CN113221954A - BP (Back propagation) classification algorithm based on improved bat algorithm - Google Patents
BP (Back propagation) classification algorithm based on improved bat algorithm Download PDFInfo
- Publication number
- CN113221954A CN113221954A CN202110403277.9A CN202110403277A CN113221954A CN 113221954 A CN113221954 A CN 113221954A CN 202110403277 A CN202110403277 A CN 202110403277A CN 113221954 A CN113221954 A CN 113221954A
- Authority
- CN
- China
- Prior art keywords
- bat
- network
- weight
- algorithm
- loudness
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a method for training the weight and the threshold of a neural network by using an improved bat algorithm, which greatly improves the accuracy of image classification. The method comprises the following algorithm steps: the method comprises the following steps: and inputting an original image and processing the original image. Step two: the network is initialized. Step three: and assigning the initial parameters of the invention. Step four: and (4) calculating a weight empirical factor, moving the bats by using an equation, and updating loudness and pulse rate. Step five: and recording the global optimal position and the local optimal position of the current population, updating the speed by using a formula, and obtaining the signal position of the bats of the population according to the formula. Step six: and the optimal solution X respectively corresponds to the weight and the threshold of the network and outputs a result. Step seven: and judging whether the maximum iteration times is reached, and if so, outputting the result. If not, returning to the step four. Compared with other algorithms, the method has higher convergence rate, development capability and stability.
Description
The technical field is as follows:
the invention relates to the technical field of strip steel surface detection, in particular to an algorithm for matching images and identifying six defects of inclusion, plaque, cracking, pitting corrosion, rolling scale and scratch.
Background art:
with the rapid development of image recognition technology in recent years, machine vision technology is gradually penetrating into various aspects of production processing. Among them, surface quality inspection of strip steel is one of the most suitable applications.
The steel plate is easy to have the problems of pitting, cracking, pits, scratches and the like in the production process. Therefore, the surface defect detection is widely concerned by steel plate production enterprises.
The classification problem of the steel plate defects can be solved by applying an image identification technology, and the image identification and technology can be divided into two steps of image feature extraction and image classification. For feature extraction for metal surface defect detection, the texture features of the image are typically selected.
Besides texture features, the selection of an image classifier has a great influence on the classification of images, and the most widely applied neural network at present is a BP neural network. The BP network is a feedforward neural network that uses gradient descent to find the optimum value of the function by error back propagation. However, the BP network has the disadvantages of low convergence rate, easy falling into a local minimum value, poor generalization capability and the like.
In order to improve the defects of the BP network, scholars apply a bat algorithm to an image recognition technology, the algorithm is inspired by the behavior that bats locate predators through echoes, the bat algorithm is a novel swarm intelligence algorithm based on iterative optimization, is superior to algorithms such as GA and PSO in the aspect of algorithm performance, and has the characteristics of high convergence rate and strong global search capability. Therefore, the BA algorithm is widely applied to various aspects of industrial production, image processing, aerospace and the like.
However, the bat algorithm also has its limitations, and is prone to problems of local optimization, low convergence accuracy and the like. In order to overcome the defects of the bat algorithm, the invention provides a method for training the weight and the threshold of the neural network by using the improved bat algorithm (WG-BA), and the method can overcome the defects of long convergence time and easy falling into local optimum of an error back propagation network, thereby improving the accuracy of image classification.
The invention content is as follows:
the invention aims to provide a method for training the weight and the threshold of a neural network by using an improved bat algorithm (WG-BA), which can overcome the defects of long convergence time of an error back propagation network and easy falling into local optimum, thereby greatly improving the accuracy of image classification and solving the problems mentioned in the background story.
In order to achieve the purpose, the invention provides the following technical scheme: a BP neural network optimization based on an improved bat algorithm is applied to a band steel surface defect image classification method.
The method comprises the following steps: inputting an original image, carrying out LBP (local binary pattern) processing on the original image, extracting characteristic vectors of information such as inclusions, plaques, cracks, pitting corrosion, rolling scale, scratches and the like, and inputting the characteristic vectors into a network.
Step two: initializing the network, and setting the number of layers of the network and the number of nodes of each layer of the network.
Step three: the initial parameters of the invention are assigned, including initial position, speed, loudness, frequency, maximum iteration number, etc.
