CN115828088A - High-voltage shunt reactor vibration abnormity fault diagnosis method based on deep learning - Google Patents
High-voltage shunt reactor vibration abnormity fault diagnosis method based on deep learning Download PDFInfo
- Publication number
- CN115828088A CN115828088A CN202211421751.1A CN202211421751A CN115828088A CN 115828088 A CN115828088 A CN 115828088A CN 202211421751 A CN202211421751 A CN 202211421751A CN 115828088 A CN115828088 A CN 115828088A
- Authority
- CN
- China
- Prior art keywords
- fault diagnosis
- network
- deep learning
- incep
- method based
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention relates to the technical field of power system fault diagnosis, in particular to a high-voltage parallel reactor vibration abnormity fault diagnosis method based on deep learning, which comprises the following steps: acquiring a vibration signal of the reactor; data preprocessing, namely converting the time sequence vibration signal into an image signal; establishing a data set; performing Incep-DenseNet network fault diagnosis training and testing; the beneficial effects are that: according to the high-voltage parallel reactor vibration abnormity fault diagnosis method based on deep learning, disclosed by the invention, through the advantages of multi-scale feature extraction and multi-level feature fusion, deep abstract data features hidden in a reactor vibration signal can be mined, the mapping relation between the fault data features and fault types is accurately established, and high-precision high-voltage reactor vibration abnormity fault diagnosis is realized. When massive data are faced, the method based on the Incep-DenseNet deep learning network can quickly extract the characteristics of good robustness and high fault sensitivity from complex data, and ensures the real-time performance and reliability of fault diagnosis.
Description
Technical Field
The invention relates to the technical field of power system fault diagnosis, in particular to a high-voltage parallel reactor vibration abnormity fault diagnosis method based on deep learning.
Background
The high-voltage shunt reactor is commonly used for compensating capacitive charging power of an extra-high voltage line, is beneficial to limiting the rise of power frequency voltage in a power system and reducing the insulation level of the extra-high voltage system, and is key equipment for ensuring the safe and reliable operation of the power system. Similar to a transformer, the reactor also has a winding iron core structure and a special structure with a multi-air-gap iron core cake, and various types of faults are caused by large vibration and strong noise in the long-term operation process. For example, the grading ring of the inner sleeve is broken due to excessive vibration, the oil leakage phenomenon is caused by the fact that the oil tank is pulled to be broken due to excessive vibration of the pulling rib, and the radiating fins are separated from the connecting bolts due to excessive vibration, so that the radiating fins are cracked. Therefore, if the abnormal vibration phenomenon cannot be found in time, the vibration is intensified, and in the long past, vicious circle finally causes the shutdown of the reactor due to the overlarge vibration, and even causes serious accidents. Therefore, the method accurately detects and identifies the abnormal vibration phenomenon of the high-voltage shunt reactor, and performs predictive maintenance on equipment in time, and is the key for improving the safety and reliability of the power system.
In the prior art, oil immersed high voltage reactor fault diagnosis technologies are mainly divided into two main categories, including experience-based diagnosis methods and data-based diagnosis methods. When the high-voltage reactor is diagnosed by adopting the experience-based fault diagnosis method, a large amount of expert knowledge needs to be accumulated, and then more accurate fault judgment can be realized. However, when the actual working condition of the reactor changes, the expert knowledge may have recognition deviation due to incomplete comprehensiveness, so that the robustness and reliability of the fault diagnosis method based on experience are poor. The data-based fault diagnosis method generally adopts a signal processing technology to analyze data acquired during the operation of the reactor, extracts a characteristic vector capable of representing the operation state of the reactor, and adopts a mode recognition method to determine the operation state of the reactor and a corresponding fault type.
However, compared with an experience-based reactor fault diagnosis method, the data-based fault diagnosis method does not need any prior information or expert knowledge when diagnosing the operating state of the reactor, and therefore receives much attention. However, the detection of the abnormal vibration of the high-voltage reactor is performed by signal processing, such as fourier transform and empirical mode decomposition. However, with the long-term operation of the reactor, a large amount of various health state data is accumulated, and when these large amounts of data are analyzed and processed, the real-time performance and reliability of the signal processing-based method are greatly challenged.
Disclosure of Invention
The invention aims to provide a method for diagnosing abnormal vibration faults of a high-voltage shunt reactor based on deep learning, so as to solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme: the method for diagnosing the abnormal vibration fault of the high-voltage shunt reactor based on deep learning comprises the following steps of:
acquiring a vibration signal of the reactor;
data preprocessing, namely converting the time sequence vibration signal into an image signal;
establishing a data set;
and performing Incep-DenseNet network fault diagnosis training and testing.
Preferably, when the time sequence vibration signal is converted into an image signal, the original one-dimensional data is equally divided into N segments in sequence, and the continuous N segments of data are respectively used as the first row pixel point to the Nth row pixel point of the image; s (i), i =1,2, …, N2, representing one-dimensional raw data, F (j, k), j =1,2, …, N, k =1,2, …, N, representing pixel points in an image; the conversion process was according to the following formula:
where min represents the minimum function, max represents the maximum function, and round represents the floor function.
Preferably, a multi-scale convolutional layer and a dense connection network architecture are built based on the Incep-DenseNet network, so that multi-scale feature extraction and multi-level feature fusion are realized; the Incep-DenseNet network includes an inclusion module.
Preferably, the inclusion module consists of 4 different feature extraction modes, including a 1x1 convolution kernel, a 3x3 convolution kernel, a 5x5 convolution kernel, and a 3x3 pooling kernel.
Preferably, different scale feature graphs are spliced in parallel through a Concat fusion operation, and multi-scale feature extraction and fusion of input data of the inclusion module are achieved.
Preferably, in the inclusion module, the input data is subjected to batch normalization and Relu function activation, and then convolution operation is performed; normalization enables the distribution of each layer of input data in the network to be relatively stable, so that the model learning speed is accelerated, the sensitivity of the model to network parameters is reduced, the parameter adjusting process is simplified, and the network learning is more stable; the Relu activation function has no saturation problem, and avoids the disappearance of the seed gradient in the training process, so that the deep network is accelerated to converge.
Preferably, within the Incep-Dense module, the output of each network layer is directly linked to all subsequent network layer inputs.
Preferably, each network layer input in the Incep-Dense module is subjected to Concat operation of splicing output feature maps of all the previous network layers in parallel in a Dense connection mode, so that feature multiplexing and multi-level feature fusion are realized.
Preferably, the Incep-DenseNet is comprised of 4 Incep-Dense modules, with adjacent Incep-Dense modules connected by a bottleneck layer.
Preferably, in the data processing process, feature extraction is performed on input data through a convolutional layer, a feature map is subjected to multi-scale feature extraction and multi-level feature fusion successively through 4 Incep-Dense modules, and feature classification is performed through a full connection layer and a softmax layer.
Compared with the prior art, the invention has the beneficial effects that:
according to the high-voltage parallel reactor vibration abnormity fault diagnosis method based on deep learning, disclosed by the invention, through the advantages of multi-scale feature extraction and multi-level feature fusion, deep abstract data features hidden in a reactor vibration signal can be mined, the mapping relation between the fault data features and fault types is accurately established, and high-precision high-voltage reactor vibration abnormity fault diagnosis is realized. In addition, when massive data are faced, the method based on the Incep-DenseNet deep learning network can quickly extract the characteristics of good robustness and high fault sensitivity from complex data, and the real-time performance and the reliability of fault diagnosis are guaranteed.
Drawings
FIG. 1 is a schematic diagram of a data preprocessing method;
FIG. 2 is a schematic view of an Incep-Dense module;
FIG. 3 is a schematic diagram of an Incep-DenseNet network architecture;
FIG. 4 is a flow chart of reactor vibration abnormity fault diagnosis based on an Incep-DenseNet deep learning network;
FIG. 5 is a diagram of a reactor vibration abnormality fault detection apparatus;
FIG. 6 is a diagram illustrating vibration waveforms in different states (g is gravity acceleration);
FIG. 7 is a graph of the frequency spectrum of a vibration signal in different states;
fig. 8 is a schematic diagram of the diagnosis result of the abnormal vibration fault of the reactor based on the Incep-DenseNet model.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clear and fully described, embodiments of the present invention are further described in detail below with reference to the accompanying drawings. It is to be understood that the specific embodiments described herein are merely illustrative of some embodiments of the invention and are not limiting of the invention, and that all other embodiments obtained by those of ordinary skill in the art without the exercise of inventive faculty are within the scope of the invention.
Example one
Referring to fig. 1 to 4, the present invention provides a technical solution: the method for diagnosing the abnormal vibration fault of the high-voltage shunt reactor based on deep learning comprises the following steps of:
acquiring a vibration signal of the reactor;
data preprocessing, namely converting the time sequence vibration signal into an image signal;
establishing a data set;
and performing Incep-DenseNet network fault diagnosis training and testing.
Data pre-processing
In fig. 1, a reactor time sequence vibration signal is in a left coordinate system, and a two-dimensional image matrix converted from a one-dimensional time sequence vibration signal is in a right coordinate system. In the transformation process, the original one-dimensional data is equally divided into N segments in sequence. Then, the continuous N segments of data are respectively used as the pixel points of the first row to the pixel points of the Nth row of the image. S (i), i =1,2, …, N2, representing one-dimensional raw data, F (j, k), j =1,2, …, N, k =1,2, …, N, representing pixel points in an image. The conversion process is according to the following formula.
Where min represents the minimum function, max represents the maximum function, and round represents the floor function.
Incep-DenseNet network architecture
In order to excavate richer depth abstract characteristics, the novel Incep-DenseNet network is provided, and multi-scale characteristic extraction and multi-level characteristic fusion are realized by building a multi-scale convolution layer and a dense connection network framework, so that the reliability and robustness of the abnormal vibration fault of the high-voltage reactor are greatly improved.
Multi-scale convolution
As shown in fig. 2, the inclusion module consists of 4 different feature extraction modes, including a 1x1 convolution kernel, a 3x3 convolution kernel, a 5x5 convolution kernel, and a 3x3 pooling kernel. The convolution kernels or pooling kernels with different scales can extract data features with different scales, feature maps with different scales are spliced in parallel through Concat fusion operation (as shown by x0, x1, x2 and x3 in the graphs), and multi-scale feature extraction and fusion of input data of the inclusion module are finally achieved.
In the inclusion module, input data are subjected to Batch Normalization (BN) and Relu function activation, and then convolution operation is performed. On one hand, the BN can not only make the distribution of each layer of input data in the network relatively stable so as to accelerate the learning speed of the model, but also can reduce the sensitivity of the model to network parameters and simplify the parameter adjusting process so as to make the network learning more stable. On the other hand, the Relu activation function has no saturation problem, and solves the problem that the gradient disappears in the training process, so that the deep network is converged at an accelerated speed.
Dense connection architecture
In order to facilitate data communication between different network layers, a new network connection architecture (as shown in fig. 2) is utilized in the method: inside the Incep-Dense module, the output of each network layer is directly linked with the input of all subsequent network layers, namely a Dense connection architecture. And (3) parallelly splicing output feature graphs of all the previous network layers by each network layer input in the Incep-Dense module in a Dense connection mode (Concat operation), so that feature multiplexing and multi-level feature fusion are realized. On one hand, the network can perform feature relearning through feature reuse, and more abundant deep-level abstract features are mined. On the other hand, the multi-level feature fusion can fuse low-level features and high-level features, and the complementarity between different-level features is utilized to fuse the advantages between the features, so that the performance of the model is improved.
As shown in fig. 3, the inclusion-densnet network consists of 4 inclusion-Dense modules, with adjacent inclusion-Dense modules connected by a bottleneck layer (convolution + pooling). In the data processing process, firstly, feature extraction is carried out on input data through a convolutional layer, then, a feature diagram successively carries out multi-scale feature extraction and multi-level feature fusion through 4 Incep-Dense modules, and finally, feature classification is carried out through a full connecting layer and a softmax layer. Specific structural parameters of the Incep-DenseNet network are shown in Table 1.
In summary, the Incep-DenseNet network architecture has the outstanding advantages of multi-scale feature extraction and multi-level feature fusion, can extract abundant diagnostic information from weak reactor diagnostic signals, establishes an accurate mapping relation between data features and reactor fault states, and is beneficial to realizing high-precision diagnosis of reactor vibration abnormity faults.
TABLE 1 Incep-DenseNet network architecture parameters
Example two
By utilizing the technical scheme of the patent, the fault diagnosis is carried out on the abnormal vibration fault of the high-voltage reactor, the implementation process of the patent method is introduced, and the advantages of the patent method are highlighted.
Experimental data:
the reactor vibration test adopts a mode that a piezoelectric acceleration vibration sensor is used for collecting vibration signals on the surface of an oil tank, the acceleration sensor is adsorbed on the surface of the oil tank through a magnet, the resolution is 1000Mv/g (wherein g is the gravity acceleration of 9.8m/s < 2 >), and the reactor vibration test has the characteristics of light weight and large measuring range. The main parameters of the high-voltage reactor are shown in table 2.
TABLE 2 main parameters of oil-immersed high-voltage shunt reactor
The vibration of different positions of the oil tank is generated by the common coupling and superposition of various vibration sources such as an iron core and a winding, and due to the difference of the contribution rates of the vibration sources and the vibration, the vibration signals acquired by different measuring points are different. If multiple phases and multiple vibration sources act simultaneously, the linearity relation between the vibration acceleration and the voltage square is reduced, and the position measuring point causes difficulty in analysis of mechanical faults inside the reactor. If only a single-phase vibration source acts independently, the vibration acceleration and the voltage square of the single-phase vibration source are in a linear relation, the vibration signal at the position of the measuring point is easy to change obviously and is more sensitive to faults, and therefore the position with good linearity can be selected as the optimal measuring point.
According to literature research, most measuring points on the oil tank vibrate slightly, and the vibration signals in the middle area directly propagate through transformer oil to form attenuation. The sensitivity to slight internal mechanical faults is reduced due to the fact that the vibration is too large, and fault diagnosis of the reactor is not facilitated. Therefore, the central area of the surface of the oil tank with relatively small vibration is selected as a measuring point, and abnormal vibration faults inside the reactor can be reflected more sensitively. The experimental acceleration sensor arrangement position is shown in fig. 5.
In a fault experiment, a core winding loosening fault is simulated through a loosening screw rod, the screw rod loosening faults of 4N m (fault 1), 8N m (fault 2) and 12N m (fault 3) are taken as three reactor vibration abnormal faults with different degrees, and the screw rod is set to be a rated pretightening force in a normal state. The vibration data for the four different states are shown in fig. 6.
It can be obtained from the results of fig. 6 that when the reactor vibration abnormal fault occurs, the vibration signal is weakly different from the signal under the normal working condition. Only when the vibration failure is serious (failure 3), the vibration acceleration amplitude thereof is increased by a small amount. On the basis of the original signal analysis, the fourier transform of the vibration signal is performed to obtain a signal spectrum, as shown in fig. 7. As can be seen from the figure, the vibration signal is mainly composed of main frequencies of 100Hz, 200Hz and 300Hz, and contains a small amount of high-frequency components. When the vibration fault is gradually serious, the main frequency component is slightly reduced. However, in general, the frequency spectrums of the four working conditions are very similar, and valuable information is difficult to provide for fault judgment.
In order to complete training and testing of the Incep-DenseNet model, vibration data of the reactor under 4 working conditions are divided into a training group and a testing group, as shown in Table 3.
TABLE 3 statistical table of image-based fault diagnosis results
Network training process as shown in fig. 8 (a) and (b), the network successfully converges after 200 training passes. The diagnostic accuracy on the training set and the test set is as high as 94.7%, which indicates that the network overfitting problem is solved. To further prove the efficiency of the reactor vibration fault diagnosis of the network, fig. 8 (c) calculates a confusion matrix of network diagnosis results. The state information can be obtained through visual analysis in the confusion matrix, and for each reactor health state, the Incep-DenseNet model can realize high-precision diagnosis.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (10)
1. The method for diagnosing the abnormal vibration fault of the high-voltage shunt reactor based on deep learning is characterized by comprising the following steps of:
acquiring a vibration signal of the reactor;
data preprocessing, namely converting the time sequence vibration signal into an image signal;
establishing a data set;
and performing Incep-DenseNet network fault diagnosis training and testing.
2. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 1, characterized in that: when the time sequence vibration signal is converted into an image signal, the original one-dimensional data is equally divided into N sections in sequence, and the continuous N sections of data are respectively used as pixel points in the first row to pixel points in the Nth row of the image; s (i), i =1,2, …, N2, representing one-dimensional raw data, F (j, k), j =1,2, …, N, k =1,2, …, N, representing pixel points in an image; the conversion process was according to the following formula:
where min represents the minimum function, max represents the maximum function, and round represents the floor function.
3. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 1, characterized in that: based on the Incep-DenseNet network, a multi-scale convolutional layer and a dense connection network architecture are built, and multi-scale feature extraction and multi-level feature fusion are realized; the Incep-DenseNet network includes an inclusion module.
4. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 3, characterized in that: the inclusion module consists of 4 different feature extraction modes, including a 1x1 convolution kernel, a 3x3 convolution kernel, a 5x5 convolution kernel, and a 3x3 pooling kernel.
5. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 4, characterized in that: and parallelly splicing feature graphs of different scales through Concat fusion operation, and realizing multi-scale feature extraction and fusion of input data of the inclusion module.
6. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 3, characterized in that: in the increment module, input data are subjected to batch normalization and Relu function activation, and then convolution operation is executed; normalization enables the distribution of each layer of input data in the network to be relatively stable, so that the model learning speed is accelerated, the sensitivity of the model to network parameters is reduced, the parameter adjusting process is simplified, and the network learning is more stable; the Relu activation function has no saturation problem, and avoids the disappearance of the seed gradient in the training process, so that the deep network is accelerated to converge.
7. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 3, characterized in that: inside the Incep-Dense module, the output of each network layer is directly linked with the input of all the subsequent network layers.
8. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 3, characterized in that: and (3) inputting each network layer in the Incep-Dense module, and parallelly splicing Concat operation of output feature graphs of all the previous network layers in a Dense connection mode to realize feature multiplexing and multi-level feature fusion.
9. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 3, characterized in that: the Incep-DenseNet is composed of 4 Incep-Dense modules, with adjacent Incep-Dense modules connected by bottleneck layers.
10. The high-voltage shunt reactor vibration abnormality fault diagnosis method based on deep learning according to claim 9, characterized in that: in the data processing process, feature extraction is carried out on input data through a convolutional layer, a feature map carries out multi-scale feature extraction and multi-level feature fusion successively through 4 Incep-Dense modules, and feature classification is carried out through a full connection layer and a softmax layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211421751.1A CN115828088A (en) | 2022-11-14 | 2022-11-14 | High-voltage shunt reactor vibration abnormity fault diagnosis method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211421751.1A CN115828088A (en) | 2022-11-14 | 2022-11-14 | High-voltage shunt reactor vibration abnormity fault diagnosis method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115828088A true CN115828088A (en) | 2023-03-21 |
Family
ID=85528021
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211421751.1A Pending CN115828088A (en) | 2022-11-14 | 2022-11-14 | High-voltage shunt reactor vibration abnormity fault diagnosis method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115828088A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117232577A (en) * | 2023-09-18 | 2023-12-15 | 杭州奥克光电设备有限公司 | Optical cable distributing box bearing interior monitoring method and system and optical cable distributing box |
-
2022
- 2022-11-14 CN CN202211421751.1A patent/CN115828088A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117232577A (en) * | 2023-09-18 | 2023-12-15 | 杭州奥克光电设备有限公司 | Optical cable distributing box bearing interior monitoring method and system and optical cable distributing box |
| CN117232577B (en) * | 2023-09-18 | 2024-04-05 | 杭州奥克光电设备有限公司 | Optical cable distributing box bearing interior monitoring method and system and optical cable distributing box |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109765053B (en) | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index | |
| CN110849626B (en) | Self-adaptive sparse compression self-coding rolling bearing fault diagnosis system | |
| CN112036547B (en) | Rolling bearing residual life prediction method combining automatic feature extraction with LSTM | |
| CN112200244B (en) | Intelligent detection method for anomaly of aerospace engine based on hierarchical countermeasure training | |
| CN110702411B (en) | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis | |
| CN111523081B (en) | Aeroengine fault diagnosis method based on enhanced gate control circulating neural network | |
| CN112257528B (en) | Wind power gear box fault diagnosis method based on wavelet transformation and dense connection expansion convolutional neural network | |
| CN109858352A (en) | A kind of method for diagnosing faults based on compressed sensing and the multiple dimensioned network of improvement | |
| CN112378660A (en) | Intelligent fault diagnosis method for aero-engine bearing based on data driving | |
| CN110632484A (en) | ELM-based GIS partial discharge defect diagnosis and classification system and method | |
| CN112070104A (en) | Main transformer partial discharge identification method | |
| CN112926728B (en) | Small sample turn-to-turn short circuit fault diagnosis method for permanent magnet synchronous motor | |
| CN112113755A (en) | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network | |
| CN115828088A (en) | High-voltage shunt reactor vibration abnormity fault diagnosis method based on deep learning | |
| CN112949402A (en) | Fault diagnosis method for planetary gear box under minimum fault sample size | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN116465628A (en) | Rolling bearing fault diagnosis method based on improved multi-source domain heterogeneous model parameter transmission | |
| CN114881071A (en) | Synchronous motor rotor winding turn-to-turn short circuit fault diagnosis method based on multi-source information | |
| Chen et al. | An adversarial learning framework for zero-shot fault recognition of mechanical systems | |
| CN112881879A (en) | High-voltage cable terminal partial discharge mode identification method, device and equipment | |
| CN116843662A (en) | Non-contact fault diagnosis method based on dynamic vision and brain-like calculation | |
| CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
| CN116298725A (en) | Fault arc detection method, system and storage medium | |
| Yuan et al. | Multi-sourced monitoring fusion diagnosis for rotating machinery faults | |
| CN115356599A (en) | Multi-mode urban power grid fault diagnosis method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |