CN118070682A

CN118070682A - Spiral bolt hanging ring damage assessment method and device based on artificial intelligence

Info

Publication number: CN118070682A
Application number: CN202410473020.4A
Authority: CN
Inventors: 王秀刚; 杜大平; 张体学; 仇恒臣; 周长磊; 张来星; 秦威; 张科; 袁伟华; 孙传东; 高琰; 王涛
Original assignee: Shandong Shenli Rigging Co ltd
Current assignee: Shandong Shenli Rigging Co ltd
Priority date: 2024-04-19
Filing date: 2024-04-19
Publication date: 2024-05-24

Abstract

The invention provides a spiral bolt hanging ring damage assessment method and device based on artificial intelligence, which relate to the technical field of data processing and comprise the following steps: and after the target vector is determined by identifying multi-dimensional vector data of the target spiral bolt hanging ring, classifying the target vector by a pre-constructed damage evaluation model, and determining the damage condition of the target spiral bolt hanging ring based on a corresponding classification result. The multi-dimensional vector data at least comprises lifting ring use data and lifting ring self-sensing monitoring data, the model is trained through a multi-dimensional training sample, and the classifier is updated in parameters based on an acceleration multi-direction whale optimization algorithm.

Description

Spiral bolt hanging ring damage assessment method and device based on artificial intelligence

Technical Field

The invention relates to the technical field of data processing, in particular to a screw bolt hanging ring damage assessment method and device based on artificial intelligence.

Background

In modern industrial applications, the integrity of the screw bolt slings is critical to ensure safe operation of equipment and structures. The occurrence of damage may lead to a decrease in structural performance and even to serious safety accidents. Thus, timely and accurate detection and assessment of the presence and severity of damage is an important task to ensure industrial safety. However, the traditional damage detection method often depends on manpower, and is not only low in efficiency, but also limited in accuracy, and the method is difficult to meet the requirements of the modern industry on efficient and automatic damage detection.

With the advent of the 4.0 era of industry, artificial intelligence technology, particularly rapid development of machine learning and deep learning, has provided new possibilities for damage detection. By utilizing a large amount of data collected by the high-precision sensor, the artificial intelligence model can learn the influence of damaged features and corresponding influence, so that automatic evaluation and classification of damage are realized.

However, artificial intelligence still faces challenges in processing high-dimensional vector-type data, particularly multi-dimensional sensor data related to equipment damage detection.

Disclosure of Invention

Therefore, the invention aims to provide the artificial intelligence-based spiral bolt hanging ring damage assessment method and the artificial intelligence-based spiral bolt hanging ring damage assessment device, which can process multidimensional data, improve identification accuracy and meet damage condition detection requirements.

In a first aspect, an embodiment of the present invention provides an artificial intelligence-based method for evaluating damage to a spiral bolt lifting ring, including collecting multidimensional vector data of a target spiral bolt lifting ring; the multidimensional vector data at least comprises lifting ring use data and lifting ring self-sensing monitoring data; identifying the multi-dimensional vector data and determining a target vector in the multi-dimensional vector data; inputting the target vector into a pre-constructed damage evaluation model, classifying the target vector, and determining a classification result; the damage evaluation model is constructed based on a preset classifier, and a training sample set for training the classifier comprises a multi-dimensional training sample and a sample label; the classifier updates parameters based on an acceleration multi-direction whale optimization algorithm; and determining the damage condition of the target spiral bolt hanging ring based on the classification result.

In a second aspect, an embodiment of the present invention provides an artificial intelligence based device for evaluating damage to a lifting ring of a screw bolt, including: the data acquisition module is used for acquiring multidimensional vector data of the target spiral bolt hanging ring; the multidimensional vector data at least comprises lifting ring use data and lifting ring self-sensing monitoring data; the data processing module is used for identifying the multi-dimensional vector data and determining a target vector in the multi-dimensional vector data; the execution module is used for inputting the target vector into a pre-constructed damage evaluation model, classifying the target vector and determining a classification result; the damage evaluation model is constructed based on a preset classifier, and a training sample set for training the classifier comprises a multi-dimensional training sample and a sample label; the classifier updates parameters based on an acceleration multi-direction whale optimization algorithm; and the output module is used for determining the damage condition of the target spiral bolt hanging ring based on the classification result.

The embodiment of the invention has the following beneficial effects: according to the artificial intelligence-based screw bolt lifting ring damage assessment method and device, the damage assessment model is used for classifying multidimensional vector data to determine the damage condition of the target screw bolt lifting ring, the multidimensional vector data at least comprise lifting ring use data and lifting ring self-sensing monitoring data, a classifier for constructing the damage assessment model is used for updating parameters based on an acceleration multidirectional whale optimization algorithm, classification accuracy and robustness can be improved, and based on the method, the multidimensional data can be processed, recognition accuracy can be improved, and damage condition detection requirements can be met.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flowchart of a method for evaluating damage to a lifting ring of a screw bolt based on artificial intelligence according to an embodiment of the present invention;

FIG. 2 is a flowchart of another artificial intelligence based method for evaluating damage to a lifting ring of a screw bolt according to an embodiment of the present invention;

FIG. 3 is a flowchart of a third artificial intelligence based method for evaluating damage to a lifting ring of a screw bolt according to an embodiment of the present invention;

FIG. 4 is a flowchart of a fourth artificial intelligence based method for evaluating damage to a lifting ring of a screw bolt according to an embodiment of the present invention;

FIG. 5 is a flowchart of a fifth artificial intelligence based method for evaluating damage to a lifting ring of a screw bolt according to an embodiment of the present invention;

Fig. 6 is a schematic structural diagram of an artificial intelligence-based screw bolt hanging ring damage assessment device according to an embodiment of the present invention;

FIG. 7 is a schematic structural diagram of another artificial intelligence-based screw bolt hanging ring damage assessment device according to an embodiment of the present invention;

Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

For the purposes of clarity, technical solutions, and advantages of the embodiments of the present disclosure, the following description describes embodiments of the present disclosure with specific examples, and other advantages and effects of the present disclosure will be readily apparent to those skilled in the art from the disclosure herein. It will be apparent that the described embodiments are merely some, but not all embodiments of the present disclosure. The disclosure may be embodied or practiced in other different specific embodiments, and details within the subject specification may be modified or changed from various points of view and applications without departing from the spirit of the disclosure. It should be noted that the following embodiments and features in the embodiments may be combined with each other without conflict. All other embodiments, which can be made by one of ordinary skill in the art without inventive effort, based on the embodiments in this disclosure are intended to be within the scope of this disclosure.

It is noted that various aspects of the embodiments are described below within the scope of the following claims. It should be apparent that the aspects described in this disclosure may be embodied in a wide variety of forms and that any specific structure and/or function described in this disclosure is illustrative only. Based on the present disclosure, one skilled in the art will appreciate that one aspect described in this disclosure may be implemented independently of any other aspects, and that two or more of these aspects may be combined in various ways. For example, apparatus may be implemented and/or methods practiced using any number of the aspects set forth in this disclosure. In addition, such apparatus may be implemented and/or such method practiced using other structure and/or functionality in addition to one or more of the aspects set forth in the disclosure.

