WO2021227406A1 - 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法 - Google Patents

一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法 Download PDF

Info

Publication number
WO2021227406A1
WO2021227406A1 PCT/CN2020/127970 CN2020127970W WO2021227406A1 WO 2021227406 A1 WO2021227406 A1 WO 2021227406A1 CN 2020127970 W CN2020127970 W CN 2020127970W WO 2021227406 A1 WO2021227406 A1 WO 2021227406A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
mode
quality
trend prediction
adaptive
Prior art date
Application number
PCT/CN2020/127970
Other languages
English (en)
French (fr)
Inventor
初红艳
赵凯林
程强
刘宸菲
李�瑞
Original Assignee
北京工业大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 北京工业大学 filed Critical 北京工业大学
Publication of WO2021227406A1 publication Critical patent/WO2021227406A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06395—Quality analysis or management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • 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

Definitions

  • the invention belongs to the field of intelligent manufacturing and intelligent quality monitoring, and particularly relates to a quality trend prediction method based on adaptive feature selection and an improved thinking evolution algorithm.
  • Control charts have been widely used in the production process as an auxiliary means of quality control and prediction.
  • the upper and lower limits of the control chart control quality data capture product quality fluctuations and abnormalities. With the development of technology and the acceleration of the production rhythm, only the upper and lower limits of the control chart are used to judge quality fluctuations. Large quality problems often occur, which cannot adapt to modern processing. Quality control is carried out by modern means such as data collection.
  • many scholars have also begun to study the control chart mode in pursuit of precise control of quality, but most of the research modes are few, do not study the mixed mode or cannot adapt to the dynamic changes of the data. , Forecast and mostly use offline recognition, and the degree of intelligence is not enough.
  • a quality trend prediction method based on adaptive feature selection and improved thinking evolution algorithm is proposed.
  • the invention can realize intelligent quality prediction and control, promptly propose measures for correction and improve product quality.
  • the purpose of the present invention is to propose a quality trend prediction method based on adaptive feature selection and improved thinking evolution algorithm based on the multiple modes of the control chart in view of the dynamic and time-varying characteristics of quality data.
  • the time-varying self-adaptive selection feature of the data predicts the online product quality trend in the dynamic production process, and at the same time, the improved thinking evolution algorithm is used to optimize the MLPNN network to improve the recognition accuracy of the quality trend.
  • the present invention proposes a quality trend prediction method based on adaptive feature selection and improved thinking evolution algorithm to solve the above problems and construct a trend prediction model to be able to predict abnormal states.
  • the present invention adopts the following technical solutions:
  • a quality trend prediction method based on adaptive feature selection and improved thinking evolution algorithm includes the following steps:
  • Step 1 Generate the data needed for model building.
  • the data generated in 9 modes are normal mode (NOR), periodic mode (CYC), system mode (SYS), hierarchical mode (STR), upward trend mode (IT), downward trend mode (DT), and upward step Mode (US), Step Down Mode (DS), Mixed Mode (MIX).
  • Step 2 The feature adaptive processing module is established.
  • the establishment of this step can select the characteristics reasonably and further improve the degree of intelligence.
  • the module establishment is divided into two steps:
  • Step 1 Establish a feature extraction model.
  • the extracted data features include the following statistical features and shape features that are more convincing; among them, the statistical features of quality data include: MEAN, VS, STD, SKEW, KURT, A; quality
  • the shape features of the data include: SL, NC1, NC2, APML, APLS, AASL, ACLPI, SRANGE, SB, PSMLSC, REAE, ABDPE.
  • MEAE mean value of quality data
  • A Autocorrelation coefficient of quality data
  • SL the slope of the least squares regression line fitted to the quality data
  • NC1 The number of intersection points between the curve formed by the quality data and the line formed by the average value
  • NC2 The number of intersection points between the curve formed by the quality data and the least square regression line
  • APML The area between the curve formed by the quality data and the average line
  • APLS The area between the curve formed by the quality data and the least square regression line
  • AASL The curve formed by the quality data is divided into four areas, and the average value of the slope of the curve formed by the combination of points in each area
  • ACLPI The ratio of APML to the standard deviation of quality data
  • SRANGE The curve formed by the quality data is divided into four areas, the difference between the maximum and minimum slopes formed by the midpoints of each area;
  • PSMLSC The average value of the intersection of the quality data and the center line, and the sum of the intersection of the least squares regression line;
  • REAE The ratio of the error between the MSE of the quality data and the average value of the MSE of the data divided into four regions;
  • ABDPE When the curve formed by the quality data is divided into two regions, the absolute value of the difference between the slope of the least square regression line of the curve formed by the overall quality data and the average value of the slope of the least square regression line of the two regions.
  • Step 2 Establish an adaptive feature selection model, which establishes a method for introducing errors to affect calculations. If there is new data, the module can be used to select suitable features according to the data characteristics. Based on different types of quality data, different features are selected according to their importance.
  • the method to build the model is as follows:
  • Step 3 Data feature fusion module.
  • the data dimensionality reduction method is divided into linear data dimensionality reduction and non-linear data dimensionality reduction.
  • the module uses the KPCA data dimensionality reduction method to fuse the original data with the feature data. The method steps for establishing the module are as follows:
  • Step 4 Establish a quality trend prediction module.
  • a 3-layer perceptron MLPNN neural network model is established based on the data after adaptive feature fusion, and the weights and thresholds of the MLPNN neural network are optimized using an improved mind evolution algorithm.
  • the general thinking evolutionary algorithm is mainly through the iterative optimization learning method, all individuals in the evolution process are called groups, and a group is divided into several subgroups. Subgroups include winning subgroups and temporary subgroups. During the evolution of thinking, the winning subgroups and temporary subgroups are randomly generated based on the optimal particle as the center during the flight evolution. There is no restriction on the degree of information contained between the particles in the subgroup. Particles that have no meaning to evolution. Here, mutual information theory is introduced to determine the degree of subgroup evolution. When the information level of the particles generated in the subgroup and the center particle is greater than 85%, the particle is considered to be an invalid particle. At the same time, if the score of a particle is greater than the center particle, it will be retained, otherwise the particle will be regenerated.
  • step c Convergence operator: Calculate the individual particle scores in all subgroups, and select the winner as the center to regenerate the subgroups. At the same time, step c is performed. If the individual particles and the central particle in the subgroup contain the appropriate information, continue to the next step, otherwise, regenerate particles and proceed to step c.
  • the entropy change theory is introduced when generating a new subgroup with a new center particle in step d to increase the degree of chaos generated by the particles and increase the entropy change inertia coefficient.
  • the inertia coefficient can improve the search ability for excellent particles.
  • Analyze the abnormal conditions in the production process formulate a method for adaptively selecting suitable features according to the characteristics of the quality data, and merge the adaptively selected features with the original data to reduce the dimensionality while increasing the recognition accuracy, and introduce information judgment operators and entropy increase
  • the theory-improved thinking evolution algorithm improves the accuracy and search ability of the algorithm.
  • the beneficial effects of the present invention are: can adaptively select suitable features according to the characteristics of quality data to ensure the accuracy of recognition; the use of data fusion methods can enhance the performance of quality trend prediction , To ensure that the recognizer has good training efficiency; the improved algorithm can ensure that the classification has good fault tolerance and stronger classification ability.
  • FIG. 1 is a flowchart of this method
  • Figure 2 is an image of 9 modes; (a) normal mode; (b) mixed mode; (c) periodic mode; (e) system mode; (f) layered mode; (g) rising trend; (h) falling Trend; (i) Up-step trend; (j) Down-step mode;
  • Figure 3 is a distribution diagram of the corresponding features of the 9 modes
  • Figure 4 is a schematic diagram of feature fusion
  • FIG. 5 is a flowchart of the improved thinking evolution algorithm
  • Figure 6 is a schematic diagram of the particle simplification of the mind evolution algorithm
  • the invention takes into account the dynamic and time-varying characteristics of quality data during production and introduces an adaptive feature selection method, and uses the KPCA data fusion method to fuse dynamic data and adaptive features for accurate identification, and finally uses the MLPNN network combined with an improved thinking evolution algorithm , Established an accurate recognition model.
  • Figure 1 shows the overall flow chart of the method, and the following is a specific expansion for each module.
  • Step 1 Generate the data needed for model building.
  • FIG. 1 shows graphs with 9 modes: normal mode, mixed mode, periodic mode, hierarchical mode, system mode, upward trend mode, downward trend mode, and upper step
  • the specific methods for data of mode and step mode are as follows:
  • ⁇ is the mean value of the quality data
  • R(t) is the random deviation of the normal distribution at time t
  • d(t) is the deviation caused by abnormal factors in the production process.
  • ⁇ is the mean value of the quality data
  • r(t) is the standard normal random distribution function
  • the minimum value of ⁇ is recommended to be 0.05
  • the maximum is 0.5;
  • MIX Mixed mode of production process
  • d(t) is the fluctuation caused by abnormal factors. It is recommended that the minimum value of ⁇ is 0.05 and the maximum value is 0.5; W is the disturbance factor of the production process 0 or 1, and m is the disturbance amplitude of the production process. It is recommended to take [1.5,2.5 ].
  • d(t) is the fluctuation caused by abnormal factors.
  • d(t) a ⁇ sin(2 ⁇ t/T) ⁇ .
  • ⁇ is 0.05 and the maximum value is 0.5; it is recommended that periodic production process disturbances
  • the minimum value range of a is [1,1.5], the maximum value range of a is [2.5, 3]; the minimum value range of the cycle T of the production process disturbance is recommended to be [4,8], the maximum value range of T Is [10,16].
  • d(t) is the fluctuation caused by abnormal factors. It is recommended that the minimum value of ⁇ is 0.05 and the maximum value is 0.5; it is recommended that the minimum value range of the disturbance proportional relationship K of the production process data is [0.1, 0.3], and the maximum value of K The range is [0.4, 0.6].
  • d(t) is the fluctuation caused by abnormal factors.
  • d(t) d ⁇ (-1) t ⁇
  • the recommended value of ⁇ is the smallest 0.05 and the largest 0.5
  • the deviation degree of the recommended quality data is d
  • the minimum value is 1
  • the maximum value of d is 3.
  • d(t) is the fluctuation caused by abnormal factors.
  • d(t) ⁇ g ⁇ t ⁇
  • the recommended value of ⁇ is the smallest 0.05 and the largest 0.5
  • the slope g of the quality data is recommended to be in the range [0.05,0.1];
  • d(t) is the fluctuation caused by abnormal factors.
  • Step 2 Establishment of feature adaptive processing module
  • Step 1 Establish a feature extraction model, and extract statistical features of quality data including: MEAN, VS, STD, SKEW, KURT, A; shape features of quality data include: SL, NC1, NC2, APML, APLS, AASL, ACLPI, SRANGE, SB, PSMLSC, REAE, ABDPE.
  • the feature extraction method is as follows:
  • t i is the distance from the origin of the inspection quality data point at the i-th time, It is the average value of the distance from the origin of the P inspection quality data points.
  • y i is the ith quality data.
  • S jk is the slope of the quality data area composed of the j-th midpoint and the k-th midpoint.
  • the coordinates of the midpoint of each area are
  • the least squares regression line slope identifier of the curve formed by the quality data If the least squares slope of the curve formed by the quality data is greater than 0, SB is 1, otherwise SB is 0;
  • the absolute value of the difference between the slope of the overall least squares regression line and the average of the slopes of the least squares regression line of the two regions Where B is the slope of the least squares regression line of the overall curve formed by the quality data, It is the average of the slope of the least squares regression line of the two regions.
  • Step 2 Establish an adaptive feature selection model, which establishes a method for introducing errors to affect calculations.
  • different types of quality data imply different important features. Based on different data, important features are selected. The following are the specific method steps of the model establishment:
  • the hidden layer is Nodes
  • m is the number of input nodes
  • n is the number of output nodes and used to select training data from the above quality data features for MLPNN neural network training
  • Step 3 Data feature fusion module
  • This module uses the KPCA data dimensionality reduction method to fuse the original data and feature data as shown in Figure 4.
  • the specific implementation steps are as follows:
  • the load factor of the combined data is selected and the fusion data is generated.
  • K(F a1 ,F a2 ) ⁇ (F a1 ) ⁇ (F a2 ) T where (a1 ⁇ (1,H),a2 ⁇ (1,H));
  • K is the H ⁇ H matrix.
  • Step 4 Establish a quality trend prediction module
  • the data after adaptive feature fusion establishes a three-layer perceptron MLPNN neural network model, uses an improved mind evolution algorithm to optimize the weights and thresholds of the MLPNN neural network, and the training data uses the data fused in the third step.
  • the general thinking evolutionary algorithm is mainly through the iterative optimization learning method, all individuals in the evolution process are called groups, and a group is divided into several subgroups. Subgroups include winning groups and temporary groups.
  • the superior particle swarm and the temporary particle swarm are based on the optimal particle as the center and randomly generated subgroups during the flight evolution process.
  • the mutual information theory is introduced here to determine the degree of subgroup evolution. When the information level of the particles generated in the subgroup and the leader particle is greater than 85%, the particle is considered as an invalid particle.
  • Fig. 6 is a schematic diagram of the process of particle evolution.
  • the degree of mutual information between the central particle and individual particles can be expressed as:
  • step c Convergence operation operator: A winning subgroup and B temporary subgroup are generated with the selected particle as the winning center and the temporary center obeys the normal distribution. The individual particle scores in all subgroups are calculated and the winner is selected. Regenerate subgroups as the center. At the same time, step c is performed. If the individual particles and the central particle in the subgroup contain the appropriate information, continue to the next step, otherwise, regenerate particles and proceed to step c.
  • Alienation operator exchange of information between all mature winning subgroups and temporary subgroups, the winning subgroups M 1 , M 2 ... M A , and the temporary subgroups G 1 , G 2 , G 3 ... G B In competition, if the score of a mature temporary subgroup is greater than the score of a mature winning subgroup, the release of the winning particles is replaced by the temporary subgroup and the temporary subgroup is refilled.
  • step d the subgroup is generated with the optimal particle as the center, and the generation of new individual particles is required.
  • the generation of new individual particles introduces entropy change theory to increase the degree of chaos of particle generation, and adds entropy change inertia coefficient to the particle generation formula.
  • the inertia coefficient can improve the search ability for excellent particles.
  • R D m c + ⁇ (2 ⁇ Z 1 ⁇ L′ -1), where ⁇ is 0.5-1, Z 1 ⁇ L′ is the normal distribution of 1 ⁇ L′ 0- Numerical matrix of 1, m c is the central particle.
  • This process is a chaotic search process in this range with a certain particle as the center, to prevent the particle search from over-fitting search, start to expand the range and gradually approach the optimal particle to improve the search accuracy around the central particle.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Educational Administration (AREA)
  • Development Economics (AREA)
  • Software Systems (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Game Theory and Decision Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Primary Health Care (AREA)

