A method, system, device, and storage medium for scheduling slope monitoring resources.

Abstract

This invention relates to a method, system, device, and storage medium for scheduling slope monitoring resources. The scheduling method first constructs a graph structure based on spatiotemporal conditions, integrating sensors, monitoring equipment, professionals, and auxiliary tools. Then, it transforms this graph into a tree structure, making resource combination schemes more hierarchical and clearer, resulting in several initial resource combination sequences. These sequences undergo crossover, mutation, and fitness learning processes to evolve towards stable monitoring quality, strong correlation between resources, and low total resource cost, yielding several optimized resource combination sequences. This improves the monitoring stability, rationality, and economy of the resource combination scheme. Finally, based on the characteristic information of the slope to be monitored, the optimal resource combination sequence with the highest matching degree is selected from the optimized resource combination sequences as the recommended monitoring scheme for the slope to be monitored, demonstrating strong rationality and scientific validity.

Description

Technical Field

[0001] This invention relates to the field of slope monitoring technology, specifically to a method, system, device, and storage medium for scheduling slope monitoring resources. Background Technology

[0002] Slope monitoring is of great significance for engineering construction and operation maintenance, personnel safety protection, property safety protection, and environmental protection. The selection of slope monitoring resources directly affects the accuracy of slope monitoring results and slope safety. Rational allocation of various slope monitoring resources is a prerequisite for completing high-quality slope monitoring.

[0003] Currently, slope resource allocation is often determined based on human experience. However, human experience relies too heavily on past experience, and when slope problems are complex or the slope area is large, human experience methods are often inefficient and result in low-quality allocation schemes. Machine learning algorithms can rapidly improve computational efficiency, but they require a large amount of training data. When faced with new slopes and new problems, they tend to lack interpretability and controllability, resulting in poor rationality and scientific rigor, and unstable generation of allocation schemes. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, device, and storage medium for scheduling slope monitoring resources.

[0005] The technical solution of this invention is as follows:

[0006] A method for scheduling slope monitoring resources includes the following operations:

[0007] S1. Based on sensors, monitoring equipment, professionals, and auxiliary tools, construct a slope monitoring resource structure map; the nodes in the slope monitoring resource structure map are sensors, monitoring equipment, professionals, and auxiliary tools. If there is overlap in available time between nodes, the nodes are connected by edges. The edge strength between nodes is based on the spatial distance and the duration of overlap in available time.

[0008] S2. Transform the slope monitoring resource structure diagram into a tree structure to obtain the slope monitoring resource structure tree. Specifically: Sensors, monitoring equipment, professional personnel, and auxiliary tools are designated as primary, secondary, tertiary, and quaternary resources, respectively. Nodes in the primary resources are randomly combined to obtain several combined nodes. These combined nodes, along with nodes in the primary resources, serve as root nodes, which are then sequentially connected to nodes with edge relationships in the secondary, tertiary, and quaternary resources, respectively, to obtain the slope monitoring resource structure tree. Based on the slope monitoring resource structure tree, several optimized resource combination sequences are obtained. Specifically: The root node in the slope monitoring resource structure tree is obtained... The path from the node to the tail node yields several initial resource combination sequences. Initial resource combination sequences with a fitness greater than a first fitness threshold are selected as initial preferred sequences. All initial preferred sequences form an initial sequence preferred set. Initial resource combination sequences with a fitness not greater than the first fitness threshold are cross-processed to obtain several cross-processed resource combination sequences. Based on the initial sequence preferred set, these cross-processed resource combination sequences are mutated to obtain several mutated resource combination sequences. The mutated resource combination sequences and the initial sequence preferred set are then subjected to fitness learning to obtain several optimized resource combination sequences. Fitness is obtained based on node cost and edge strength between nodes.

[0009] S3. Based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored, obtain the feature vector of the slope to be monitored; obtain the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vector of the slope to be monitored to obtain several matching degrees; take the optimized resource combination sequence corresponding to the maximum matching degree as the recommended monitoring scheme for the slope to be monitored.

[0010] The edge strengths of corresponding edges between sensor nodes in S1, and / or the edge strengths of corresponding edges between sensor nodes and monitoring equipment nodes, and / or the edge strengths of corresponding edges between monitoring equipment nodes, are obtained based on spatial distance, available time overlap duration, functional relevance, and historical co-occurrence frequency. The edge strengths of corresponding edges between sensor nodes and / or monitoring equipment nodes and professional personnel nodes are obtained based on spatial distance, available time overlap duration, and the dependence of sensors and / or monitoring equipment on professional personnel. The edge strengths of corresponding edges between sensor nodes and / or monitoring equipment nodes and auxiliary tool nodes are obtained based on spatial distance, available time overlap duration, and the necessity of auxiliary tools for sensors and / or monitoring equipment. The edge strengths of corresponding edges between professional personnel nodes are obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency. The edge strengths of corresponding edges between professional personnel nodes and auxiliary tool nodes are obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency. The edge strengths of corresponding edges between auxiliary tool nodes are obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency.

[0011] The crossover process in S2 is as follows: Initial resource combination sequences with fitness no greater than the first fitness threshold are used as crossover sequences; the maximum and minimum fitness values ​​are obtained from all crossover sequences; based on the maximum and minimum fitness values, the crossover probability of each crossover sequence is obtained; all crossover sequences are sorted according to their crossover probabilities from smallest to largest and from largest to largest, resulting in a forward-sorted sequence set and a reverse-sorted sequence set; two crossover sequences with the same sort number in the forward-sorted sequence set and the reverse-sorted sequence set undergo partial sequence swapping to obtain several crossover resource combination sequences.

[0012] The mutation process in S2 is as follows: Based on the sensor failure probability, monitoring equipment failure probability, professional error probability, and auxiliary tool damage probability, obtain the mutation probability of each cross-resource combination sequence; determine whether the mutation probability of the current cross-resource combination sequence is greater than the mutation probability threshold; if it is greater, replace the resources at the corresponding positions of the non-root nodes in the current cross-resource combination sequence with resources from any initial resource combination sequence in the initial sequence selection set to obtain the current initial mutation sequence; determine whether the weighted difference between the fitness and mutation probability of the initial mutation sequence is greater than the comprehensive mutation probability threshold; if it is greater, use the current initial mutation sequence as the current mutated resource combination sequence; if it is not greater... The initial mutation sequence undergoes resource replacement and comparison with the comprehensive mutation probability threshold until the weighted difference between fitness and mutation probability is greater than the comprehensive mutation probability threshold, resulting in the current mutated resource combination sequence. If the difference is not greater, an initial resource combination sequence with a fitness greater than the second fitness threshold is randomly selected from the initial sequence selection set as the target sequence. The resources at the non-root nodes in the current crossover resource combination sequence are replaced with any resource in the target sequence, resulting in the current mutated resource combination sequence. The second fitness threshold is greater than the first fitness threshold. After all crossover resource combination sequences have undergone mutation probability comparison, several mutated resource combination sequences are obtained.

[0013] The fitness learning process in S2 is as follows: The mutated resource combination sequence and the initial sequence optimization set form a set of sequences to be learned; the fitness of each sequence to be learned in the set is obtained, and the sequences with fitness greater than the third fitness threshold are used as target learning sequences, forming a target learning sequence set; each target learning sequence in the target learning sequence set is processed by a first fixed-length combination resource segmentation to obtain several resource combination fragments; the resource combination fragment with the maximum fitness at the same position among these fragments is selected as the target learning resource combination fragment at the corresponding position; each sequence to be learned in the set is traversed using a first fixed-length sliding window, and the resource combination fragment with the minimum fitness in each sequence to be learned is replaced with the target learning resource combination fragment at the same position, resulting in several optimized resource combination sequences.

[0014] In S2, fitness is obtained using the following formula:

[0015]

[0016] F represents fitness, E ij Let C be the edge strength between node i and node j, and let node i and node j be resources, including sensors, monitoring equipment, professionals, and auxiliary tools. s C m C p C a Let S, M, P, and A represent the costs of the s-th sensor, m-th monitoring equipment, p-th professional personnel, and a-th auxiliary tool, respectively. Let S, M, P, and A represent the total number of sensors, monitoring equipment, professional personnel, and auxiliary tools in the current resource sequence, respectively. Let ε be the compensation amount.

[0017] In S3, if the slope risk value of the slope to be monitored is greater than the risk threshold, the optimal resource combination sequence with a matching degree greater than the matching degree threshold and the minimum overlap with the recommended monitoring scheme will be used as a supplementary monitoring scheme for the slope to be monitored, and will be used to form the optimal monitoring scheme with the recommended monitoring scheme for the slope to be monitored.

[0018] A slope monitoring resource scheduling system, used to implement the aforementioned slope monitoring resource scheduling method, includes:

[0019] The slope monitoring resource structure map generation module is used to construct a slope monitoring resource structure map based on sensors, monitoring equipment, professionals, and auxiliary tools. The nodes in the slope monitoring resource structure map are sensors, monitoring equipment, professionals, and auxiliary tools. If there is overlap in the available time between nodes, the nodes are connected by edges. The edge strength between nodes is based on the spatial distance and the duration of overlap in available time.

[0020] The optimized resource combination sequence generation module transforms the slope monitoring resource structure map into a tree structure, resulting in a slope monitoring resource structure tree. Specifically, sensors, monitoring equipment, professional personnel, and auxiliary tools are designated as primary, secondary, tertiary, and quaternary resources, respectively. Nodes in the primary resources are randomly combined to obtain several combined nodes. These combined nodes, along with nodes in the primary resources, serve as root nodes, which are then sequentially connected to nodes with edge relationships in the secondary, tertiary, and quaternary resources, respectively, to obtain the slope monitoring resource structure tree. Based on this structure tree, several optimized resource combination sequences are generated. Specifically, the slope monitoring resource structure is obtained by... The path from the root node to the tail node in the tree is used to obtain several initial resource combination sequences. Initial resource combination sequences with a fitness greater than a first fitness threshold are used as initial preferred sequences. All initial preferred sequences form an initial sequence preferred set. Initial resource combination sequences with a fitness not greater than the first fitness threshold are cross-processed to obtain several cross-processed resource combination sequences. Based on the initial sequence preferred set, several cross-processed resource combination sequences are mutated to obtain several mutated resource combination sequences. The mutated resource combination sequences and the initial sequence preferred set are then subjected to fitness learning to obtain several optimized resource combination sequences. Fitness is obtained based on node cost and edge strength between nodes.

[0021] The recommended monitoring scheme generation module is used to obtain the feature vector of the slope to be monitored based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored; obtain the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vector of the slope to be monitored to obtain several matching degrees; and take the optimized resource combination sequence corresponding to the maximum matching degree as the recommended monitoring scheme for the slope to be monitored.

[0022] A slope monitoring resource scheduling device includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the above-mentioned slope monitoring resource scheduling method.

[0023] A computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the above-described method for scheduling slope monitoring resources.