It should also be noted that the illustrations provided in the following embodiments merely illustrate the basic concepts of the disclosure by way of illustration, and only the components related to the disclosure are shown in the illustrations, rather than being drawn according to the number, shape and size of the components in actual implementation, and the form, number and proportion of the components in actual implementation may be arbitrarily changed, and the layout of the components may be more complex. In addition, in the following description, specific details are provided in order to provide a thorough understanding of the examples. However, it will be understood by those skilled in the art that aspects may be practiced without these specific details.

Machine learning is a sub-area of Artificial Intelligence (AI) that gives computer systems the ability to learn empirically and improve their performance without requiring explicit programming. This means that the machine learning model can automatically identify patterns and rules in the data and use this knowledge to make decisions or predictions. The following are some technical data of machine learning relevant to the present invention:

Data driving: machine learning relies on large amounts of data to train a model. The data may be labeled (supervised learning), unlabeled (unsupervised learning), or obtained through interaction with the environment (reinforcement learning).

And (3) model: in machine learning, models are mathematical representations of real world problems that can be learned from data. Common models include decision trees, neural networks, support vector machines, and the like.

Learning algorithm: the learning algorithm defines how the model is adjusted or "learned". The algorithm runs on the data and optimizes according to the behavior of the model, for example, by minimizing the difference between the predicted and actual results.

Evaluation: the performance of the model needs to be measured by some form of evaluation. This typically involves dividing the data into a training set for learning and a test set for evaluating the ability of the model to generalize to unseen data.

Overfitting and underfilling: overfitting means that the model performs well on training data, but cannot be generalized to new data. Under-fitting means that the model does not learn enough features of the data to be able to predict efficiently.

And (3) supervised learning: this type of machine learning involves training a model using a set of labeled samples, i.e., input data paired with expected output. Common supervised learning tasks include classification and regression.

The spiral bolt hanging ring damage assessment method and device based on artificial intelligence provided by the embodiment of the invention can process multidimensional data, can improve identification accuracy and can meet equipment damage detection requirements.

For the understanding of the present embodiment, first, a detailed description is given of a method for evaluating damage to a screw bolt lifting ring based on artificial intelligence disclosed in the embodiment of the present invention, and fig. 1 shows a flowchart of the method for evaluating damage to a screw bolt lifting ring based on artificial intelligence provided in the embodiment of the present invention, and as shown in fig. 1, the method includes the following steps:

And S102, collecting multidimensional vector data of the target spiral bolt hanging ring.

Step S104, the multi-dimensional vector data are identified, and target vectors in the multi-dimensional vector data are determined.

And S106, inputting the target vector into a pre-constructed damage evaluation model, classifying the target vector, and determining a classification result.

And step S108, determining the damage condition of the target spiral bolt hanging ring based on the classification result.

In specific implementation, the multi-dimensional vector data at least comprises lifting ring use data and lifting ring self-sensing monitoring data, and the embodiment of the invention classifies the multi-dimensional vector data through a damage evaluation model so as to determine the damage condition of the target spiral bolt lifting ring. The damage evaluation model is constructed based on a preset classifier, and a training sample set for training the classifier comprises a multi-dimensional training sample and a sample label; moreover, the classifier updates parameters based on an acceleration multi-direction whale optimization algorithm, so that the classification precision and robustness can be improved; therefore, the embodiment of the invention not only can process multidimensional data, but also can improve the identification precision and meet the detection requirement of the damaged state.

Further, in the embodiment of the invention, the characteristic extraction is carried out on the multidimensional vector data through a pre-constructed characteristic extraction model, and the target characteristic is determined; and performing dimension reduction on the target features through a pre-constructed dimension reduction model to obtain target vectors in the multi-dimension vector data. Among other things, the prior art may employ traditional feature extraction and dimension reduction methods that may not adequately capture the complexity and subtle differences of damaged features. Moreover, traditional model training and optimization methods may be too simple or static, difficult to cope with complex impairment detection tasks, and prone to falling into locally optimal solutions. The feature extraction model of the embodiment of the invention is constructed after updating network parameters based on a dynamic sequence optimization algorithm.

In specific implementation, the embodiment of the invention utilizes a preset training sample set to train the feature extraction model, and adopts a dynamic sequence optimization neural network model to extract the features. Traditional neural network optimization methods, such as gradient descent algorithms, rely on gradient information of the loss function to update network parameters. This approach tends to fall into a local optimum in the face of complex loss planes or non-convex optimization problems. The invention adopts a dynamic sequence optimization neural network to extract characteristics, is inspired by the dynamic balance principle of an ecological system in nature, and in the ecological system, the number and the distribution of species can be self-regulated according to the change of environmental conditions and resources so as to achieve a dynamic balance state. The dynamic sequence optimization neural network algorithm simulates the natural phenomenon, and the optimization of network performance is realized by dynamically adjusting the sequence of the neural network parameters. The dynamic sequence optimization neural network algorithm not only enables the update of network parameters to depend on the current gradient information through simulating a dynamic adjustment mechanism of an ecological system, but also considers the historical sequence of parameter update and the performance of the network under different parameter configurations, thereby increasing the diversity of parameter space exploration and reducing the possibility of sinking into local optimum.

Specifically, the training steps of the dynamic sequence optimization neural network model in the embodiment of the invention comprise an initialization stage, an evaluation stage, a sequence adjustment stage, a parameter updating stage, an environment disturbance stage, a termination condition checking and outputting stage. On the basis of the above embodiment, the embodiment of the present invention further provides another method for evaluating damage to a screw bolt lifting ring based on artificial intelligence, which describes a method for constructing a feature extraction model, and fig. 2 shows a flowchart of another method for evaluating damage to a screw bolt lifting ring based on artificial intelligence, provided by the embodiment of the present invention, as shown in fig. 2, and the method includes the following steps:

step S202, a preset neural network is obtained, and network parameters and parameter updating sequences of the neural network are initialized randomly.

When the model is specifically implemented, the model is initialized first, and then the model training is performed by using a training sample set. An initialization stage: network parameters and a parameter update sequence are randomly initialized, wherein the parameter update sequence is a parameter set arranged in time sequence and is used for recording and simulating a historical track of parameter update. Specifically, parameters of the neural networkAnd parameter update sequence/>Is randomly initialized, parameter/>Including the weight/>And bias/>Can be expressed as:

further, parameter update sequences Initialized to contain/>The set of randomly generated parameters can be expressed as:

wherein, For initialized neural network parameters,/>And/>For the weight and bias of the initialized neural network, each/>Is a randomly selected parameter state.

Step S204, forward propagation is carried out on a preset training sample set through a neural network, and a network performance index is calculated.

Further, the initialized neural network is used for forward propagation of the training sample set, and the current neural network is evaluated. Evaluation phase: and under the current network parameters, forward propagation is carried out on the training set, and the performance index of the current network is calculated. In one embodiment, the performance index uses the loss function value, then the current network parameters are calculatedLower loss functionThe manner of (a) can be expressed as:

wherein, Is the number of samples,/>Is the actual tag,/>Is the predicted output of the neural network model, and is obtained by a preset Softmax function.

Step S206, based on the network performance index and the historical performance of the parameter updating sequence, the parameters in the parameter updating sequence are adjusted by adopting a rule-based method.

Further, model optimization is performed according to the evaluation result, wherein the embodiment of the invention takes the historical sequence of parameter updating and the performance of the network under different parameter configurations into consideration for model optimization. In a specific implementation, the parameter update sequence is updated by: determining a plurality of parameter sets of a parameter updating sequence, and calculating the resonance intensity of the parameter sets and the oscillator based on the quantum entropy of the parameter state of each parameter set; and updating the parameter updating sequence based on the resonance intensity and a preset dynamic adjustment function. The dynamic adjustment function comprises an exclusion strategy and an introduction strategy, wherein the exclusion strategy excludes the parameter set based on the diversity of the parameter set, and the introduction strategy generates a new parameter set by adding random disturbance into the parameter set.

Specifically, the sequence adjustment phase: according to the current performance index and parameter updating sequence, a rule-based method is adopted to adjust parameters in the sequence, a mechanism of species in an ecological system for adjusting self behavior according to environmental changes is simulated, and the parameter sequence is updatedTo reflect the current environment. Specifically, a dynamic adjustment function/>, is definedAccording to the current loss/>And sequence/>To update the sequence, may be expressed as:

wherein, For the parameter sequence before adjustment,/>For the adjusted parameter sequence,/>Dynamically adjusting the function/>, for the neural network parameters of the t-th iterationPossibly including excluding the worst performing parameter set, employing a new random parameter set, etc., to increase diversity. /(I)Is a super-parameter controlling the intensity of the influence of resonance effect,/>Is the average of all parameter sets in the sequence. /(I)Is the resonance intensity.

Further, for each set of parameters in the sequenceCalculate its resonant intensity with the oscillator/>The manner of (a) can be expressed as:

wherein, Is the parameter set/>Performance index of/>Is a super-parameter for regulating resonance sensitivity,/>Is the oscillator state variable at the t-th iteration. /(I)Is a super parameter controlling the influence of quantum entropy,/>Is the quantum entropy.

Further, the training process of the neural network is guided by defining the quantum entropy of the parameter states, enhancing the ability of the algorithm to explore unknown parameter spaces while maintaining efficient global searching. Specifically, each parameter is assembledConsidered as a quantum state, its quantum entropy/>The calculation of (2) can be expressed as:

wherein, Is the parameter set/>In/>The probability amplitude in each dimension simulates the probability nature of a quantum state.

Further, for each parameter set, the way its probability amplitude in each dimension is calculated can be expressed as:

wherein, Is a normalization factor, ensuring that the sum of all probability magnitudes is 1; /(I)And/>All the parameters are respectively assembled in the/>Mean and variance across the dimensions.

Further, oscillator state variablesAs network performance dynamically adjusts, the adjustment mode can be expressed as:

wherein, For the adjusted oscillator state variable,/>In order to adjust the state variable of the oscillator before,Representing the change in the loss function of two successive iterations,/>Is the learning rate for adjusting the response speed of the oscillator.

In one embodiment, the function is dynamically adjustedIs decomposed into two main parts: exclusion strategy/>And import strategy/>. Specifically, for the exclusion strategy/>For each set of parameters in the sequence/>Calculate an exclusion score/>As the probability that it is excluded, it can be expressed as:

wherein, And/>Is a hyper-parameter that balances loss and sequence diversity. In one embodiment,/>AndIs set to 0.7 and 0.3. /(I)Representation sequence/>The variance of the medium parameters is used to measure diversity.

Further, for the introduction strategyThe process of introducing a new set of parameters can be accomplished by adding random perturbations to a subset of the existing parameters in the sequence, which can be expressed as:

wherein, Is from/>Is a randomly selected set of parameters,/>Is a super-parameter that controls the intensity of the disturbance,Mean value is 0, variance is/>Is a normal distribution of (c). In one embodiment,/>Is set to 0.1.

Further, sequence diversityThe calculation of (2) can be expressed as:

wherein, Is the sequence/>The mean value of all parameter sets in (a) is calculated as follows:

Step S208, selecting a target parameter set from the parameter updating sequence, and performing iterative training by taking the target parameter set as a new network parameter of the neural network; and introducing random variation in the iterative training process to update the network parameters of the neural network.

Further, after updating the parameter updating sequence, updating the network parameter, wherein for the parameter updating stage, the embodiment of the invention is realized by the following steps: a set of parameters is selected from the adjusted sequence of parameters as new parameters for the network. Specifically, from the adjusted sequenceIn selecting a parameter set/>As a next round of network parameters, the selection is based on a hybrid strategy, and the historical performance and diversity of the parameter set are considered, so that the updating mode of the parameters can be expressed as follows:

wherein, The function may select a set of parameters based on the weighted scores for performance and diversity.

In one embodiment, the function is selectedSelecting optimal parameter set/>By calculating fitness score/>, for each set of parametersAnd selecting the highest scoring parameter set to achieve, can be expressed as:

wherein, Is a superparameter for balancing the effects of loss and diversity, in one embodiment,/>May be set to 0.5. /(I)Representing parameter set/>And sequence/>Average differences among other parameter sets.

Further, the differenceRepresentation/>Average distance from other parameter sets in the sequence for evaluation/>The way in which it is calculated can be expressed as:

wherein, Is the L1 norm.

Further, the embodiment of the invention also introduces random variation to update network parameters, wherein the random variation corresponds to an environmental disturbance stage: specifically, small random changes are introduced in each iteration, and random events in the natural environment are simulated to increase the exploration range of the parameter space. In order to simulate the environmental disturbance, the invention introduces a random noiseTo the current parameters/>The way the parameters are further updated can be expressed as:

wherein, For simulating the neural network parameters updated by the environmental disturbance,/>Is a small random vector used to increase exploratory properties. /(I)Is the resonance intensity.

Further, random noiseThe calculation of (2) can be expressed as:

wherein, Is a super parameter controlling noise intensity,/>Expressed at/>Random variables uniformly distributed over the range. In one embodiment,/>Set to 0.05.

Further, random variablesThe calculation of (2) can be expressed as:

wherein, Is/>Uniform random variable within range, i.e./>。

And step S210, constructing a feature extraction model based on the neural network until the neural network meets preset iteration conditions.

Further, the condition check is terminated: it is determined whether an optimized termination condition is met and in one embodiment, the iteration is stopped when the iteration reaches 1000 times. Further, the optimized neural network parameters and performance metrics are output, i.e., the optimized parameters are finally outputAnd minimum loss/>. Based on this, a feature extraction model is constructed. The embodiment of the invention also identifies the multidimensional vector data of the target spiral bolt hanging ring through the characteristic extraction model to obtain the characteristic extraction parameters, so that the damage evaluation model classifies the data to perform damage evaluation on the target spiral bolt hanging ring.

According to the artificial intelligence-based spiral bolt hanging ring damage assessment method provided by the embodiment of the invention, the dynamic balance principle of a natural ecological system is used for reference, a dynamic sequence optimization strategy is adopted, a self-adjustment mechanism of the ecological system is simulated, the sequence of the neural network parameters is dynamically adjusted to optimize the model, the diversity of parameter space exploration is increased, the risk of sinking into local optimum is reduced, the adaptability and the optimization efficiency of the model on complex data are improved, and in addition, the diversity of parameter space exploration is increased and the risk of sinking into local optimum is reduced by introducing the resonance intensity and quantum entropy concepts of the parameter sequence.