Step four: calculating weight value empirical factor, the bat position is Xi(t+1)=Xi(t)+Vi(t +1) and the velocity is determined by equation Vik(t+1)=ω·Vik(t)+(Xik(t)-Pk(t))·fi(t) determining and updating loudness and pulse rate.
Step five: recording the global optimal position and the local optimal position of the current population, updating the speed by using a formula, obtaining the new position of the bat of the population according to the formula and recording the new position as
Step six: and the optimal solution X respectively corresponds to the weight and the threshold of the network and outputs a result.
Step seven: and judging whether the maximum iteration times is reached, and if so, outputting the result. If not, returning to the step four.
The invention has the technical effects and advantages that:
first, in the classification problem of metal surface defects, the accuracy, precision, sensitivity, specificity and F1 value of the invention are better than those of BP networks optimized by other group intelligent algorithms. Compared with other algorithms, the method has higher convergence speed, and the added weight empirical factor can be changed according to the change of the iteration times and the fitness function, so that the flying speed of the bats is controlled; the gamma distribution is added in the calculation of the bat loudness, so that the local searching capability of the bats can be enhanced, and the algorithm has development capability and is more stable.
Compared with BP, GA-BP, PSO-BP, SSA-BP and BA-BP, the invention can reach the minimum error more quickly, the accuracy, precision, sensitivity and specificity of the invention are improved, and F1 values are respectively improved by 2.52-5%, 3.57-5.32%, 3.89-6.33%, 0.8-1.26% and 3.92-6.66%.
The third invention solves the classification problem of six defect images (including objects, patches, cracks, pitting corrosion, rolling scale and scratches) with the accuracy reaching 97.37 percent, and experiments prove that the invention is an effective method for classifying the surface defects of the strip steel.
In conclusion, the invention can well improve the classification precision of the surface defects of the belt steel.
Description of the drawings:
FIG. 1 is a flow chart of the algorithm of the present invention in conjunction with a BP neural network.
The specific implementation mode is as follows:
the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example (b):
the method comprises the following steps: inputting an original image, carrying out LBP (local binary pattern) processing on the original image, extracting characteristic vectors of information such as inclusions, plaques, cracks, pitting corrosion, rolling scale, scratches and the like, and inputting the characteristic vectors into a network.
Step two: initializing the network, and setting the number of layers of the network and the number of nodes of each layer of the network.
Step three: the initial parameters of the invention are assigned, including initial position, speed, loudness, frequency, maximum iteration number, etc.
Step four: calculating weight value empirical factor, the bat position is Xi(t+1)=Xi(t)+Vi(t +1) and the velocity is determined by equation Vik(t+1)=ω·Vik(t)+(Xik(t)-Pk(t))·fi(t) determining and updating loudness and pulse rate.
Step five: recording the global optimal position and the local optimal position of the current population, updating the speed by using a formula, obtaining the new position of the bat of the population according to the formula and recording the new position as
Step six: and the optimal solution X respectively corresponds to the weight and the threshold of the network and outputs a result.
Step seven: and judging whether the maximum iteration times is reached, and if so, outputting the result. If not, returning to the step four.
The embodiment of the invention has the following specific principles:
since the classic bat algorithm only depends on the global optimal position for speed updating, the bat is easy to fall into the local optimal condition in the search process of the bat. The text proposes a way of adding two new gravitations to a velocity update formula, and the calculation method of the weight empirical factor is as follows:
Vik(t+1)=ω·Vik(t)+(Xik(t)-Pk(t))·fi(t)
wherein ω ismaxIs the maximum weight empirical factor, ωminIs a minimum weight empirical factor, t is the number of iterations, fmaxTo an optimum fitness value, fminWorst fitness value, τ andis constant and 0<τ<1,
While the advantages of linear decrement of the classic bat algorithm are retained, the formula (1) and the formula (2) can calculate a new position according to the iteration times and the change of the fitness function. Therefore, the algorithm has good global optimization performance in the initial stage of calculation, can avoid falling into a local optimum point, and avoids too fast convergence. And local optimization performance is enhanced in the later period of iteration. The method plays an important role in improving the solving quality of the algorithm and accelerating the optimizing speed.
The gamma distribution is a continuous probability function, and if the random variable x is the time required for the occurrence of the a-th event, the density function is
Where α is called the shape parameter and β is called the inverse scale parameter
Among the basic bat algorithms, the global search bat quantity can be increased and the local search bat quantity can be reduced through the reduction of loudness along with the superposition of iteration times, the proportion of the global search and the local search of each generation of bat algorithms is a probability problem, and the classic bat algorithm is realized through an equationThe loudness is updated, the gamma function is introduced into the loudness updating formula, so that the bat can adjust the change of the loudness along with the increase of iteration, the local searching capability is enhanced at the initial stage of finding the target by the bat, and the local searching capability is reduced after finding the target.
The combination of the method and the BP network is to take the weight and the threshold of the BP network as the position vector of the method, each bat is equivalent to each node of the network, the position of each bat represents the weight or the threshold in the network, the input layer, the hidden layer and the output layer of the three-layer neural network are respectively provided with m, n and q neurons, and then the position vector calculation method represented by the ith bat is as the following formula xi=(xi1,xi2,xi3,…xid)=(w11,…w1q,wn1,…wnq,W11,…W1m,Wq1,…Wqm,θ1,…θq,θ1′,…θ′m)
Wherein xiRepresenting the location of the bat, d nq qm q m, wij(i-1, 2 … n, j-1, 2 … q) represents the weight of the connection between the input layer and the hidden layer, wjk(j-1, 2 … q; k-1, 2 … m) represents the weight of the connection between the hidden layer and the output layer, θj,θkThreshold values representing layers, using a batThe fitness function of the weight and the threshold of the position training BP network of the bat isWhere n denotes the number of learned features, OihRepresents the actual output, T, of the ith bat in the networkihRepresenting the expected output of the ith bat in the network, the weight and threshold of the network are updated once every time the bats in the population update a position, so as to achieve the effect of optimizing the weight and threshold of the network by using an improved bat algorithm.
The invention applies a novel inertia weight to a basic bat algorithm, improves the global search capability in the early period, adaptively adjusts the flying speed according to the fitness value, slows down the bat flying speed along with the increase of the iteration times, gradually enhances the local search capability, and better adjusts the relationship between the global search and the local search.
The invention also provides a new method for updating the loudness of the bat algorithm, and the searching capability of the invention in the local searching process is improved.
The invention also combines the improved bat algorithm with the BP network, trains the weight and the threshold of the BP network by the improved bat algorithm, and can improve the accuracy of the BP network for identifying the metal surface defects.
Claims (6)
1. An improved bat algorithm, comprising the following algorithm steps:
the method comprises the following steps: inputting an original image, carrying out LBP (local binary pattern) processing on the original image, extracting characteristic vectors of information such as inclusions, plaques, cracks, pitting corrosion, rolling scale, scratches and the like, and inputting the characteristic vectors into a network.
Step two: initializing the network, and setting the number of layers of the network and the number of nodes of each layer of the network.
Step three: the initial parameters of the invention are assigned, including initial position, speed, loudness, frequency, maximum iteration number, etc.
Step four: and (4) calculating a weight empirical factor, moving the bats by using an equation, and updating loudness and pulse rate.
Step five: and recording the global optimal position and the local optimal position of the current population, updating the speed by using a formula, and obtaining the signal position of the bats of the population according to the formula.
Step six: and the optimal solution X respectively corresponds to the weight and the threshold of the network and outputs a result.
Step seven: and judging whether the maximum iteration times is reached, and if so, outputting the result. If not, returning to the step four.
2. A method and apparatus for improved bat algorithm as claimed in claim 1, wherein: the invention provides a method for improving loudness by introducing a gamma function into an update formula of loudness, so that the bat can adjust the change of loudness along with the increase of iteration, the local search capability is enhanced at the initial stage of finding a target by the bat, and the local search capability is reduced after finding the target. The new loudness update formula is:
3. a method and apparatus for improved bat algorithm as claimed in claim 1, wherein: since the classic bat algorithm only depends on the global optimal position for speed updating, the bat is easy to fall into the local optimal condition in the search process of the bat. The text provides a mode of adding two new gravitations into a speed updating formula, and a calculation method of a weight empirical factor comprises the following steps:
Vik(t+1)=ω·Vik(t)+(Xik(t)-Pk(t))·fi(t)
wherein ω ismaxIs the maximum weight empirical factor, ωminIs a minimum weight empirical factor, and t is the number of iterationsNumber fmaxTo an optimum fitness value, fminWorst fitness value, τ andis a constant andthe two formulas can calculate a new position according to the change of the iteration times and the fitness function while keeping the advantage of linear decrement of the classic bat algorithm. Therefore, the algorithm has good global optimization performance in the initial stage of calculation, can avoid falling into a local optimum point, and avoids too fast convergence. And local optimization performance is enhanced in the later period of iteration. The method plays an important role in improving the solving quality of the algorithm and accelerating the optimizing speed.