[0024] The beneficial effects of this invention are as follows:

[0025] This invention provides a method for scheduling slope monitoring resources. First, based on spatiotemporal conditions, the currently available sensors, monitoring equipment, professionals, and auxiliary tools are constructed into a graph structure to intuitively reflect the distribution of available monitoring resources and the correlation between resources. Then, the slope monitoring resource structure graph is transformed into a tree structure, making the resource combination scheme more hierarchical and clearer, resulting in several initial resource combination sequences. These initial resource combination sequences are then subjected to crossover, mutation, and fitness learning processes to evolve towards the goals of stable monitoring quality, strong correlation between resources, and low total resource cost, resulting in several optimized resource combination sequences. This improves the monitoring stability, rationality, and economy of the resource combination scheme. Finally, based on the characteristic information of the slope to be monitored, the optimal resource combination sequence with the highest matching degree value is selected from the several optimized resource combination sequences as the recommended monitoring scheme for the slope to be monitored, which has strong rationality and scientific validity. Detailed Implementation

[0026] This embodiment provides a method for scheduling slope monitoring resources, including the following operations:

[0027] S1. Based on sensors, monitoring equipment, professionals, and auxiliary tools, construct a slope monitoring resource structure map; the nodes in the slope monitoring resource structure map are sensors, monitoring equipment, professionals, and auxiliary tools. If there is overlap in available time between nodes, the nodes are connected by edges. The edge strength between nodes is based on the spatial distance and the duration of overlap in available time.

[0028] S2. Transform the slope monitoring resource structure diagram into a tree structure to obtain the slope monitoring resource structure tree. Specifically: Sensors, monitoring equipment, professional personnel, and auxiliary tools are designated as primary, secondary, tertiary, and quaternary resources, respectively. Nodes in the primary resources are randomly combined to obtain several combined nodes. These combined nodes, along with nodes in the primary resources, serve as root nodes, which are then sequentially connected to nodes with edge relationships in the secondary, tertiary, and quaternary resources, respectively, to obtain the slope monitoring resource structure tree. Based on the slope monitoring resource structure tree, several optimized resource combination sequences are obtained. Specifically: The root node in the slope monitoring resource structure tree is obtained... The path from the node to the tail node yields several initial resource combination sequences. Initial resource combination sequences with a fitness greater than a first fitness threshold are selected as initial preferred sequences. All initial preferred sequences form an initial sequence preferred set. Initial resource combination sequences with a fitness not greater than the first fitness threshold are cross-processed to obtain several cross-processed resource combination sequences. Based on the initial sequence preferred set, these cross-processed resource combination sequences are mutated to obtain several mutated resource combination sequences. The mutated resource combination sequences and the initial sequence preferred set are then subjected to fitness learning to obtain several optimized resource combination sequences. Fitness is obtained based on node cost and edge strength between nodes.

[0029] S3. Based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored, obtain the feature vector of the slope to be monitored; obtain the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vector of the slope to be monitored to obtain several matching degrees; take the optimized resource combination sequence corresponding to the maximum matching degree as the recommended monitoring scheme for the slope to be monitored.

[0030] S1. Construct a slope monitoring resource structure map based on sensors, monitoring equipment, professionals, and auxiliary tools. The nodes in the slope monitoring resource structure map are sensors, monitoring equipment, professionals, and auxiliary tools. If there is overlap in available time between nodes, the nodes are connected by edges. The edge strength between nodes is obtained based on spatial distance and the overlap duration of available time.

[0031] Based on spatiotemporal conditions, the currently available sensors, monitoring equipment, professionals, and auxiliary tools are constructed into a graphical structure to intuitively reflect the distribution of currently available monitoring resources and the correlation between resources (the ease with which resources can be acquired simultaneously and the degree of functional connection between resources).

[0032] Based on currently available sensors with both temporal and spatial attributes (sensors directly installed inside the slope body that can directly acquire slope conditions, such as fiber optic displacement sensors, stress sensors, pore water pressure sensors, etc.), monitoring equipment (drones, cameras, ground monitoring stations, etc. installed outside the slope to acquire slope conditions), professionals (such as geotechnical engineers, surveying engineers, and geological engineers), and auxiliary tools (such as mounting brackets, drilling equipment, anchoring tools, etc.), a slope monitoring resource structure map is constructed.

[0033] In the slope monitoring resource structure diagram, nodes represent sensors, monitoring equipment, professional personnel, and auxiliary tools. If there is overlap in the available time between nodes, it means that the resources corresponding to the two nodes can be obtained simultaneously. In this way, the nodes are connected by edges, thus treating the slope monitoring resources as a system. This clearly shows the temporal correlation of different resources and helps to quickly and comprehensively understand the distribution of currently available resources.

[0034] The edge strength between nodes is based on the overlap of spatial distance and available time. The greater the edge strength, the shorter the edge length, which can intuitively show the ease with which resources can be obtained simultaneously in time and space. The edge strength between resources is defined as follows.

[0035] The edge strength of corresponding edges between sensor nodes, and / or between sensor nodes and monitoring device nodes, and / or between monitoring device nodes, is derived based on spatial distance, available time overlap duration, functional relevance, and historical co-occurrence frequency. Smaller spatial distances, longer available time overlap durations, higher functional relevance, and higher historical co-occurrence frequency result in stronger edge strengths, greater resource correlation, and lower difficulty in simultaneous scheduling. Functional relevance is based on the number of overlapping functional keywords between sensors, and / or between sensors and monitoring devices, and / or between monitoring device nodes.