On the basis of the embodiment, the embodiment of the invention also carries out data feature dimension reduction on the feature extraction parameters through the dimension reduction model so as to improve the efficiency and effect of feature extraction. Specifically, the invention adopts a self-coding neural network algorithm based on random mapping to perform data feature dimension reduction so as to fully capture the complexity and subtle difference of damaged features. In one embodiment, the data after feature extraction can be input into the feature dimension reduction model to train the feature dimension reduction model. The conventional self-encoder structure consists of two parts: an encoder and a decoder. The function of the encoder is to map the high-dimensional data to a low-dimensional feature space, and the decoder tries to reconstruct the original data from the low-dimensional space. In addition, the importance of each node in the network can be dynamically adjusted through the participation degree limiting mechanism, and the efficiency and effect of feature extraction are further improved.

Based on this, the embodiment of the invention further provides a third artificial intelligence-based screw bolt lifting ring damage assessment method to explain a method for constructing a dimension reduction model, and fig. 3 shows a flowchart of the third artificial intelligence-based screw bolt lifting ring damage assessment method provided by the embodiment of the invention, as shown in fig. 3, and the method includes the following steps:

Step S302, introducing a random mapping layer to a preset self-encoder, and initializing parameters of the random mapping layer.

Step S304, carrying out random nonlinear transformation on a preset training sample set through a random mapping layer to obtain mapping layer output.

Before training begins, parameters of the random mapping layer are initialized, and the parameters remain unchanged during training. The random mapping layer is used for carrying out random nonlinear transformation on the data before the data enters the encoder, and setting a random mapping matrix asWherein/>Representing the feature dimension after random mapping,/>Is a feature dimension symbol. For each pre-processed sample/>Its output after passing through the random mapping layer is/>The calculation method is as follows:

wherein, Representing the Tanh activation function.

In one embodiment, the matrix is randomly mappedThe computation of (a) is based on random initialization of gaussian distribution for a matrixEach element/>The calculation formula is as follows:

wherein, Mean value is 0, variance is/>Is a gaussian distribution of (c).

In step S306, the mapping layer output is input to the self-encoder, and the self-encoder is trained.

The embodiment of the invention dynamically adjusts the participation degree of the hidden layer according to the output of the hidden layer node of the self-encoder. In a specific implementation, the data passes through the random mapping layer and then enters the encoder, which maps the data to a low-dimensional feature space, and then is reconstructed by the decoder. The purpose of the encoder is to randomly map the dataEncoding as low-dimensional feature space/>Wherein/>Is the encoded feature dimension. Let the weight matrix and bias vector of the encoder be respectivelyAnd/>The encoding process may be expressed as:

The task of the decoder is to extract the low-dimensional feature space from Reconstruction data/>To be as close as possible to the original data/>, after random mapping. Let the weight matrix and bias vector of decoder be/>, respectivelyAnd/>The decoding process can be expressed as:

Step S308, calculating the corresponding current loss of the self-encoder, and updating the network parameters of the self-encoder based on the current loss.

Specifically, the loss between the reconstructed data from the encoder and the original data is calculated and the network parameters are updated by a back propagation algorithm. The self-encoder of the present invention trains the network by minimizing reconstruction errors and regularization terms, loss functionsThe definition is as follows:

wherein, Representing reconstruction errors,/>Representing regularization term,/>Is a regularization coefficient.

Further, encoder weightsAnd bias/>Is achieved by a back propagation algorithm by which the network parameters/>Participation weight/>Is iteratively updated to minimize the loss function/>. In one embodiment, regularization coefficient/>Is set to 0.01.

To be used forFor example, the update method can be expressed as:

wherein, To update the post-weight,/>To update the pre-weights,/>Is learning rate,/>Is a loss function/>For a pair ofCan be further developed by the chain law as: /(I)

Similarly, biasThe updated formula of (2) is:

wherein, The calculation can be performed by a similar chain law.

In each iteration, the participation degree of the hidden layer node is dynamically adjusted according to the output of the hidden layer node, and the influence of unimportant nodes is reduced in a manner of participation degree limiting adjustment. Specifically, the participation limitation is realized by dynamically adjusting the weight of each node, and the node is setThe output of (2) is/>Its participation weight is/>The adjusted node outputs are:

in one embodiment, the update of engagement weights depends on the statistical properties of the node outputs, which are the variances of the outputs The manner in which the engagement weight is calculated can be expressed as:

wherein, And/>Is a super parameter for controlling the adjustment sensitivity and threshold of engagement. In one embodiment,/>And/>Set to 5 and 0.1.

Further, in the present embodiment, the variance of the node outputCan be expressed as:

wherein, Is the number of samples in a batch,/>Is node/>For/>Output of individual samples,/>Is node/>The average of all sample outputs in the current batch can be expressed as:

Step S310, until the self-encoder meets the preset training requirement, a dimension reduction model is built based on the self-encoder.

And repeating the steps until the preset iteration times are met, constructing a dimension reduction model, and further, processing the characteristic extraction parameters by using the dimension reduction model to determine a target vector. The feature extraction parameters are obtained by identifying the multi-dimensional vector data by the feature extraction model.

In summary, according to the third artificial intelligence-based spiral bolt hanging ring damage assessment method provided by the invention, a random mapping layer is introduced in front of a self-encoder, so that the dependence of an initial structure of data is broken, the generalization capability of a model on unknown data is enhanced, meanwhile, the importance of network nodes is dynamically adjusted through a participation degree limiting mechanism, and the efficiency and effect of feature extraction are optimized.

Further, based on the above embodiment, the embodiment of the present invention further provides a fourth artificial intelligence-based method for evaluating damage to a lifting ring of a screw bolt, which describes steps for constructing a training sample set. It can be understood that in the task of the invention, acquisition, labeling and preprocessing of training data are time-consuming and labor-consuming, and insufficient training samples easily result in poor generalization capability of the model, and simultaneously influence the accuracy of the model.

The conventional method may face the problem of insufficient training data, especially in terms of damaged type diversity and sample number, limiting generalization ability and accuracy of the model. The invention adopts the generation countermeasure network based on projection quantum coding to generate samples, thereby realizing data expansion. In the traditional generation countermeasure network, the generator and the discriminator improve the quality of data generation through mutual competition, and in the generation countermeasure network based on projection quantum coding, input data is firstly converted into quantum bits through a quantum coding process, so that the data obtains multi-state expression capability in a quantum space, and the expression and processing capability of the data are greatly enhanced. In addition, the data after quantum coding effectively maps high-dimensional quantum information to a low-dimensional space which can be processed by a generation network through specially designed quantum projection operation, so that the efficiency and quality of a data generation process are ensured. In order to improve the performance of the generation countermeasure network based on projection quantum coding, the invention provides a loss function based on quantum state similarity, which can evaluate the difference between generated data and real data more accurately. The generation countermeasure network based on projection quantum coding not only can generate more various and real spiral bolt hanging ring damage data, but also has obvious improvement on the efficiency and quality of data generation.

Specifically, fig. 4 shows a flowchart of a fourth artificial intelligence-based method for evaluating damage to a screw bolt hanging ring according to an embodiment of the present invention, as shown in fig. 4, including the following steps:

and step S402, collecting multi-dimensional data of the spiral bolt hanging ring in a preset environment to obtain a multi-dimensional training sample.

And step S404, labeling the multi-dimensional training samples, and constructing an initial sample set.

