Harbor machinery winding module health state evaluation method based on step flow type residual error link layer network
Through the method based on the step flow residual link layer network, the problem that traditional evaluation methods are difficult to identify the timing characteristics of the port machine winch module is solved, and accurate and efficient evaluation of the health status is achieved, and the real-time and adaptability of the evaluation is improved.
Patent Information
- Application Number
- CN202510258819.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-04
- Publication Date
- 2025-06-24
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The traditional port machine hoisting module health status evaluation method is difficult to effectively identify the timing characteristics of the equipment under high-intensity continuous operation conditions, resulting in insufficient timeliness and accuracy of health status evaluation.
The health status evaluation method based on the step flow residual link layer network is adopted, and the health status of the port machine winch module is accurately evaluated through time-series fatigue information distillation, damage step threshold determination, flow residual link layer network establishment and sparrow optimization algorithm optimization parameters.
This method can efficiently capture the key health characteristics of the winch module, identify significant turning points in equipment state changes, reduce noise data interference, achieve accurate and efficient health status assessment, and improve the real-time and adaptability of the assessment.
Smart Images
Figure CN120197474A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for evaluating health status, and in particular to a method for evaluating the health status of the winch module of a port crane based on a step-flow residual link layer network. Background Art
[0002] With the increase in the complexity of the operating environment of port machinery and the operating intensity of equipment, the health status evaluation of the winch module of a port crane is facing new challenges. Traditional health status evaluation methods usually ignore the time-series dependent signals generated during the long-term healthy operation of the winch module of a port crane, and it is difficult to comprehensively reflect the dynamic degradation characteristics of the winch module under continuous operation. Especially under the condition of high-intensity continuous operation, the time-series characteristics of the equipment state change cannot be effectively identified. This limitation not only affects the timeliness of health status evaluation, but also may lead to misjudgment or missed judgment of equipment failures. Therefore, there is an urgent need for a method for evaluating the health status of the winch module of a port crane to solve the deficiencies of traditional methods in obtaining long-term time-series dependent signals and provide a more accurate and reliable decision-making support tool for the intelligent operation and maintenance of port machinery. Summary of the Invention
[0003] Object of the Invention: The object of the present invention is to provide a method for evaluating the health status of the winch module of a port crane based on a step-flow residual link layer network.
[0004] Technical Solution: The method for evaluating the health status of the winch module of a port crane based on a step-flow residual link layer network according to the present invention includes the following steps:
[0005] (1) Time-series fatigue information distillation of the winch module of a port crane;
[0006] (2) Determination of the damage step threshold of the winch module of a port crane;
[0007] (3) Establishment of a flow residual link layer network;
[0008] (4) Solving the layer network parameters based on sparrow optimization;
[0009] (5) Numerical evaluation of the health status of the winch module of a port crane.
[0010] Further, the step (1) includes:
[0011] Arrange vibration sensors at the fault-prone parts of the rotating components of the winch module of a port crane, and obtain the time-series fatigue information of the rotating components according to the operating state and degradation characteristics of the winch module of a port crane, and use the recursive method to distill the signals.
[0012] For the obtained time-series fatigue signal x(t), construct a transformation function W(a, b):
[0013]
[0014] Among them, Ψ(t) is the mother transformation function, a is the frequency resolution influence factor, b is the position translation factor, and Ψ * is the conjugate of the mother transformation function;
[0015] For the obtained sub-signals, information distillation is performed through a screening function. The screening function is set as:
[0016]
[0017] where λ is the screening threshold, sign is the general sign function, and d j (t) is the sub-signal after screening,
[0018] The screened signal is reconstructed, and the reconstructed signal is expressed as:
[0019] u(t) = ∑d j (t)
[0020] d j (t) is the sub-signal after screening.
[0021] Furthermore, the obtained time-series fatigue signal x(t) is expressed as:
[0022]
[0023] Among them, Dj(t) represents the sub-signal of the j-th layer, and W(a, b) represents the constructed transformation function.
