the visibility threshold is excluded from the current feasible detection region set. This is to ensure the data acquisition quality of the optical sensor of the mobile robot and avoid navigation and collision risks caused by unclear vision.

[0083] (5) Task integrity constraint: ensure that the path covers all key detection points, and the number of completed key devices Nc must satisfy: Nc≥Nt×η, where η is the task completion rate threshold, and Nt is the total number of key devices.

[0084] Through the above quantification process, the path planning algorithm can be optimized in a model that truly reflects the task requirements, resource limitations and environmental constraints, thereby generating a global detection path with high efficiency, safety and reliability. Further, the current environmental data is not limited to wind speed and visibility. In the preferred embodiment, other environmental parameters including temperature, humidity, wave height, and rainfall can also be accessed in S102 to cope with more complex port working conditions and trigger corresponding adaptive strategies accordingly. For example:

[0085] When the real-time wave height data exceeds the preset safety threshold, the area near the shore can be determined as a high-risk area, and the path planning is avoided to prevent the equipment from being wetted by the spray or involved in the water.

[0086] When the environmental humidity value exceeds the preset humidity threshold, the moisture-proof protection mode of the sensor can be triggered, for example, the wiping and cleaning function of the exposed optical lens is suspended, or the heating and dehumidifying power of the internal circuit is increased, to ensure the long-term reliability of the sensing unit in harsh conditions.

[0087] Those skilled in the art should understand that the definition of the current environmental data of the present application is open, and the core is to introduce external working condition information that affects mobile safety and detection quality. Therefore, any quantifiable environmental parameter that can reflect such influence falls within the protection scope of the present application.

[0088] S103, an optimization algorithm is used to solve the global detection path that meets the constraint conditions in the port environment map, with the optimization goal of maximizing the total expected detection accuracy level and minimizing the total task energy consumption.

[0089] Specifically, this step uses an intelligent optimization algorithm combining ant colony algorithm and genetic algorithm to find the approximately optimal global detection path in a complex multi-constraint solution space. The intelligent optimization algorithm combining ant colony algorithm and genetic algorithm comprehensively utilizes the positive feedback convergence characteristics of ant colony algorithm and the global search ability of genetic algorithm, and its optimization process iteratively performs the following steps:

[0090] S1031, path population generation step: generate a path population that meets the constraint conditions (such as mobile robot power, communication load, environmental avoidance, etc.). In the first iteration, the initial path population is created by random generation or based on heuristic rules (such as the nearest neighbor method). In subsequent iterations, the new path population is generated by the previous step (S1033, population evolution step).

[0091] S1032, fitness evaluation step: evaluate each path in the current path population based on a fitness function to obtain a corresponding quantitative fitness evaluation result (i.e. fitness value). The fitness function is configured to perform weighted calculation on the total expected detection accuracy, total energy consumption and the number of completed key devices of the path, that is, the core expression of the fitness function is as follows: Fitness = Wp x ∑Pi + We x (1 / ET) + Wc x Nc. Wherein: n i=1 P i + We x (1 / ET) + Wc x Nc. Wherein:

[0092] Fitness represents the fitness value of the path, and the higher the value, the better the comprehensive performance of the path.

[0093] P i Pi represents the sum of the expected detection accuracy levels of the i-th detection point on the path, i is 1 to n, n is the number of detection points on the path, and the sum of the expected detection accuracy levels of all detection points on the path is obtained, that is, the total expected detection accuracy, which reflects the pursuit of overall detection quality.

[0094] ET represents the total energy consumption estimate consumed along the path and performing all detection tasks. It comprehensively calculates the movement energy consumption based on distance, terrain and environment, and the detection energy consumption based on sensor working mode and collection time. In the formula, the reciprocal form is used, aiming to equivalently convert the optimization goal of "minimizing energy consumption" into "maximizing function value", which is convenient for unified processing.

[0095] Nc represents the number of completed key device detections of the path. The key device refers to a device whose fault risk level or production criticality level is determined to be higher than a preset threshold, to ensure the achievement of high-priority tasks.

[0096] Wp, We and Wc are preset non-negative weight coefficients corresponding to detection accuracy, energy consumption and key task completion degree, respectively. By adjusting the values of the three coefficients, the relative importance of different optimization objectives in the overall evaluation can be flexibly adjusted to adapt to different task strategies. For example, when the remaining power of the mobile robot is low, the weight of We can be increased to preferentially select an energy-saving path; when it is necessary to ensure the operation of key equipment, the weight of Wc can be increased to ensure that key equipment is detected preferentially.

[0097] S1033, population evolution step: based on the fitness evaluation result obtained in S1032, selection, crossover and mutation operations are performed on the current path population to generate a new path population for the next iteration.

[0098] The selection operation adopts roulette wheel selection method. This method assigns each path in the population a selection probability proportional to its fitness value. The path with higher fitness value has a higher probability to be selected and reserved to the next generation "mating pool", thus ensuring the inheritance of good genes.

[0099] The crossover operation randomly selects a pair of parent paths from the "mating pool" and adopts sequential crossover method. This operation randomly selects two crossover points, exchanges the path segments between the crossover points in the parent paths, and generates two new and legal child paths through conflict resolution mechanisms (such as deleting duplicate nodes and supplementing missing nodes). This process can effectively explore different combinations of detection points and is the main means of generating new path schemes.

[0100] The mutation operation randomly adjusts individual detection points in the path with a preset small probability. Specific mutation methods include but are not limited to: randomly exchanging the access order of two detection points in the path, or randomly inserting or deleting a non-critical detection point under the premise of meeting the constraint conditions. This operation injects new genes into the population, which helps to maintain population diversity and avoid premature convergence into local optimal solutions.

[0101] The optimization process is iteratively executed through the above operations, driving the path population to continuously evolve and eventually converge to a global detection path sequence with excellent comprehensive performance.

[0102] S1034, convergence judgment step:

[0103] The optimization process continues (S1031→S1032→S1033), and the termination condition is that the fitness of the path population meets the convergence condition, for example, the fitness value of the best path no longer improves significantly (the improvement amplitude is less than a threshold) in the last N generations (such as 50 generations), or reaches a preset maximum iteration number.

[0104] Finally, the algorithm outputs the path with the highest fitness value obtained during the iteration process as the global detection path sequence.

[0105] S200, the mobile robot moves to each detection point in the path sequence corresponding to the global detection path in turn according to the global detection path. Taking the case of moving to a detection point where one of the target driving devices is located as an example, this process specifically includes:

[0106] S2001, path following and real-time perception:

[0107] The navigation control system of the mobile robot decomposes the global detection path into a series of continuous motion instructions, controlling its mobile platform (such as a four-legged walking mechanism) to travel along the global detection path. During the movement, the mobile robot continuously perceives the surrounding environment through its carried laser radar, depth camera and inertial measurement unit, and matches and locates the real-time data with the port environment map constructed in S101, realizing simultaneous localization and map maintenance, so as to realize accurate self-positioning in the known environment framework.

[0108] S2002, dynamic obstacle avoidance:

[0109] During the travel, if the mobile robot detects through the sensor that a dynamic or unknown obstacle (such as temporarily stacked tools, moving vehicles or personnel) not recorded in the port environment map appears on the global detection path, the local processor of the mobile robot will immediately trigger local path re-planning. This process takes the current position of the mobile robot as the temporary starting point, and takes the first reachable point on the original global detection path that crosses the obstacle and is marked as safe in the map as the target, and uses a local planning algorithm to calculate a safe local detour path in real time. After successful avoidance, the mobile robot will automatically return to the original global detection path.

[0110] S2003, state monitoring and emergency handling:

[0111] During the entire movement, the remaining power and core system state of the mobile robot are continuously monitored. If the remaining power is lower than the safe return threshold, or the target point cannot be reached within the specified time, the current task will be aborted, and based on the charging base location marked in the port environment map, the emergency path planning and process of returning will be immediately executed.

[0112] S2004, accurate positioning of detection points:

[0113] When the mobile robot approaches the target detection point according to the path planning, it activates its vision sensor in combination with the approximate position of the target driving device corresponding to the target detection point pre-stored in the port environment map, scans and identifies the identification code (such as a two-dimensional code) on the target driving device. By identifying the identification code, the mobile robot can finally calibrate its own position and the accurate position of the target driving device, ensuring that it is parked at the correct pose that can perform the best detection action.

[0114] S300, after the mobile robot arrives at the detection point, it identifies the identification code of the detection point and uploads the identification information to the cloud server, and then receives the detection instruction set dynamically generated by the cloud server based on the device image information of the target driving device associated with the identification code.

[0115] Specifically, this step includes the following processes:

[0116] Identification association and data request:

[0117] The mobile robot reads the identification code on the target drive equipment through its identification code recognition module (such as a two-dimensional code scanner). The identification code uniquely associates a specific target drive equipment, and the exact location of the equipment has been pre-stored in the port environment map as its attribute information (i.e. equipment coordinates). The mobile robot uploads the identification code to the cloud server as a keyword for data retrieval through its communication module to request a customized detection scheme for the equipment.

[0118] Cloud intelligent decision-making and instruction generation:

[0119] After the cloud server receives the identification code, it can combine the mobile robot's uploaded self-pose (based on the environment map) to start the dynamic instruction generation process, which specifically includes:

[0120] S301, data retrieval: the cloud server retrieves the equipment profile information of the target drive equipment that exactly corresponds to the received identification code in its equipment management database.

[0121] Among them, the equipment profile information is a multi-dimensional data model constructed for fault prediction, which at least includes the following four types of data:

[0122] Historical state time series data: refers to the equipment operating parameters recorded in chronological order, used to analyze performance degradation trends, such as vibration signal time series data (such as time domain waveforms collected by acceleration sensors, or frequency domain spectrum sequences after Fourier transform), temperature time series data (such as temperature curves recorded by key part temperature sensors over time), current / voltage time series data (such as three-phase current and voltage waveform data of the drive motor).

[0123] Equipment maintenance information: refers to the maintenance records of the equipment, used to evaluate the impact of maintenance status on health, such as the time of previous maintenance, the list of replaced parts, and performance calibration data after maintenance.

[0124] Equipment static attributes: refers to the inherent identity and design parameters of the equipment, providing personalized baseline for fault prediction, such as equipment model, factory number, rated power, design life, structural drawings, etc.

[0125] Contextual environment data: refers to the working condition information when the equipment was running in the past, used to identify abnormalities caused by external factors, such as historical detection environment temperature, humidity, and equipment load rate.

[0126] S302, intelligent prediction: based on the retrieved device profile information, real-time analysis is performed through the built-in machine learning model to predict the potential fault type (such as bearing wear, insulation aging) and risk position (such as the non-driving end of the drive motor, the output shaft of the reduction box) of the target drive device, and the corresponding prediction result is generated.

[0127] Among them, the machine learning model is a fault prediction model based on deep learning, which is constructed, input and output as follows:

[0128] Model training: the model is trained offline on the cloud server based on a large number of historical device archives. The training data comes from the annotated device data accumulated by the terminal in the past few years, the input is the state data sequence of the drive device, and the output label is the actual final fault type and fault position. Through continuous iteration optimization (such as minimizing the prediction error by gradient descent method), the model learns to identify sensitive features and patterns that cause device failure from complex multi-source historical data.

[0129] Model input: when making predictions, the input of the model is a feature vector that integrates multi-dimensional information. The vector at least includes real-time and historical state time series data of the target drive device, device maintenance information, device attribute information and context environment data.

[0130] Model output: the output of the model is the prediction result for the current target drive device, which specifically includes:

[0131] Probability distribution of potential fault type: output a list of various possible faults (such as "bearing wear", "insulation aging", "gear tooth breakage") and their corresponding occurrence probabilities.

[0132] Risk position positioning information: output one or more riskiest components or positions (such as "drive motor non-driving end bearing", "three-stage gear of reduction box").

[0133] S303, instruction set construction: generate a highly targeted detection instruction set according to the prediction result. The detection instruction set at least includes the following two types of core instructions:

[0134] (1) Detection parameter instruction: accurately specifies the technical parameters required to perform detection, which is used to configure the working mode of the sensor. Its specific content includes:

[0135] Specified sensor type (for example, when the prediction model shows that the device has an overheating risk, the instruction will specify to enable the infrared thermal imager);

[0136] Sensor configuration parameters (for example, set the emissivity value of the infrared thermal imager, or configure the resolution and sampling frame rate of the visual camera);

[0137] Key points of data collection (i.e. potential risk points indicated in the prediction result);

[0138] Collection duration (e.g. long enough signal sampling time is set to ensure the accuracy of vibration spectrum analysis).

[0139] (2) Spatial relationship instruction: generated based on the prior knowledge of the three-dimensional spatial structure of the equipment stored in the port environment map, used to guide the mobile robot to adjust its own detection pose or the orientation of the sensor gimbal. This spatial relationship instruction converts abstract detection points into specific spatial coordinates and actions, for example:

[0140] “Target point positioning: drive the motor non-drive end bearing, located on the north side of the equipment facade, 1.2 meters above the ground.”

[0141] “Execution action: adjust the gimbal pitch angle to +15 degrees, and focus on the above target point.”