The multidimensional data comprises lifting ring use data and lifting ring self-sensing monitoring data. In the invention, when the screw bolt hanging ring damage is detected, the data of the training screw bolt hanging ring damage assessment model is collected in the high-precision sensor, the collected data is vector data, and in one embodiment, the attribute of the data comprises: stress value (a 1): the stress of the hanging ring at a specific time point is represented; temperature value (a 2): reflecting the temperature condition of the hanging ring; vibration frequency (a 3, a 4): the cracking refers to the vibration frequency of the hanging ring; vibration amplitude (a 5): the vibration amplitude of the hanging ring; use time (a 6): reflecting the used time length of the hanging ring; load (a 7): the load born by the hanging ring is represented; corrosion degree (a 8): quantifying the corrosion condition of the surface of the lifting ring; ambient humidity (a 9): the humidity of the environment where the hanging ring is located is represented; ambient temperature (a 10): the temperature of the environment where the hanging ring is positioned is represented;

Further, the collected data is marked in a manual marking manner, and in one embodiment, the marked categories include 3 categories of micro-damage, moderate damage and serious damage, which correspond to labels 0, 1 and 2 respectively.

Step S406, the initial sample set is standardized and then converted into quantum bits, and quantum projection is carried out on the quantum bits, so that low-dimensional space quantum data is obtained.

In a specific implementation, in order to enhance the expression and processing capability of data, the input data is first converted into quantum bits through a quantum encoding process, so that the data obtains multi-state expression capability in a quantum space. Specifically, firstly, the original data of the damaged spiral bolt hanging ring is subjected to standardized processing, so that the data is ensured to be in a state suitable for training. Providing an original screw bolt flying ring damaged data setWherein/>Represents the/>The data sample is firstly subjected to data preprocessing, the data is standardized to a certain range, and the standardized mode can be expressed as follows: /(I)

Wherein,And/>Representing the maximum and minimum values in the dataset,/>, respectivelyIs a standardized function.

Data after normalizationConversion into a qubit representation by quantum encoding, let/>Is a quantum encoding function, then the quantum encoding process is expressed as:

wherein, Is/>Qubit representation of data samples,/>Representing superposition of qubits,/>For/>And/>Quantum entanglement between individual features.

Further, the firstAnd/>Quantum entanglement between individual features/>The calculation of (2) can be expressed as:

wherein, And/>Is the entanglement coefficient calculated based on the original data.

Further, entanglement coefficientAnd/>The computation of (2) involves a correlation analysis between the original data features, the computation can be expressed as:

wherein, Is a super parameter for controlling entanglement strength and is preset by human beings.

In one embodiment, the quantum encoding functionThe way classical data is converted into qubits can be expressed as:

wherein, Representing superposition of qubits,/>Is the number of qubits,/>Is the angular parameter of classical data to qubit mapping, and the calculation mode can be expressed as:

In one embodiment, the number of qubits Is set to 5.

Step S408, performing iterative training on a preset generation countermeasure network through low-dimensional space quantum data to generate an expansion sample; calculating quantum state similarity between the extended sample and the low-dimensional spatial quantum data updates network parameters of the generated countermeasure network.

In particular, iterative training of generators and discriminators is performed, in particular, in a projection quantum encoding-based generation countermeasure network, the generatorsAttempt to generate and real data/>Data similar, discriminant/>Attempting to differentiate between generated dataAnd real data/>，/>The iterative update process of the generator and the arbiter, which is random noise sampled from the normal distribution, can be expressed by the following formula:

wherein, Is a cost function of the arbiter and generator,/>Is the distribution of real data,/>Is the distribution of noise.

Further, a cost functionIn/>Items and itemsThe manner in which an item is calculated can be expressed as:

wherein, Is the number of real data samples,/>Is the number of generated data samples,/>Is/>And the generated noise samples.

In each iteration, quantum state similarity between the generated data and the real data is calculated and used to adjust parameters of the generator and the arbiter. Wherein, the invention introduces a quantum state similarity loss functionAnd the similarity between the generated data and the real data is measured.

In specific implementation, respectively calculating amplitude components of the extended sample and the low-dimensional space quantum data on the quantum bit; calculating an inner product of the extended sample and the low-dimensional spatial quantum data based on the amplitude component of the extended sample and the amplitude component of the low-dimensional spatial quantum data; and inputting the inner product into a preset quantum state similarity loss function, and determining the quantum state similarity between the extended sample and the low-dimensional space quantum data.

The quantum state similarity loss function can be expressed as:

wherein, Representing the quantum state/>And/>The inner product between the two quantum states is used for measuring the similarity of the two quantum states.

Further, the inner productThe calculation of (2) can be expressed as: /(I)

Wherein,And/>Respectively quantum state/>And/>In/>The magnitude component on the individual qubits can be calculated as:

step S410, decoding the extended samples meeting the extension requirement, and combining the decoded extended samples with the initial sample set to construct a training sample set.

After multiple iterations, the generated quantum encoded data is decoded back into classical data format to form final enhanced data set, quantum decoding functionThe definition is as follows:

wherein, Is the generator/>The resulting decoded data samples.

In one embodiment, the quantum decoding function is obtained by measuring the quantum statesRealization, measurement of ground state/>Or (b)Specifically, the measurement result is/>The probability of (1) is/>Measurement result is/>The probability of (1) is/>The quantum decoding function may be expressed as a probabilistic process, which may be expressed as:

and converting the quantum bits back to a classical data format through a decoding process, and obtaining the expanded sample.

Further, a training sample set is constructed based on the current expanded sample and the initial sample set for model training. The method can also use traditional data expansion technology, such as scaling, noise injection and the like, to perform simpler and easily-realized data expansion.

According to the fourth artificial intelligence-based spiral bolt hanging ring damage assessment method provided by the embodiment of the invention, a quantum computing principle is introduced, data is converted into quantum bits through quantum coding, the data representation is enhanced by utilizing the multi-state expression capability of a quantum space, and the expression and processing capability of the data are greatly enhanced. Meanwhile, the specially designed quantum projection operation and the loss function based on quantum state similarity are used for optimizing the generated data, high-quality data are generated, consistency between the generated data and real data is improved, effective data expansion is achieved, and the problem of mode collapse possibly faced by a traditional generation countermeasure network is effectively solved.

Furthermore, on the basis of the embodiment, the embodiment of the invention also provides a fifth artificial intelligence-based screw bolt hanging ring damage assessment method, which is used for explaining the construction steps of the damage assessment model. Wherein, the existing classifier may not fully utilize the optimization strategy to improve the accuracy and robustness of classification. The embodiment of the invention optimizes the parameters of the high-order neural network classifier based on the acceleration multi-directional whale optimization algorithm, and combines the advantages of the acceleration multi-directional whale optimization algorithm and the Riemann manifold learning, so that the classifier has higher precision and robustness when processing complex data.

In specific implementation, the embodiment of the present invention constructs the damage assessment model by using a preset training sample set, and fig. 5 shows a flowchart of a fifth artificial intelligence based screw bolt hanging ring damage assessment method provided by the embodiment of the present invention, as shown in fig. 5, including the following steps:

Step S502, a preset training sample set is obtained, and after feature extraction is carried out on the training sample set, data dimension reduction is carried out, so that a target training set is obtained.

In specific implementation, the training sample set of the embodiment can be trained by sequentially performing feature extraction and dimension reduction, and then inputting the dimension reduced data into the classifier.