[0036] The edge strength between sensor nodes and / or monitoring device nodes and professional personnel nodes is derived based on spatial distance, available time overlap duration, and the dependence of the sensor or / or monitoring device on the professional personnel. The dependence of the sensor or / or monitoring device on the professional personnel is based on the historical number of times the sensor or / or monitoring device and the professional personnel have co-occurred.

[0037] The edge strength between sensor nodes and / or monitoring device nodes and auxiliary tool nodes is determined based on spatial distance, available time overlap duration, and the necessity of the auxiliary tool for the sensor and / or monitoring device. The necessity of the auxiliary tool for the sensor and / or monitoring device is based on the historical number of times the auxiliary tool and the sensor and / or monitoring device have co-occurred.

[0038] The edge strength between professional node corresponding edges is obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency.

[0039] The edge strength between professional personnel nodes and auxiliary tool nodes is determined based on spatial distance, the duration of overlap in available time, and the number of times they co-occur in history.

[0040] The edge strength between corresponding edges between auxiliary tool nodes is obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency.

[0041] S2. Transform the slope monitoring resource structure diagram into a tree structure to obtain the slope monitoring resource structure tree; based on the slope monitoring resource structure tree, obtain several optimized resource combination sequences.

[0042] The slope monitoring resource structure map is transformed into a tree structure, making the resource combination scheme more hierarchical and clearer, and obtaining several initial resource combination sequences. The initial resource combination sequences are then subjected to crossover, mutation, and fitness learning processes to evolve the resource combination sequences toward the goals of stable monitoring quality, strong correlation between resources, and low total resource cost, resulting in several optimized resource combination sequences. This improves the monitoring stability, rationality, and economy of the resource combination scheme.

[0043] First, the slope monitoring resource structure diagram is transformed into a tree structure, resulting in a slope monitoring resource structure tree. In the slope monitoring resource structure tree, the position of each tree node (node / resource) in the tree can reasonably reflect its role in the slope monitoring resource scheme system.

[0044] The specific steps to obtain the slope monitoring resource structure tree are as follows: Based on the resource's ability to directly and quickly acquire data from the slope itself, the resources are classified into levels from high to low. Sensors buried on the slope for monitoring, monitoring equipment located a distance from the slope for monitoring the overall condition of the slope, professional personnel for placing instruments and conducting data analysis, and auxiliary tools for installing or fixing instruments are respectively classified as Level 1, Level 2, Level 3, and Level 4 resources. The nodes in the Level 1 resources are randomly combined to obtain several combined nodes. The number of nodes (resources) in each combined node varies. The number of nodes can be two or more, ensuring the richness of resource combinations. Several combination nodes and nodes in the primary resources serve as root nodes, while resources that indirectly affect slope data judgment through other intermediate data (secondary, tertiary, and quaternary resources) can serve as child nodes. The root node is then sequentially connected to nodes with edge relationships in the secondary, tertiary, and quaternary resources, respectively, to obtain the slope monitoring resource structure tree. That is, after the root node is connected to several nodes with edge relationships in the secondary resources, it is then sequentially connected to several nodes with edge relationships in the tertiary and quaternary resources. When nodes in the upper-level resources are connected to several nodes in the lower-level resources, the connections are sequential, meaning that each node has only one inbound and one outbound edge. When nodes in the upper-level resources are connected to several nodes in the lower-level resources, they can be randomly excluded from connection to reduce the number of resources in the resource combination scheme, forming a low-cost resource combination scheme and enriching the resource combination schemes.

[0045] Then, the path from the root node to the tail node in the slope monitoring resource structure tree is obtained, resulting in several initial resource combination sequences. Each initial resource combination sequence consists of several nodes / resources and edges.

[0046] Subsequently, the initial resource combination sequences with fitness greater than the first fitness threshold—that is, initial resource combination sequences that have advantages in both cost and relevance between resources—are selected as the initial preferred sequences; all initial preferred sequences form the initial sequence preferred set. Fitness is obtained based on node cost and edge strength between nodes.

[0047] The fitness values ​​mentioned above are obtained using the following formula:

[0048]

[0049] F represents fitness, E ij Let C be the edge strength between node i and node j, where node i and node j are resources. Node i and node j can be different resources of different levels, or different resources of the same level. Resources include sensors, monitoring equipment, professional personnel, and auxiliary tools. s C m C p C a Let S, M, P, and A represent the costs of the s-th sensor, m-th monitoring equipment, p-th professional personnel, and a-th auxiliary tool, respectively. S, M, P, and A represent the total number of sensors, monitoring equipment, professional personnel, and auxiliary tools in the current resource sequence (including the initial resource combination sequence), respectively. ε is the compensation amount.

[0050] Next, based on the slope monitoring resource structure tree, several optimized resource combination sequences are obtained.

[0051] One approach is to obtain the path from the root node to the tail node in the slope monitoring resource structure tree, thus obtaining several initial resource combination sequences; and then, among these initial resource combination sequences, the initial resource combination sequence with a fitness greater than the initial fitness threshold is taken as the optimized resource combination sequence, thus obtaining several optimized resource combination sequences.

[0052] Another approach is to obtain a resource combination scheme with stable monitoring quality, strong correlation between resources, and low total resource cost by acquiring the path from the root node to the tail node in the slope monitoring resource structure tree, thus obtaining several initial resource combination sequences.

[0053] To enhance the diversity of resource combination sequences, the initial resource combination sequences with fitness no greater than the first fitness threshold are cross-processed to obtain several cross-processed resource combination sequences.

[0054] The crossover process is as follows: Initial resource combination sequences with fitness no greater than the first fitness threshold are selected as crossover sequences; the maximum and minimum fitness values ​​are obtained from all crossover sequences; based on the maximum and minimum fitness values, the crossover probability of each crossover sequence is obtained; all crossover sequences are sorted twice according to their crossover probabilities, resulting in a forward-sorted sequence set and a reverse-sorted sequence set; two crossover sequences with the same sorting number in the forward-sorted and reverse-sorted sequence sets undergo partial sequence swapping (swapping resource combinations after resources at the same position), allowing two crossover sequences with higher and lower crossover probabilities to exchange some resource combinations, balancing fitness while enriching the diversity of resource combination sequences, resulting in several crossover resource combination sequences.

[0055] The aforementioned partial sequence swapping is achieved by exchanging resource combinations at the same positions in the two sequences to be crossed. For example, sequence 1 to be crossed is sensor a-sensor b-monitoring equipment a-professional a-auxiliary tool a, and sequence 2 to be crossed is sensor b-monitoring equipment b-professional b-auxiliary tool b. After partial sequence swapping (using the 3rd position as the swap point), the initial cross-resource combination sequence 1 is sensor a-sensor b-professional b-auxiliary tool b, and the initial cross-resource combination sequence 2 is sensor b-monitoring equipment b-monitoring equipment a-professional a-auxiliary tool a. If, after partial sequence swapping, there are instances where resources cannot be used simultaneously due to the absence of edges between them, then the unswapped resources are deleted to obtain a cross-filtered resource sequence, which serves as a cross-resource combination sequence. For example, in the initial cross-resource combination sequence 2 above, which is sensor b-monitoring device b-monitoring device a-professional a-auxiliary tool a, there is no edge between monitoring device b and monitoring device a, but there is an edge relationship between sensor b and monitoring device a. Therefore, monitoring device b is deleted, and sensor b is connected to monitoring device a, resulting in the cross-resource combination sequence sensor b-monitoring device a-professional a-auxiliary tool a.

[0056] The crossover probability of the current sequence to be crossed is the ratio of the difference between the maximum fitness and the fitness of the current sequence to be crossed to the difference between the maximum fitness and the minimum fitness.

[0057] Next, in order to maximize the monitoring quality of resources in the resource combination sequence, and to ensure that the resource combinations in the resource combination sequence are not only of high quality but also have strong correlations between resources, thereby improving the rationality and scientific nature of the resource combination sequence, based on the initial sequence optimization set with high fitness, several cross-resource combination sequences are mutated to obtain several mutated resource combination sequences.

[0058] The mutation handling process is as follows.

[0059] Based on the probability of sensor failure, monitoring equipment failure, professional error, and auxiliary tool damage, the mutation probability of each cross-resource combination sequence is obtained. A higher probability of these probabilities indicates a greater likelihood of failure during implementation of the resource combination scheme within the cross-resource combination sequence, resulting in unstable monitoring quality and a greater need for quality optimization of the resource combination. The probability of these probabilities is derived from historical data on sensor failure frequency and repair frequency, monitoring equipment failure frequency and repair frequency, professional error frequency and tenure, and auxiliary tool damage frequency and repair frequency, respectively.

[0060] The probability of variation can be a weighted sum of the probabilities of sensor failure, monitoring equipment failure, professional error, and auxiliary tool damage. Alternatively, it can be a weighted sum of the probabilities of monitoring equipment failure, professional error, auxiliary tool damage, and sensor failure, considering the independence of these probabilities and using an "OR" logical relationship. For example, if the probability of sensor failure is p1, the probability of monitoring equipment failure is p2, the probability of professional error is p3, and the probability of auxiliary tool damage is p4, then the probability of variation P = w1p1 + w2p2 + w3p3 + w4p4, or the probability of variation P = w1p1 + w2p2 + w3p3 + w4p4 - w5(p1p2 - p1p3 - p1p4 - p2p3 - p2p4 - p3p4) + w6(p1p2p3 + p1p2p4 + p2p3p4), where w1, w2, w3, w4, w5, and w6 are the first, second, third, fourth, fifth, and sixth weights, respectively.

[0061] Determine whether the mutation probability of the current cross-resource combination sequence is greater than the mutation probability threshold.

[0062] If the difference is greater than the threshold, it indicates that the resource combination scheme in the current cross-resource combination sequence has low monitoring quality and unstable monitoring results during subsequent implementation. To improve the quality of the resource combination scheme and ensure strong correlation between resources and low total resource cost, the resources at the non-root node positions in the current cross-resource combination sequence are replaced with resources from any initial resource combination sequence in the initial sequence optimization set, resulting in the current initial mutation sequence. Then, the weighted difference between the fitness and mutation probability of the initial mutation sequence is used as the mutation quality after the initial mutation. It is then determined whether the weighted difference between the fitness and mutation probability of the initial mutation sequence is greater than the comprehensive mutation probability threshold. If it is greater, it indicates that the mutation operation has reached the target state, and the current initial mutation sequence is used as the current mutated resource combination sequence. If it is not greater, the current initial mutation sequence performs resource replacement at the non-root node positions and performs a comparison with the comprehensive mutation probability threshold until the weighted difference between the fitness and mutation probability is greater than the comprehensive mutation probability threshold. The mutation operation has reached the target state, and the resource combination scheme corresponding to the resource sequence satisfies high combination scheme quality, strong correlation between resources, and low total resource cost, thus obtaining the current mutated resource combination sequence.

[0063] If the fitness is not greater than the second fitness threshold, it proves that the resource combination scheme in the current cross-resource combination sequence has high monitoring quality and stable monitoring results during subsequent implementation. In order to further improve the correlation between resources and further reduce the total resource cost on the basis of high quality of resource combination scheme, an initial resource combination sequence with a fitness greater than the second fitness threshold is randomly selected from the initial sequence optimization set as the target sequence. The resources at the non-root node positions in the current cross-resource combination sequence are replaced with any resource in the target sequence to obtain the current mutated resource combination sequence.

[0064] The second fitness threshold mentioned above is greater than the first fitness threshold.

[0065] After all cross-resource combination sequences have undergone the mutation probability determination operation, several mutated resource combination sequences are obtained.

[0066] Finally, to further improve the fitness of resource combination schemes in the resource sequence and further reduce the total resource cost by increasing the correlation between resources in the resource sequence, the mutated resource combination sequence and the initial sequence optimization set are subjected to fitness learning to obtain several optimized resource combination sequences.

[0067] The fitness learning process is as follows: A set of sequences to be learned is formed by the mutated resource combination sequence and the initial sequence optimization set; the fitness of each sequence to be learned is obtained, and the sequences with fitness greater than a third fitness threshold are used as target learning sequences, forming a target learning sequence set; each target learning sequence in the target learning sequence set is processed by a first fixed-length resource combination segmentation, dividing each target learning sequence into many resource combination fragments, resulting in several resource combination fragments; the resource combination fragment with the maximum fitness at the same position among these fragments is selected as the target learning resource combination fragment at that position; simultaneously, each sequence to be learned is traversed using a first fixed-length sliding window, and the resource combination fragment with the minimum fitness in each sequence to be learned is replaced with the target learning resource combination fragment at the same position, further optimizing the fitness of each sequence to be learned locally, resulting in several optimized resource combination sequences.

[0068] S3. Based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored, obtain the feature vector of the slope to be monitored; obtain the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vector of the slope to be monitored to obtain several matching degrees; take the optimized resource combination sequence corresponding to the maximum matching degree as the recommended monitoring scheme for the slope to be monitored.

[0069] Based on the characteristic information of the slope to be monitored, the optimal resource combination sequence with the highest matching degree value is selected from several optimal resource combination sequences as the recommended monitoring scheme for the slope to be monitored, so as to achieve the correspondence, rationality and scientific nature of the recommended monitoring scheme.

[0070] First, the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored are spliced ​​together and then embedded to obtain the feature vector of the slope to be monitored.

[0071] Simultaneously, each optimized resource combination sequence is embedded separately to obtain the optimized resource combination sequence feature vector for each optimized resource combination sequence.

[0072] Next, the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vectors of the slope to be monitored is obtained. That is, the product of the feature vector of each optimized resource combination sequence and the feature vector of the slope to be monitored is obtained to obtain several matching degrees.

[0073] Finally, the optimized resource combination sequence corresponding to the maximum matching degree is used as the recommended monitoring scheme for the slope to be monitored.

[0074] In addition, if the slope risk value of the slope to be monitored is greater than the risk threshold, the optimized resource combination sequence with a matching degree greater than the matching degree threshold and the minimum overlap with the optimal monitoring scheme will be used as a supplementary monitoring scheme for the slope to be monitored. This will complement the recommended monitoring scheme and form the optimal monitoring scheme with the recommended monitoring scheme for the slope to be monitored, so as to ensure the quality of slope monitoring for slopes with high risk values.

[0075] The aforementioned slope risk values ​​can be obtained based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored; the aforementioned repeatability can be obtained based on the number of times the resource category is repeated and the number of times the resource is repeated in two resource combination sequences.

[0076] This embodiment also provides a slope monitoring resource scheduling system for implementing the above-mentioned slope monitoring resource scheduling method, including:

[0077] The slope monitoring resource structure map generation module is used to construct a slope monitoring resource structure map based on sensors, monitoring equipment, professionals, and auxiliary tools. The nodes in the slope monitoring resource structure map are sensors, monitoring equipment, professionals, and auxiliary tools. If there is overlap in the available time between nodes, the nodes are connected by edges. The edge strength between nodes is based on the spatial distance and the duration of overlap in available time.

[0078] The optimized resource combination sequence generation module transforms the slope monitoring resource structure map into a tree structure, resulting in a slope monitoring resource structure tree. Specifically, sensors, monitoring equipment, professional personnel, and auxiliary tools are designated as primary, secondary, tertiary, and quaternary resources, respectively. Nodes in the primary resources are randomly combined to obtain several combined nodes. These combined nodes, along with nodes in the primary resources, serve as root nodes, which are then sequentially connected to nodes with edge relationships in the secondary, tertiary, and quaternary resources, respectively, to obtain the slope monitoring resource structure tree. Based on this structure tree, several optimized resource combination sequences are generated. Specifically, the slope monitoring resource structure is obtained by... The path from the root node to the tail node in the tree is used to obtain several initial resource combination sequences. Initial resource combination sequences with a fitness greater than a first fitness threshold are used as initial preferred sequences. All initial preferred sequences form an initial sequence preferred set. Initial resource combination sequences with a fitness not greater than the first fitness threshold are cross-processed to obtain several cross-processed resource combination sequences. Based on the initial sequence preferred set, several cross-processed resource combination sequences are mutated to obtain several mutated resource combination sequences. The mutated resource combination sequences and the initial sequence preferred set are then subjected to fitness learning to obtain several optimized resource combination sequences. Fitness is obtained based on node cost and edge strength between nodes.

[0079] The recommended monitoring scheme generation module is used to obtain the feature vector of the slope to be monitored based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored; obtain the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vector of the slope to be monitored to obtain several matching degrees; and take the optimized resource combination sequence corresponding to the maximum matching degree as the recommended monitoring scheme for the slope to be monitored.

[0080] This embodiment also provides a slope monitoring resource scheduling device, including a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the above-described slope monitoring resource scheduling method.

[0081] This embodiment also provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the above-described method for scheduling slope monitoring resources.

[0082] This embodiment provides a method for scheduling slope monitoring resources. First, based on spatiotemporal conditions, the currently available sensors, monitoring equipment, professionals, and auxiliary tools are constructed into a graph structure to intuitively reflect the distribution of available monitoring resources and the correlation between resources. Then, the slope monitoring resource structure graph is transformed into a tree structure, making the resource combination scheme more hierarchical and clearer, resulting in several initial resource combination sequences. These initial resource combination sequences are then subjected to crossover, mutation, and fitness learning processes to evolve towards the goals of stable monitoring quality, strong correlation between resources, and low total resource cost, resulting in several optimized resource combination sequences. This improves the monitoring stability, rationality, and economy of the resource combination scheme. Finally, based on the characteristic information of the slope to be monitored, the optimal resource combination sequence with the highest matching degree value is selected from the several optimized resource combination sequences as the recommended monitoring scheme for the slope to be monitored, which has strong rationality and scientific validity. This method, when applied to slope resource scheduling, generates high-quality schemes with good stability.

Claims

1. A method for scheduling slope monitoring resources, characterized in that, This includes the following operations: S1. Based on sensors, monitoring equipment, professionals, and auxiliary tools, construct a slope monitoring resource structure map; the nodes in the slope monitoring resource structure map are sensors, monitoring equipment, professionals, and auxiliary tools. If there is overlap in available time between nodes, the nodes are connected by edges. The edge strength between nodes is based on the spatial distance and the duration of overlap in available time. S2. Transform the slope monitoring resource structure diagram into a tree structure to obtain the slope monitoring resource structure tree. Specifically, sensors, monitoring equipment, professional personnel, and auxiliary tools are respectively designated as primary, secondary, tertiary, and quaternary resources. Nodes in the primary resources are randomly combined to obtain several combined nodes. These combined nodes and nodes in the primary resources are all used as root nodes and are sequentially connected to nodes with edge relationships in the secondary, tertiary, and quaternary resources to obtain the slope monitoring resource structure tree. Based on the slope monitoring resource structure tree, several optimized resource combination sequences are obtained. Specifically: the paths from the root node to the tail node in the slope monitoring resource structure tree are obtained, resulting in several initial resource combination sequences; the initial resource combination sequences with fitness greater than a first fitness threshold are used as initial preferred sequences; all initial preferred sequences form an initial sequence preferred set; the initial resource combination sequences with fitness not greater than the first fitness threshold are cross-processed to obtain several cross-processed resource combination sequences; based on the initial sequence preferred set, the cross-processed resource combination sequences are mutated to obtain several mutated resource combination sequences; the mutated resource combination sequences and the initial sequence preferred set are subjected to fitness learning to obtain several optimized resource combination sequences; fitness is obtained based on node cost and edge strength between nodes. The fitness learning process involves: the mutated resource combination sequence and the initial sequence optimization set form the set of sequences to be learned; The fitness of each learning sequence in the learning sequence set is obtained. The learning sequences with fitness greater than the third fitness threshold are used as target learning sequences, and all target learning sequences form a target learning sequence set. Each target learning sequence in the target learning sequence set is processed by a first fixed-length combination resource segmentation to obtain several resource combination segments. The resource combination segment with the maximum fitness at the same position is selected from several resource combination segments and used as the target learning resource combination segment at the corresponding position. Each learning sequence in the learning sequence set is traversed with a first fixed-length sliding window, and the resource combination segment with the minimum fitness in each learning sequence is replaced with the target learning resource combination segment at the same position to obtain several optimized resource combination sequences. S3. Based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored, obtain the feature vector of the slope to be monitored; obtain the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vector of the slope to be monitored to obtain several matching degrees; take the optimized resource combination sequence corresponding to the maximum matching degree as the recommended monitoring scheme for the slope to be monitored.

2. The method for scheduling slope monitoring resources according to claim 1, characterized in that, In S1, The edge strength of corresponding edges between sensor nodes, and / or the edge strength of corresponding edges between sensor nodes and monitoring device nodes, and / or the edge strength of corresponding edges between monitoring device nodes, is obtained based on spatial distance, available time overlap duration, functional relevance, and historical co-occurrence frequency. The edge strength between sensor nodes and / or monitoring equipment nodes and professional personnel nodes is obtained based on spatial distance, available time overlap duration, and the dependence of sensors and / or monitoring equipment on professional personnel. The edge strength between the corresponding edge of the sensor node and / or monitoring device node and the auxiliary tool node is obtained based on the spatial distance, the available time overlap duration, and the necessity of the auxiliary tool for the sensor and / or monitoring device. The edge strength between professional node corresponding edges is obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency; The edge strength between professional personnel nodes and auxiliary tool nodes is obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency. The edge strength between corresponding edges between auxiliary tool nodes is obtained based on spatial distance, available time overlap duration, and historical co-occurrence frequency.

3. The method for scheduling slope monitoring resources according to claim 1, characterized in that, In S2, the cross-processing operation is specifically as follows: The initial resource combination sequence with fitness no greater than the first fitness threshold is taken as the crossover sequence; the maximum fitness and minimum fitness values ​​are obtained among all the crossover sequences. Based on the maximum and minimum fitness values, the crossover probability of each sequence to be crossed is obtained. According to the order of crossover probabilities from smallest to largest and from largest to largest, all sequences to be crossed are sorted to obtain a forward sorted sequence set and a reverse sorted sequence set. By partially swapping two sequences with the same sorting number in the forward and reverse sorting sequence sets, several cross-resource combination sequences are obtained.

4. The method for scheduling slope monitoring resources according to claim 1, characterized in that, In S2, the mutation processing operation is specifically as follows: Based on the probability of sensor failure, the probability of monitoring equipment failure, the probability of professional personnel error, and the probability of auxiliary tool damage, the mutation probability of each cross-resource combination sequence is obtained; Determine whether the mutation probability of the current cross-resource combination sequence is greater than the mutation probability threshold; If the difference is greater than the threshold, replace the resources at the non-root node positions in the current cross-resource combination sequence with resources from any initial resource combination sequence in the initial sequence preference set to obtain the current initial mutation sequence; determine whether the weighted difference between the fitness and mutation probability of the initial mutation sequence is greater than the comprehensive mutation probability threshold; if it is greater, use the current initial mutation sequence as the current mutation resource combination sequence; if it is not greater, perform resource replacement and comparison with the comprehensive mutation probability threshold on the current initial mutation sequence until the weighted difference between the fitness and mutation probability is greater than the comprehensive mutation probability threshold to obtain the current mutation resource combination sequence. If the fitness is not greater than the second fitness threshold, randomly select an initial resource combination sequence from the initial sequence preference set whose fitness is greater than the second fitness threshold as the target sequence; replace the resources at the non-root node positions in the current crossover resource combination sequence with any resource in the target sequence to obtain the current mutated resource combination sequence; the second fitness threshold is greater than the first fitness threshold. After all cross-resource combination sequences have undergone the mutation probability determination operation, several mutated resource combination sequences are obtained.

5. The method for scheduling slope monitoring resources according to claim 1, characterized in that, In S2, the fitness is obtained using the following formula: ， For fitness, For nodes i With nodes j Edge strength between nodes i With nodes j Resources include sensors, monitoring equipment, professional personnel, and auxiliary tools. , , , The first s The cost of the first sensor, m Cost of each monitoring device, p Cost per professional and the first a Cost of each auxiliary tool S , M , P , A These represent the total number of sensors, monitoring devices, professional personnel, and auxiliary tools in the current resource sequence. This is the amount of compensation.

6. The method for scheduling slope monitoring resources according to claim 1, characterized in that, In step S3, if the slope risk value of the slope to be monitored is greater than the risk threshold, the optimized resource combination sequence with a matching degree greater than the matching degree threshold and the least overlap with the recommended monitoring scheme will be used as a supplementary monitoring scheme for the slope to be monitored, and will be used to form the optimal monitoring scheme with the recommended monitoring scheme for the slope to be monitored.

7. A slope monitoring resource scheduling system, used to implement the slope monitoring resource scheduling method described in claim 1, characterized in that, include: The slope monitoring resource structure map generation module is used to construct a slope monitoring resource structure map based on sensors, monitoring equipment, professionals, and auxiliary tools. In the slope monitoring resource structure diagram, nodes represent sensors, monitoring equipment, professional personnel, and auxiliary tools. If there is overlap in available time between nodes, the nodes are connected by edges. The edge strength between nodes is based on the spatial distance and the duration of overlap in available time. The optimized resource combination sequence generation module transforms the slope monitoring resource structure map into a tree structure, resulting in a slope monitoring resource structure tree. Specifically, sensors, monitoring equipment, professional personnel, and auxiliary tools are designated as primary, secondary, tertiary, and quaternary resources, respectively. Nodes in the primary resources are randomly combined to obtain several combined nodes. These combined nodes, along with nodes in the primary resources, serve as root nodes, which are then sequentially connected to nodes with edge relationships in the secondary, tertiary, and quaternary resources, respectively, to obtain the slope monitoring resource structure tree. Based on this structure tree, several optimized resource combination sequences are generated. Specifically, the slope monitoring resource structure is obtained by... The path from the root node to the tail node in the tree is used to obtain several initial resource combination sequences. Initial resource combination sequences with a fitness greater than a first fitness threshold are used as initial preferred sequences. All initial preferred sequences form an initial sequence preferred set. Initial resource combination sequences with a fitness not greater than the first fitness threshold are cross-processed to obtain several cross-processed resource combination sequences. Based on the initial sequence preferred set, several cross-processed resource combination sequences are mutated to obtain several mutated resource combination sequences. The mutated resource combination sequences and the initial sequence preferred set are then subjected to fitness learning to obtain several optimized resource combination sequences. Fitness is obtained based on node cost and edge strength between nodes. The recommended monitoring scheme generation module is used to obtain the feature vector of the slope to be monitored based on the slope height, slope angle, soil and rock properties, groundwater information, and crack development information of the slope to be monitored; obtain the product of the feature vectors corresponding to several optimized resource combination sequences and the feature vector of the slope to be monitored to obtain several matching degrees; and take the optimized resource combination sequence corresponding to the maximum matching degree as the recommended monitoring scheme for the slope to be monitored.

8. A scheduling device for slope monitoring resources, characterized in that, It includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the slope monitoring resource scheduling method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the method for scheduling slope monitoring resources as described in any one of claims 1-6.

Images