[0142] By decomposing the detection task into specific detection parameter instructions and spatial relationship instructions, a precise leap from high-level task-level instructions to action-level execution that can be directly driven is achieved. This enables the mobile robot to simulate an experienced engineer, not only clearly indicating what needs to be detected, but also autonomously and accurately completing how to arrive and perform the detection, thereby significantly improving the automation level and reliability of the entire detection process.

[0143] Subsequently, the cloud server issues the generated personalized detection instruction set to the mobile robot at the detection point through a wireless network. The data processing and communication unit of the mobile robot receives and verifies the instruction set, preparing for the next execution.

[0144] Further, to ensure the robustness of the task in a poor communication environment, S300 also includes a communication interruption processing flow: if the mobile robot cannot establish a stable communication connection with the cloud server at the detection point, a local emergency detection mode is enabled, and the pre-stored basic detection instruction set in the local storage is executed to complete data collection, and the collected data is cached, and uploaded to the cloud server after the communication is restored. The basic instruction set includes a data collection process for general driven equipment that ensures basic safety, such as using a visible light camera to take a global photo and collecting a fixed duration of vibration signal.

[0145] In the local emergency detection mode, the mobile robot autonomously navigates to one or more preset key detection points based on a simplified map stored in the local memory, where historical data confirms that the communication signal strength of the key detection points is above a preset strength threshold for more than 80% of the task cycles in the last 10 task cycles, and then attempts to execute the basic detection instruction set again to maximize the communication window to complete the detection task of the key equipment.

[0146] The cloud server maintains a historical communication quality database that records the average signal strength (unit: dBm) of each detection point in past tasks. Only detection points with a consistently stable average signal strength above a preset strength threshold (e.g., -85 dBm) are marked as high signal strength key points and pre-stored in the mobile robot's simplified map. This key point list is not fixed and will be updated regularly by the cloud server based on new communication data to ensure that the mobile robot always navigates to the location with the best current network conditions in emergency mode.

[0147] S400, the mobile robot parses and executes the detection instruction set to control the sensors on board to collect data from the target drive equipment and generate equipment state data.

[0148] This step is the specific implementation of cloud-based intelligent decision-making at the terminal, and its core is to convert abstract detection tasks into a series of precise mechanical and sensor actions. Specifically, it includes the following sub-steps:

[0149] S401, parse the detection instruction set into a series of bottom-level executable instructions.

[0150] The data processing unit of the mobile robot receives the detection instruction set issued by the cloud and parses it. The instruction set is translated into a series of bottom-level executable instructions. These instructions are specifically divided into two categories:

[0151] Mechanical control instructions: used to directly drive the mechanical structure of the mobile robot, such as controlling the torque and speed of the joint motor, planning the foot trajectory.

[0152] Sensor control instructions: used to directly configure and drive specific sensors, such as setting the focal length and aperture of the camera, setting the temperature measurement range and emissivity of the infrared thermal imager, configuring the sampling frequency and range of the vibration sensor.

[0153] S402, adjust the movement gait of the mobile robot, the relative distance to the target drive equipment, and the detection posture according to the mechanical control instructions.

[0154] According to the parsed mechanical control instructions, the bottom-level motion controller of the mobile robot performs corresponding actions:

[0155] Adjusting the moving gait: According to the final distance from the target, switch from the marching gait to a more refined adjustment gait to achieve smooth and precise approach.

[0156] Adjusting the relative distance: Control the mobile robot to move to the optimal observation position specified by the instruction relative to the target device.

[0157] Adjusting the detection posture: Adjust the body posture and gimbal angle so that the mounted sensors can align with the key data collection points specified in the instruction at the optimal viewing angle and distance.

[0158] S403, according to the sensor control instruction, activate the specified sensor and collect data according to the set parameters and collection time.

[0159] After the mobile robot reaches the predetermined position and stabilizes the posture, according to the analyzed sensor control instruction, the following operations are performed:

[0160] Activate the specified sensor: accurately activate one or more sensors according to the sensor control instruction (such as "activate only the infrared thermal imager" or "synchronously activate the visible light camera and vibration sensor").

[0161] Configure sensor parameters: set the working parameters of the sensor according to the sensor control instruction (such as set the image resolution of the infrared thermal imager to 640x480 and the sampling frame rate to 25Hz).

[0162] Perform data collection: control the sensor to collect data within the collection time set by the sensor control instruction (such as continuously collect vibration signals for 10 seconds).

[0163] Generate device state data: encapsulate and timestamp the collected raw data (such as images, temperature matrices, and vibration waveforms) to generate structured device state data, preparing for subsequent data upload.

[0164] S500, the mobile robot uploads the collected device state data to the cloud server; the cloud server compares the device state data with the preset normal operation parameter threshold range, and sends subsequent action instructions to the mobile robot based on the comparison result.

[0165] Specifically, S500 can include the following sub-steps:

[0166] S501, data upload and reception:

[0167] The mobile robot uploads the device state data generated in S400 to the cloud server through its communication unit. The cloud server receives and verifies the integrity of the data packet.

[0168] S502, cloud intelligent analysis and comparison:

[0169] The cloud server calls its internal data analysis engine to perform the following operations:

[0170] Data parsing and feature extraction: From the received device state data, extract key feature indicators (such as maximum temperature value, vibration frequency amplitude, crack density, etc.).

[0171] Threshold comparison: Compare the extracted key feature indicators with the preset normal operation parameter threshold range determined based on the device model and historical baseline.

[0172] The normal operation parameter threshold range determined based on the device model and historical baseline is pre-set and stored in the cloud server by one of the following methods:

[0173] Method 1: Statistical model method

[0174] Based on a large amount of historical data (historical baseline) of the model device under normal state, it is calculated by statistical analysis model. For example, for a vibration characteristic value, its threshold range can be set as [μ-3σ, μ+3σ] within the range of three times standard deviation of historical mean value, where μ is the historical mean value and σ is the historical standard deviation.

[0175] Method 2: Combination of industry standards and experience

[0176] First, set a basic threshold according to the technical specifications provided by the equipment manufacturer and industry standards, and then fine-tune it based on the historical operation data of the device itself. For example, for temperature parameters, based on the upper limit of the rated temperature given by the manufacturer, a floating correction of ±5% can be made according to the stability of the long-term operation of the device.

[0177] As understood by those skilled in the art, any feasible method based on historical normal operation data of the device to quantify the boundaries of "normal" and "abnormal" falls within the scope of protection of the present application.

[0178] S503, decision making and instruction issuing:

[0179] The cloud server generates specific control instructions based on the comparison results and issues them to the mobile robot:

[0180] If the device state data is within the preset normal operation parameter threshold range, it is determined that the device state is normal, and the mobile robot is instructed to go directly to the next detection point.

[0181] If the device status data exceeds the preset normal operating parameter threshold range, the device status is determined to be abnormal. At this time, the decision logic of the cloud server is not singular, but further makes a comprehensive judgment based on the urgency of the repair of the target driven device and the remaining power of the mobile robot, so as to achieve the optimal balance between task reliability, equipment safety and mobile robot endurance.

[0182] The urgency level (or abnormal urgency level) of the repair is calculated by the cloud server based on the following factors:

[0183] Severity of abnormal data: that is, the extent to which the current data deviates from the normal threshold.

[0184] Historical fault records of the equipment: the frequency and severity of past faults of the equipment.

[0185] Criticality level of equipment in production: The importance of the equipment in the dock operation process.

[0186] Specifically, the urgency level E of the maintenance meets the following condition: E = g1 × Sc + g2 × H + g3 × C, where Sc is the current severity score of the anomaly. This is calculated based on the extent to which the current test data exceeds the normal operating threshold. For example, it can be defined as: Sc = |Vm - Vt| / Vt, where Vm is the measured value of the current test data, and Vt is the threshold value of the current test data.

[0187] H is the historical failure tendency score, calculated based on the equipment's historical failure records and maintenance files. For example, it can be defined as: H = (Nf / T) × (1 + (∑...) m j=1 SL j ) / Nf), where Nf is the total number of faults recorded for the device within a preset historical statistical period (e.g., the past year). T is the total runtime (in hours) within the same statistical period, used to normalize the fault frequency. SL j To quantify the historical fault level of the j-th historical fault, in a preferred embodiment, the severity level can be quantified as follows: 0.5 for general faults, 1.0 for severe faults, and 1.5 for emergency faults. This design ensures that equipment with higher fault frequency and greater average severity has a higher historical fault tendency score, thus gaining higher weight in the maintenance urgency assessment. The value of j ranges from 1 to m, where m is the total number of historical faults.

[0188] C represents the production criticality score, a static parameter pre-set based on the importance of the driving equipment in the dock operation process. For example, the core main tower crane can be set to 10, the auxiliary tower crane to 6, and the general conveying equipment to 3.

[0189] g1 to g3 are weight coefficients, g1 + g2 + g3 = 1, for adjusting the relative importance of the three dimensions. As a preferred embodiment, g1 = 0.5, g2 = 0.3, g3 = 0.2 can be set to highlight the primary status of the current anomaly.

[0190] After calculating the emergency score E, it is mapped to an emergency level used by the decision logic by comparing it with preset level thresholds:

[0191] High emergency (first emergency level): ≥ T1 (e.g., T1 = 8);

[0192] Medium emergency (second emergency level): T2 ≤ E < T1 (e.g., T2 = 5);

[0193] Low emergency (third emergency level): E < T2.

[0194] Based on the above factors, the output of the decision logic is as follows:

[0195] When the repair emergency level is rated as high (first emergency level), and the mobile robot has more than a first power threshold (e.g., 40% of the total power) remaining, the cloud server determines an emergency confirmation scenario. At this time, the cloud server instructs the mobile robot to immediately re-detect to obtain more reliable data to confirm the anomaly and provide a basis for decision-making for potential emergency repair.

[0196] When the repair emergency level is rated as medium (second emergency level), or it is determined through environmental data that there may be transient interference at present, the cloud server determines a temporary avoidance for detection scenario. At this time, the cloud server instructs the mobile robot to wait for a preset time (e.g., 5 minutes) and then re-detect. The mobile robot can move to a nearby safe area to temporarily avoid during this period, and then try to detect again after the preset time to exclude transient interference or confirm whether the anomaly persists.

[0197] When the repair emergency level is rated as low (third emergency level), or the mobile robot has less than a second power threshold (e.g., 25% of the total power) remaining, the cloud server determines a record and retreat scenario. At this time, priority is given to ensuring that the mobile robot can safely return and complete subsequent key tasks. The cloud server generates a detailed anomaly alarm log and instructs the mobile robot to go to the next detection point to ensure the completion rate of the overall inspection task.

[0198] Further, the method further comprises a dynamic re-planning step:

[0199] S600, during the execution of the global detection path by the mobile robot, the actual power consumption and communication signal strength of the mobile robot are continuously monitored.

[0200] In the process of the mobile robot executing the global detection path, the local processor of the mobile robot continuously monitors key real-time parameters, mainly including:

[0201] Actual power consumption: by monitoring the battery management system data, the actual unit distance energy consumption is calculated.

[0202] Communication signal strength: real-time monitoring of the signal strength (such as RSRP value) of the communication link with the cloud server.

[0203] S700, abnormality evaluation and path re-planning triggering.

[0204] In the embodiment of the application, the specific determination criteria of the triggering condition are as follows:

[0205] Energy consumption abnormality triggering: when the average value of the actual unit distance energy consumption in a continuous first preset time period (for example, 5 minutes) is greater than the expected unit distance energy consumption in the path planning by a preset percentage (for example, 20%), that is, the ratio of the difference between the average value of the actual unit distance energy consumption and the expected unit distance energy consumption in the path planning to the expected unit distance energy consumption reaches or exceeds the preset percentage, the path re-planning is triggered. This situation indicates that the current terrain or environmental resistance (such as strong wind, steep slope) is much higher than the estimate, and continuous execution of the original path will result in rapid consumption of power, and there is a risk of task failure.

[0206] Communication interruption triggering: when the communication signal strength continuously falls below the minimum signal strength threshold required to maintain the transmission of real-time control instructions and state feedback data at a preset packet loss rate for a continuous second preset time period (for example, 30 seconds), the path re-planning is triggered. This situation indicates that the mobile robot is about to enter or has entered a communication blind area, and cannot guarantee normal interaction with the cloud, and needs to adjust the path in time to restore communication or avoid the blind area. Among them, the real-time control instructions mainly include the detection instruction set and subsequent action instructions issued by the cloud server to the mobile robot; the state feedback data mainly includes the device state data and its own state information uploaded by the mobile robot. In specific implementation, the preset packet loss rate can be set according to the requirement for control real-time, for example, set to 1%. The minimum signal strength threshold is the critical signal strength determined through experiments or experience, which can meet the requirement of the preset packet loss rate. In an exemplary embodiment, for example, it can be -100 dBm.

[0207] Once any of the above triggering conditions is met, the re-planning process will be started, and the updated global detection path will be generated and issued to the mobile robot. The re-planning process includes:

[0208] Data upload and request: the mobile robot sends its current location, remaining power, and monitored abnormal data to the cloud server as new constraints, and requests path re-planning if communication is available.

[0209] Cloud re-planning: the cloud server re-executes the path planning steps (S100) based on the received latest state of the mobile robot and global environment data, and generates an updated global inspection path that meets the current actual situation.

[0210] Instruction issuance and execution: the new global path is issued to the mobile robot through the wireless network. After receiving and verifying, the mobile robot immediately suspends the original path and continues to perform the inspection task according to the updated global inspection path.