Step S504, network parameters of the higher order neural network are initialized randomly, and points on the Riemann manifold are initialized based on the target training set.

Step S506, mapping the target training set to the Riemann manifold, and extracting deep features of the target training set by utilizing the geometric structure on the Riemann manifold.

In specific implementation, the embodiment of the invention uses the deep features to train the classifier after extracting the deep features of the target training set. Wherein deep feature extraction is performed using geometry on the Riemann manifold by mapping the target training set onto the Riemann manifold.

Firstly, initializing a model, wherein the weight and the bias parameters of a higher-order neural network are randomly initialized, and simultaneously, initializing points on a Riemann manifold according to the reduced-dimension data. Let the data set after dimension reduction be，/>Is the number of samples. Each data/>By mapping function/>Mapped to Riemann manifold/>Can be expressed as:

wherein, 。

In one embodiment, the higher order neural network employs three hidden layers, with the number of neurons in each layer set to 128, 64, and 32, respectively.

Further, riemann feature extraction is performed, the data is mapped onto a Riemann manifold, and deep features are extracted using geometry on the manifold. Specifically, it is provided withTo be from/>The extracted feature vector is:

wherein, Is a feature extraction function on the Riemann manifold.

And step S508, classifying and training the deep features through the high-order neural network, calculating a loss function, and optimizing network parameters of the high-order neural network by using an acceleration multi-direction whale optimizing algorithm.

In a specific implementation, the method comprises the following steps: 1) Calculating a candidate solution corresponding to the optimal solution based on the current optimal solution of the higher-order neural network so as to simulate the behavior of whale surrounding the prey and obtain a first updated parameter; 2) Simulating whale to approach the first updating parameters through spiral ascending, and determining second updating parameters; 3) Randomly selecting a search point to obtain a third updating parameter; 4) And performing performance evaluation on the first updating parameter, the second updating parameter and the third updating parameter, and determining the parameter with the best performance as the final updating parameter.

Specifically, the embodiment of the invention optimizes the parameters of the neural network by using an accelerated multidirectional whale optimization algorithm which simulates the surrounding prey, spiral rising and random search behaviors of whales to find the optimal weight and bias. Wherein, let whale population beEach/>A set of weights and bias parameters representing a neural network,/>Is the number of whales. In one embodiment, whale population size/>Is set to 50. In a specific implementation, the above step of determining the final updated parameters is implemented by the following steps 1) -4).

1) Surrounding the prey:

The algorithm simulates whale behavior around the prey by computing candidate solutions around the current optimal solution, thus updating the parameters of the neural network, which can be expressed as:

wherein, Is the current optimal whale position,/>And/>Is a coefficient that is dynamically adjusted according to the number of iterations.

Further, dynamic coefficientsAnd/>The surrounding prey behavior for simulating whales can be calculated as:

wherein, Is a reduced linear coefficient,/>Is the current iteration number,/>Is the maximum number of iterations that can be performed,Generating a random number in the range of 0, 1. /(I)Gradually decreasing with increasing iteration number,/>Randomly during each iteration.

2) Spiral rise:

The behavior of whales approaching a prey through a spiral rise is simulated, and parameters of the neural network are further adjusted, which can be expressed as:

wherein, Is/>And/>Distance between/>And/>Is a coefficient in the algorithm.

Further, the distance in the spiral ascending behaviorSum coefficient/>Can be expressed as:

wherein, Is a coefficient scaled around the optimal solution,/>Is a constant controlling the shape of the helix,/>Generating a random number in the range of 0, 1.

3) Random search:

The algorithm simulates behavior of whales searching for prey in a large range by randomly selecting search points, increases diversity of the algorithm, avoids sinking into a local optimal solution, and can be expressed as:

/>

wherein, Is a whale position randomly selected in the current iteration.

4) Evaluation and selection:

Using loss functions Evaluate each/>Of (3), wherein/>Is the corresponding set of real tags. At the same time, select the best performing/>As/>The next iteration is entered.

Further, training the neural network through forward propagation and backward propagation according to the optimized parameters, calculating a loss function and feeding the loss function back to an acceleration multi-directional whale optimization algorithm to guide the optimization of the next round. For each iterationUpdate whale position/>To simulate the predatory behaviour of whales.

And step S510, constructing a damage evaluation model based on the higher-order neural network until the higher-order neural network meets the preset training requirement.

Specifically, the above steps are repeated until a termination condition is satisfied, in one embodiment, the termination condition is that the maximum number of iterations is reached, in this embodiment, the maximum number of iterationsSet to 100. Based on this, a damage assessment model is constructed to classify the above-mentioned target vector using the damage assessment model to determine the damage condition of the target screw eye. The classification result of the damage evaluation model comprises labels of damage conditions, such as 0, 1 and 2, corresponding to the labels, and the labels are used for indicating 3 categories of micro damage, moderate damage and serious damage, and the damage condition of the target screw bolt hanging ring can be determined according to the labels output by the model.

The invention provides a fifth artificial intelligence-based spiral bolt hanging ring damage assessment method, which is used for carrying out parameter optimization on a classifier by combining a high-order neural network and an acceleration multi-direction whale optimization algorithm, so that the classification accuracy and the robustness are improved. And meanwhile, deep feature extraction is performed by utilizing Riemann manifold learning, so that the deep feature is extracted by utilizing the geometric structure of data better, and the classification performance is enhanced.

In summary, the embodiment of the invention can obviously improve the precision of detecting the damage of the screw bolt hanging ring, and particularly obviously improve the identification of fine damage. By the application of the technologies such as random mapping and dynamic sequence optimization, the model has better generalization capability, and unseen data samples can be effectively processed. In addition, the data expansion is performed through the generation countermeasure network based on quantum coding, so that the problem of insufficient training samples is effectively solved, and the utilization rate of the data and the training efficiency of the model are improved.

Further, on the basis of the above method embodiment, the embodiment of the present invention further provides an artificial intelligence based screw bolt lifting ring damage assessment device, and fig. 6 shows a schematic structural diagram of the artificial intelligence based screw bolt lifting ring damage assessment device provided by the embodiment of the present invention, as shown in fig. 6, where the device includes: the data acquisition module 100 is used for acquiring multidimensional vector data of the target spiral bolt hanging ring; the multidimensional vector data at least comprises lifting ring use data and lifting ring self-sensing monitoring data; the data processing module 200 is configured to identify multi-dimensional vector data and determine a target vector in the multi-dimensional vector data; the execution module 300 is configured to input a target vector into a pre-constructed damage evaluation model, classify the target vector, and determine a classification result; the damage evaluation model is constructed based on a preset classifier, and a training sample set for training the classifier comprises a multi-dimensional training sample and a sample label; the classifier updates parameters based on an acceleration multi-direction whale optimization algorithm; and the output module 400 is used for determining the damage condition of the target spiral bolt hanging ring based on the classification result.

The screw bolt hanging ring damage assessment device based on the artificial intelligence provided by the embodiment of the invention has the same technical characteristics as the screw bolt hanging ring damage assessment method based on the artificial intelligence provided by the embodiment, so that the same technical problems can be solved, and the same technical effects can be achieved.

