Humanoid robot control method and system

Abstract

The invention provides a humanoid robot control method and system, and relates to the technical field of robot control, and the method comprises the steps: obtaining a scanning image of a large storage tank; determining a plurality of first corrosion degradation areas and a corrosion degradation degree distribution diagram of each first corrosion degradation area based on the scanning image of the large storage tank; determining a plurality of first detection point locations of each first corrosion degradation area based on the plurality of first corrosion degradation area scanning images; determining a plurality of second detection point locations of each first corrosion degradation area based on the state detection data of each first detection point location; generating a second corrosion degradation degree distribution diagram of each first corrosion degradation area; determining a plurality of second corrosion degradation regions; and determining the high-risk corrosion and degradation area based on the state detection data of the plurality of first detection points of each second corrosion and degradation area and the state detection data of the plurality of second detection points of the second corrosion and degradation area. According to the method, the high-risk corrosion and degradation area of the large storage tank can be rapidly and accurately determined.

Description

Technical Field

[0001] The present invention relates to the technical field of robot control, and in particular to a humanoid robot control method and system. Background Art

[0002] As a core infrastructure in industries such as petrochemical and energy storage, large storage tanks store flammable, explosive or toxic media for a long time. Their structural integrity is directly related to production safety and environmental safety. Corrosion of storage tanks may cause major safety accidents such as fires and explosions. Timely and accurately detecting the corrosion state of storage tanks and identifying high-risk corrosion degradation areas is crucial for ensuring industrial production safety. Traditional storage tank corrosion detection mainly relies on means such as manual visual inspection and sampling inspection with ultrasonic thickness gauges. Manual detection requires point-by-point measurement, which takes too long to complete a comprehensive inspection of large storage tanks. Limited by manpower and time, usually only some points can be sampled for inspection, making it difficult to detect potential local corrosion. Moreover, when manually observing and detecting, it is easily affected by environmental factors such as light and subjective factors such as the experience of inspectors, resulting in a high missed detection rate and inaccurate detection results.

[0003] Therefore, how to quickly and accurately determine the high-risk corrosion degradation areas of large storage tanks is an urgent problem to be solved at present. Summary of the Invention

[0004] The main technical problem to be solved by the present invention is how to quickly and accurately determine the high-risk corrosion degradation areas of large storage tanks.

[0005] According to a first aspect, the present invention provides a humanoid robot control method, including: obtaining a scanned image of a large storage tank; determining a plurality of first corrosion degradation areas and a corrosion degradation degree distribution map of each first corrosion degradation area based on the scanned image of the large storage tank; determining a plurality of first detection points of each first corrosion degradation area based on the scanned images of the plurality of first corrosion degradation areas; controlling the humanoid robot to perform detections at the plurality of first detection points of each first corrosion degradation area and obtaining status detection data of each first detection point; determining a plurality of second detection points of each first corrosion degradation area based on the status detection data of each first detection point; controlling the humanoid robot to the plurality of second detection points of each first corrosion degradation area and obtaining status detection data of each second detection point; generating a second corrosion degradation degree distribution map of each first corrosion degradation area based on the status detection data of each first detection point and the status detection data of each second detection point of each first corrosion degradation area; determining a plurality of second corrosion degradation areas based on the first corrosion degradation degree distribution map of each first corrosion degradation area and the second corrosion degradation degree distribution map of each first corrosion degradation area; determining high-risk corrosion degradation areas based on the status detection data of the plurality of first detection points and the status detection data of the plurality of second detection points of each second corrosion degradation area.

[0006] In a possible implementation, the state detection data includes structural state detection data and corrosion feature detection data.

[0007] In a possible implementation, determining a high-risk corrosion deterioration area based on the state detection data of multiple first detection points in each second corrosion deterioration area and the state detection data of multiple second detection points in the second corrosion deterioration area includes: constructing a storage tank map, where the storage tank map includes multiple state detection nodes and edges between the multiple state detection nodes. The feature of each state detection node is the position of the state detection and the state detection data, and the features of the edges are the spatial orientation relationship, distance, and similarity of the state detection data between the state detection nodes; processing the storage tank map based on a graph convolutional network to determine multiple verification points in each second corrosion deterioration area; and determining the high-risk corrosion deterioration area based on the detection and verification results of the multiple verification points in each second corrosion deterioration area.

[0008] In a possible implementation, the input of the graph convolutional network is the storage tank map, and the output of the graph convolutional network is multiple verification points in each second corrosion deterioration area.