[0211] Optionally, if re-planning is triggered due to communication interruption and cannot be connected with the cloud server, the mobile robot can enable local emergency re-planning logic. This re-planning logic generates a safe emergency path based on the simplified map stored in the local memory and the preset rules (such as "prefer to return to the area with the highest signal strength" or "directly plan a path to return to the base").

[0212] Further, S700 also includes an emergency task response mechanism: when the cloud server receives an inspection instruction from the operation terminal with the highest priority label, the mobile robot will respond to the instruction immediately regardless of whether it is currently performing a regular inspection task, charging, or in standby state. The cloud server will analyze the instruction to determine the target drive device, and then insert the detection point corresponding to the target drive device as the highest priority target at the front of the current path planning queue. After the insertion, the path planning process described in S100 is re-executed to generate a new global inspection path that incorporates the emergency inspection task, and is immediately issued to the mobile robot for execution.

[0213] Specifically, the emergency task response mechanism includes:

[0214] Instruction reception and analysis: the cloud server receives an emergency inspection instruction from the operation personnel, which specifies the target drive device that needs to be immediately inspected.

[0215] Task insertion: the cloud server inserts the detection point corresponding to the target drive device as the highest priority target that must be accessed into the sequence of detection points that the mobile robot has not completed.

[0216] Emergency re-planning: the cloud server re-executes the path planning step (i.e., S100) with the current position as the starting point and ensures that the new global detection path is prioritized to go to the emergency detection point, thereby generating a new global detection path that integrates the emergency task.

[0217] Instruction issuing and execution: the new global detection path is immediately issued to the mobile robot, the mobile robot interrupts the current action, and goes to perform the emergency detection task according to the new global detection path.

[0218] The embodiment of the present application also provides an electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are configured to execute the method described in the embodiment of the present application.

[0219] The embodiment of the present application also provides a computer readable storage medium, which stores computer executable instructions, and the computer executable instructions are used for executing the method described in the embodiment of the present application.

[0220] It should be understood that the various forms of flow shown above can be used to reorder, add or delete steps. For example, the steps described in the present application can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in the present application can be achieved, which is not limited herein.

[0221] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A failure detection method of a drive device, characterized by, The method is executed by a mobile robot in cooperation with a cloud server, and comprises the following steps: S100, generating a global detection path based on the remaining power of the mobile robot, the expected data communication load, the mapping relationship between the power and the detection accuracy level, and the number of devices to be detected; S200, the mobile robot moves to each detection point in the global detection path in turn according to the global detection path; S300, after the mobile robot arrives at the detection point, the mobile robot identifies the identification code of the detection point and uploads the identification information to the cloud server, and then receives a detection instruction set dynamically generated by the cloud server based on the device image information of the target drive device associated with the identification code; S400, the mobile robot analyzes and executes the detection instruction set, controls the sensor carried by the mobile robot to collect data of the target drive device, and generates device state data; S500, the mobile robot uploads the collected device state data to the cloud server; the cloud server compares the device state data with the preset normal operation parameter threshold range, and sends a subsequent action instruction to the mobile robot based on the comparison result.

2. The method of claim 1, wherein, S100 specifically comprises: S101, constructing a terminal environment map containing the positions of all devices to be detected, obstacle areas and safe channels; S102, based on the terminal environment map, quantifying the remaining power of the mobile robot, the expected data communication load, and the mapping relationship between the power and the detection accuracy level as constraint conditions for path planning; S103, taking maximizing the total expected detection accuracy level and minimizing the total task energy consumption as the optimization goal, using an optimization algorithm to solve the global detection path that meets the constraint conditions in the terminal environment map.

3. The method of claim 2, wherein, In S102, the mapping relationship between the power and the detection accuracy level is configured to dynamically assign appropriate detection accuracy levels to each detection point from a plurality of preset detection accuracy levels according to the real-time remaining power of the mobile robot; wherein different detection accuracy levels are associated with different sensor operating modes and data collection durations.

4. The method of claim 3, wherein, The mapping relationship specifically comprises: When the remaining power is higher than a first threshold, a first accuracy level is assigned to the detection point, and a first detection strategy is called, the first detection strategy is configured to enable a specified sensor type combination, set a first group of sensor operating parameters, and use a first collection duration; When the remaining power is lower than the first threshold but higher than a second threshold, a second accuracy level is assigned to the detection point, and a second detection strategy is called, the second detection strategy is configured to enable a subset of the sensor type combination, set a second group of sensor operating parameters with lower power consumption than the first group of sensor operating parameters, and / or use a second collection duration shorter than the first collection duration; When the remaining power is lower than the second threshold, a third accuracy level is assigned to the detection point or the detection point is excluded from the current path, and a path back to the base is planned; wherein the third accuracy level corresponds to a third detection strategy, and the third detection strategy is configured to enable only a single basic sensor and / or use the shortest collection duration to ensure basic data acquisition.

5. The method of claim 2, wherein, In S102, current environment data is also introduced, which at least includes wind speed and visibility; the path planning dynamically corrects the mobile energy consumption and feasible region based on the current environment data.

6. The method of claim 2, wherein, In S103, the optimization algorithm adopted is a hybrid ant colony algorithm and genetic algorithm, and the optimization process of the hybrid ant colony algorithm and genetic algorithm iteratively performs the following steps: A path population generation step: a path population satisfying the constraint condition is generated, and an initial path population is generated at the first iteration, and a new path population is generated in subsequent iterations from the previous step; An adaptability evaluation step: each path in the current path population is evaluated based on an adaptability function to obtain a corresponding adaptability evaluation result; the adaptability function is configured to perform weighted calculation on the total expected detection accuracy of the path, the total energy consumption, and the number of completed key devices; A population evolution step: based on the adaptability evaluation result, selection, crossover and mutation operations are performed on the current path population to generate a new path population for the next iteration; The optimization process continues until the adaptability of the path population meets the convergence condition.

7. The method of claim 1, wherein, In S100, the expected data communication load is estimated based on the data volume, data type of the detection data to be transmitted at each detection point, and the historical communication signal strength of the corresponding detection point.

8. The method of claim 2, wherein, In S300, the detection instruction set is obtained by the following steps: S301, the cloud server receives an identification code and calls the device image information of the target driving device corresponding to the identification code, and the device image information includes historical state time series data, device maintenance information, device attribute information and context environment data; S302, based on the device image information of the target driving device, the potential fault type and risk position of the target driving device are predicted by a machine learning model to generate a corresponding prediction result; S303, according to the prediction result, a targeted detection instruction set is generated, which includes detection parameter instructions and spatial relationship instructions, wherein the detection parameter instructions are used to configure the sensor working mode, and the spatial relationship instructions are used to guide the mobile robot to adjust the detection pose based on the wharf environment map.

9. The method of claim 1, wherein, In S400, the mobile robot analyzes and executes the detection instruction set, specifically including: S401, the detection instruction set is parsed into a series of bottom layer executable instructions, which include mechanical control instructions for directly controlling the mechanical structure of the mobile robot, and sensor control instructions for directly configuring and driving the sensor; S402, according to the mechanical control instructions, adjust the moving gait, relative distance to the target driving device and detection pose of the mobile robot; S403, according to the sensor control instructions, activate the specified sensor and perform data acquisition according to the set parameters and acquisition time.

10. The method of claim 1, wherein, The method further includes a dynamic re-planning step: S600, in the process of executing the global detection path by the mobile robot, the actual power consumption and communication signal strength of the mobile robot are continuously monitored; S700, when any of the following conditions is met, trigger path re-planning: Condition 1: within a continuous first preset time period, the average value of the actual unit distance energy consumption exceeds the expected unit distance energy consumption by a preset proportion; Condition 2: within a continuous second preset time period, the communication signal strength is continuously lower than the minimum signal strength threshold required for maintaining the transmission of real-time control instructions and state feedback data at a preset packet loss rate; When the above conditions are triggered, an updated global detection path is generated and sent to the mobile robot.

11. The method of claim 10, wherein, S700 further comprises: when the cloud server receives a detection instruction from the operation terminal and the instruction priority label is the highest, inserting the detection point associated with the target drive device corresponding to the highest detection instruction as the highest priority target into the current path planning, and re-executing S100 to generate a new global detection path.

12. The method of claim 1, wherein, In S500, if the equipment state data is outside the preset normal operation parameter threshold range, the mobile robot is instructed to perform re-detection or go to the next detection point. If the equipment state data is within the preset normal operation parameter threshold range, the mobile robot is instructed to go directly to the next detection point.

13. An electronic device, comprising: comprise a processor and a memory; The processor is configured to execute the steps of the method according to any one of claims 1 to 12 by invoking programs or instructions stored in the memory.

14. A computer-readable storage medium, characterized in that, The computer readable storage medium is configured to store programs or instructions for enabling a computer to execute the steps of the method according to any one of claims 1 to 12.

Images