Further, based on the above embodiment, the embodiment of the present invention further provides another artificial intelligence-based screw bolt lifting ring damage assessment device, fig. 7 shows a schematic structural diagram of another artificial intelligence-based screw bolt lifting ring damage assessment device provided by the embodiment of the present invention, and as shown in fig. 7, the data processing module 200 is further configured to perform feature extraction on multidimensional vector data through a pre-built feature extraction model, so as to determine a target feature; the feature extraction model is constructed after updating network parameters based on a dynamic sequence optimization algorithm; and performing dimension reduction on the target features through a pre-constructed dimension reduction model to obtain target vectors in the multi-dimension vector data.

The data processing module 200 is further configured to obtain a preset neural network, and randomly initialize network parameters and a parameter update sequence of the neural network; forward propagation is carried out on a preset training sample set through a neural network, and a network performance index is calculated; based on the network performance index and the historical performance of the parameter updating sequence, adjusting the parameters in the parameter updating sequence by adopting a rule-based method; selecting a target parameter set from the parameter updating sequence, and taking the target parameter set as a new network parameter of the neural network for iterative training; and introducing random variation in the iterative training process to update the network parameters of the neural network; and constructing a feature extraction model based on the neural network until the neural network meets preset iteration conditions.

The data processing module 200 is further configured to determine a plurality of parameter sets of the parameter update sequence, and calculate a resonance intensity of the parameter set and the oscillator based on a quantum entropy of a parameter state of each parameter set; updating the parameter updating sequence based on the resonance intensity and a preset dynamic adjustment function; the dynamic adjustment function comprises an exclusion strategy and an introduction strategy, wherein the exclusion strategy is used for excluding the parameter set based on the diversity of the parameter set, and the introduction strategy is used for generating a new parameter set by adding random disturbance into the parameter set.

The data processing module 200 is further configured to introduce a random mapping layer to a preset self-encoder, and initialize parameters of the random mapping layer; carrying out random nonlinear transformation on a preset training sample set through a random mapping layer to obtain mapping layer output; the mapping layer output is input into a self-encoder, and the self-encoder is trained; wherein, the participation degree of the hidden layer is dynamically adjusted according to the output of the hidden layer node of the self-encoder; calculating the current loss corresponding to the self-encoder, and updating the network parameters of the self-encoder based on the current loss; and constructing a dimension reduction model based on the self-encoder until the self-encoder meets the preset training requirement.

The execution module 300 is further configured to obtain a preset training sample set, perform feature extraction on the training sample set, and perform data dimension reduction to obtain a target training set; randomly initializing network parameters of a high-order neural network, and initializing points on a Riemann manifold based on a target training set; mapping the target training set onto a Riemann manifold, and extracting deep features of the target training set by utilizing a geometric structure on the Riemann manifold; classifying and training deep features through a high-order neural network, calculating a loss function, and optimizing network parameters of the high-order neural network by utilizing an acceleration multi-direction whale optimizing algorithm; and constructing a damage evaluation model based on the high-order neural network until the high-order neural network meets the preset training requirement.

The execution module 300 is further configured to calculate a candidate solution corresponding to the optimal solution based on a current optimal solution of the higher-order neural network, so as to simulate behavior of whale surrounding a prey, and obtain a first updated parameter; simulating whale to approach the first updating parameters through spiral ascending, and determining second updating parameters; randomly selecting a search point to obtain a third updating parameter; and performing performance evaluation on the first updating parameter, the second updating parameter and the third updating parameter, and determining the parameter with the best performance as the final updating parameter.

Further, the device further comprises a construction module 500, configured to collect multidimensional data of the spiral bolt hanging ring in a preset environment, so as to obtain a multidimensional training sample; the multidimensional data comprise lifting ring use data and lifting ring self-sensing monitoring data; labeling the multi-dimensional training samples to construct an initial sample set; the method comprises the steps of carrying out standardization on an initial sample set, converting the initial sample set into quantum bits, and carrying out quantum projection on the quantum bits to obtain low-dimensional space quantum data; performing iterative training on a preset generation countermeasure network through low-dimensional space quantum data to generate an expansion sample; calculating quantum state similarity between the extended sample and the low-dimensional space quantum data, and updating network parameters for generating an countermeasure network; and decoding the extended samples meeting the extension requirements, and combining the decoded extended samples with the initial sample set to construct a training sample set.

The above construction module 500 is further configured to calculate amplitude components of the extended sample and the low-dimensional spatial quantum data on the qubit, respectively; calculating an inner product of the extended sample and the low-dimensional spatial quantum data based on the amplitude component of the extended sample and the amplitude component of the low-dimensional spatial quantum data; and inputting the inner product into a preset quantum state similarity loss function, and determining the quantum state similarity between the extended sample and the low-dimensional space quantum data.

The embodiment of the invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps of the method shown in the figures 1 to 5. The embodiments of the present invention also provide a computer readable storage medium having a computer program stored thereon, which when executed by a processor performs the steps of the method shown in fig. 1 to 5 described above.

The embodiment of the invention further provides a schematic structural diagram of an electronic device, as shown in fig. 8, where the electronic device includes a processor 81 and a memory 80, where the memory 80 stores computer executable instructions that can be executed by the processor 81, and the processor 81 executes the computer executable instructions to implement the methods shown in fig. 1 to 3.

In the embodiment shown in fig. 8, the electronic device further comprises a bus 82 and a communication interface 83, wherein the processor 81, the communication interface 83 and the memory 80 are connected by the bus 82. The memory 80 may include a high-speed random access memory (RAM, random Access Memory), and may further include a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. The communication connection between the system network element and at least one other network element is implemented via at least one communication interface 83 (which may be wired or wireless), and may use the internet, a wide area network, a local network, a metropolitan area network, etc. Bus 82 may be an ISA (Industry Standard Architecture ) Bus, PCI (PERIPHERAL COMPONENT INTERCONNECT, peripheral component interconnect standard) Bus, EISA (Extended Industry Standard Architecture ) Bus, etc., or AMBA (Advanced Microcontroller Bus Architecture, standard for on-chip buses) Bus, where AMBA defines three buses, including an APB (ADVANCED PERIPHERAL Bus) Bus, an AHB (ADVANCED HIGH-performance Bus) Bus, and a AXI (Advanced eXtensible Interface) Bus. The bus 82 may be classified as an address bus, a data bus, a control bus, or the like. For ease of illustration, only one bi-directional arrow is shown in FIG. 8, but not only one bus or type of bus.

The processor 81 may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in the processor 81 or by instructions in the form of software. The processor 81 may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU for short), a network processor (Network Processor, NP for short), and the like; but may also be a digital signal Processor (DIGITAL SIGNAL Processor, DSP), application Specific Integrated Circuit (ASIC), field-Programmable gate array (FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present application may be embodied directly in the execution of a hardware decoding processor, or in the execution of a combination of hardware and software modules in a decoding processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory, and the processor 81 reads the information in the memory and, in combination with its hardware, performs the method shown in any of the foregoing figures 1 to 5.

The embodiment of the invention provides a computer program product of a method and a device for evaluating damage to a lifting ring of a screw bolt based on artificial intelligence, which comprises a computer readable storage medium storing program codes, wherein the program codes comprise instructions for executing the method described in the previous method embodiment, and specific implementation can be seen in the method embodiment and will not be repeated here.

It will be clear to those skilled in the art that, for convenience and brevity of description, reference may be made to the corresponding process in the foregoing method embodiment for the specific working process of the above-described system, which is not described herein again. In addition, in the description of embodiments of the present invention, unless explicitly stated and limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood by those skilled in the art in specific cases.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

In the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings, are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.

Finally, it should be noted that: the above examples are only specific embodiments of the present invention for illustrating the technical solution of the present invention, but not for limiting the scope of the present invention, and although the present invention has been described in detail with reference to the foregoing examples, it will be understood by those skilled in the art that the present invention is not limited thereto: any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or perform equivalent substitution of some of the technical features, while remaining within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.

Claims

1. The artificial intelligence-based screw bolt hanging ring damage assessment method is characterized by comprising the following steps of:

Collecting multi-dimensional vector data of a target spiral bolt hanging ring; the multi-dimensional vector data at least comprises lifting ring use data and lifting ring self-sensing monitoring data;

identifying the multi-dimensional vector data and determining a target vector in the multi-dimensional vector data;

Inputting the target vector into a pre-constructed damage evaluation model, classifying the target vector, and determining a classification result; the damage assessment model is constructed based on a preset classifier, and a training sample set for training the classifier comprises a multi-dimensional training sample and a sample label; the classifier updates parameters based on an acceleration multi-direction whale optimization algorithm;

And determining the damage condition of the target spiral bolt hanging ring based on the classification result.

2. The method of claim 1, wherein the step of identifying the multi-dimensional vector data and determining the target vector in the multi-dimensional vector data comprises:

Performing feature extraction on the multi-dimensional vector data through a pre-constructed feature extraction model to determine target features; the feature extraction model is constructed after updating network parameters based on a dynamic sequence optimization algorithm;

And performing dimension reduction on the target features through a pre-constructed dimension reduction model to obtain target vectors in the multi-dimension vector data.

3. The method according to claim 2, wherein the method for constructing the feature extraction model includes:

Acquiring a preset neural network, and randomly initializing network parameters and a parameter updating sequence of the neural network;

forward propagation is carried out on a preset training sample set through the neural network, and a network performance index is calculated;

based on the network performance index and the historical performance of the parameter updating sequence, adopting a rule-based method to adjust parameters in the parameter updating sequence;

Selecting a target parameter set from the parameter updating sequence, and taking the target parameter set as a new network parameter of the neural network for iterative training; and introducing random variation in the iterative training process so as to update network parameters of the neural network;

And constructing a feature extraction model based on the neural network until the neural network meets preset iteration conditions.

4. A method according to claim 3, wherein the step of adjusting parameters in the parameter update sequence using a rule-based method based on the network performance metrics and the historical performance of the parameter update sequence comprises:

Determining a plurality of parameter sets of the parameter updating sequence, and calculating the resonance intensity of the parameter sets and an oscillator based on the quantum entropy of the parameter state of each parameter set;

Updating the parameter updating sequence based on the resonance intensity and a preset dynamic adjustment function; the dynamic adjustment function comprises an exclusion strategy and an introduction strategy, wherein the exclusion strategy excludes the parameter set based on the diversity of the parameter set, and the introduction strategy generates a new parameter set by adding random disturbance to the parameter set.

5. The method according to claim 2, wherein the method for constructing the dimension-reduction model comprises:

introducing a random mapping layer to a preset self-encoder, and initializing parameters of the random mapping layer;

Carrying out random nonlinear transformation on a preset training sample set through the random mapping layer to obtain mapping layer output;

Inputting the mapping layer output into the self-encoder, and training the self-encoder; wherein, the participation degree of the hidden layer is dynamically adjusted according to the output of the hidden layer node of the self-encoder;

calculating the current loss corresponding to the self-encoder, and updating the network parameters of the self-encoder based on the current loss;

and constructing a dimension reduction model based on the self-encoder until the self-encoder meets preset training requirements.

6. The method of claim 1, wherein the method of constructing the damage assessment model comprises:

Acquiring a preset training sample set, extracting features of the training sample set, and then performing data dimension reduction to obtain a target training set;

randomly initializing network parameters of a high-order neural network, and initializing points on a Riemann manifold based on the target training set;

Mapping the target training set onto a Riemann manifold, and extracting deep features of the target training set by utilizing a geometric structure on the Riemann manifold;

classifying and training the deep features through the high-order neural network, calculating a loss function, and optimizing network parameters of the high-order neural network by utilizing an acceleration multi-direction whale optimization algorithm;

And constructing a damage evaluation model based on the higher-order neural network until the higher-order neural network meets a preset training requirement.

7. The method of claim 6, wherein optimizing network parameters of the higher order neural network using an accelerated multi-directional whale optimization algorithm comprises:

calculating a candidate solution corresponding to the optimal solution based on the current optimal solution of the higher-order neural network so as to simulate the behavior of whale surrounding a prey and obtain a first updated parameter;

Simulating whale to approach the first updating parameters through spiral ascending, and determining second updating parameters;

randomly selecting a search point to obtain a third updating parameter;

And performing performance evaluation on the first updating parameter, the second updating parameter and the third updating parameter, and determining the parameter with the best performance as a final updating parameter.

8. The method according to claim 1, wherein the method for constructing the training sample set comprises:

the method comprises the steps of collecting multidimensional data of a spiral bolt hanging ring in a preset environment to obtain a multidimensional training sample; the multidimensional data comprise lifting ring use data and lifting ring self-sensing monitoring data;

labeling the multi-dimensional training samples to construct an initial sample set;

The initial sample set is standardized and then converted into quantum bits, and quantum projection is carried out on the quantum bits, so that low-dimensional space quantum data are obtained;

Performing iterative training on a preset generation countermeasure network through the low-dimensional space quantum data to generate an expansion sample; calculating quantum state similarity between the extended sample and the low-dimensional space quantum data, and updating network parameters of the generated countermeasure network;

Decoding the extended samples meeting the extension requirements, and combining the decoded extended samples with the initial sample set to construct a training sample set.

9. The method of claim 8, wherein the step of updating the network parameters of the generated countermeasure network by calculating quantum state similarities between the extended samples and the low-dimensional spatial quantum data comprises:

Respectively calculating amplitude components of the extended sample and the low-dimensional space quantum data on a quantum bit;

calculating an inner product of the extended sample and the low-dimensional spatial quantum data based on the amplitude component of the extended sample and the amplitude component of the low-dimensional spatial quantum data;

and inputting the inner product into a preset quantum state similarity loss function, and determining the quantum state similarity between the extended sample and the low-dimensional space quantum data.

10. Spiral bolt rings impaired evaluation device based on artificial intelligence, characterized by comprising:

the data acquisition module is used for acquiring multidimensional vector data of the target spiral bolt hanging ring; the multi-dimensional vector data at least comprises lifting ring use data and lifting ring self-sensing monitoring data;

The data processing module is used for identifying the multi-dimensional vector data and determining a target vector in the multi-dimensional vector data;

The execution module is used for inputting the target vector into a pre-constructed damage evaluation model, classifying the target vector and determining a classification result; the damage assessment model is constructed based on a preset classifier, and a training sample set for training the classifier comprises a multi-dimensional training sample and a sample label; the classifier updates parameters based on an acceleration multi-direction whale optimization algorithm;

And the output module is used for determining the damage condition of the target spiral bolt hanging ring based on the classification result.