CN113011479A - Multi-source information fusion method for intelligent manufacturing - Google Patents
Multi-source information fusion method for intelligent manufacturing Download PDFInfo
- Publication number
- CN113011479A CN113011479A CN202110246821.3A CN202110246821A CN113011479A CN 113011479 A CN113011479 A CN 113011479A CN 202110246821 A CN202110246821 A CN 202110246821A CN 113011479 A CN113011479 A CN 113011479A
- Authority
- CN
- China
- Prior art keywords
- groups
- difference
- sensor
- fusion
- data
- 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
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 23
- 238000007500 overflow downdraw method Methods 0.000 title claims abstract description 19
- 230000004927 fusion Effects 0.000 claims abstract description 49
- 238000000034 method Methods 0.000 claims abstract description 43
- 238000005516 engineering process Methods 0.000 claims abstract description 4
- 238000001228 spectrum Methods 0.000 claims description 37
- 238000011156 evaluation Methods 0.000 claims description 16
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000000513 principal component analysis Methods 0.000 claims description 8
- 238000000491 multivariate analysis Methods 0.000 claims description 6
- 238000002329 infrared spectrum Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 claims description 2
- 238000004817 gas chromatography Methods 0.000 claims 2
- 238000004811 liquid chromatography Methods 0.000 claims 2
- 238000001237 Raman spectrum Methods 0.000 claims 1
- 238000000862 absorption spectrum Methods 0.000 claims 1
- 238000000295 emission spectrum Methods 0.000 claims 1
- 238000002189 fluorescence spectrum Methods 0.000 claims 1
- 238000004949 mass spectrometry Methods 0.000 claims 1
- 238000000655 nuclear magnetic resonance spectrum Methods 0.000 claims 1
- 238000002371 ultraviolet--visible spectrum Methods 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 3
- 239000000843 powder Substances 0.000 description 22
- TYAQXZHDAGZOEO-KXQOOQHDSA-N 1-myristoyl-2-stearoyl-sn-glycero-3-phosphocholine Chemical compound CCCCCCCCCCCCCCCCCC(=O)O[C@@H](COP([O-])(=O)OCC[N+](C)(C)C)COC(=O)CCCCCCCCCCCCC TYAQXZHDAGZOEO-KXQOOQHDSA-N 0.000 description 15
- 208000014446 corneal intraepithelial dyskeratosis-palmoplantar hyperkeratosis-laryngeal dyskeratosis syndrome Diseases 0.000 description 15
- 206010004542 Bezoar Diseases 0.000 description 13
- 239000006187 pill Substances 0.000 description 8
- 230000002159 abnormal effect Effects 0.000 description 7
- 238000010521 absorption reaction Methods 0.000 description 7
- 238000013441 quality evaluation Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 238000003070 Statistical process control Methods 0.000 description 3
- 239000003814 drug Substances 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 230000003595 spectral effect Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000011425 standardization method Methods 0.000 description 2
- 238000001429 visible spectrum Methods 0.000 description 2
- 230000002547 anomalous effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 239000009528 niuhuang qingxin Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000003359 percent control normalization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 229940126680 traditional chinese medicines Drugs 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The invention discloses a multi-source information fusion method for intelligent manufacturing, and belongs to the technical field of intelligent manufacturing. The invention provides a weight distribution method based on inter-group difference and intra-group difference, and establishes a multi-source information fusion method suitable for intelligent manufacturing. The method comprises the following steps: acquiring multidimensional information of a sample by adopting a multi-sensor technology; calculating the difference between groups and the difference between groups of different sensor information of the sample; determining the weight of each sensor according to the difference between sample groups and the ratio of the difference in the sample groups; and the information fusion strategy is adopted, and the multi-source information fusion is realized by combining the distribution strategy of the weight of each sensor. The invention creatively provides a weight distribution method based on Mahalanobis distance, Euclidean distance and variance analysis, and in addition, the weight distribution method of the difference between groups and the difference in groups is introduced for the first time, so that the multi-source information fusion of the intelligent manufacturing production process is realized.
Description
Technical Field
The invention belongs to the technical field of intelligent manufacturing, and particularly relates to a multi-source information fusion method based on a weight distribution strategy of difference between groups and difference between groups in an intelligent manufacturing production process.
Background
Smart manufacturing is a manufacturing process that employs artificial intelligence and integrates automation and manufacturing techniques. Traditional Chinese medicines, agricultural products, food and the like have the characteristics of complex components, multiple targets and multiple action ways, a comprehensive and reliable evaluation system is an important link of intelligent manufacturing and application, and the information of multiple sensors is fused, cooperated and complemented by multi-source information, so that the reliability of the system can be improved.