Abstract

一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法，该方法主要包括三个模块：特征自适应处理模块、数据融合模块、质量趋势预测模块。该方法的实现主要包括以下几个步骤：(1)设计相应参数生成建立该模型的数据；(2)应用误差影响程度算法建立特征自适应选择模块；(3)应用KPCA数据融合方法建立数据融合模块；(4)应用改进的思维进化算法优化多层感知器(MLPNN)网络建立质量趋势预测模块。通过建立该方法，能够实施在质量趋势预测领域，能够自适应根据不同类型的数据选择不同的特征进行预测，并且应用数据融合、算法改进提高产品质量趋势预测的精度，及时采取适当的方式进行修正。

Description

一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法 技术领域
本发明属于智能制造及智能质量监控领域，特别涉及一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法。
背景技术
在航天、航空、船舶、汽车等领域应用广泛，由于在这些领域的产品大部分需要的质量数据精度相对较高，在生产过程中产品会时常会受到人、机、料、法、环、测等多种因素的影响，同时质量数据常常具有时变性、非线性、相关性和动态等特性常常会导致产品产生较大的质量问题无法及时预测趋势提出修改措施，严重影响产品的使用性能和质量。
一直以来控制图作为质量控制及预测的辅助手段广泛用于生产过程中。控制图控制质量数据的上下限捕捉产品质量的波动与异常，随着技术的发展与生产节奏的加快只通过控制图的上下限来判断质量波动经常会出现较大的质量问题，无法适应现代加工数据采集等现代化手段进行质量控制。目前有很多学者也开始研究控制图的模式以追求质量的精准控制，但是大多数研究的模式较少、不对混合模式进行研究或是无法适应数据的动态变化不能对数据进行自适应精准控制、识别、预测并且大多采用线下识别，智能化程度不够。
为实现产品智能化质量控制、提高产品质量提出一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法。本发明可实现智能化质量预测与控制，及时提出措施进行修正并提高产品质量。
发明内容
本发明的目的：针对质量数据的动态性、时变性等特性，基于控制图的多种模式提出一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法实现多种模式识别并根据质量数据的时变性自适应选择特征预测动态生产过程中的线上产品质量趋势，同时还应用改进思维进化算法优化MLPNN网络提高质量趋势的识别精度。解决现有质量趋势智能化识别模式少、精度低、不能自适应变化、质量控制力度不够的问题，提高良品率达到智能化生产及智能化趋势预测。
本发明提出来一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法解决上述问题，构建趋势预测模型以便能够预测异常状态。为实现上述目的，本发明采用以下技术方案：
一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法包括以下步骤：
步骤1：生成模型建立所需数据。
生成9种模式的数据分别为正常模式(NOR)、周期模式(CYC)、系统模式(SYS)、分层模式(STR)、上升趋势模式(IT)、下降趋势模式(DT)、向上阶跃模式(US)、向下阶跃模式(DS)、混合模式(MIX)。
步骤2：特征自适应处理模块建立。
在质量趋势预测的过程中，一般的常用方法直接应用数据进行趋势的预测精度不高，该步骤的建立可将特征合理选择并进一步提高智能化程度。该模块建立分为两步：
第一步：建立特征提取模型，根据研究表明提取的数据特征包括以下统计特征和形状特征更有说服力；其中，质量数据的统计特征包括：MEAN、VS、STD、SKEW、KURT、A；质量数据的形状特征包括：SL、NC1、NC2、APML、APLS、AASL、ACLPI、SRANGE、SB、PSMLSC、REAE、ABDPE。
下列特征为质量数据在观测窗口所提取的，各个符号的含义如下：
MEAE：质量数据的均值；
VS：质量数据的均方值；
STD：质量数据的标准差；
SKEW：质量数据的偏态系数；
KURT：质量数据的峰态系数；
A：质量数据的自相关系数；
SL：质量数据拟合的最小二乘回归线斜率；
NC1：质量数据所成曲线与平均值所成线的交叉点数；
NC2：质量数据所成曲线与最小二乘回归线的交叉点数；
APML：质量数据所成曲线与平均线之间的面积；
APLS：质量数据所成曲线与最小二乘回归线之间的面积；
AASL：质量数据所成曲线分为四个区域，各区域中点组合连线所形成曲线斜率的平均值
ACLPI：APML与质量数据的标准差之比；
SRANGE：质量数据所成曲线分为四个区域，各区域中点连线所成斜率的最大值与最小值之差；
SB：质量数据最小二乘回归线的斜率标识符；
PSMLSC：质量数据与中心线交点、最小二乘回归线交点和的平均值；
REAE：质量数据的MSE与分成四个区域数据MSE平均值的误差之比；
ABDPE：将质量数据所成曲线分为两个区域时，整体质量数据所成曲线最小二乘回归线斜率与两个区域最小二乘回归线斜率平均值之差的绝对值。
第二步：建立自适应特征选取模型，该模型建立引入误差影响计算的方法。若有新的数据可根据数据特性应用该模块选择合适的特征。基于不同种类的质量数据，根据特征重要的程度选取不同的特征。
建立该模型的方法如下：
a)设质量数据特征有N个，将质量数据特征进行预处理；
b)建立初始MLPNN神经网络，并应用处理好的质量数据特征进行训练；
c)若有新数据输入，预处理新质量数据特征并将每种质量数据特征数值分别增加10％、减少10％共生成2N组数据，将每组数据分别带入MLPNN中进行识别得到2N个误差值，然后将每种质量数据特征对应的增加10％、减少10％所得误差取平均值得到N个误差值；
d)将误差由大到小排序选取前85％的特征规定为影响该质量数据程度较高的特征。
e)完成自适应质量数据特征选择模型。
步骤3：数据特征融合模块。
为了能更加精准的进行趋势的预测，将原始数据与自适应选择的数据进行融合。若将原始数据直接与选择的特征进行质量趋势的预测，输入数据太过庞大、大大增加了模型的计算复杂性，在数据降维方法中分为线性数据降维、非线性数据降维，该模块采用KPCA数据降维方法将原始数据与特征数据进行融合，建立该模块的方法步骤如下:
a)将组合数据进行标准化、中心化；
b)构造组合数据的核函数，将数据映射到高维度中并计算核矩阵；
c)计算的特征值、选择特征向量；
d)进行数据降维、融合；
步骤4：建立质量趋势预测模块。
基于自适应特征融合后的数据建立3层感知器MLPNN神经网络模型，使用改进的思维进化算法对MLPNN神经网络权值、阈值进行优化。
一般的思维进化算法主要是通过迭代优化的学习方式，进化过程中所有的个体叫做群体，一个群体分为若干个子群。子群包括优胜子群和临时子群。在思维进化时优胜子群和临时子群在飞行进化过程中是以最优的粒子为中心随机产生的子群没有任何的限制，子群内粒子之间包含的信息程度也无法判断，从而产生对进化没有意义的粒子。在这里引入互信息理论来 判定子群进化的优劣程度。当子群内产生的粒子与中心粒子包含的信息程度大于85％则认为此粒子为无效粒子，同时若某粒子的得分大于中心粒子则保留，否则就重新生成该粒子。
建立改进思维进化算法方法步骤如下：
a)初始化种群产生，在空间内生成种群；
b)在初始化种群中选择得分较高的粒子分别作为优胜子群中心、临时子群中心并产生子群。
c)引入信息判断算子：分别计算各优胜子群、临时子群的粒子与本身子群的中心粒子之间的互信息程度，若个体粒子与中心粒子的互信息程度大于85％则认为是相似粒子，同时若个体粒子比中心粒子得分高则保留，否则释放；
d)趋同操作算子：计算所有子群中的个体粒子得分，选出优胜者作为中心重新生成子群。同时进行步骤c操作，若子群中个体粒子与中心粒子包含信息程度合适则继续下一步操作，否则重新生成粒子并进行步骤c操作。
e)判断是否各子群是否成熟，若子群成熟继续下一步操作，否则继续步骤d操作。
f)异化操作算子：将所有的成熟优胜子群和临时子群进行信息的交流，得分较高的临时子群将代替得分较低的临时子群。
g)判断若成熟临时子群中没有得分超过成熟优胜子群则跳出循环，否则重复进行步骤c-步骤g操作。
同时上述操作中在步骤d中以新的中心粒子生成新的子群时引入熵变理论增加粒子生成的混沌程度，增加熵变惯性系数。在搜索前期要求范围较大且与中心粒子的信息保持一定关系，后期则要求收敛较快，而引入惯性系数可提高对优秀粒子的搜索能力。
通过分析生产过程中的异常情况，根据质量数据的特点制定自适应选择合适特征的方法，并将自适应选择的特征与原始数据融合降低维度的同时增加识别精度，引入信息判断算子和熵增理论改进的思维进化算法提高算法的精度及搜索能力，本发明的有益效果是：能够根据质量数据的特性自适应选择合适的特征保证识别的精度；数据融合方法的使用可增强质量趋势预测的性能，保证识别器具有良好的训练效率；改进的算法能够保证分类时具有良好的容错性、分类能力更强。
附图说明
下面结合附图和实施例对本发明进行进一步说明
图1是本方法的流程图；
图2是9种模式的图像；(a)正常模式；(b)混合模式；(c)周期模式；(e)系统模式；(f)分层模式；(g)上升趋势；(h)下降趋势；(i)上阶跃趋势；(j)下阶跃模式；
图3是9种模式对应特征的分布图；
图4是特征融合的示意图
图5是改进思维进化算法的流程图
图6是思维进化算法粒子简化的示意图
具体实施方式
本发明考虑生产时质量数据的动态性、时变特性引入自适应特征选择方法，并且为准确识别应用KPCA数据融合方法将动态数据与自适应特征进行融合，最终应用MLPNN网络结合改进的思维进化算法，建立了精准的识别模型。下面结合所示附图以及具体的方法建立的实施方式，对本发明的建立做进一步的描述：
如图1所示为该方法实现的总体流程图，下面是针对每一个模块进行具体的展开。
步骤1：生成模型建立所需数据。
分析在生产过程中会出现的模式。受实际生产的影响，产品质量的波动会呈一定的规律出现并影响接下来的生产。设计不同的参数来生成不同类型的趋势图如图2所示为9种模式的图形：正常模式、混合模式、周期模式、分层模式、系统模式、上趋势模式、下趋势模式、上阶跃模式、下阶跃模式的数据，具体方法如下：
仿真公式：y＝μ+R(t)+d(t)
其中：μ为质量数据的均值，R(t)为t时刻的正态分布随机偏差，d(t)为生产过程出现异常因素导致的偏差，下面是生产时各模式公式以及参数说明:
(1)生产过程的正常模式(NOR)：
y＝μ+r(t)×σ+d(t)
其中：μ为质量数据的均值，r(t)为标准正态随机的分布函数，d(t)为异常因素引起的波动，正常模式下d(t)＝鲀，建议σ取值最小为0.05，最大为0.5；
(2)生产过程的混合模式(MIX)：
y＝μ+r(t)×σ+(-1) W×m×σ
其中：d(t)为异常因素引起的波动，建议σ取值最小为0.05，最大为0.5；W为生产过程的扰动因子0或1，m为生产过程的扰动幅值建议取[1.5,2.5]。
(3)生产过程的周期模式(CYC)：
y＝μ+r(t)×σ+a×sin(2πt/T)×σ
其中：d(t)为异常因素引起的波动，周期模式下d(t)＝a×sin(2πt/T)×σ，建议σ取值最小为0.05，最大为0.5；建议周期性生产过程扰动的幅值a最小取值范围[1,1.5]，a最大取值范围为[2.5，3]；建议生产过程扰动的周期T的最小取值范围为[4,8],T最大取值范围为[10,16]。
(4)生产过程的分层模式(STR)：
y＝μ+r(t)×σ×K
其中：d(t)为异常因素引起的波动，建议σ取值最小为0.05，最大为0.5；建议生产过程数据的扰动比例关系K的最小取值范围为[0.1,0.3]，K最大取值范围为[0.4,0.6]。
(5)生产过程的系统模式(SYS)：
y＝μ+r(t)×σ+d×(-1) t×σ
其中：d(t)为异常因素引起的波动，系统模式下d(t)＝d×(-1) t×σ，建议σ取值最小为0.05，最大为0.5；建议质量数据的偏离程度d最小取值为1；d最大取值为3。
(6)生产过程的上下趋势模式(IT/DT)：
y＝μ+r(t)×σ±g×t×σ
其中：d(t)为异常因素引起的波动，趋势模式下d(t)＝±g×t×σ，建议σ取值最小为0.05，最大为0.5；建议质量数据的斜率g取值范围为[0.05,0.1]；
(7)生产过程的上下阶跃模式(US/DS)：
y＝μ+r(t)×σ±b×s×σ
其中：d(t)为异常因素引起的波动，阶跃模式下d(t)＝±b×s×σ，建议σ取值最小为0.05，最大为0.5；当t<P时b＝0，t>P时b＝1；建议质量数据的随机阶跃位置P最小取值为[4,9],P最大取值范围为[13,19]；建议阶跃幅值s最小取值范围为[0.5,1.5],s最大取值范围为[2.5,3.5]。
步骤2：特征自适应处理模块建立
在质量趋势预测的过程中，一般的常用方法直接应用数据进行趋势的预测，研究表明合理的使用质量数据特征将大大提高预测精准。提取质量数据特征，具体特征分布如图3所示。建立特征自适应处理模块分为两步：
第一步：建立特征提取模型，提取质量数据的统计特征包括：MEAN、VS、STD、SKEW、KURT、A；质量数据的形状特征包括：SL、NC1、NC2、APML、APLS、AASL、ACLPI、SRANGE、SB、PSMLSC、REAE、ABDPE。设质量数据为Y＝(y 1,y 2,y 3….y P)(i＝1,2,3…P)，特征提取如下方法：
质量数据的均值公式：
Figure PCTCN2020127970-appb-000001
质量数据的均方值公式：
Figure PCTCN2020127970-appb-000002
质量数据的标准差
Figure PCTCN2020127970-appb-000003
质量数据的偏态系数公式：
Figure PCTCN2020127970-appb-000004
质量数据的峰态系数公式：
Figure PCTCN2020127970-appb-000005
质量数据的自相关系数：
Figure PCTCN2020127970-appb-000006
质量数据所成曲线的最小二乘回归线斜率：
Figure PCTCN2020127970-appb-000007
t i为第i次时间的检测质量数据点到原点的距离，
Figure PCTCN2020127970-appb-000008
为P次检测质量数据点距离原点的平均值。y i为第i次质量数据。
质量数据所成曲线与平均值所成曲线的交叉点数：如果(y i-MEAN)(y i+1-MEAN)<0，则NC1加1；
质量数据所成曲线与最小二乘回归线的交叉点数：如果(y i-L(y i))(y i+1-L(y i+1))<0，则NC2加1；
质量数据所成曲线与平均线之间所成的面积：APML；
质量数据所成曲线与最小二乘回归线所成的面积：APLS；
将质量数据分成四段区域，两两区域中点连线得到斜率平均值：
Figure PCTCN2020127970-appb-000009
Figure PCTCN2020127970-appb-000010
其中，S jk为第j个中点和第k个中点组成的质量数据区域斜率。各区域中点坐标为
Figure PCTCN2020127970-appb-000011
质量数据与中心线组成的面积与质量数据的标准差之比：
Figure PCTCN2020127970-appb-000012
质量数据分为四个区域，各区域中点连线所成斜率的最大值与最小值之差：SRANGE＝max(S jk)-min(S jk)；(j＝123；k＝234；J<k)
质量数据所成曲线的最小二乘回归线斜率标识符：如果质量数据所成曲线的最小二乘斜率大于0则SB为1，否则SB为0；
质量数据所成曲线与中心线交点、最小二乘回归线交点和的平均值：
Figure PCTCN2020127970-appb-000013
质量数据的MSE与分成四个区域数据的MSE平均值的误差之比：
Figure PCTCN2020127970-appb-000014
Figure PCTCN2020127970-appb-000015
质量数据分为两个区域时，整体的最小二乘回归线斜率与两个区域最小二乘回归线斜率的平均值之差的绝对值：
Figure PCTCN2020127970-appb-000016
其中B为质量数据所成曲线整体的最小二乘回归线斜率，
Figure PCTCN2020127970-appb-000017
为两个区域的最小二乘回归线斜率平均值。
第二步：建立自适应特征选取模型，该模型建立引入误差影响计算的方法。在生产过程中不同种类的质量数据隐含的重要的特征不同，基于不同的数据，选取重要的特征。如下为该模型建立的具体方法步骤：
a)将质量数据特征进行预处理；
设特征矩阵F＝(f ab)(a＝1,2,3…H；b＝1,2,3…N)，其中有H个样本，N类特征，对质量数据特征进行标准化处理:
Figure PCTCN2020127970-appb-000018
其中：
Figure PCTCN2020127970-appb-000019
为第b类质量数据特征的均值，sb为第b类质量数据特征的标准差
Figure PCTCN2020127970-appb-000020
b)建立3层感知器神经网络，隐含层为
Figure PCTCN2020127970-appb-000021
个节点，m为输入节点数，n为输出节点数并用在上述质量数据特征中选取训练数据进行MLPNN神经网络训练；
c)若新输入质量数据特征[f 1,f 2,f 3….f N]，将质量数据特征预处理得到F′＝(f b′)(b＝1,2,3…N)，并使每种质量数据特征数值分别增加10％、减少10％共生成2N组数据[f 1 *,f 2 *,f 3 *….f 2N *]，其中f 1 *-f N *为每种质量数据特征数值增加10％数据，f N+1 *-f 2N *为将每种质量数据特征数值减少10％数据，将两组质量数据特征分别带入MLPNN中进行识别得到2N个趋势预测误差分别为[E 1 U,E 2 U,E 3 U….E N U]、[E 1 D,E 2 D,E 3 D….E N D]，然后将每种质量数据特征对应增加10％、减少10％所得趋势预测误差取平均值E i＝(E i U+E i D)/2(i＝123….N)得到N个趋势预测误差值[E 1,E 2,E 3….E N]；
d)将[E 1,E 2,E 3….E N]误差由大到小排序选取前85％的特征规定为影响该质量数据程度较高的特征。
e)完成自适应质量数据特征选择得到L(L<N)个特征，完成该模型的建立。为后面特征融合做准备。
步骤3：数据特征融合模块
建立数据特征融合模块，为了能更加精准的进行趋势的预测，将原始数据与自适应选择的数据进行融合。该模块采用KPCA数据降维方法将原始数据与特征数据进行融合如图4所示，具体实现的步骤方法如下：
a)将组合数据进行标准化、中心化；
设数据矩阵F＝(f ab)(a＝1,2,3…H；b＝1,2,3…(L+P))，其中有H个样本，每个样本有L+P个数据，对组合数据进行标准化处理:
Figure PCTCN2020127970-appb-000022
其中：
Figure PCTCN2020127970-appb-000023
组合数据中第b维数据的均值，s b为质量数据与质量数据特征组合后第b维数据的标准差
Figure PCTCN2020127970-appb-000024
Figure PCTCN2020127970-appb-000025
得到F a＝[f a1″,f a2″,f a3″….f a(L+P)″]，F″＝[F 1,F 2….F H]。
b)构造组合数据核函数，将处理的数据映射到高维度中，并计算组合数据的核矩阵；
根据组合数据的特征值λ及特征向量W选择组合数据的载荷因子并生成融合数据，这里引非线性映射Φ，可以将问题转换成Φ(F″)Φ(F″) TW＝λW。
引入组合数据的映射函数得到组合数据的向量矩阵W＝(w 1,w 2,w 3….w H)，w a为组合数据的基向量，映射数据的关系可表示为
Figure PCTCN2020127970-appb-000026
可得
Figure PCTCN2020127970-appb-000027
令
Figure PCTCN2020127970-appb-000028
得α＝[α 1,α 2….α H]
Figure PCTCN2020127970-appb-000029
最终，得到Φ(F″)Φ(F″) TΦ(F″)α＝λΦ(F″)α
这里引入核技巧方程两边同时乘Φ(F″) T得到：
Figure PCTCN2020127970-appb-000030
c)计算组合数据的特征值、选择特征向量；
引入组合数据的核函数并进行核技巧转换K(F a1,F a2)＝Φ(F a1)Φ(F a2) T其中(a1∈(1,H),a2∈(1,H))；此时可以定义当特征值累计贡献率大于85％时的前g个主成分可表示原数据矩阵，此时的K为H×H的矩阵选择后的载荷因子矩阵为A＝[α 1,α 2….α g]。
d)进行数据降维；
将组合数据进行核变换后数据和选择的载荷向量相乘得到融合后的数据X *＝K×A即为：
Figure PCTCN2020127970-appb-000031
步骤4：建立质量趋势预测模块
自适应特征融合后的数据建立3层感知器MLPNN神经网络模型，使用改进的思维进化算法对MLPNN神经网络权值、阈值进行优化，训练数据采用第3步融合的数据。
一般的思维进化算法主要是通过迭代优化的学习方式，进化过程中所有的个体叫做群体，一个群体分为若干个子群。子群包括优胜群体和临时群体。在思维进化时优胜粒子群和临时粒子群在飞行进化过程中是以最优的粒子为中心随机产生的子群没有任何的限制，子群内粒子之间包含的信息程度也无法判断，从而很可能产生很多对进化没有任何意义的粒子，对进化没有意义。在这里引入互信息理论来判定子群进化的有优劣程度。当子群内产生的粒子与领头粒子包含的信息程度大于85％则认为此粒子为无效粒子，同时若某粒子的得分大于领头粒子则保留，否则就以领头粒子为中心生成粒子，如图5所示为具体操作的步骤，图6为粒子进化的过程示意图。
建立改进思维进化算法方法步骤如下：
a)初始化种群产生，在空间内生成具有T个粒子的种群，并根据适应函数计算各个粒子的得分；
b)在初始化种群中选择A个得分较高的粒子作为优胜子群中心，选择B个得分较高的粒子作为临时子群中心。设定各子群大小为T *，其中T *＝T/(A+B)。
c)引入信息判断算子：设粒子长度为L′分别计算优胜子群各粒子中心与相对应
Figure PCTCN2020127970-appb-000032
(t 1＝1,2,3,…A)、临时子群各粒子中心与
Figure PCTCN2020127970-appb-000033
(t 2＝1,2,3,…B)之间的互信息
Figure PCTCN2020127970-appb-000034
及
Figure PCTCN2020127970-appb-000035
其中(i＝1,2,3….T *,l＝123…..L′t 1＝1,2,3,…A，t 2＝1,2,3,…B)，若两个子群的个体粒子与中心粒子互信息程度大于85％则认为粒子为相似粒子，同时制定规则若粒子得分大于中心粒子保留，否则释放；
以优胜子群的互信息程度计算为例：
首先，计算优胜子群的信息熵公式如下：
Figure PCTCN2020127970-appb-000036
其中
Figure PCTCN2020127970-appb-000037
表示个体粒子
Figure PCTCN2020127970-appb-000038
的信息量，这里通常对数所用的底为2，e或10。
对于中心粒子和个体粒子互信息程度可以表达为：
Figure PCTCN2020127970-appb-000039
同理对于临时子群G也应用上述公式进行计算。
d)趋同操作算子：分别以选定的粒子为优胜中心、临时中心服从正态分布产生A个优胜子群与B个临时子群，计算所有子群中的个体粒子得分，选出优胜者作为中心重新生成子群。同时进行步骤c操作，若子群中个体粒子与中心粒子包含信息程度合适则继续下一步操作，否则重新生成粒子并进行步骤c操作。
e)判断是否各子群是否成熟，若子群成熟继续下一步操作，否则继续步骤d操作。
f)异化操作算子：将所有的成熟优胜子群和临时子群进行信息的交流，优胜子群M 1,M 2…M A,临时子群G 1,G 2,G 3…G B进行竞争，若存在成熟的临时子群得分大于成熟的优胜子群得分，则优胜粒子的释放被临时子群替代并且重新补充临时子群。
g)判断若成熟临时子群中得分没有超过成熟优胜子群得分则跳出循环，否则重复进行步骤c-步骤g操作。
同时上述操作中在步骤d中以最优粒子为中心生成子群，则需要新个体粒子的生成。新个体粒子生成引入熵变理论增加粒子生成的混沌程度，将熵变惯性系数加入粒子生成公式中。在搜索前期要求范围较大且与中心粒子的信息保持一定关系，后期则要求收敛较快，而引入惯性系数可提高对优秀粒子的搜索能力。
对于生成新的个体粒子的方法给出以下方式进行：
a)为扩大范围搜索，根据公式生成初始粒子矩阵R D×L′，其中粒子有D个、每个粒子长度为L′。粒子生成的公式为：R d＝m c+×θ×(2×Z 1×L′-1)，其中θ取0.5-1，Z 1×L′为1×L′的正态分布0-1的数值矩阵，m c为中心粒子。
b)对R D×L′矩阵进行粒子的信息熵计算并进行粒子的混沌迭代。设置迭代次数为K max次，k为当次迭代次数的值，最终选择最佳的粒子。将粒子矩阵转换成S 1×(D×L′)的矩阵，整体粒子的信息熵计算公式为：
Figure PCTCN2020127970-appb-000040
应用公式得到变化的权值:
Figure PCTCN2020127970-appb-000041
c)改变公式为R d+1＝wR d+(1-w)×θ×(2×Z 1×L′-1)，循环K max次后得到新的R D×L′，同时在循环过程中对于每个粒子得分高的会保留，循环结束后最终选择矩阵中得分最高的粒子为新粒子。
此过程为在某一粒子为中心在此范围内混沌搜索的过程，防止粒子搜索过拟合搜索，开始扩大范围最后逐渐向最优粒子靠近，提高在中心粒子周围的搜索精度。

Claims (5)

  1. 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法，其特征在于：该方法的建模分为以下四个步骤：
    步骤1：生成产品质量模型建立所需数据，模拟产品的质量数据；该模拟产品的质量数据的模式包括正常模式、周期模式、混合模式、系统模式、上升趋势模式、下降趋势模式、向上阶跃模式、向下阶跃模式；
    步骤2：建立特征自适应处理模块，根据步骤1所述产品数据进行预处理，该特征自适应处理模块建立分为两步：提取产品质量数据特征及应用初始化MLPNN网络建立自适应特征选择模型；
    步骤3：建立数据特征融合模块，应用KPCA方法实现特征的融合、数据的降维，从而简化后续产品质量趋势预测模块；
    步骤4：建立产品质量趋势预测模块，利用改进的思维进化算法优化MLPNN神经网络，该模型的优化目标为预测准确率，通过增加互信息判断算子同时引入熵变理论使算法得到预测模型。
  2. 根据权利要求1所述一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法，其特征在于，步骤1中生成9种模式的数据分别为：正常模式NOR、周期模式CYC、系统模式SYS、分层模式STR、上升趋势模式IT、下降趋势模式DT、向上阶跃模式US、向下阶跃模式DS、混合模式MIX。
  3. 根据权利要求1所述一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法，其特征在于，步骤2中第一步采用的质量数据的统计特征包括：MEAN、VS、STD、SKEW、KURT、A；质量数据的形状特征包括：SL、NC1、NC2、APML、APLS、AASL、ACLPI、SRANGE、SB、PSMLSC、REAE、ABDPE；第二步采用初始化MLPNN网络应用误差影响程度算法自适应选择特征。
  4. 根据权利要求3所述一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法，其特征在于，步骤3中应用KPCA法将所述自适应选择的特征与原始产品质量数据进行融合。
  5. 根据权利要求1所述一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法，其特征在于，步骤4中设计互信息判断算子改进思维进化算法，应用熵变理论改进思维进化算法的粒子生成方式。
PCT/CN2020/127970 2020-05-14 2020-11-11 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法 WO2021227406A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010405648.2 2020-05-14
CN202010405648.2A CN111598435B (zh) 2020-05-14 2020-05-14 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法

Publications (1)

Publication Number Publication Date
WO2021227406A1 true WO2021227406A1 (zh) 2021-11-18

Family

ID=72190810

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/127970 WO2021227406A1 (zh) 2020-05-14 2020-11-11 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法

Country Status (2)

Country Link
CN (1) CN111598435B (zh)
WO (1) WO2021227406A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114997070A (zh) * 2022-07-15 2022-09-02 合肥中科迪宏自动化有限公司 控制图模式识别模型的训练方法及控制图模式识别方法
CN117310118A (zh) * 2023-11-28 2023-12-29 济南中安数码科技有限公司 一种地下水污染可视化监测方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111598435B (zh) * 2020-05-14 2023-08-04 北京工业大学 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法
CN114496209B (zh) * 2022-02-18 2022-09-27 青岛市中心血站 一种献血智能决策方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105139093A (zh) * 2015-09-07 2015-12-09 河海大学 基于Boosting算法和支持向量机的洪水预报方法
CN106769030A (zh) * 2016-11-10 2017-05-31 浙江工业大学 一种基于mea‑bp神经网络算法的轴承状态跟踪与预测方法
CN110047015A (zh) * 2019-04-22 2019-07-23 水利部信息中心 一种融合kpca和思维优化bp神经网络的用水总量预测方法
CN111027745A (zh) * 2019-11-08 2020-04-17 广东财经大学 一种基于自适应特征提取的股票指数预测方法
CN111598435A (zh) * 2020-05-14 2020-08-28 北京工业大学 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105139093A (zh) * 2015-09-07 2015-12-09 河海大学 基于Boosting算法和支持向量机的洪水预报方法
CN106769030A (zh) * 2016-11-10 2017-05-31 浙江工业大学 一种基于mea‑bp神经网络算法的轴承状态跟踪与预测方法
CN110047015A (zh) * 2019-04-22 2019-07-23 水利部信息中心 一种融合kpca和思维优化bp神经网络的用水总量预测方法
CN111027745A (zh) * 2019-11-08 2020-04-17 广东财经大学 一种基于自适应特征提取的股票指数预测方法
CN111598435A (zh) * 2020-05-14 2020-08-28 北京工业大学 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
CHU HONG-YAN,LI PENG,CAI LI-GANG,LI FENG-GUANG: "Pattern Recognition for Printing Quality Control Chart Based on MC Method and BP Neural Network", JOURNAL OF BEIJING UNIVERSITY OF TECHNOLOGY, vol. 37, no. 6, 15 June 2011 (2011-06-15), pages 816 - 821, XP055866955, ISSN: 0254-0037, DOI: 10.13423/j.cnki.cjcmi.007654 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114997070A (zh) * 2022-07-15 2022-09-02 合肥中科迪宏自动化有限公司 控制图模式识别模型的训练方法及控制图模式识别方法
CN117310118A (zh) * 2023-11-28 2023-12-29 济南中安数码科技有限公司 一种地下水污染可视化监测方法
CN117310118B (zh) * 2023-11-28 2024-03-08 济南中安数码科技有限公司 一种地下水污染可视化监测方法

Also Published As

Publication number Publication date
CN111598435A (zh) 2020-08-28
CN111598435B (zh) 2023-08-04

Similar Documents

Publication Publication Date Title
WO2021227406A1 (zh) 一种基于自适应特征选择及改进思维进化算法的质量趋势预测方法
CN106599797B (zh) 一种基于局部并行神经网络的红外人脸识别方法
CN109815886A (zh) 一种基于改进YOLOv3的行人和车辆检测方法及系统
CN110470259A (zh) 基于lstm的滑坡位移动态预测方法
CN109490814A (zh) 基于深度学习和支持向量数据描述的计量自动化终端故障诊断方法
CN110009108A (zh) 一种全新的量子超限学习机
He et al. Improved particle swarm optimization algorithms for aerodynamic shape optimization of high-speed train
Tung et al. Falcon: neural fuzzy control and decision systems using FKP and PFKP clustering algorithms
CN108647772A (zh) 一种用于边坡监测数据粗差剔除的方法
CN109546648A (zh) 一种考虑预防策略的大规模电力系统自动发电控制方法
CN106295691A (zh) 单分类svm的特征选择和参数同步优化方法
Li Computer network connection enhancement optimization algorithm based on convolutional neural network
Wang et al. Parameters optimization of SVM based on the swarm intelligence
Lizhe et al. BP network model optimized by adaptive genetic algorithms and the application on quality evaluation for class teaching
Han et al. Endpoint prediction model of basic oxygen furnace steelmaking based on PSO-ICA and RBF neural network
CN116565877A (zh) 基于谱聚类分析的电压自动分区控制方法
CN110288002A (zh) 一种基于稀疏正交神经网络的图像分类方法
CN113420492A (zh) 一种基于gan与gru神经网络的风光火耦合系统频率响应模型建模方法
Guo et al. Short-Term Photovoltaic Power-Forecasting based on Machine Learning
CN113054653A (zh) 基于VGGNet-SVM的电力系统暂态稳定评估方法
Yin et al. Forecasting the intrinsic viscosity of polyester based on improved extreme learning machine
Dingxue et al. An adaptive particle swarm optimization algorithm and simulation
Zhou et al. Neural network-based optimization algorithm for intelligent manufacturing
Wang et al. Research on prediction and evaluation algorithm of sports athletes performance based on neural network
Zhou et al. Safety helmet recognition based on deep convolution neural networks

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20935184

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 26.04.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20935184

Country of ref document: EP

Kind code of ref document: A1