4. A method and apparatus for improved bat algorithm as claimed in claim 1, wherein: the invention combines with BP network, weight and threshold value of BP network are regarded as the position vector of the invention, each bat is equivalent to each node of the network, the position of each bat represents the weight or threshold value in the network, set up the input layer of the neural network of the three-layer, the hidden layer, the output layer has m, n, q neurons separately, then the position vector calculation method that the ith bat represents is:
xi=(xi1,xi2,xi3,…xid)
=(w11,…w1q,wn1,…wnq,W11,…W1m,Wq1,…Wqm,θ1,…θq,θ′1,…θ′m)
wherein xiDenotes the location of the bat, d ═ nq + qm + q + m, wij(i-1, 2 … n, j-1, 2 … q) represents the weight of the connection between the input layer and the hidden layer, wjk(j-1, 2 … q; k-1, 2 … m) represents the weight of the connection between the hidden layer and the output layer, θj,θkRepresenting the threshold value for each layer.
5. A method and apparatus for improved bat algorithm as claimed in claim 1, wherein: the fitness function of the weight and the threshold value of the BP network trained by using the bat position is as follows:
where n denotes the number of learned features, OihRepresents the actual output, T, of the ith bat in the networkihRepresenting the expected output of the ith bat in the network, the weight and threshold of the network are updated once every time the bats in the population update a position, so as to achieve the effect of optimizing the weight and threshold of the network by using an improved bat algorithm.
6. A method and apparatus for improved bat algorithm as claimed in claim 1, wherein: the equations stated in step four are respectively the equation X for determining the positioni(t+1)=Xi(t)+Vi(t +1) and equation V for determining velocityik(t+1)=ω·Vik(t)+(Xik(t)-Pk(t))·fi(t) the frequency at which each bat is given should be subject to a uniform distribution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110403277.9A CN113221954B (en) | 2021-04-15 | 2021-04-15 | Method for detecting surface defects of strip steel by using improved bat algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110403277.9A CN113221954B (en) | 2021-04-15 | 2021-04-15 | Method for detecting surface defects of strip steel by using improved bat algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113221954A true CN113221954A (en) | 2021-08-06 |
CN113221954B CN113221954B (en) | 2022-09-27 |
Family
ID=77087360
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110403277.9A Active CN113221954B (en) | 2021-04-15 | 2021-04-15 | Method for detecting surface defects of strip steel by using improved bat algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113221954B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115034121A (en) * | 2022-08-11 | 2022-09-09 | 太原科技大学 | Strip steel process regulation and control method based on organization performance intelligent prediction model |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107886158A (en) * | 2017-10-30 | 2018-04-06 | 中国地质大学(武汉) | A kind of bat optimized algorithm based on Iterated Local Search and Stochastic inertia weight |
CN109543572A (en) * | 2018-11-07 | 2019-03-29 | 北京交通大学 | A kind of traveling pavement condition evaluation method |
CN109887035A (en) * | 2018-12-27 | 2019-06-14 | 哈尔滨理工大学 | Based on bat algorithm optimization BP neural network binocular vision calibration |
CN110782460A (en) * | 2019-09-25 | 2020-02-11 | 哈尔滨理工大学 | Image segmentation method based on FCM fusion improved bat algorithm |
CN110991721A (en) * | 2019-11-26 | 2020-04-10 | 国网山东省电力公司电力科学研究院 | Short-term wind speed prediction method based on improved empirical mode decomposition and support vector machine |
CN111368900A (en) * | 2020-02-28 | 2020-07-03 | 桂林电子科技大学 | Image target object identification method |
-
2021
- 2021-04-15 CN CN202110403277.9A patent/CN113221954B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107886158A (en) * | 2017-10-30 | 2018-04-06 | 中国地质大学(武汉) | A kind of bat optimized algorithm based on Iterated Local Search and Stochastic inertia weight |
CN109543572A (en) * | 2018-11-07 | 2019-03-29 | 北京交通大学 | A kind of traveling pavement condition evaluation method |
CN109887035A (en) * | 2018-12-27 | 2019-06-14 | 哈尔滨理工大学 | Based on bat algorithm optimization BP neural network binocular vision calibration |
CN110782460A (en) * | 2019-09-25 | 2020-02-11 | 哈尔滨理工大学 | Image segmentation method based on FCM fusion improved bat algorithm |
CN110991721A (en) * | 2019-11-26 | 2020-04-10 | 国网山东省电力公司电力科学研究院 | Short-term wind speed prediction method based on improved empirical mode decomposition and support vector machine |
CN111368900A (en) * | 2020-02-28 | 2020-07-03 | 桂林电子科技大学 | Image target object identification method |
Non-Patent Citations (5)
Title |
---|
HUANYU DONG等: "A Novel Method for Power Transformer Fault Diagnosis Based on Bat-BP Algorithm", 《2018 INTERNATIONAL CONFERENCE ON SENSING,DIAGNOSTICS, PROGNOSTICS, AND CONTROL (SDPC)》 * |
XIAOFENG YUE等: "Research on image classification method of strip steel surface defects based on improved Bat algorithm optimized BP neural network", 《JOURNAL OF INTELLIGENT & FUZZY SYSTEMS》 * |
牛颖超等: "GABA算法的遥感图像分类", 《测绘通报》 * |
简琤峰等: "面向边缘计算的改进混沌蝙蝠群协同调度算法", 《小型微型计算机系统》 * |
贾鹤鸣等: "基于改进共生生物搜索算法的林火图像多阈值分割", 《计算机应用》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115034121A (en) * | 2022-08-11 | 2022-09-09 | 太原科技大学 | Strip steel process regulation and control method based on organization performance intelligent prediction model |
CN115034121B (en) * | 2022-08-11 | 2022-10-25 | 太原科技大学 | Strip steel process regulation and control method based on organization performance intelligent prediction model |
Also Published As
Publication number | Publication date |
---|---|
CN113221954B (en) | 2022-09-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110807365B (en) | Underwater target identification method based on fusion of GRU and one-dimensional CNN neural network | |
CN109800778B (en) | Faster RCNN target detection method based on difficultly-divided sample mining | |
US11816183B2 (en) | Methods and systems for mining minority-class data samples for training a neural network | |
CN107392919B (en) | Adaptive genetic algorithm-based gray threshold acquisition method and image segmentation method | |
CN110309771B (en) | GBDT-INSGAII-based EAS (Acoustic magnetic System) label identification algorithm | |
CN109934158B (en) | Video emotion recognition method based on local enhanced motion history map and recursive convolutional neural network | |
CN113706581B (en) | Target tracking method based on residual channel attention and multi-level classification regression | |
CN115131347B (en) | Intelligent control method for processing zinc alloy parts | |
CN108665322A (en) | The construction method of grain ration Consumption forecast model, Consumption forecast method and device | |
CN113221954B (en) | Method for detecting surface defects of strip steel by using improved bat algorithm | |
CN112257847A (en) | Method for predicting geomagnetic Kp index based on CNN and LSTM | |
CN112287592A (en) | Industrial equipment fault diagnosis method and system based on deep confidence network | |
CN111368900A (en) | Image target object identification method | |
CN115358305A (en) | Incremental learning robustness improving method based on iterative generation of boundary samples | |
CN113743474A (en) | Digital picture classification method and system based on cooperative semi-supervised convolutional neural network | |
CN113536939B (en) | Video duplication removing method based on 3D convolutional neural network | |
CN109492816B (en) | Coal and gas outburst dynamic prediction method based on hybrid intelligence | |
CN113516713B (en) | Unmanned aerial vehicle self-adaptive target tracking method based on pseudo twin network | |
CN113627075A (en) | Projectile aerodynamic coefficient identification method based on adaptive particle swarm optimization extreme learning | |
CN111860265B (en) | Multi-detection-frame loss balanced road scene understanding algorithm based on sample loss | |
CN110779526B (en) | Path planning method, device and storage medium | |
CN117079120A (en) | Target recognition model optimization method based on improved GA algorithm | |
CN116689503A (en) | Strip steel full-length thickness prediction method based on memory function network | |
CN116542139A (en) | Method and device for predicting roughness of liquid jet polishing surface | |
CN113837254B (en) | Method for detecting image target in computer vision |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |