A data-driven robot dynamic inspection path planning method

Abstract

The present application relates to the technical field of path planning, in particular to a robot dynamic inspection path planning method based on data driving. The method comprises the following steps: using the dynamic surface convection heat transfer coefficient, the stator real-time current, the stator resistance and the body real-time temperature to calculate the transient temperature increment corresponding to the robot crossing a single discrete grid node; calculating the expected total temperature based on the transient temperature increment and the body real-time temperature; generating the dynamic resistance expansion ratio through a nonlinear penalty function based on the expected total temperature and the rated maximum tolerance temperature; weighting the physical distance of the robot movement by using the dynamic resistance expansion ratio to construct a directed dynamic thermal impedance optimization cost function. The inspection path is generated by searching the path based on the directed dynamic thermal impedance optimization cost function. The present application can make the robot avoid the dangerous path of wind heating, and effectively prevent the thermal shutdown failure of the robot in the complex high-temperature flow field.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, and more specifically to a data-driven method for dynamic inspection path planning of robots. Background Technology

[0002] In complex industrial manufacturing environments such as thermal power plants and large boiler rooms, automated inspection robots are commonly used to perform equipment inspection tasks. Under these physical conditions characterized by strong heat radiation and complex fluid wind fields, path planning is a core technology for ensuring the safe and efficient execution of tasks by robots. Currently, the A* path planning algorithm based on grid maps is widely used in this field for global path search. To avoid thermal protection shutdowns or damage to robots in high-temperature areas, existing technologies have improved upon the traditional A* algorithm. For example, they use environmental sensors to collect historical temperature data of the plant area to generate a two-dimensional environmental temperature field, and forcibly assign infinite impedance to physical grids exceeding a preset static safety temperature threshold in the algorithm, thereby forcing the robot to detour around static high-temperature areas during path planning.

[0003] However, applying the aforementioned existing technologies to real-world, complex fluid physics conditions presents significant safety hazards. The root of the problem lies in the fact that real industrial environments are not static air environments, but rather high-speed, highly convective fluid fields created by exhaust fans, high-temperature exhaust gases, and the chimney effect. In fluid mechanics and heat transfer, the intensity of convective heat transfer on an object's surface is highly dependent on the relative velocity between the fluid and the object's surface. Existing dynamic thermodynamic models merely treat the temperature impedance of the spatial environment as an undirected scalar, completely ignoring the asymmetry of vector aerodynamics. Because the robot's motion vector is not physically coupled with the environmental wind speed vector, existing technologies cannot distinguish the vast difference in heat flux generated when the robot is "moving against the hot wind" versus "moving with the hot wind and towards the wind." When the robot is moving against the wind in a high-temperature area, the rapid superposition of relative wind speeds triggers extremely intense forced surface convection heat transfer, causing environmental heat to flow back into the drive motor at several times the rate; while when moving with the wind, heat exchange is significantly reduced. Because existing optimization algorithms cannot identify directional hazards in such high heat flux flow fields, they are very likely to plan a fatal path with the shortest physical distance under the optimization rules, but which is located on the windward heated surface. As a result, when the robot penetrates the strong convection area against the wind, the actual accumulated heat caused by the combination of ambient temperature and frictional Joule heat far exceeds the expectation, leading to sudden thermal shutdown and mechanical seizure damage. Summary of the Invention

[0004] To address the technical problem of existing technologies treating thermal impedance as an undirected scalar, which fails to quantify the heat transfer asymmetry caused by relative motion in complex flow fields and leads to robots easily planning dangerous paths into the wind and causing thermal shutdown in high-temperature, high-convective environments, this invention provides a data-driven dynamic inspection path planning method for robots. The specific technical solution adopted is as follows:

[0005] One embodiment of the present invention provides a data-driven method for dynamic inspection path planning of robots, comprising the following steps:

[0006] The discrete grid environmental temperature field and discrete grid environmental wind speed vector field of the inspection area are obtained, as well as the robot's body operation status data. Each discrete grid node contains the real-time environmental temperature and local wind speed vector. The body operation status data includes at least the yaw speed vector of the robot's drive motor, the real-time body temperature, and the real-time stator current.

[0007] The relative wind speed modulus is obtained by performing vector operations on the heading velocity vector and the local wind speed vector. A dynamic surface convection heat transfer coefficient is constructed based on the relative wind speed modulus. The transient temperature increment corresponding to the robot crossing a single discrete grid node is calculated using the dynamic surface convection heat transfer coefficient, the stator real-time current and the body real-time temperature.

[0008] The expected total temperature is calculated based on the transient temperature increment and the real-time temperature of the body. Based on the expected total temperature and the rated maximum tolerance temperature, a dynamic resistance expansion ratio is generated through a nonlinear penalty function. The dynamic resistance expansion ratio is then used to weight the physical distance the robot moves, and a directed dynamic thermal impedance optimization cost function is constructed.

[0009] Inspection paths are generated based on the directed dynamic thermal impedance optimization cost function.

[0010] Preferably, the method for obtaining the relative wind speed modulus includes:

[0011] The global positioning and environment matching coordinate transformation matrix configured in the robot's internal memory is invoked to transform the local wind speed vector to the robot's body coordinate system. The vector difference between the heading velocity vector and the transformed local wind speed vector is calculated by vector subtraction, and the relative wind speed modulus is obtained by taking the modulus of the vector difference.

[0012] Preferably, the method for obtaining the dynamic surface convective heat transfer coefficient includes:

[0013] The characteristic length and fluid kinematic viscosity are read from the robot's non-volatile configuration table and calculated based on the relative wind speed modulus, characteristic length, and fluid kinematic viscosity. The Nusselt number is calculated based on the Reynolds number, empirical constant, convective heat transfer index, and Prandtl number. The surface forced convection heat transfer coefficient is calculated based on the Nusselt number, thermal conductivity, and characteristic length. The result of adding the surface forced convection heat transfer coefficient to the basic natural convection heat transfer coefficient is recorded as the dynamic surface convection heat transfer coefficient.

[0014] Preferably, the Reynolds number expression is: The expression for the Nusselt number is as follows: The expression for the surface forced convection heat transfer coefficient is as follows: , For convective heat transfer form factor, The expression is , The heading velocity vector, Let be the local wind speed vector, D be the characteristic length, v be the fluid kinematic viscosity, C0 be the empirical constant, m be the convective heat transfer index, Pr be the Prandtl number, and k be the thermal conductivity.

[0015] Preferably, the method for obtaining the transient temperature increment corresponding to the robot traversing a single discrete grid node includes:

[0016] Based on the magnitude of the heading velocity vector and the side length of the discrete grid, the time window for the robot to traverse a single discrete grid node is obtained; the intrinsic heating power of the robot's drive motor is calculated based on the stator real-time current and stator resistance; the convective heat transfer power is calculated based on the dynamic surface convective heat transfer coefficient, effective heat dissipation area, real-time ambient temperature within the discrete grid node, and real-time body temperature; based on the traversal time window, the algebraic sum of the intrinsic heating power and the convective heat transfer power is integrated to obtain the cumulative net heat generated when traversing a single discrete grid node; the cumulative net heat is divided by the physical heat capacity of the drive motor to obtain the transient temperature increment corresponding to the robot traversing a single discrete grid node.

[0017] Preferably, spanning a time window The expression is The intrinsic heating power of the drive motor The expression is The convective heat transfer power The expression is The cumulative net heat The expression is , Let be the side length of the discrete grid. Let be the magnitude of the heading velocity vector. For stator real-time current, For stator resistance, The dynamic surface convective heat transfer coefficient, For effective heat dissipation area, Real-time ambient temperature within discrete grid nodes. This is the real-time temperature of the body.

[0018] Preferably, the expected total temperature from the preceding node to the current node is the sum of the robot's real-time body temperature when it is at the preceding node and the transient temperature increment corresponding to the robot's transition from the preceding node to the current node; the preceding node is the node preceding the current node.

[0019] Preferably, the method for obtaining the dynamic resistance expansion ratio includes:

[0020] The difference between the rated maximum withstand temperature and the expected total temperature when the robot crosses from the previous node to the current node is calculated to obtain the remaining heat capacity safety margin. The ratio of the expected total temperature when the robot crosses from the previous node to the current node to the remaining heat capacity safety margin is multiplied by a preset nonlinear expansion coefficient and recorded as the dynamic resistance expansion ratio corresponding to the robot crossing from the previous node to the current node.

[0021] Preferably, the expression for the cost function of directed dynamic thermal impedance optimization is:

[0022]

[0023] in, Let n be the cost function for directed dynamic thermal impedance optimization, and also the actual cumulative cost from the starting point of the robot path planning to the current node. Let n represent the current node and n-1 represent the previous node. This represents the actual cumulative cost from the starting point to the preceding node. This represents the single-step cost of the combined thermal impedance when the robot traverses from the previous node to the current node. for , This represents the dynamic drag expansion ratio when the robot moves from the previous node to the current node.

[0024] Preferably, the discrete grid environmental temperature field and discrete grid environmental wind speed vector field of the inspection area are realized based on a fixed infrared thermal imager array and an ultrasonic anemometer array deployed in the inspection area, and the robot's body operation status data are synchronously collected through the robot's industrial fieldbus system at a preset sampling period.

[0025] Beneficial effects: (1) It overcomes the defect of existing technology in being unable to quantify asymmetric heat transfer and realizes accurate prediction of absolute physical temperature rise under extreme flow fields. This invention extracts the heading velocity vector characterizing the robot's current direction and speed, as well as the local wind speed vector in each discrete grid, and uses vector subtraction to extract the relative wind speed modulus. Using this as the core calculation parameter, it introduces the Reynolds number and Nusselt number equations to reconstruct the unidirectional asymmetric dynamic surface convection heat transfer coefficient. This substantial specific technical feature breaks the technical bias of existing technology that only relies on static temperature scalars to evaluate heat, and integrates the relative motion law of fluid passing over the surface of an object in fluid mechanics into the electromechanical heating model of the robot. When the robot travels against the wind, the relative wind speed modulus increases, resulting in an increase in the surface forced convection heat transfer coefficient, accurately restoring the intense heat transfer process of hot air being forced into the motor; when traveling with the wind, it automatically degenerates into basic natural convection. By integrating the asymmetric heat transfer data with the motor's internal heating power, the objective physical temperature rise caused by the robot crossing the discrete grid along a specific vector direction can be accurately predicted, providing an extremely reliable data base for subsequent high-safety path planning. (2) It breaks the limitations of existing static obstacle avoidance algorithms and realizes automatic avoidance of high-risk windward paths from the underlying mathematical causal logic. This invention calculates the remaining heat capacity safety margin between the expected total temperature containing transient temperature increments and the rated maximum tolerance temperature, and uses it as the denominator of the rational fractional obstacle function. It also constructs the dynamic resistance expansion ratio by combining the nonlinear expansion coefficient, and then reconstructs the single-step cost of composite thermal impedance and the directed dynamic thermal impedance optimization cost function. The mechanism of this substantial specific technical feature is that it transforms the safety margin near the thermal shutdown limit into a directed penalty weight in the graph search network. When the robot's heading velocity vector cuts into the high-heat flow field path head-on, the expected total temperature rises sharply, causing the denominator to approach zero, making the single-step cost of composite thermal impedance diverge exponentially towards positive infinity. This mathematically logical directed weight distribution completely changes the blind spot of the traditional A-star algorithm that relies solely on physical distance for optimization. It forces the algorithm network to reduce the priority of high-heat windward paths and actively search for yaw paths with smaller actual cumulative costs under crosswind or tailwind conditions. This enables the robot to have the intelligent avoidance capability of actively seeking "safe hot shear angles" like an aerodynamic aircraft. (3) It fills the gap in transient protection under dynamic fluid environments and greatly improves the inherent safety of continuous inspection operations. This invention deploys a path planning algorithm containing a directed dynamic thermal impedance optimization cost function to the main control unit. During the execution of the physical motion trajectory, it performs high-frequency cyclical execution of relative wind speed extraction and actual cumulative cost update with a preset control cycle. When a step change occurs in the fluid field (such as a sudden high-temperature steam leak exhaust) causing the actual cumulative cost to exceed the system's safe optimization threshold, an interrupt command is immediately triggered and a second calculation is initiated.This control mechanism fully utilizes the characteristic of the aforementioned directional dynamic thermal impedance model being extremely sensitive to relative wind speed, ensuring that before unpredictable fluid mutations cause severe heat backflow, the continued execution of dangerous tasks is blocked at the control level using data causal correlation logic, fundamentally eliminating the risk of equipment damage caused by sudden thermal shutdown. Attached Figure Description

[0026] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of a data-driven robot dynamic inspection path planning method according to the present invention. Detailed Implementation

[0028] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the protection scope of the embodiments of the present invention.

[0029] 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.

[0030] This embodiment provides a data-driven method for dynamic inspection path planning of robots, which is described in detail below:

[0031] like Figure 1 As shown, the robot's dynamic inspection path planning method includes the following steps:

[0032] Step S001: Obtain the discrete grid environmental temperature field and discrete grid environmental wind speed vector field of the inspection area, as well as the robot's body operation status data. Each discrete grid node includes the real-time environmental temperature and local wind speed vector. The body operation status data includes at least the robot's drive motor heading speed vector, the body's real-time temperature, and the stator real-time current.

[0033] It should be noted that the coupling relationship between the environmental thermal field and physical heat capacity is a key physical factor affecting the safety of robot inspection. In order to solve the problem of missing data interaction caused by the separation of environmental physical state and equipment body state in traditional static planning, this embodiment of the invention acquires multi-dimensional environmental physical field and the robot's own dynamic operating parameters in real time through a sensor array deployed inside the factory area. This data serves as the underlying foundation for the subsequent construction of a directed dynamic thermal impedance optimization cost function that includes fluid dynamics characteristics.

[0034] Specifically, in this embodiment of the invention, a fixed array of infrared thermal imagers and an ultrasonic anemometer array deployed within the factory area are used to acquire real-time global thermal radiation data and air velocity and direction data. The acquired global thermal radiation data and air velocity and direction data are converted into a standard grid matrix by the factory's central control system. After three-dimensional spatial projection and dimensionality reduction data processing, a spatial resolution of [missing information] is constructed. Furthermore, the discrete grid environment temperature field and environmental wind speed vector field, represented in meters, represent the current inspection physical space or the current inspection area. The real-time environmental temperature and local wind speed vector within each discrete grid are extracted. One discrete grid unit corresponds to one node, and the node corresponding to each discrete grid unit is recorded as a discrete grid node. Through the industrial fieldbus system inside the robot, with a data sampling period of 20 milliseconds, the real-time temperature of the drive motor body, the real-time current of the motor stator, and the heading velocity vector representing the current motion direction and speed are collected synchronously.

[0035] Step S002: Perform vector operation on the heading velocity vector and the local wind speed vector to obtain the relative wind speed modulus. Based on the relative wind speed modulus, construct the dynamic surface convection heat transfer coefficient. Using the dynamic surface convection heat transfer coefficient, the stator real-time current, and the body real-time temperature, calculate the transient temperature increment corresponding to when the robot crosses a single discrete grid node.

[0036] It should be noted that existing dynamic thermodynamic models ignore the asymmetry of vector aerodynamics, which makes it impossible for existing technologies to distinguish the difference in high heat flux between headwind and tailwind. In order to accurately quantify the objective physical temperature rise caused by the robot taking one step forward in a complex flow field, this embodiment of the invention is based on a forced convection relative velocity model. The relative wind speed modulus is extracted by subtracting the heading velocity vector from the local wind speed vector, so as to construct a convective heat transfer coefficient prediction model with aerodynamic characteristics and a motor transient temperature increment prediction model.

[0037] Specifically, in this embodiment of the invention, the local wind speed vector is transformed to the robot's body coordinate system by calling the global positioning and environment matching coordinate transformation matrix configured in the robot's internal memory, and the heading speed vector is calculated by vector subtraction. With local wind speed vector The vector difference is used to extract the relative wind speed modulus, which is the modulus of the vector difference between the heading velocity vector and the transformed local wind speed vector. The feature length is read from the robot's non-volatile configuration table. and fluid kinematic viscosity Calculate the Reynolds number corresponding to the relative wind speed modulus. Introducing empirical constants convective heat transfer index and Prandtl number Nusselt numbers containing forced convection heat transfer characteristics are constructed using relative wind speed modulus. Nusel number The calculation equation is .

[0038] Furthermore, since the relationship between the Nusselt number and the surface heat transfer coefficient originates from the definition of the dimensionless criterion number in thermodynamics, its professional expression is: In this embodiment, the surface forced convection heat transfer coefficient is extracted by inversely deforming this relationship. Introducing the thermal conductivity measured in prior experiments Thermal conductivity, characteristic length, fluid kinematic viscosity, Prandtl number, and empirical constants are combined and packaged into a convective heat transfer shape factor. The surface forced convection heat transfer coefficient was obtained. Finally, the calculated surface forced convection heat transfer coefficient is linearly added to the foundation natural convection heat transfer coefficient under windless conditions to obtain the adaptive dynamic surface convection heat transfer coefficient covering all operating conditions. The calculation equation for the dynamic surface convection heat transfer coefficient is as follows:

[0039]

[0040] In the formula, The dynamic surface convective heat transfer coefficient is expressed in watts per square kelvin (W / m²). ); The basic natural convection heat transfer coefficient, with dimensions of watts per square kelvin (W / m²). ); The form factor for convective heat transfer is expressed in watts per square kelvin divided by the power of m in units of velocity. ); The modulo operator for vectors, dimensionless; The yaw velocity vector has the dimension of meters per second (m / s). ); This is a local wind speed vector, with dimensions in meters per second (m / s). ); The characteristic length is measured in meters (m). ); The viscosity of a fluid is expressed in square meters per second (m² / s). ); These are empirical constants, dimensionless. The convective heat transfer index is dimensionless. It is the Prandtl number, which is dimensionless; These are Nusselt numbers, dimensionless; The Reynolds number is dimensionless. Thermal conductivity, with dimensions in watts per meter Kelvin (W / m). ); The surface forced convection heat transfer coefficient is expressed in watts per square kelvin (W / m²). ).

[0041] The calculation formula extracts the true relative wind speed modulus by subtracting the heading velocity vector from the local wind speed vector using vector arithmetic. Since the degree of convective heat transfer depends on the Reynolds number, and the core of the Reynolds number is the relative velocity of the fluid passing over the surface of an object, when the robot travels against the wind, the relative wind speed modulus increases, leading to an increase in the Reynolds number and the surface forced convection heat transfer coefficient. This increase in the heat transfer coefficient characterizes the physical process of forced convection heat transfer generated by the hot wind environment on the drive motor. Conversely, when traveling with the wind, the relative wind speed modulus decreases and approaches zero, leading to a decrease in the surface forced convection heat transfer coefficient. The dynamic surface convection heat transfer coefficient degenerates into the basic natural convection heat transfer coefficient. This calculation logic realizes the numerical quantification of heat transfer asymmetry under headwind and tailwind conditions.

[0042] Furthermore, in this embodiment of the invention, the magnitude of the heading velocity vector is used as a reference. and the side length of the discrete grid Determine the time window spanning the robot within each discrete grid. Based on Joule's law, the real-time current of the motor stator is utilized. and stator resistance Constructing the intrinsic heating power of the drive motor Based on Newton's law of cooling, the dynamic surface convective heat transfer coefficient is utilized. Effective heat dissipation area Real-time ambient temperature and the body's real-time temperature Convection heat transfer power The effective heat dissipation area is the surface area of ​​the robot drive motor that actually participates in convective heat exchange with the ambient air. The cumulative net heat, including endogenous heat and external convective heat transfer, is obtained by integrating across time windows. Finally, the accumulated net heat is divided by the inherent physical heat capacity of the drive motor to obtain the transient temperature increment that includes aerodynamic characteristics. This is the transient temperature increment corresponding to the robot crossing a single discrete grid node. The calculation equation is as follows:

[0043]

[0044] In the formula, The transient temperature increment is expressed in Kelvin (K). ); The physical heat capacity is expressed in joules per kelvin (J / K). ); This is the real-time stator current of the motor, measured in amperes (A). ); The stator resistance has the dimension of ohms (Ω). ); The dynamic surface convective heat transfer coefficient is expressed in watts per square kelvin (W / m²). ); The effective heat dissipation area is measured in square meters (m²). ); The ambient real-time temperature is expressed in Kelvin (K). ); The real-time temperature of the body is expressed in Kelvin (K). ); Let be the side length of the discrete grid, with the dimension of meter (m). ); Let be the magnitude of the heading velocity vector, with dimensions in meters per second (m / s). ); To span a time window, the unit of measurement is seconds ( ); The power of endogenous heating is expressed in watts (W). ); The power of convective heat transfer is expressed in watts (W). ); To accumulate net heat, the dimension is joule (J). ).

[0045] The calculation formula incorporates the unidirectional asymmetric dynamic surface convective heat transfer coefficient into the framework of Newton's law of cooling. The calculated convective heat transfer power value characterizes the heat exchange rate from the environment to the motor input or stripping through convection. Combining the endogenous heating power and the time window crossing, the cumulative net heat accumulated by the robot when crossing a single discrete grid under specific flow field and thermal radiation temperature is derived through algebraic superposition. Under the premise that the physical heat capacity is constant, the increase in the cumulative net heat value leads to a proportional increase in the final transient temperature increment value, enabling the system to numerically predict the objective physical temperature rise caused by the robot crossing the discrete grid along a specific vector direction.

[0046] Step S003: Calculate the expected total temperature based on the transient temperature increment and the real-time temperature of the body; generate a dynamic resistance expansion ratio based on the expected total temperature and the rated maximum tolerance temperature through a nonlinear penalty function; and use the dynamic resistance expansion ratio to weight the physical distance the robot moves to construct a directed dynamic thermal impedance optimization cost function.

[0047] It should be noted that in a physical environment of strong thermal radiation and strong convection, once the real-time temperature of the body approaches the rated maximum withstand temperature, the probability of motor shutdown will increase non-linearly. In order to enable the path planning network to avoid high heat and windward heating surfaces through computational logic, this embodiment introduces the obstacle function theory in the field of constraint optimization, and transforms the remaining heat capacity safety margin into the directed resistance expansion weight in the graph search network, so as to reconstruct the directed dynamic thermal impedance optimization cost function.

[0048] Specifically, this embodiment calculates the real-time body temperature obtained from the industrial fieldbus system using addition. With transient temperature increment The sum of these values ​​is used to obtain the expected total temperature. The rated maximum withstand temperature is preset and stored in the firmware for the insulation class of the insulating material. Subtract the expected total temperature and calculate the remaining thermal capacity safety margin of the drive motor before it reaches the physical limit of thermal failure. A rational fractional barrier function structure is selected, with the expected total temperature as the numerator and the remaining heat capacity safety margin as the denominator, combined with the nonlinear expansion coefficient experimentally calibrated based on the thermistor characteristics of motor insulation materials. By performing multiplication operations, the dynamic drag expansion ratio under nonlinear constraint conditions is constructed. Furthermore, the dynamic drag expansion ratio corresponding to the robot crossing from the previous node to the current node is the ratio of the expected total temperature corresponding to the robot crossing from the previous node to the current node to the remaining heat capacity safety margin, multiplied by a preset nonlinear expansion coefficient. Here, the remaining heat capacity safety margin is the difference between the rated maximum withstand temperature and the expected total temperature when crossing from the previous node to the current node. The expected total temperature when crossing from the previous node to the current node is the result of adding the robot's real-time body temperature when it is located at the previous node and the transient temperature increment corresponding to the robot crossing from the previous node to the current node. The previous node is the previous adjacent node of the current node on the generated path or the node before the current node. The current node refers to the discrete grid position that the path planning algorithm (specifically the A* algorithm) is evaluating or that the robot is actually located in during path search or robot movement.

[0049] Then, the side lengths of the discrete grid are calculated using multiplication and superposition operations. The single-step physical cost represented by the calculated dynamic drag expansion ratio is combined with the calculated single-step thermal resistance cost, which is then reconstructed into a composite thermal resistance cost that includes physical distance and thermodynamic penalty. , This also represents the single-step cost of the combined thermal impedance when the robot moves from the previous node to the current node. Essentially, it's the dynamic drag expansion factor added by 1 when the robot moves from the previous node to the current node, and then... The product of these factors. Finally, based on the fundamental recursive law of the cost function in the A* optimization algorithm, the single-step cost of the composite thermal impedance is accumulated into the actual accumulated cost corresponding to the preceding node, thus deriving the final directed dynamic thermal impedance optimization cost function used for the planning solution. Its calculation equation is:

[0050]

[0051] In the formula, From the initial node to the current node The actual cumulative cost, i.e., the value of the directed dynamic thermal impedance optimization cost function, is in meters (m). ); For the initial node to reach the previous node The actual cumulative cost, in meters (m). ); The single-step physical cost is always equal to the side length of the discrete grid, with the dimension of meters (m). ); The coefficient of expansion is nonlinear and dimensionless. The real-time temperature of the body is expressed in Kelvin (K). ); The transient temperature increment is expressed in Kelvin (K). ); The rated maximum withstand temperature, with dimensions in Kelvin ( ); The expected total temperature, in Kelvin ( ) ); The remaining heat capacity safety margin is expressed in Kelvin (K). ); The dynamic resistance expansion ratio is dimensionless. The single-step cost of composite thermal resistance, in units of meters (m). The initial node refers to the starting point of the robot's path planning, that is, the discrete grid corresponding to the starting position where the robot begins to perform the inspection task.

[0052] The calculation formula constructs a causal penalty mechanism by combining the algebraic expression of the nonlinear expansion coefficient and the barrier function structure. When the expected total temperature rises to approach the rated maximum tolerance temperature, the remaining heat capacity safety margin in the denominator tends to zero, causing the dynamic drag expansion ratio to diverge towards positive infinity. Since the transient temperature increment calculated in the previous step is nested with unidirectional asymmetric relative wind speed modulus and fluid convection characteristic parameters, the directional dynamic thermal impedance optimization cost function is expressed as a directional weight distribution in the spatial coordinate network. When the heading velocity vector points to the high heat flow field, the single-step cost value of the composite thermal impedance increases nonlinearly, which mathematically reduces the priority of the windward path and guides the algorithm optimization network to automatically turn to search for a safe yaw path with a smaller actual cumulative cost value.

[0053] Step S004: Based on the directed dynamic thermal impedance optimization cost function, perform path search to generate an inspection path.

[0054] It should be noted that the global thermal radiation data and air velocity and direction data in the industrial manufacturing environment are not static constants. The start-up, shutdown or exhaust actions of mechanical equipment will cause a step change in the local fluid field. In order to ensure the safety of continuous inspection operations in a dynamic and ever-changing physical environment, this embodiment implements closed-loop feedback on the robot's motion control system based on the reconstructed directed dynamic thermal impedance optimization cost function.

[0055] Specifically, in this embodiment of the invention, a path planning algorithm based on a directed dynamic thermal impedance optimization cost function is configured inside the robot's main control unit. A global traversal search based on cost weights is performed to generate physical motion trajectories connecting discrete grids at the starting point and the ending point. As the robot moves along these physical motion trajectories, parameter acquisition and actual cumulative cost updates are performed cyclically within a preset control cycle. The actual cumulative cost of the current node n is compared in real-time with a pre-configured system safety optimization threshold. When a step change in the local wind speed vector causes an increase in the relative wind speed modulus, resulting in the actual cumulative cost of the current node n exceeding the system safety optimization threshold, the robot's main control unit issues a control command via the real-time control local area network bus to interrupt the current motion control task. At the current node, a new trajectory cost calculation is initiated towards the ending point. The updated real-time environmental temperature and local wind speed vector are used to re-converge and generate a safe yaw inspection path that avoids high-heat-flux discrete grids. The ending point refers to the final target position set by the robot's inspection task.

[0056] It should be further explained that, by constructing a data derivation link and using a forced convection relative velocity model as the calculation kernel, this embodiment of the invention couples aerodynamics and dynamic thermodynamic parameters, thus solving the defect of traditional optimization algorithms that only treat the spatial environment cost as an undirected scalar. This calculation logic uses the vector difference between the heading velocity vector and the local wind speed vector to extract the relative wind speed modulus, calculates the asymmetric difference in heat transfer, and converts the expected total temperature into a dynamic drag expansion ratio. This enables numerical prediction of the expected physical temperature rise under continuous dynamic conditions, and reduces the priority of entering high-risk convection areas at the planning and control level. By using mathematical correlation logic, it avoids the risk of downtime caused by heat accumulation, ensuring the operational safety of the robot during continuous operation in extreme thermal flow fields.

[0057] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A data-driven method for dynamic inspection path planning of robots, characterized in that, The method includes the following steps: The discrete grid environmental temperature field and discrete grid environmental wind speed vector field of the inspection area are obtained, as well as the robot's body operation status data. Each discrete grid node contains the real-time environmental temperature and local wind speed vector. The body operation status data includes at least the yaw speed vector of the robot's drive motor, the real-time body temperature, and the real-time stator current. The relative wind speed modulus is obtained by performing vector operations on the heading velocity vector and the local wind speed vector. A dynamic surface convection heat transfer coefficient is constructed based on the relative wind speed modulus. The transient temperature increment corresponding to the robot crossing a single discrete grid node is calculated using the dynamic surface convection heat transfer coefficient, the stator real-time current and the body real-time temperature. The expected total temperature is calculated based on the transient temperature increment and the real-time temperature of the body. Based on the expected total temperature and the rated maximum tolerance temperature, a dynamic resistance expansion ratio is generated through a nonlinear penalty function. The dynamic resistance expansion ratio is then used to weight the physical distance the robot moves, and a directed dynamic thermal impedance optimization cost function is constructed. Inspection paths are generated based on the directed dynamic thermal impedance optimization cost function.

2. The data-driven robot dynamic inspection path planning method as described in claim 1, characterized in that, Methods for obtaining relative wind speed modulus include: The global positioning and environment matching coordinate transformation matrix configured in the robot's internal memory is invoked to transform the local wind speed vector to the robot's body coordinate system. The vector difference between the heading velocity vector and the transformed local wind speed vector is calculated by vector subtraction, and the relative wind speed modulus is obtained by taking the modulus of the vector difference.

3. The data-driven robot dynamic inspection path planning method as described in claim 1, characterized in that, Methods for obtaining the dynamic surface convective heat transfer coefficient include: The characteristic length and fluid kinematic viscosity are read from the robot's non-volatile configuration table and calculated based on the relative wind speed modulus, characteristic length, and fluid kinematic viscosity. The Nusselt number is calculated based on the Reynolds number, empirical constant, convective heat transfer index, and Prandtl number. The surface forced convection heat transfer coefficient is calculated based on the Nusselt number, thermal conductivity, and characteristic length. The result of adding the surface forced convection heat transfer coefficient to the basic natural convection heat transfer coefficient is recorded as the dynamic surface convection heat transfer coefficient.

4. The data-driven robot dynamic inspection path planning method as described in claim 3, characterized in that, The Reynolds number expression is: The expression for the Nusselt number is as follows: The expression for the surface forced convection heat transfer coefficient is as follows: , For convective heat transfer form factor, The expression is , The heading velocity vector, Let be the local wind speed vector, D be the characteristic length, v be the fluid kinematic viscosity, C0 be the empirical constant, m be the convective heat transfer index, Pr be the Prandtl number, and k be the thermal conductivity.

5. The data-driven robot dynamic inspection path planning method as described in claim 1, characterized in that, Methods for obtaining the transient temperature increment corresponding to a robot traversing a single discrete grid node include: Based on the magnitude of the heading velocity vector and the side length of the discrete grid, the time window for the robot to cross a single discrete grid node is obtained; the intrinsic heating power of the robot drive motor is calculated based on the stator real-time current and stator resistance; the convective heat transfer power is calculated based on the dynamic surface convective heat transfer coefficient, effective heat dissipation area, real-time ambient temperature within the discrete grid node, and real-time body temperature; based on the crossing time window, the algebraic sum of the intrinsic heating power and the convective heat transfer power is integrated to obtain the cumulative net heat generated when crossing a single discrete grid node; the cumulative net heat is divided by the physical heat capacity of the drive motor to obtain the transient temperature increment corresponding to the robot crossing a single discrete grid node.

6. The data-driven robot dynamic inspection path planning method as described in claim 5, characterized in that, Crossing the Time Window The expression is The intrinsic heating power of the drive motor The expression is The convective heat transfer power The expression is The cumulative net heat The expression is , Let be the side length of the discrete grid. Let be the magnitude of the heading velocity vector. For stator real-time current, For stator resistance, The dynamic surface convective heat transfer coefficient, For effective heat dissipation area, Real-time ambient temperature within discrete grid nodes. This is the real-time temperature of the body.

7. The data-driven robot dynamic inspection path planning method as described in claim 1, characterized in that, The expected total temperature from the preceding node to the current node is the sum of the robot's real-time body temperature when it is at the preceding node and the transient temperature increment corresponding to the robot's transition from the preceding node to the current node; the preceding node is the node before the current node.

8. The data-driven robot dynamic inspection path planning method as described in claim 7, characterized in that, Methods for obtaining the dynamic drag expansion ratio include: The difference between the rated maximum withstand temperature and the expected total temperature when the robot crosses from the previous node to the current node is calculated to obtain the remaining heat capacity safety margin. The ratio of the expected total temperature when the robot crosses from the previous node to the current node to the remaining heat capacity safety margin is multiplied by a preset nonlinear expansion coefficient and recorded as the dynamic resistance expansion ratio corresponding to the robot crossing from the previous node to the current node.

9. The data-driven robot dynamic inspection path planning method as described in claim 8, characterized in that, The expression for the cost function of directed dynamic thermal impedance optimization is: ； in, Let n be the cost function for directed dynamic thermal impedance optimization, and also the actual cumulative cost from the starting point of the robot path planning to the current node. Let n represent the current node and n-1 represent the previous nodes. This represents the actual cumulative cost from the starting point to the preceding node. This represents the single-step cost of the combined thermal impedance when the robot traverses from the previous node to the current node. for , This represents the dynamic drag expansion ratio corresponding to the robot's transition from the previous node to the current node. is the side length of the discrete grid.

10. The data-driven robot dynamic inspection path planning method as described in claim 1, characterized in that, The discrete grid environment temperature field and discrete grid environment wind speed vector field of the inspection area are realized based on the fixed infrared thermal imager array and ultrasonic anemometer array deployed in the inspection area. The robot's body operation status data are synchronously collected through the robot's industrial fieldbus system at a preset sampling period.

Images