The determination of the weight of the sensor is a key problem of a multi-source information fusion algorithm, and the traditional weight determination method is a weighted average method, namely different sensors are endowed with the same weight factor. However, the weighted average method does not consider the properties of each sensor, and actually, even the same type of sensor may have different weights in the measurement process, and a weight distribution strategy based on the properties of the sample itself is urgently needed to be established.
The invention creatively provides a weight distribution method based on Mahalanobis distance, Euclidean distance and variance analysis, and in addition, a weight distribution strategy based on inter-group difference and intra-group difference is introduced for the first time, so that multi-source information fusion in the intelligent manufacturing production process is realized.
Disclosure of Invention
The invention aims to provide a weight distribution strategy based on the difference between groups and the difference in groups, and establish a multi-source information fusion method suitable for intelligent manufacturing.
To achieve the above object, wherein the method comprises:
step 1: acquiring information of a sample by a multi-sensor technology, and calculating the difference between groups and the difference between groups of different sensor information of the sample;
step 2: determining the weight of each sensor according to the difference between sample groups and the ratio of the difference in the sample groups;
and step 3: and a multi-source information fusion strategy is adopted, and the distribution weight of each sensor is combined to realize multi-source information fusion.
The evaluation of the difference between different sensor information groups and the difference between different sensor information groups comprises the following methods:
(1) the multi-dimensional space Euclidean/Mahalanobis distance evaluation method of the original data of the sample comprises the following steps:
step 1: detecting the attribute of each batch of samples, repeatedly measuring each sample at least three times, and taking an average value;
step 2: calculating the maximum Euclidean/Mahalanobis distance of each batch of sample attributes, and representing the intra-group difference of each batch of samples;
and step 3: calculating the maximum Euclidean/Mahalanobis distance of the center of each batch, and representing the difference among groups of samples of different batches;
and 4, step 4: establishing a multidimensional space Euclidean/Mahalanobis distance evaluation method of original data to obtain the difference between groups and in groups of samples.
(2) The Euclidean/Mahalanobis distance evaluation method of the sample principal component space comprises the following steps:
step 1: detecting the attribute of each batch of samples, repeatedly measuring each sample at least three times, and taking an average value;
step 2: adopting principal component analysis to reduce the dimension and obtaining the characteristic variable of each sensor;
and step 3: calculating the maximum Euclidean/Mahalanobis distance of the characteristic variables of the samples of each batch, representing the intra-group difference of the samples of each batch,
and 4, step 4: calculating the maximum Euclidean/Mahalanobis distance of the characteristic variable center of each batch, and representing the difference among the groups of samples in different batches;
and 5: and establishing a principal component analysis space Euclidean/Mahalanobis distance evaluation method to obtain the difference between groups and in-group of the samples.
(3) The multivariate analysis of variance evaluation method for the sample comprises the following steps:
step 1: detecting the attribute of each batch of samples, repeatedly measuring each sample at least three times, and taking an average value;
step 2: adopting principal component analysis to reduce the dimension and obtaining the characteristic variable of each sensor;
and step 3: calculating the within-group variance and the between-group variance of each batch of samples;
and 4, step 4: and establishing a multivariate analysis of variance evaluation method to obtain the difference between groups and in groups of the samples.
The ratio of the difference between groups and the difference in groups represents the capability of the sensor for distinguishing the samples between the groups, the larger the ratio is, the better the distinguishing capability of the sensor is, the larger the weight of the sensor is, and the ratio of the difference between the groups and the difference in groups is in direct proportion to the weight of the sensor.
The multi-source information fusion of the invention comprises the following steps: data layer fusion, feature layer fusion and decision layer fusion.
(1) The method for fusing the data layer and combining the weight strategy comprises the following steps:
step 1: preprocessing the spectrum by adopting one or more of derivative, smoothing, standard normal transformation (SNV), Multivariate Scattering Correction (MSC), standardization and normalization methods;
step 2: obtaining information of different dimensions of k different sensors based on spectrum preprocessing, and taking original data as input variables of data layer fusion/taking characteristic data obtained by Principal Component Analysis (PCA) as input variables of characteristic layer fusion;
and step 3: standardizing input variables fused with a data layer/a feature layer to obtain data of different sensors in the same dimension;
and 4, step 4: obtaining the weights of k sensors according to the ratio of the difference between the sensor groups to the difference in the sensor groups;
and 5: and establishing a mathematical relation model of the data layer fusion input variable based on the weight.