[0009] According to a second aspect, the present invention provides a humanoid robot control system, including: an acquisition module for acquiring a scanned image of a large storage tank; a deterioration analysis module for determining multiple first corrosion deterioration areas and a corrosion deterioration degree distribution map of each first corrosion deterioration area based on the scanned image of the large storage tank; a point position planning module for determining multiple first detection points in each first corrosion deterioration area based on the scanned images of the multiple first corrosion deterioration areas; a first detection module for controlling the humanoid robot to perform detections at the multiple first detection points in each first corrosion deterioration area and acquiring the state detection data of each first detection point; a second planning module for determining multiple second detection points in each first corrosion deterioration area based on the state detection data of each first detection point; a second detection module for controlling the humanoid robot to the multiple second detection points in each first corrosion deterioration area and acquiring the state detection data of each second detection point; a generation module for generating a second corrosion deterioration degree distribution map of each first corrosion deterioration area based on the state detection data of each first detection point and the state detection data of each second detection point in each first corrosion deterioration area; a region determination module for determining multiple second corrosion deterioration areas based on the first corrosion deterioration degree distribution map of each first corrosion deterioration area and the second corrosion deterioration degree distribution map of each first corrosion deterioration area; and a high-risk identification module for determining a high-risk corrosion deterioration area based on the state detection data of multiple first detection points in each second corrosion deterioration area and the state detection data of multiple second detection points in the second corrosion deterioration area.

[0010] In a possible implementation, the state detection data includes structural state detection data and corrosion feature detection data.

[0011] In a possible implementation, the high-risk identification module is further configured to: construct a storage tank graph, where the storage tank graph includes multiple state detection nodes and edges between the multiple state detection nodes. The feature of each state detection node is the position of state detection and the state detection data, and the feature of the edge is the spatial orientation relationship, distance, and similarity of the state detection data between the state detection nodes; process the storage tank graph based on a graph convolutional network to determine multiple verification points for each second corrosion deterioration area; determine the high-risk corrosion deterioration area based on the detection and verification results of the multiple verification points in each second corrosion deterioration area.

[0012] In a possible implementation, the input of the graph convolutional network is the storage tank graph, and the output of the graph convolutional network is multiple verification points for each second corrosion deterioration area.

[0013] A humanoid robot control method and system provided by the present invention. The method includes obtaining a scanned image of a large storage tank; determining multiple first corrosion deterioration areas and the corrosion deterioration degree distribution map of each first corrosion deterioration area based on the scanned image of the large storage tank; determining multiple first detection points for each first corrosion deterioration area based on the scanned images of the multiple first corrosion deterioration areas; controlling the humanoid robot to perform detections at the multiple first detection points of each first corrosion deterioration area and obtaining the state detection data of each first detection point; determining multiple second detection points for each first corrosion deterioration area based on the state detection data of each first detection point; controlling the humanoid robot to the multiple second detection points of each first corrosion deterioration area and obtaining the state detection data of each second detection point; generating a second corrosion deterioration degree distribution map of each first corrosion deterioration area based on the state detection data of each first detection point and the state detection data of each second detection point in each first corrosion deterioration area; determining multiple second corrosion deterioration areas based on the first corrosion deterioration degree distribution map of each first corrosion deterioration area and the second corrosion deterioration degree distribution map of each first corrosion deterioration area; determining the high-risk corrosion deterioration area based on the state detection data of the multiple first detection points and the state detection data of the multiple second detection points in each second corrosion deterioration area. This method can quickly and accurately determine the high-risk corrosion deterioration area of a large storage tank. Description of the Drawings

[0014] Figure 1 It is a schematic flowchart of a humanoid robot control method provided by an embodiment of the present invention;

[0015] Figure 2 It is a schematic diagram of a large storage tank provided by an embodiment of the present invention;

[0016] Figure 3 Schematic diagram of a humanoid robot provided by an embodiment of the present invention;

[0017] Figure 4 Schematic diagram of a process for determining a high-risk corrosion deterioration area provided by an embodiment of the present invention;

[0018] Figure 5 Schematic diagram of constructing a storage tank atlas provided by an embodiment of the present invention;

[0019] Figure 6 Schematic diagram of a control system of a humanoid robot provided by an embodiment of the present invention. Detailed implementation manners

[0020] In an embodiment of the present invention, there is provided a humanoid robot control method as shown in Figure 1 the following, and the humanoid robot control method includes steps S1 to S9:

[0021] Step S1, obtaining a scanned image of a large storage tank.

[0022] A large storage tank is a large industrial container that can be used to store liquid or gas raw materials. Figure 2 Schematic diagram of a large storage tank provided by an embodiment of the present invention. The large storage tank is made of metal materials and has the characteristics of strong sealing performance and large capacity. The large storage tank can be applied to fields such as petrochemical industry, energy storage, and food processing.

[0023] The scanned image of the large storage tank is an image that can reflect the surface morphological characteristics of the large storage tank obtained after scanning the surface and internal structure of the large storage tank by a high-resolution industrial camera.

[0024] Step S2, determining a plurality of first corrosion deterioration areas and a corrosion deterioration degree distribution map of each first corrosion deterioration area based on the scanned image of the large storage tank.

[0025] In some embodiments, a first deterioration analysis model can be used to determine a plurality of first corrosion deterioration areas and a corrosion deterioration degree distribution map of each first corrosion deterioration area based on the scanned image of the large storage tank. The first deterioration analysis model is a convolutional neural network model. The input of the first deterioration analysis model is the scanned image of the large storage tank, and the output of the first deterioration analysis model is a plurality of first corrosion deterioration areas and a corrosion deterioration degree distribution map of each first corrosion deterioration area.

[0026] The convolutional neural network model includes a convolutional neural network (CNN), which is a deep learning model that can process grid-structured data. The convolutional neural network can automatically extract local features in an image through multiple layers of convolutional operations and then combine them layer by layer to form a high-level semantic representation. The structure of the convolutional neural network includes a convolutional layer, a pooling layer, and a fully connected layer. The convolutional neural network can effectively capture the spatial hierarchical features in an image and can handle recognition tasks with obvious visual patterns.

[0027] The first corrosion deterioration area is the potential corrosion area on a large storage tank output by the first deterioration analysis model. The first corrosion deterioration area shows abnormal texture, color, or geometric features in the scanned image of the large storage tank. The first corrosion deterioration area can indicate that corrosion damage may occur and exist in the material to varying degrees.

[0028] The corrosion deterioration degree distribution map of the first corrosion deterioration area is a visual representation map of the corrosion characteristics of the first corrosion deterioration area generated by extracting features from the scanned image of the large storage tank output by the first deterioration analysis model. The deterioration degree distribution map is presented in the form of a pseudo-color image. Through the deterioration degree distribution map, the corrosion deterioration feature intensity and the evaluation confidence at each position can be represented.

[0029] In the corrosion deterioration degree distribution map of the first corrosion deterioration area, each pixel point contains two precisely quantified optical features: the hue value (H value) of the color channel and the lightness value (L value) of the lightness channel. Among them, the hue value (H value) of the pixel point can represent the corrosion deterioration feature intensity, and the hue range of 0 - 360 degrees corresponds to the corrosion deterioration feature intensity of 0 - 1. The lightness value (L value) of the pixel point can represent the evaluation confidence, and the lightness range of 0% - 100% corresponds to the confidence of 0 - 1.

[0030] The scanned image of the large storage tank records specific visual feature changes caused by metal corrosion on the surface of the large storage tank. When the metal of the large storage tank corrodes, unique texture patterns, color anomalies, or geometric deformations will form on the surface, such as the brownish-red change of rust spots, the irregular edges of coating peeling, or the local shadow differences caused by pits. These visual features are manifested as changes in pixel-level brightness, chromaticity, and texture features in the image. The convolutional neural network can cluster continuous pixel regions with similar corrosion features into independent first corrosion deterioration areas through an image segmentation algorithm, and each area corresponds to an image block that is spatially continuous and has similar corrosion features. During the training process, the convolutional neural network can learn the feature differences between the corrosion area and the non-corrosion area, so that it can perform pixel-level analysis on the scanned image of a new large storage tank in the inference stage and output the corrosion deterioration degree distribution map of each first corrosion deterioration area.

[0031] Step S3: Determine multiple first detection points for each first corrosion deterioration area based on multiple scanned images of the first corrosion deterioration areas.

[0032] The scanned image of the first corrosion deterioration area is the scanned image corresponding to the part of the first corrosion deterioration area extracted from the scanned image of the large storage tank.

[0033] In some embodiments, multiple first detection points for each first corrosion deterioration area can be determined based on multiple scanned images of the first corrosion deterioration areas using a first point determination model. The first point determination model is a Transformer model. The input of the first point determination model is multiple scanned images of the first corrosion deterioration areas, and the output of the first point determination model is multiple first detection points for each first corrosion deterioration area.

[0034] The Transformer model is a deep learning model based on the self-attention mechanism. The Transformer model includes an encoder and a decoder. The encoder can perform representation learning on the input sequence, which includes the self-attention mechanism and a feed-forward neural network. Based on the encoder, the decoder additionally introduces the multi-head attention mechanism (Multi-HeadAttention). The decoder can decode the output of the encoder and generate the target sequence.

[0035] Multiple first detection points for each first corrosion deterioration area are a set of key detection positions planned for each first corrosion deterioration area output by the first point determination model.

[0036] Visual features generated by the corrosion of the inner and outer metal surfaces of the large storage tank, such as texture anomalies, color changes, or geometric deformations, can be extracted from the scanned image of the first corrosion deterioration area. These visual features have an identifiable spatial distribution pattern. By analyzing the distribution law of these features, the Transformer model can determine which areas require denser detection coverage and which positions may represent the key transition zones of corrosion development. For example, the edges of the corrosion area, the junctions of different texture features, and the core areas where abnormal features gather may all be selected as the first detection points.

[0037] The self-attention mechanism of the Transformer model can consider both the local performance and the global distribution of the corrosion features simultaneously. The Transformer model can evaluate the degree of association between each area in multiple scanned images of the first corrosion deterioration areas and the surrounding areas, and preferentially select those positions that can represent both the typicality of the corrosion features and the comprehensiveness of the detection range as the first detection points. Through this method of selecting detection points based on global feature association, the detection can be made efficient and targeted.

[0038] In some embodiments, the first point determination model includes a visual extraction layer, an anomaly assessment layer, and a point analysis layer. The input of the visual extraction layer is multiple scanned images of the first corrosion deterioration areas, and the output of the visual extraction layer is the abnormal positions of the corrosion area textures, the color change distribution sequences, and the geometric deformation areas. The input of the anomaly assessment layer is the abnormal positions of the corrosion area textures, the color change distribution sequences, and the geometric deformation areas, and the output of the anomaly assessment layer is multiple abnormal areas, the corrosion risk scores of each abnormal area, the feature correlation degree index between the abnormal areas, and the importance ranking of the abnormal areas. The input of the point analysis layer is multiple abnormal areas, the corrosion risk scores of each abnormal area, the feature correlation degree index between the abnormal areas, and the importance ranking of the abnormal areas, and the output of the point analysis layer is multiple first detection points for each first corrosion deterioration area.

[0039] Different layers are responsible for information abstraction and processing at different levels. For example, the visual extraction layer is responsible for extracting corrosion-related visual features from the scanned images of the first corrosion deterioration areas, the anomaly assessment layer is responsible for evaluating the risk level and correlation relationship of each abnormal area based on these features, and the point analysis layer is responsible for determining specific multiple first detection points according to the evaluation results. Through such hierarchical processing, complex image information can be processed more effectively. Constructing multiple layers can improve the modularity of the system and can process information more effectively, thereby improving the accuracy and efficiency of the system.

[0040] Step S4: Control the humanoid robot to perform detections at multiple first detection points in each first corrosion deterioration area, and obtain the status detection data of each first detection point.

[0041] A humanoid robot is a mobile robot with a humanoid form. Figure 3 It is a schematic diagram of a humanoid robot provided by an embodiment of the present invention. The humanoid robot is equipped with a multi-degree-of-freedom robotic arm, high-precision sensors, and an autonomous navigation system. The humanoid robot can perform close-range detection tasks in complex and high-risk industrial environments.

[0042] The status detection data includes structural status detection data and corrosion feature detection data.

[0043] The structural status detection data is the data obtained by detecting the changes in the electrical conductivity and magnetic permeability of the metal surface of the storage tank with an eddy current detector, and the structural status detection data can be used to detect corrosion or structural defects.

[0044] The corrosion feature detection data is the corrosion product composition data obtained by scanning each first detection point with the X-ray fluorescence spectrometer equipped on the humanoid robot.

[0045] As an example, the corrosion feature detection data can be obtained by emitting X-rays to the corrosion area through an X-ray fluorescence spectrometer and then analyzing the energy spectral lines of the characteristic X-rays detected.

[0046] The corrosion feature detection data includes the percentage content of elements such as sulfur and chlorine in the corrosion products.

[0047] Step S5: Based on the status detection data of each of the first detection points, determine multiple second detection points for each of the first corrosion deterioration areas.

[0048] In some embodiments, based on the status detection data of each of the first detection points, a second point determination model can be used to determine multiple second detection points for each of the first corrosion deterioration areas. The second point determination model is a deep neural network model. The input of the second point determination model is the status detection data of each of the first detection points, and the output of the second point determination model is multiple second detection points for each of the first corrosion deterioration areas.

[0049] The deep neural network model includes a deep neural network (DNN). A deep neural network is a machine learning model composed of multiple hidden layers. A deep neural network can automatically learn complex features and patterns in the input data through hierarchical processing. The deep neural network model can be trained through forward propagation and backpropagation algorithms and can establish a non-linear mapping relationship between the input data and the output results.

[0050] The multiple second detection points for each of the first corrosion deterioration areas are multiple supplementary detection positions calculated through the second point determination model based on the status detection data of each of the first detection points.

[0051] The second detection points are more targeted. The second detection points can be distributed in the surrounding areas of each of the first detection points where the status detection data of the first detection points is abnormal, and in the transition areas where the status detection data of the first detection points undergoes sudden changes.

[0052] The status detection data of the first detection points includes structural status detection data and corrosion feature detection data, and these data can reflect the actual situation of corrosion development. For example, when the status detection data of a certain first detection point is abnormal, it indicates that there may be more serious corrosion problems in this area, and then it is necessary to add second detection points in the surrounding positions for verification. If there are significant differences between the status detection data of the current first detection point and the adjacent first detection points, it indicates that this transition area also needs to add second detection points for key detection.

[0053] By analyzing the spatial correlation and numerical variation law among the state detection data of each first detection point, the deep neural network can automatically identify the areas that need to be focused on for detection. The hidden layer of the deep neural network can extract the non-linear features in the state detection data, and the output layer can calculate the optimal distribution schemes of multiple second detection points for each first corrosion degradation area based on these features, so as to ensure that the second detection points can accurately cover all the risk areas that need to be verified.

[0054] Step S6: Control the humanoid robot to the multiple second detection points in each first corrosion degradation area, and obtain the state detection data of each second detection point.

[0055] When the multiple second detection points in each first corrosion degradation area are determined, control the humanoid robot to the multiple second detection points in each first corrosion degradation area, and detect each second detection point to obtain the state detection data of each second detection point. The state detection data includes structural state detection data and corrosion feature detection data.

[0056] Step S7: Generate the second corrosion degradation degree distribution map of each first corrosion degradation area based on the state detection data of each first detection point and the state detection data of each second detection point in each first corrosion degradation area.

[0057] In some embodiments, the second corrosion degradation degree distribution map of each first corrosion degradation area can be generated by a generative adversarial network based on the state detection data of each first detection point and the state detection data of each second detection point in each first corrosion degradation area. The input of the generative adversarial network is the state detection data of each first detection point and the state detection data of each second detection point in each first corrosion degradation area, and the output of the generative adversarial network is the second corrosion degradation degree distribution map of each first corrosion degradation area.

[0058] A generative adversarial network is a deep learning framework composed of a generator network and a discriminator network. The generator network can learn to map the input data to the target distribution, and the discriminator network can distinguish between the generated data and the real data. The two are continuously optimized through adversarial training, and finally the generator can output new samples that conform to the real data distribution.

[0059] The second corrosion degradation degree distribution map of the first corrosion degradation area is a more accurate visualization representation map of the corrosion characteristics for the first corrosion degradation area generated by the generative adversarial network. The second corrosion degradation degree distribution map is presented in the form of a pseudo-color image, and the second corrosion degradation degree distribution map can represent the corrosion degradation feature intensity and the evaluation confidence at each position.

[0060] In the second corrosion degradation degree distribution map of the first corrosion degradation area, each pixel contains two precisely quantified optical features: the hue value (H value) of the color channel and the lightness value (L value) of the lightness channel. Among them, the hue value (H value) of the pixel can represent the intensity of the corrosion degradation feature, and the hue range of 0 - 360 degrees corresponds to the corrosion degradation feature intensity of 0 - 1. The lightness value (L value) of the pixel can represent the evaluation confidence, and the lightness range of 0% - 100% corresponds to the confidence of 0 - 1.

[0061] The status detection data of each first detection point and each second detection point in each first corrosion degradation area together constitute the spatial sampling point set of the corrosion characteristics. These data include the measured values of the states at each discrete position on the inner and outer surfaces of the large storage tank and the spatial correlation between each point, such as the trend of corrosion spreading along the weld. By analyzing the spatial autocorrelation characteristics and numerical gradient changes of these data points through a model, the distribution law of the corrosion characteristics in the continuous space can be established.

[0062] The generator of the generative adversarial network can automatically learn the spatial interpolation rules between the detection points. The generator can convert the status detection data of the discrete detection points into a distribution map of continuous features. The discriminator can compare the generated distribution map with the statistical characteristics of real corrosion cases to ensure that the generated distribution map conforms to the physical laws. In addition, the gradient penalty mechanism in the adversarial training process can ensure the smooth transition of the generated results in space, thus avoiding the emergence of mutation regions that do not conform to the actual corrosion expansion law. In this way, the generative adversarial network can reconstruct the second corrosion degradation degree distribution map that not only conforms to the measured value constraints but also has physical rationality from the status detection data of limited points.

[0063] Step S8, determining a plurality of second corrosion degradation areas based on the first corrosion degradation degree distribution map of each first corrosion degradation area and the second corrosion degradation degree distribution map of each first corrosion degradation area.

[0064] In some embodiments, a second degradation determination model can be used to determine a plurality of second corrosion degradation areas based on the first corrosion degradation degree distribution map of each first corrosion degradation area and the second corrosion degradation degree distribution map of each first corrosion degradation area. The second degradation determination model is a convolutional neural network model. The input of the second degradation determination model is the first corrosion degradation degree distribution map of each first corrosion degradation area and the second corrosion degradation degree distribution map of each first corrosion degradation area, and the output of the second degradation determination model is the second corrosion degradation area.

[0065] The second corrosion degradation area is an area with high corrosion degradation feature intensity and high confidence selected by fusing and analyzing the first corrosion degradation degree distribution map and the second corrosion degradation degree distribution map through the second degradation determination model on the basis of each first corrosion degradation area.

[0066] The first corrosion degradation degree distribution map and the second corrosion degradation degree distribution map of each first corrosion degradation area together constitute a dual evidence system for corrosion risk assessment. Both the first corrosion degradation degree distribution map and the second corrosion degradation degree distribution map can present the corrosion characteristic intensity and confidence at each position in the first corrosion degradation area, but the data sources and analysis dimensions of the two can form a complementary relationship. The generation of the first corrosion degradation degree distribution map is based on image visual feature recognition. The first corrosion degradation degree distribution map can provide the corrosion probability distribution from a macroscopic level, while the generation of the second corrosion degradation degree distribution map is based on the measured data of the detection points and can verify and correct the corrosion characteristic intensity and confidence from a microscopic level.

[0067] The convolutional neural network can effectively fuse the spatial information of the first corrosion degradation degree distribution map and the second corrosion degradation degree distribution map of the first corrosion degradation area through its unique hierarchical feature extraction ability. The convolutional neural network can synchronously process the data of the two types of distribution maps through multi-scale convolutional kernels and automatically learn the spatial association rules between the two. Shallow convolution can identify the consistency of local features, such as the spatial overlap between the abnormal image area and the measured high-value area. Deep convolution can establish the mapping relationship of global features, such as the matching between the corrosion propagation direction and the gradient of the measured data. Through this feature learning process, the convolutional neural network can accurately identify the areas that simultaneously meet the significance of image features and the reliability of measured data, and thus output the determination result of the second corrosion degradation area that meets the standard.

[0068] Step S9, determine the high-risk corrosion degradation area based on the status detection data of multiple first detection points in each second corrosion degradation area and the status detection data of multiple second detection points in the second corrosion degradation area.

[0069] In some embodiments, it can be through Figure 4 the following process to determine the high-risk corrosion degradation area. Figure 4 FIG. is a schematic flow chart of a process for determining a high-risk corrosion degradation area provided by an embodiment of the present invention. The determination of the high-risk corrosion degradation area includes steps S41 to S43:

[0070] Step S41, construct a storage tank map, where the storage tank map includes multiple status detection nodes and the edges between the multiple status detection nodes. The feature of each status detection node is the position of the status detection and the status detection data, and the feature of the edge is the spatial orientation relationship, distance, and similarity of the status detection data between the status detection nodes.

[0071] The storage tank atlas is a data structure in the form of a networked model that can represent the inspection data of large storage tanks. The storage tank atlas consists of state inspection nodes and the edges between the state inspection nodes. The state inspection nodes correspond to the specific inspection positions on the inner and outer surfaces of the storage tank, and each node records the spatial coordinate information and state inspection data at that point; the edges representing the connection relationships between the nodes can describe the spatial correlation and data correlation between different inspection points, and the features of the edges include the spatial orientation relationship, physical distance between the state inspection nodes, and the similarity of the state inspection data. Figure 5 This is a schematic diagram of constructing a storage tank atlas provided by an embodiment of the present invention. As Figure 5 shown, Figure 5 It includes state inspection node A, state inspection node B, state inspection node C, and state inspection node D. The features of the edges between two state inspection nodes are the spatial orientation relationship, distance, and similarity of the state inspection data between every two state inspection nodes.

[0072] The inspection positions of the state inspections of multiple state inspection nodes include multiple first inspection points and multiple second inspection points in multiple second corrosion deterioration areas.

[0073] In some embodiments, the similarity between the state inspection data in each inspection node can be determined through a deep neural network model.

[0074] By constructing this networked model of the storage tank atlas, discrete inspection data can be organized into an interconnected network system, thereby realizing the quantitative expression of the spatial distribution law of the corrosion characteristics and structural state of the storage tank.

[0075] Step S42: Process the storage tank atlas based on the graph convolutional network to determine multiple verification points for each second corrosion deterioration area.

[0076] The graph convolutional network (GCN) is a deep learning model that can be used to process networked data. The graph convolutional network can extract node features through a local neighborhood aggregation mechanism. The graph convolutional network is a neural network that can directly act on atlas data. The input of the graph convolutional network is the storage tank atlas, and the output of the graph convolutional network is multiple verification points for each second corrosion deterioration area.

[0077] The multiple verification points for each second corrosion deterioration area are a set of decisive inspection position points determined through the analysis of the graph convolutional network for the final determination of high-risk corrosion deterioration areas.

[0078] The tank map represents the state detection nodes and the relationships between them as a graph structure. The tank map integrates various information (such as the location of state detection and state detection data) into one structure. The characteristics of the edge include the spatial orientation relationship, distance and similarity of the state detection data between the detection point nodes. The edge can represent the spatial position relationship and data association degree between the detection points. This information can help the graph convolutional network better understand the propagation path and distribution pattern of corrosion characteristics on the surface of the tank. The tank map can integrate the feature information of multiple nodes, including the location of state detection and state detection data. By integrating these multi-source information into the tank map, more comprehensive input data can be provided, so that the graph convolutional network can more accurately learn and determine multiple verification points of each second corrosion degradation zone.

[0079] When the graph convolutional network is processing the tank graph, the spatial correlation features of each state detection node and its surrounding nodes can be extracted through the neighborhood aggregation operation, and then the multi-layer convolution kernel can be used to learn the propagation law of corrosion features in space. Finally, the high-risk areas that need to be verified can be identified through the node classification task.

[0080] Step S43: determining a high-risk corrosion degradation area based on the detection and verification results of the multiple verification points of each second corrosion degradation area.

[0081] The verification results of the verification points are obtained by using a humanoid robot equipped with an ultrasonic phased array thickness gauge and a laser confocal microscope to perform two inspection tasks: metal wall thickness measurement and surface morphology scanning on the verification points in the second corrosion degradation zone. After data filtering, noise reduction, feature extraction and evaluation, a standardized data set including the actual wall thickness measured value, pitting characteristic parameters and residual strength coefficient is formed.

[0082] In some embodiments, a high-risk corrosion degradation zone can be determined based on a deep neural network model based on the detection and verification results of multiple verification points of each second corrosion degradation zone, the input of the deep neural network model is the detection and verification results of multiple verification points of each second corrosion degradation zone, and the output of the deep neural network model is the high-risk corrosion degradation zone.

[0083] High-risk corrosion degradation areas are areas that are severely corroded and pose a direct threat to the safety of large tank structures, as determined by deep neural network assessment. High-risk corrosion degradation areas are characterized by metal remaining thickness significantly lower than the design standard, the presence of penetrating defects or cracks, and corrosion rates exceeding the safety threshold. Immediate repair or replacement measures are required to avoid safety accidents such as tank leakage and structural failure.

[0084] The detection and verification results of the verification points include multi-dimensional data such as the remaining metal thickness, pitting characteristic parameters, and remaining strength coefficients. The deep neural network can perform non-linear transformation and feature extraction on these data through multiple layers of neurons, and automatically mine the complex correlation relationships and potential rules among the detection and verification result data. The deep neural network can learn the corrosion risk level patterns corresponding to different data combinations, so as to determine the high-risk corrosion deterioration areas based on the detection and verification results of multiple verification points.

[0085] Based on the same inventive concept, Figure 6 The present invention provides a schematic diagram of a humanoid robot control system according to an embodiment of the present invention. The humanoid robot control system includes:

[0086] An acquisition module 61, configured to acquire a scanned image of a large storage tank;

[0087] A deterioration analysis module 62, configured to determine a plurality of first corrosion deterioration areas and a corrosion deterioration degree distribution map of each first corrosion deterioration area based on the scanned image of the large storage tank;

[0088] A point position planning module 63, configured to determine a plurality of first detection point positions of each first corrosion deterioration area based on the scanned images of the plurality of first corrosion deterioration areas;

[0089] A first detection module 64, configured to control the humanoid robot to perform detections at the plurality of first detection point positions of each first corrosion deterioration area, and acquire the status detection data of each first detection point position;

[0090] A second planning module 65, configured to determine a plurality of second detection point positions of each first corrosion deterioration area based on the status detection data of each first detection point position;

[0091] A second detection module 66, configured to control the humanoid robot to the plurality of second detection point positions of each first corrosion deterioration area, and acquire the status detection data of each second detection point position;

[0092] A generation module 67, configured to generate a second corrosion deterioration degree distribution map of each first corrosion deterioration area based on the status detection data of each first detection point position and the status detection data of each second detection point position of each first corrosion deterioration area;

[0093] A region determination module 68, configured to determine a plurality of second corrosion deterioration areas based on the first corrosion deterioration degree distribution map of each first corrosion deterioration area and the second corrosion deterioration degree distribution map of each first corrosion deterioration area;

[0094] A high-risk identification module 69, configured to determine a high-risk corrosion deterioration area based on the status detection data of the plurality of first detection point positions and the status detection data of the plurality of second detection point positions of each second corrosion deterioration area.

[0095] Similarly, it should be noted that, in order to simplify the description disclosed in this specification and thus assist in the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of this specification, sometimes multiple features are merged into one embodiment, drawing, or description thereof. However, this method of disclosure does not mean that the features required by the subject matter of this specification are more than those mentioned in the claims. In fact, the features of the embodiments are fewer than all the features of the single embodiments disclosed above.

[0096] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be regarded as consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to those explicitly introduced and described in this specification.

Claims

1. A humanoid robot control method, characterized in that, Including: Obtaining a scanned image of a large storage tank; Determining a plurality of first corrosion deterioration areas and a corrosion deterioration degree distribution map of each first corrosion deterioration area based on the scanned image of the large storage tank; Determining a plurality of first detection points of each first corrosion deterioration area based on the scanned images of the plurality of first corrosion deterioration areas; Controlling a humanoid robot to perform detections at the plurality of first detection points of each first corrosion deterioration area and obtaining status detection data of each first detection point; Determining a plurality of second detection points of each first corrosion deterioration area based on the status detection data of each first detection point; Controlling the humanoid robot to the plurality of second detection points of each first corrosion deterioration area and obtaining status detection data of each second detection point; Generating a second corrosion deterioration degree distribution map of each first corrosion deterioration area based on the status detection data of each first detection point and the status detection data of each second detection point of each first corrosion deterioration area; Determining a plurality of second corrosion deterioration areas based on the first corrosion deterioration degree distribution map of each first corrosion deterioration area and the second corrosion deterioration degree distribution map of each first corrosion deterioration area; Determining a high-risk corrosion deterioration area based on the status detection data of the plurality of first detection points of each second corrosion deterioration area and the status detection data of the plurality of second detection points of the second corrosion deterioration area.

2. The humanoid robot control method according to claim 1, characterized in that, The status detection data includes structural status detection data and corrosion feature detection data.

3. The humanoid robot control method according to claim 1, characterized in that The determining a high-risk corrosion deterioration area based on the status detection data of the plurality of first detection points of each second corrosion deterioration area and the status detection data of the plurality of second detection points of the second corrosion deterioration area includes: Constructing a storage tank map, where the storage tank map includes a plurality of status detection nodes and edges between the plurality of status detection nodes. The feature of each status detection node is the position of the status detection and the status detection data, and the feature of the edge is the spatial orientation relationship, distance, and similarity of the status detection data between the status detection nodes; Processing the storage tank map based on a graph convolutional network to determine a plurality of verification points of each second corrosion deterioration area; Determining a high-risk corrosion deterioration area based on the detection and verification results of the plurality of verification points of each second corrosion deterioration area.

4. The humanoid robot control method according to claim 2, wherein The input of the graph convolutional network is the storage tank map, and the output of the graph convolutional network is a plurality of verification points of each second corrosion deterioration area.

5. A humanoid robot control system, characterized in that, Including: An acquisition module for obtaining a scanned image of a large storage tank; A deterioration analysis module for determining a plurality of first corrosion deterioration areas and a corrosion deterioration degree distribution map of each first corrosion deterioration area based on the scanned image of the large storage tank; A point position planning module for determining a plurality of first detection points of each first corrosion deterioration area based on the scanned images of the plurality of first corrosion deterioration areas; A first detection module for controlling a humanoid robot to perform detections at the plurality of first detection points of each first corrosion deterioration area and obtaining status detection data of each first detection point; A second planning module for determining a plurality of second detection points of each first corrosion deterioration area based on the status detection data of each first detection point; The second detection module is used to control the humanoid robot to multiple second detection points of each of the first corrosion deterioration areas, and obtain the status detection data of each second detection point; The generation module is used to generate a second corrosion deterioration degree distribution map of each first corrosion deterioration area based on the status detection data of each first detection point and the status detection data of each second detection point of each first corrosion deterioration area; The area determination module is used to determine multiple second corrosion deterioration areas based on the first corrosion deterioration degree distribution map of each first corrosion deterioration area and the second corrosion deterioration degree distribution map of each first corrosion deterioration area; The high-risk identification module is used to determine the high-risk corrosion deterioration area based on the status detection data of multiple first detection points of each second corrosion deterioration area and the status detection data of multiple second detection points of the second corrosion deterioration area.

6. The humanoid robot control system according to claim 5, characterized in that, The status detection data includes structural status detection data and corrosion feature detection data.

7. The humanoid robot control system according to claim 5, wherein The high-risk identification module is further used for: Constructing a storage tank atlas, which includes multiple status detection nodes and the edges between the multiple status detection nodes. The feature of each status detection node is the position of the status detection and the status detection data, and the feature of the edge is the spatial orientation relationship, distance and similarity of the status detection data between the status detection nodes; Processing the storage tank atlas based on a graph convolutional network to determine multiple verification points of each second corrosion deterioration area; Determining the high-risk corrosion deterioration area based on the detection and verification results of multiple verification points of each second corrosion deterioration area.

8. The humanoid robot control system according to claim 6, characterized in that, The input of the graph convolutional network is the storage tank atlas, and the output of the graph convolutional network is multiple verification points of each second corrosion deterioration area.

Images