[0024] Furthermore, it is characterized in that the step (2) includes:
[0025] When determining the healthy operation state of the port crane hoisting module according to the vibration signal u(t) after information distillation obtained in step (1), a fatigue threshold still needs to be set. The step method can avoid misjudgment of the health state.
[0026] For the distilled vibration signal u(t), its damage cumulative distribution function F(t) is expressed as:
[0027]
[0028] Among them, θ is the signal energy mean value, and α and β are control parameters;
[0029] Thus, the fatigue step threshold T of the port crane hoisting module can be predicted:
[0030] T = α + βu(t)
[0031] Among them, α and β are control parameters, and u(t) is the time-series fatigue signal after distillation and reconstruction;
[0032] Determine the step fatigue threshold of the winch module of the port crane through the above derivation process.
[0033] Furthermore, the α and β are estimated by the method of maximum likelihood estimation.
[0034]
[0035] Among them, C(t) is the normal operation time, k is the moment of fatigue failure, n is the time series length of the signal, and f represents the normalization operation.
[0036] Furthermore, the step (3) includes:
[0037] When constructing the streaming residual link layer network based on the convolutional neural network, first define the causal convolutional layer and the dilated convolutional layer as:
[0038]
[0039] Among them, w represents the convolutional kernel, u(t) represents the time series signal as the input, y t and r t are the convolutional output and the dilated output respectively, d represents the dilation factor, and m represents the position information of the convolutional kernel.
[0040] Realize the construction of the streaming layer network by stacking multiple convolutional layers and dilated layers, and add residual links between layers to reduce the information loss in data transmission to obtain the streaming residual link layer network based on residual links. The output of the network is expressed as:
[0041]
[0042] Among them, l represents the number of layers of the streaming link layer network, y and r are the convolutional operation and the dilated operation respectively. is the output of the l-th layer of the streaming residual link network.
[0043] Furthermore, the multi-layer streaming residual link layer network imports the output of the previous layer as input data into the next layer network, which is expressed as:
[0044]
[0045] Among them, redul represents the residual activation function.
[0046] Furthermore, the step (4) includes introducing the sparrow optimization algorithm into the streaming residual link layer network that has been constructed for the health state assessment of the winch module of the port crane in step (3), and performing optimal parameter search for the control factors and the number of layers of the streaming residual link layer network.
[0047] Furthermore, the optimal parameter search includes:
[0048] Construct a multi-dimensional solution space according to the length T of the time-series fatigue signal u(t), and use the optimization result in the solution space as the discoverer in the sparrow search optimization algorithm. The remaining solutions are used as followers to modify the position of the discoverer. Then, there is an update formula for the position of the discoverer:
[0049]
[0050] where u t represents the optimal solution of the t-th dimension, rand and R2 represent random numbers between 0 and 1, Q represents a random number from a standard normal distribution, ST represents the parameter alarm value, exp represents the empirical function, and iter max represents the maximum value in the multi-dimensional solution.
[0051]
[0052] where u′ t represents the corrected solution of the t-th dimension, rand represents a random number between 0 and 1, Q represents a random number from a standard normal distribution, and T represents the total number of dimensions. The optimal parameters of the streaming residual link layer network are obtained by correcting the solution through multiple iterations.
[0053] Furthermore, the step (5) includes:
[0054] According to the streaming residual link layer network established in steps (3) and (4), evaluate the health status of the winch module of the port crane. The input time-series fatigue information undergoes a linear transformation through the first fully connected layer of the streaming residual link layer network to obtain the life evaluation result:
[0055]
[0056] where Θ is the life evaluation result of the high-speed heavy machine drive system, and sigmoid() represents the activation function.
[0057] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: The present invention combines the determination of the step fatigue threshold and the residual link layer network modeling technology, can efficiently capture the key health features of the winch module from multi-source heterogeneous data, and through the enhancement mechanism of the sparrow search optimization algorithm, identify the significant turning points of the equipment state change, thereby effectively reducing the interference of noise data; at the same time, through the multi-level information screening and transmission of the residual link layer network, it can accurately depict the local mutation characteristics and global trend characteristics during the equipment degradation process, and accurately and efficiently realize the health status evaluation of the port crane winch module; through the online update mechanism of the model, the real-time performance and adaptability of the health status evaluation are further improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] Figure 1 is a flowchart of the present invention. Detailed implementation mode
[0059] The technical solution of the present invention will be further described below with reference to the accompanying drawings.
[0060] As Figure 1 shown, the present invention includes the following steps:
[0061] (1) Time series fatigue information distillation of the port crane hoisting module
[0062] Arrange vibration sensors at the parts of the rotating components of the port crane hoisting module that are prone to failure. According to the operating state of the port crane hoisting module and its degradation characteristics, obtain the time series fatigue information of the rotating components. However, when the port crane is operating normally, various vibration signals will inevitably be generated, making it difficult to accurately identify the fault signals. Therefore, use a recursive method to distill the signals to achieve the fine-graining of the original signals, thereby improving the accuracy of the health state assessment.
[0063] For the obtained time series fatigue signal x(t), construct the transformation function W(a, b):
[0064]
[0065] where Ψ(t) is the mother transformation function, a is the frequency resolution influence factor, b is the position translation factor, and Ψ * is the conjugate of the mother transformation function.
[0066] The original time series fatigue signal x(t) can be expressed as:
[0067]
[0068] where Dj(t) represents the jth layer of sub-signals, and W(a, b) represents the constructed transformation function.
[0069] For the obtained sub-signals, perform information distillation through a screening function. Set the screening function as:
[0070]
[0071] where λ is the screening threshold, sign is the general sign function, and d j (t) is the screened sub-signal.
[0072] Reconstruct the screened signal. The reconstructed signal can be expressed as:
[0073] u(t) = ∑d j (t)
[0074] d j (t) is the screened sub-signal.
[0075] S2. Determination of the Damage Step Threshold of the Harbour Crane Hoisting Module
[0076] When determining the healthy operating state of the harbour crane hoisting module based on the vibration signal u(t) obtained after information distillation in the above S1 step, it is still necessary to set the fatigue threshold, and the step method can avoid misjudgment of the health state.
[0077] For the distilled vibration signal u(t), its damage cumulative distribution function F(t) can be expressed as:
[0078]
[0079] where θ is the mean value of the signal energy, and α and β are control parameters. For α and β, the maximum likelihood estimation method is used for estimation.
[0080]
[0081] where C(t) is the normal operation time, k is the moment of fatigue failure, n is the time series length of the signal, and f represents the normalization operation.
[0082] Thus, the fatigue step threshold T of the harbour crane hoisting module can be predicted as:
[0083] T = α + βu(t)
[0084] where α and β are control parameters, and u(t) is the time series fatigue signal after distillation and reconstruction.
[0085] Through the above derivation process, the step fatigue threshold of the harbour crane hoisting module can be determined to facilitate the subsequent establishment of the streaming link layer network.
[0086] (2) Determination of the Damage Step Threshold of the Harbour Crane Hoisting Module
[0087] When determining the healthy operating state of the harbour crane hoisting module based on the vibration signal u(t) obtained after information distillation in the above step (1), it is still necessary to set the fatigue threshold, and the step method can avoid misjudgment of the health state.
[0088] For the distilled vibration signal u(t), its damage cumulative distribution function F(t) can be expressed as:
[0089]
[0090] where θ is the mean value of the signal energy, and α and β are control parameters. For α and β, the maximum likelihood estimation method is used for estimation.
[0091]
[0092] Among them, C(t) is the normal operation time, k is the moment of fatigue failure, nn is the time series length of the signal, and f represents the normalization operation.
[0093] From this, the fatigue step threshold T of the winch module of the port crane can be predicted as follows:
[0094] T = α + βu(t)
[0095] Among them, α and β are control parameters, and u(t) is the time series fatigue signal after distillation reconstruction.
[0096] Through the above derivation process, the step fatigue threshold of the winch module of the port crane can be determined to facilitate the subsequent establishment of the streaming link layer network.
[0097] (3) Establishment of the streaming residual link layer network
[0098] When constructing the streaming residual link layer network based on the convolutional neural network, first define the causal convolutional layer and the dilated convolutional layer as follows:
[0099]
[0100] Among them, w represents the convolutional kernel, u(t) represents the time series signal as the input, and y t and r t are the convolutional output and the dilated output respectively, d represents the dilation factor, and m represents the position information of the convolutional kernel.
[0101] By stacking multiple convolutional layers and dilated layers, the construction of the streaming layer network can be realized, and adding residual links between layers can reduce the information loss in data transmission. The streaming residual link layer network based on residual links can be obtained, and the output of this network can be expressed as:
[0102]
[0103] Among them, l represents the number of layers of the streaming link layer network, y and r are the convolutional operation and the dilated operation respectively, is the output of the l-th layer of the streaming residual link network. By importing the output of the previous layer as the input data into the next layer network, the formula for obtaining the multi-layer streaming residual link layer network is:
[0104]
[0105] Among them, redul represents the residual activation function.
[0106] (4) Solving the network parameters based on sparrow search optimization
[0107] In step (3) above, a streaming residual link layer network for the health state assessment of the hoisting module of a port crane has been constructed. However, due to the significant differences in the characteristics presented by signals under different working conditions, determining the parameters of the network model through experience will result in a large error in the health state assessment results. Therefore, the sparrow optimization algorithm is introduced to optimize the best parameters of the control factors and the number of layers of the streaming residual link layer network.
[0108] Construct a multi-dimensional solution space according to the length T of the time-series fatigue signal u(t). Take the optimization result in the solution space as the discoverer in the sparrow optimization algorithm, and use the remaining solutions as followers to modify the position of the discoverer. Then, there is an update formula for the position of the discoverer:
[0109]
[0110] where u t represents the optimization solution of the t-th dimension, rand and R2 represent random numbers between 0 and 1, Q represents a random number from a standard normal distribution, ST represents the parameter warning value, exp represents the empirical function, and iter max represents the maximum value in the multi-dimensional solution.
[0111]
[0112] where u′ t represents the corrected solution of the t-th dimension, rand represents a random number between 0 and 1, Q represents a random number from a standard normal distribution, and T represents the total number of dimensions.
[0113] The optimal parameters of the streaming residual link layer network can be obtained by repeatedly iterating the corrected solution.
[0114] (5) Numerical assessment of the health state of the hoisting module of a port crane
[0115] Based on the streaming residual link layer network established in steps (3) and (4), the health state of the hoisting module of a port crane is evaluated. The input time-series fatigue information undergoes a linear transformation through the first fully connected layer of the streaming residual link layer network, and the life assessment result can be obtained:
[0116]
[0117] where Θ is the life assessment result of the high-speed heavy machine drive system, and sigmoid() represents the activation function.
Claims
1. A method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network, characterized in that: The steps include: (1) Distillation of time series fatigue information of port machinery winch module; (2) Determination of damage step threshold of port machinery winch module; (3) Establishment of streaming residual link layer network; (4) Solving the layer network parameters based on sparrow optimization; (5) Numerical evaluation of the health status of port machinery winch modules.
2. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 1 is characterized in that: The step (1) comprises: The vibration sensor is placed at the failure-prone part of the rotating component of the port machinery winch module. The timing fatigue information of the rotating component is obtained according to the operating status and degradation characteristics of the port machinery winch module. The signal is distilled using a recursive method. For the obtained time series fatigue signal x(t), construct the transformation function W(a, b): Among them, Ψ(t) is the mother transformation function, a is the frequency resolution influencing factor, b is the position translation factor, and Ψ* is the conjugate of the mother transformation function; For the sub-signals that have been obtained, information distillation is performed through the screening function, and the screening function is set to: Where λ is the screening threshold, sign is the general sign function, d j (t) is the sub-signal after screening, The filtered signal is reconstructed, and the reconstructed signal is expressed as: u(t)=∑d j (t) d j (t) is the sub-signal after filtering.
3. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 2 is characterized in that: The obtained time series fatigue signal x(t) is expressed as: Wherein, Dj(t) represents the j-th layer sub-signal, and W(a, b) represents the transformation function constructed.
4. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 1 is characterized in that: The step (2) comprises: When determining the healthy operating status of the port machinery winch module based on the vibration signal u(t) after the information distillation obtained in step (1), it is still necessary to set the fatigue threshold. The step method can avoid misjudgment of the health status. For the distilled vibration signal u(t), its damage cumulative distribution function F(t) is expressed as: Where θ is the mean signal energy, α and β are control parameters; From this, the fatigue step threshold T of the port machinery winch module can be predicted: T=α+βu(t) Among them, α and β are control parameters, and u(t) is the time series fatigue signal after distillation and reconstruction; The step fatigue threshold of the port machinery winch module is determined through the above derivation process.
5. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 4 is characterized in that: The α and β are estimated by maximum likelihood estimation. Among them, C(t) is the normal operation time, k is the time when fatigue failure occurs, n is the time series length of the signal, and f represents the normalization operation.
6. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 1 is characterized in that: The step (3) comprises: When building a streaming residual link layer network based on a convolutional neural network, the causal convolution layer and the dilated convolution layer are first defined as: Among them, w represents the convolution kernel, u(t) represents the time series signal as input, and y t and r t They are convolution output and expansion output respectively, d represents the expansion factor, and m represents the position information of the convolution kernel. The streaming layer network is constructed by stacking multiple convolutional layers and expansion layers, and residual links are added between layers to reduce the information loss in data transmission to obtain a streaming residual link layer network based on residual links. The output of the network is expressed as: Among them, l represents the number of layers of the streaming link layer network, y and r are convolution operations and expansion operations respectively. is the output of the l-th layer streaming residual link network.
7. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 6 is characterized in that: The multi-layer streaming residual link layer network imports the output of the previous layer as input data into the next layer network, which is expressed as: Among them, redul represents the residual activation function.
8. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 1 is characterized in that: The step (4) includes introducing a sparrow optimization algorithm into the streaming residual link layer network for health status assessment of the hoisting module of the port crane constructed in step (3), and searching for optimal parameters for the control factors and number of layers of the streaming residual link layer network.
9. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 8 is characterized in that: The optimal parameter optimization includes: According to the length T of the time series fatigue signal u(t), a multidimensional solution space is constructed. The optimal result in the solution space is used as the finder in the sparrow search algorithm, and the remaining solutions are used as followers to modify the finder's position. Then, the update formula of the finder's position is: where u t represents the optimal solution of the t-th dimension, rand, R2 represents a random number between 0 and 1, Q represents a standard normal distribution random number, ST represents a parameter warning value, exp represents an empirical function, and iter max represents the maximum value in the multidimensional solution, where u′ t represents the corrected solution of the t-th dimension, rand represents a random number between 0 and 1, Q represents a standard normal distribution random number, and T represents the total dimension. The optimal parameters of the streaming residual link layer network are obtained through multiple iterations of the corrected solution.
10. The method for evaluating the health status of a port machinery winch module based on a step flow residual link layer network according to claim 1, characterized in that: The step (5) comprises: According to the streaming residual link layer network established in step (3) and step (4), the health status of the port machinery winch module is evaluated. The input time series fatigue information is linearly transformed through the first fully connected layer of the streaming residual link layer network to obtain the life assessment result: Among them, Θ is the life assessment result of the high-speed heavy machine transmission system, and sigmoid() represents the activation function.