(2) The method for fusing and combining the decision layer with the weight strategy comprises the following steps:
step 1: one or more spectra of derivative, smoothing, standard normal transformation (SNV), Multivariate Scattering Correction (MSC), standardization and normalization methods are adopted for pretreatment;
step 2: respectively establishing a mathematical relation model of each sensor by adopting the preprocessed input variables;
and step 3: obtaining the weights of k sensors according to the ratio of the difference between the sensor groups to the difference in the sensor groups;
and 4, step 4: and combining the weight of the mathematical relation model of each sensor to represent the fusion result of the decision layer.
The mathematical relationship model included in the multi-source information fusion mathematical relationship model provided by the invention is suitable for all methods in machine learning.
In summary, the invention provides an intelligent manufacturing multi-source information fusion method based on weight distribution strategy of difference between groups and in groups.
The method of the invention has the following advantages:
the invention adopts multi-sensor technology to obtain multi-dimensional information of a sample, calculates the difference between groups and the difference in groups of different sensor information of the sample, determines the weight of each sensor according to the ratio of the difference between the groups and the difference in groups of the sample, and adopts an information fusion strategy to realize multi-source information fusion by combining the distribution strategy of the weight of each sensor. The established detection method has strong specificity and accurate and stable detection result.
The invention creatively provides a weight distribution method based on Mahalanobis distance, Euclidean distance and variance analysis, and in addition, the weight distribution method of difference between characteristic space groups and difference in groups is introduced for the first time, so that multi-source information fusion in the intelligent manufacturing production process is realized.
Drawings
FIG. 1 is a flow diagram of a multi-source information fusion method; (a) data layer/feature layer fusion, (b) decision fusion
FIG. 2 is a graph of Vis spectra of calculus bovis powder, (b) SNV pretreated Vis spectra, (c) NIR spectra of calculus bovis powder, (d) SNV pretreated NIR spectra, and (e) chromaticity values of calculus bovis powder;
FIG. 3 is a multivariate statistical process control MSPC model based on Vis, NIR and color value methods; (a) vis method Hotelling T2Statistics, (b) Vis method prediction error square sum SPE statistics, (c) NIR method Hotelling T statistics2Statistic, (d) NIR method SPE statistic, and (e) color value method Hotelling T2Statistic, (f) color method SPE statistic.
FIG. 4 is a multivariate statistical process control MSPC model based on fusion of NIR and MIR data layers; (a) hotelling T2Statistics, (b) SPE statistics.
Detailed Description
The invention is further described below with reference to the figures and the examples, but is not limited thereto.
Example 1: application of intelligent manufacturing decision layer fusion method in quality evaluation of precious fine medicine
(1) Multi-source information acquisition of bezoar powder
Fig. 1 shows a schematic diagram of a fusion method of a calculus bovis powder decision layer. Collecting three-dimensional information of visible spectrum Vis, near infrared spectrum NIR and colorimetric values of 30 batches of bezoar powder, which comprises the following steps: collecting visible spectrums of 90 samples by using a visible-near infrared spectrometer, wherein the spectrum range is 400-800 nm, the resolution is 0.5nm, scanning is carried out for 32 times, each sample is collected into 4 areas, and the average value is calculated; NIR spectra of 90 samples are collected by a near infrared spectrometer, the spectral range is 1100-2400 nm, the resolution is 0.5nm, scanning is carried out for 32 times, 4 areas are collected for each sample, and the average value is calculated; the chromaticity values of 90 samples were measured using an electronic eye, the light source was D65, the viewing angle was 10 °, 3 regions were collected for each sample, and the average was determined.
FIG. 2.a is a Vis original spectrum of calculus bovis powder, and FIG. 2.b is a spectrum of calculus bovis powder after SNV pretreatment, which has the effect of eliminating solid particle surface scattering and optical path transformation. FIG. 2.c is the NIR raw spectrum of the bezoar bovis powder, and the main characteristic bands of the NIR spectrum of the sample are assigned as follows: the absorption bands near 2300nm, 2100nm and 1900 nm are combined frequency absorption of C-H, N-H and O-H stretching vibration respectively; absorption bands near 1700nm, 1500 nm and 1480nm are first-order frequency absorption of C-H, N-H and O-H stretching vibration respectively; the absorption near 1200 nm is the double frequency absorption of C-H stretching vibration. FIG. 2.d is the spectrum of NIR spectrum of NINGRENHUANGQINGXIN pill after SNV pretreatment. Fig. 2.e are the colorimetric values of the bezoar powder sample, wherein L, a and b are the coordinates of the sample in the color space, L represents black and white, a represents red and green, and b represents yellow and blue. From fig. 2.e, it can be seen that the colorimetric values of the samples of different batches are substantially identical.
(2) Establishment of multivariate statistical process control MSPC model based on Vis, NIR and colorimetric values
And (4) establishing a model according to the data of each sensor in decision fusion, and combining a weight distribution strategy to realize fusion in a decision center. Based on the quality evaluation of the bezoar powder of the precious medicine, the MSPC method is adopted, and the establishment is based onVis, NIR and colorimetric value data. The optimum main component number of the bezoar powder is 2 based on a model established by Vis, NIR and an electronic eye sensor, and the spectrum interpretations of the first two main components of the Vis spectrum are 92.51% and 4.93% respectively; the interpretations of the first two principal components of the NIR spectrum are 90.43% and 6.03%, respectively; the interpretations of the first two principal components of the chroma value are 77.90% and 21.07%, respectively. FIG. 3 is a MSPC model based on principal component analysis of Vis, NIR and color values of calculus bovis powder, wherein Hotelling T2And SPE are two parameter indicators of the MSPC model. Out of Hotelling T2Or samples of 95% control limit of SPE, considered abnormal samples.
And (3) carrying out SNV pretreatment on the Vis spectrum of the bezoar powder to establish an MSPC model. FIGS. 3.a-3.b are schematic diagrams of the model of MSPC Hotelling T2And SPE control charts with thresholds of 6.34 and 0.05, respectively, and 3 samples exceeding the two confidence intervals, i.e., sample nos. 88, 89 and 90. FIG. 3 c-3.d shows the MSPC model of bezoar bovis powder after SNV pretreatment, Hotelling T2And SPE control plots with thresholds of 6.34 and 0.45, respectively, there were 4 samples that exceeded two confidence intervals, i.e., sample nos. 82, 88, 89 and 90. FIG. 3, e-3f are MSPC model of the colorimetric values of calculus bovis powder, T2And SPE control map thresholds of 6.34 and 0.65, respectively, 8 samples out of two confidence intervals were sample numbers 2, 31, 58, 59, 60, 67, 70 and 90.
(3) Weight strategy based on decision layer fusion method
The functions of different sensors in data fusion depend on the performances of the sensors, and the weight of each sensor is calculated by using the ratio alpha of the variance between groups to the variance in groups by adopting a multivariate analysis and evaluation method. The principal component numbers of Vis, NIR and chromaticity values were each selected to be 2. Table 1 shows that without sensor based on a weight, via multivariate analysis of variance, the Vis spectrum has a of 184.5177, with a Vis sensor centered at a corresponding weight of 0.2816; alpha of the NIR spectrum is 460.1062, alpha value of the NIR sensor is maximum, and corresponding weight is 0.7020 at most; in contrast to the Vis and NIR methods, the chromaticity values were 10.7412 minimum in α and the electronic eye sensor weight was 0.0164 minimum. In summary, the NIR sensors are weighted more, probably because they contain abundant characteristic absorption relative to Vis and electronic eye sensors on the one hand, and because the signal-to-noise ratio of the NIR sensors is higher, so that the samples between groups and the samples within groups can be distinguished more clearly.
TABLE 1 weights of different methods
(4) Calculus bovis powder quality evaluation combining decision fusion method with weight strategy
The calculus bovis powder is characterized in that 3 abnormal samples are distinguished based on an MSPC mathematical relation model established by a Vis sensor, wherein the abnormal samples are No. 88, No. 89 and No. 90 samples. The weight of the Vis sensor is 0.2816, namely the 3 samples are used as abnormal samples to obtain 0.2816 votes; the bezoar powder identified 4 abnormal samples based on the established MSPC model of the NIR sensor, which are sample nos. 82, 88, 89 and 90, respectively. The weight of the NIR sensor was 0.7020, i.e. the 4 samples above were used as outliers to get a vote of 0.7020; the calculus bovis powder distinguishes 8 abnormal samples (table 2) based on an MSPC model established by an electronic eye sensor, however, the weight of the 8 abnormal samples based on a colorimetric value is only 0.0164, the sensor data fusion does not play a critical role, and the parameter of the colorimetric value can be ignored for the quality evaluation of the calculus bovis powder. From the results of each of the Vis and NIR sensors, and their corresponding weights, it can be concluded that there are 4 samples that are anomalous based on the decision fusion strategy, No. 82, 88, 89 and 90, respectively.
TABLE 2 results of decision fusion
In summary, the Vis sensor and the NIR sensor both adopt an SNV preprocessing mode, the electronic eye sensor is not preprocessed, the three sensors all adopt 2 principal components, and MSPC models based on the Vis sensor, the NIR sensor and the electronic eye sensor are respectively established. Further, the weight of each sensor is determined by multivariate analysis of variance as the ratio of variance between groups and variance within groups, α. Due to too low weight, the electronic eye sensor does not play a critical role in data fusion of quality evaluation of the bezoar powder, the advantages of different sensor information are complementary, and an accurate and reliable quality evaluation result is obtained by means of information fusion of the Vis sensor and the NIR sensor.
Example 2: application of intelligent manufacturing data layer fusion method in quality evaluation of big honeyed pills
The schematic diagram of the data layer fusion method of the big honeyed pill is shown in fig. 1. And collecting two-dimensional information of a near infrared NIR spectrum and a mid-infrared MIR spectrum of 30 batches of the Niuhuang Qingxin pills. The measurement range of the NIR spectrum is 1100-2400 nm, the resolution is 0.5nm, and the original spectrum data consists of 2601 data points; the measurement range of the MIR spectrum is 600-4000 nm, the resolution is 2nm, and original spectrum data consist of 1701 data points. Four determinations were made for each sample and the average was calculated.
And calculating the difference between groups and in groups based on an original multi-dimensional space Euclidean distance evaluation method, and evaluating the quality of the bezoar heart-clearing pill by adopting a data layer fusion method. The NIR spectrum and the MIR spectrum are preprocessed by adopting SNV (single noise melting) to obtain spectrum information preprocessed by the two sensors. The euclidean distance for each batch of samples based on NIR and MIR spectra were first calculated separately and the maximum euclidean distance within the NIR spectral group was determined to be 2.2423 and the maximum euclidean distance within the MIR spectral group was determined to be 0.9477. Further calculation of the Euclidean distance of the center of each batch determined the maximum Euclidean distance between NIR spectra groups to be 3.7554 and the maximum Euclidean distance between MIR spectra groups to be 0.9477. The resulting ratios of the differences between the NIR and MIR sensors and within the group were 1.6748 and 2.1587, respectively, and the weights for these two sensors were obtained as 0.4369 and 0.5631.
The quality of the bezoar heart-fire clearing pill is evaluated by adopting a weight-based data layer fusion method, firstly, data of two sensors are respectively normalized to obtain data of the same dimension, the data are further combined with the weight to obtain input variables based on the weight-based data layer fusion, an MSPC mathematical relationship model of 30 batches of the bezoar heart-fire clearing pills with the same kernel is established, and the result is shown in figure 4. The result shows that the quality of 6 samples of 16, 18, 27, 28, 41 and 66 has large fluctuation through a weight-based data fusion strategy, and the quality control of the large honeyed pill of the traditional Chinese medicine is realized by adopting a multi-source information fusion strategy and the combined action of multiple sensors.
Claims (9)
1. The multi-source information fusion method for intelligent manufacturing is characterized in that distribution weights are obtained by different sensor information through the difference between groups and the difference ratio in the groups, and multi-source information fusion is achieved, and the method comprises the following specific steps:
step 1: acquiring information of a sample by a multi-sensor technology, and calculating the difference between groups and the difference between groups of different sensor information of the sample;
step 2: determining the weight of each sensor according to the difference between sample groups and the ratio of the difference in the sample groups;
and step 3: and a multi-source information fusion strategy is adopted, and the distribution weight of each sensor is combined to realize multi-source information fusion.
2. The method of claim 1, wherein the data in step 1 can be sourced as: several or more of nuclear magnetic resonance spectrum, gas chromatography, liquid chromatography, mass spectrometry, gas chromatography, liquid chromatography, near infrared spectrum, intermediate infrared spectrum, ultraviolet visible spectrum, Raman spectrum, atomic absorption spectrum, atomic emission spectrum, fluorescence spectrum, texture analyzer, electronic tongue, electronic nose, electronic eye, and infrared imager.
3. The method according to claim 1, wherein the evaluation method of the difference between groups and the difference between groups in step 1 is as follows:
(1) a multidimensional space Euclidean distance evaluation method for original data of the sample;
(2) a multi-dimensional space Mahalanobis distance evaluation method of original sample data;
(3) an Euclidean distance evaluation method of a sample principal component space;
(4) a mahalanobis distance evaluation method of a sample principal component space;
(5) multivariate analysis of variance evaluation of samples.
4. The method of any one of claims 1 to 3, wherein the ratio of the difference between the groups and the difference between the groups in step 2 represents the ability of the sensor corresponding to the method to distinguish between the samples between the groups and the samples within the groups, and wherein a larger ratio indicates a better distinguishing ability of the sensor, a larger weight of the sensor, and a ratio of the difference between the groups and the difference within the groups proportional to the weight of the sensor.
5. The method according to one of claims 1-4, wherein the multi-source information fusion of step 3 in claim 1 comprises: data layer fusion, feature layer fusion and decision layer fusion.
6. The method of claim 5, wherein the data layer/feature layer fusion method comprises the steps of:
step 1: preprocessing an original spectrum by adopting a spectrum preprocessing method to obtain information of different dimensions of different sensors/obtain characteristic variables of different sensors after Principal Component Analysis (PCA);
step 2: determining input variables of data layer/feature layer fusion;
and step 3: standardizing input variables fused with a data layer/a feature layer, and keeping information of different sensors in the same dimension;
and 4, step 4: obtaining the weight of each sensor according to the ratio of the difference between the sensor groups to the difference in the sensor groups;
and 5: and establishing a mathematical relation model by combining the normalized transformed data layer/characteristic layer fusion data with the weight of each sensor.
7. The method of claim 5, wherein the decision-level fusion method comprises the steps of:
step 1: preprocessing an original spectrum by adopting a spectrum preprocessing method to obtain information of different dimensions of different sensors;
step 2: establishing a mathematical relation model of each sensor based on information of different dimensions of different sensors;
and step 3: obtaining the weight of each sensor according to the ratio of the difference between the sensor groups to the difference in the sensor groups;
and 4, step 4: and combining the weights of the sensors to realize the fusion of decision layers.
8. The method according to claim 1, wherein the weighted multi-source information fusion strategy in step 3 specifically comprises the following steps:
step 1: selecting a data preprocessing method according to the original data of the sensor;
step 2: determining the weight of each sensor by using the preprocessed data and applying a method for calculating the ratio of the difference between the sample groups to the difference between the sample groups in claim 3;
and step 3: multiplying the preprocessed original data/feature data by respective weights to respectively serve as input variables of data layer/feature layer fusion in claims 5-6, and establishing a data layer/feature layer fusion mathematical relation model;
and 4, step 4: inputting the preprocessed data into the decision layer fusion mathematical relation model in claims 5 and 7, and multiplying the single sensor model result by the respective weight to obtain a decision layer fusion result.
9. The method of any one of claims 1-8, wherein the mathematical relationship model of multi-source information fusion is applicable to all methods in machine learning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110246821.3A CN113011479A (en) | 2021-03-05 | 2021-03-05 | Multi-source information fusion method for intelligent manufacturing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110246821.3A CN113011479A (en) | 2021-03-05 | 2021-03-05 | Multi-source information fusion method for intelligent manufacturing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113011479A true CN113011479A (en) | 2021-06-22 |
Family
ID=76407280
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110246821.3A Pending CN113011479A (en) | 2021-03-05 | 2021-03-05 | Multi-source information fusion method for intelligent manufacturing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113011479A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113052437A (en) * | 2021-03-05 | 2021-06-29 | 北京中医药大学 | Application of intelligent manufacturing multi-source information fusion method in quality evaluation of large traditional Chinese medicine honeyed pills |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030186663A1 (en) * | 2002-03-26 | 2003-10-02 | Hai-Wen Chen | Method and system for multi-sensor data fusion |
| CN101799674A (en) * | 2010-03-28 | 2010-08-11 | 华中科技大学 | Method for analyzing service state of numerical control equipment |
| US20120097554A1 (en) * | 2010-10-18 | 2012-04-26 | Medtronic Minimed, Inc. | Analyte sensors comprising electrodes having selected electrochemical and mechanical properties |
| CN103278609A (en) * | 2013-06-27 | 2013-09-04 | 山东商业职业技术学院 | Meat product freshness detection method based on multisource perceptual information fusion |
| JP2016008922A (en) * | 2014-06-25 | 2016-01-18 | 株式会社東芝 | Sensor information fusion device |
| CN108960334A (en) * | 2018-07-12 | 2018-12-07 | 中国人民解放军陆军炮兵防空兵学院郑州校区 | A kind of multi-sensor data Weighted Fusion method |
| CN108985373A (en) * | 2018-07-12 | 2018-12-11 | 中国人民解放军陆军炮兵防空兵学院郑州校区 | A kind of multi-sensor data Weighted Fusion method |
| US20190175079A1 (en) * | 2017-12-13 | 2019-06-13 | Medtronic Minimed, Inc. | Optional sensor calibration in continuous glucose monitoring |
| US20190197861A1 (en) * | 2017-07-27 | 2019-06-27 | NXT-ID, Inc. | Method and system to improve accuracy of fall detection using multi-sensor fusion |
| CN111157677A (en) * | 2019-12-19 | 2020-05-15 | 江苏大学 | Organic waste stacking fermentation decomposition degree evaluation method |
| CN111353510A (en) * | 2018-12-20 | 2020-06-30 | 长沙智能驾驶研究院有限公司 | Multi-sensor target detection method and device, computer equipment and storage medium |
| CN111986772A (en) * | 2020-07-31 | 2020-11-24 | 江苏大学 | Intelligent pharmacy optimization method based on digital twin |
| CN113052437A (en) * | 2021-03-05 | 2021-06-29 | 北京中医药大学 | Application of intelligent manufacturing multi-source information fusion method in quality evaluation of large traditional Chinese medicine honeyed pills |
-
2021
- 2021-03-05 CN CN202110246821.3A patent/CN113011479A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030186663A1 (en) * | 2002-03-26 | 2003-10-02 | Hai-Wen Chen | Method and system for multi-sensor data fusion |
| CN101799674A (en) * | 2010-03-28 | 2010-08-11 | 华中科技大学 | Method for analyzing service state of numerical control equipment |
| US20120097554A1 (en) * | 2010-10-18 | 2012-04-26 | Medtronic Minimed, Inc. | Analyte sensors comprising electrodes having selected electrochemical and mechanical properties |
| CN103278609A (en) * | 2013-06-27 | 2013-09-04 | 山东商业职业技术学院 | Meat product freshness detection method based on multisource perceptual information fusion |
| JP2016008922A (en) * | 2014-06-25 | 2016-01-18 | 株式会社東芝 | Sensor information fusion device |
| US20190197861A1 (en) * | 2017-07-27 | 2019-06-27 | NXT-ID, Inc. | Method and system to improve accuracy of fall detection using multi-sensor fusion |
| US20190175079A1 (en) * | 2017-12-13 | 2019-06-13 | Medtronic Minimed, Inc. | Optional sensor calibration in continuous glucose monitoring |
| CN108985373A (en) * | 2018-07-12 | 2018-12-11 | 中国人民解放军陆军炮兵防空兵学院郑州校区 | A kind of multi-sensor data Weighted Fusion method |
| CN108960334A (en) * | 2018-07-12 | 2018-12-07 | 中国人民解放军陆军炮兵防空兵学院郑州校区 | A kind of multi-sensor data Weighted Fusion method |
| CN111353510A (en) * | 2018-12-20 | 2020-06-30 | 长沙智能驾驶研究院有限公司 | Multi-sensor target detection method and device, computer equipment and storage medium |
| CN111157677A (en) * | 2019-12-19 | 2020-05-15 | 江苏大学 | Organic waste stacking fermentation decomposition degree evaluation method |
| CN111986772A (en) * | 2020-07-31 | 2020-11-24 | 江苏大学 | Intelligent pharmacy optimization method based on digital twin |
| CN113052437A (en) * | 2021-03-05 | 2021-06-29 | 北京中医药大学 | Application of intelligent manufacturing multi-source information fusion method in quality evaluation of large traditional Chinese medicine honeyed pills |
Non-Patent Citations (9)
| Title |
|---|
| QIANQIAN LI, 等: "High-level Fusion Coupled with Mahalanobis Distance Weighted (MDW) Method for Multivariate Calibration", SCIENTIFIC REPORTS, vol. 10, 25 March 2020 (2020-03-25), pages 1 - 9 * |
| RODRIGO R. DE OLIVEIRA 等: "Data fusion strategies to combine sensor and multivariate model outputs for multivariate statistical process control", ANALYTICAL AND BIOANALYTICAL CHEMISTRY, vol. 412, 21 January 2020 (2020-01-21), pages 2151 - 2163, XP037063303, DOI: 10.1007/s00216-020-02404-2 * |
| ZHANG Y 等: "Multi-sensor Information Fusion and its Application Analysis", INTERNATIONAL INFORMATION AND ENGINEERING ASSOCIATION.PROCEEDINGS OF 2018 2ND INTERNATIONAL CONFERENCE ON SYSTEMS, COMPUTING, AND APPLICATIONS, 27 October 2018 (2018-10-27), pages 1 - 4 * |
| 关欣 等: "基于灰关联度和距离的特征关联算法研究", 雷达科学与技术, vol. 11, no. 4, 15 August 2013 (2013-08-15), pages 363 - 367 * |
| 刘晓娜 等: "基于多源信息融合的中药质量标志物与质量评价研究模式", 中草药, vol. 50, no. 19, 12 October 2019 (2019-10-12), pages 4576 - 4581 * |
| 张娜 等: "丹参提取过程多源信息融合建模方法研究", 中草药, vol. 49, no. 06, 28 March 2018 (2018-03-28), pages 1304 - 1310 * |
| 曹煌 等: "基于电子鼻和电子舌技术的辛味中药气-味的表征研究", 中草药, vol. 47, no. 11, 12 June 2016 (2016-06-12), pages 1962 - 1967 * |
| 范丹君: "基于多传感器的辛味中药材气—味信息融合研究", 中国优秀硕士学位论文全文数据库 (信息科技辑), no. 2014, 16 October 2014 (2014-10-16), pages 140 - 127 * |
| 范丹君: "基于多传感器的辛味中药材气—味信息融合研究基于多传感器的辛味中药材气—味信息融合研究", 中国优秀硕士学位论文全文数据库 (信息科技辑), no. 2014, 15 October 2014 (2014-10-15), pages 140 - 127 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113052437A (en) * | 2021-03-05 | 2021-06-29 | 北京中医药大学 | Application of intelligent manufacturing multi-source information fusion method in quality evaluation of large traditional Chinese medicine honeyed pills |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2018010352A1 (en) | Qualitative and quantitative combined method for constructing near infrared quantitative model | |
| CN113533236B (en) | Fruit spectrum feature extraction method, model training method and detection method | |
| CN110232419A (en) | A kind of method of side slope rock category automatic identification | |
| Goicoechea et al. | A new family of genetic algorithms for wavelength interval selection in multivariate analytical spectroscopy | |
| CN106250896A (en) | The recognition methods of the positive and negative of online Nicotiana tabacum L. based on image collecting device | |
| CN105938093A (en) | Oolong tea producing area discrimination method based on combination of genetic algorithm and support vector machine | |
| CN107389596A (en) | A kind of method of fast prediction Barley straw trophic component | |
| CN113011479A (en) | Multi-source information fusion method for intelligent manufacturing | |
| Yun et al. | Identification of tea based on CARS‐SWR variable optimization of visible/near‐infrared spectrum | |
| CN111896495A (en) | Method and system for discriminating Taiping Houkui production places based on deep learning and near infrared spectrum | |
| CN108037081B (en) | Wine grape maturity monitoring method and system | |
| CN105223164B (en) | Differentiate the method and system of buckwheat or the adulterated wheat flour of oatmeal | |
| CN113418869B (en) | Polarized light-based rice grain doping or adulteration nondestructive rapid quantitative detection method | |
| Wu et al. | Determination of apple varieties by near infrared reflectance spectroscopy coupled with improved possibilistic Gath–Geva clustering algorithm | |
| CN113390806B (en) | Lettuce heavy metal content detection method based on multi-scale image | |
| Li et al. | Qualitative and quantitative analysis of the pile fermentation degree of Pu-erh tea | |
| CN114092839A (en) | Soybean harvesting period maturity judging method based on unmanned aerial vehicle remote sensing | |
| CN113052437A (en) | Application of intelligent manufacturing multi-source information fusion method in quality evaluation of large traditional Chinese medicine honeyed pills | |
| CN113570538A (en) | Blade RGB image skewed distribution parameter information acquisition and analysis method | |
| Mignani et al. | Spectral nephelometry for the geographic classification of Italian extra virgin olive oils | |
| He et al. | Green tea grades identification via Fourier transform near‐infrared spectroscopy and weighted global fuzzy uncorrelated discriminant transform | |
| CN116662751A (en) | Tobacco leaf moisture content detection method for removing abnormal samples based on principal component analysis and lever value method | |
| AU2021104058A4 (en) | Analysis Method of Nutritional Quality of Pleurotus Ostreatus | |
| CN105527241A (en) | Non-destructive method for detecting authenticity of raw cordyceps sinensis | |
| Cheng et al. | Identification of wine grape varieties based on near-infrared hyperspectral imaging |
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 |