CN117034601A - Soft measurement modeling method for overflow granularity in multi-station ore dressing process - Google Patents

Soft measurement modeling method for overflow granularity in multi-station ore dressing process Download PDF

Info

Publication number
CN117034601A
CN117034601A CN202310984717.3A CN202310984717A CN117034601A CN 117034601 A CN117034601 A CN 117034601A CN 202310984717 A CN202310984717 A CN 202310984717A CN 117034601 A CN117034601 A CN 117034601A
Authority
CN
China
Prior art keywords
window
data
beneficiation
process variable
overflow granularity
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
Application number
CN202310984717.3A
Other languages
Chinese (zh)
Inventor
阎高伟
孙建鑫
李�荣
王芳
肖舒怡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Taiyuan University of Technology
Original Assignee
Taiyuan University of Technology
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 Taiyuan University of Technology filed Critical Taiyuan University of Technology
Priority to CN202310984717.3A priority Critical patent/CN117034601A/en
Publication of CN117034601A publication Critical patent/CN117034601A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume or surface-area of porous materials
    • G01N15/02—Investigating particle size or size distribution
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Algebra (AREA)
  • Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biochemistry (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • Analytical Chemistry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Immunology (AREA)
  • Computational Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Dispersion Chemistry (AREA)
  • Pure & Applied Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention belongs to the technical field of grinding particle size measurement, and particularly relates to a soft measurement modeling method for overflow particle size in a multi-station ore dressing process. S1, collecting and storing m ore dressing process variables related to overflow granularity as an ore dressing process data set; s2, cutting the data by a sliding window method according to time sequence to obtain a plurality of window sub-data sets with the same size; s3, selecting part of ore dressing process variables as process variables to be tested; s4, establishing a regression model for predicting the process variable to be detected selected in the S3; s5, predicting the process variables to be detected in all window sub-data sets by using the regression model in S4; s6, predicting deviation according to the process variable in each window obtained in the S5, and comparing the predicted deviation with a set threshold value to perform modal division to obtain a modal identification result; s7, obtaining the characteristics of each mode according to the mode identification result, and establishing a soft measurement model; s8, completing prediction of overflow granularity.

Description

Soft measurement modeling method for overflow granularity in multi-station ore dressing process
Technical Field
The invention belongs to the technical field of grinding particle size measurement, and particularly relates to a soft measurement modeling method for overflow particle size in a multi-station ore dressing process.
Background
In the ore grinding process, the hydrocyclone can separate mixtures with different granularity by centrifugal sedimentation, and in the ore dressing process, the hydrocyclone is mainly used for classifying coarse sand and overflow, and on-site staff can adjust the ore feeding amount and water feeding amount of the ball mill according to the size of the overflow granularity. Therefore, accurate and timely acquisition of overflow granularity has important significance for the whole beneficiation process.
Because overflow granularity is difficult to directly and accurately obtain, the overflow granularity can be indirectly obtained by adopting a soft measurement technology through taking a process variable which can be acquired in the beneficiation process as an auxiliary variable. However, in the actual beneficiation process, the process variable is influenced by factors such as ore hardness, mill feeding amount, steel ball abrasion and the like, the collected data show a plurality of modal characteristics, and different modal data distribution and correlation relations are different, if the overall beneficiation process data are subjected to soft measurement modeling to predict overflow granularity, larger model prediction deviation can be caused. Therefore, accurate mode identification is required to be carried out on the beneficiation process data of multiple working conditions, a soft measurement model is established on the corresponding mode, and the prediction accuracy of the mill load is improved. Since the multi-mode process data presents different distributions, a clustering algorithm can be adopted to carry out mode division on the multi-mode process data. However, in general, it is difficult to obtain a specific number of modes, and the clustering algorithm cannot automatically identify the modes. Furthermore, there are cases where the same cluster data is discontinuous in time using a clustering algorithm, which is impractical in a continuous beneficiation process. The multi-mode division method based on the sliding window can well solve the problem.
The sliding window is utilized to cut the time sequence variable of the ore dressing process according to a certain step length, and then the distribution difference between the adjacent window data is compared, so that the ore dressing process can be realizedThe stable mode and the transition mode can be effectively divided. The above methods are all based on the similarity of data over time series for modal partitioning. The soft measurement modeling based on the statistical method requires the assumption of the same distribution of data, and after the data distribution is changed, the soft measurement model is deteriorated to different degrees, so that the degree of the deterioration of the soft measurement model can be utilized for carrying out modal division and identification. However, due to technical limitations, overflow granularity is often difficult to directly obtain and is influenced by human factors, and whether a model is deteriorated or not cannot be measured by using the overflow granularity. Starting from the data-driven soft measurement modeling principle, a soft measurement model of a dominant variable can be established, then a prediction model of other process variables (non-dominant variables) can also be established, and the process variables can obtain true values in real time, so that the deterioration degree of the model can be measured through the prediction deviation of the beneficiation process variables, and the modal classification and identification of the beneficiation process can be carried out. After the mineral separation process mode division is completed, establishing a soft measurement prediction model of overflow granularity for each mode, and calculating L between the current mode and the prediction deviation of the historical mode process variable when the online process variable is acquired 2 Distance, selecting current mode and L 2 And the soft measurement model with the minimum distance is used for completing overflow granularity prediction.
Disclosure of Invention
The invention aims to identify each mode in multi-mode beneficiation process data and perform soft side quantity modeling aiming at the corresponding mode, so that the prediction precision of a soft measurement model is improved, and provides a soft measurement modeling method for overflow granularity of a multi-mode beneficiation process.
The invention adopts the following technical scheme: a soft measurement modeling method for overflow granularity in a multi-station ore dressing process comprises the following steps:
s1: collecting and storing m ore dressing process variables related to overflow granularity as an ore dressing process data set, and carrying out standardized treatment on the ore dressing process data set;
s2: cutting the standard beneficiation process data set according to time sequence by adopting a sliding window method based on the standard beneficiation process data set in the S1 to obtain a plurality of window sub-data sets with the same size;
s3: selecting part of beneficiation process variables as process variables to be measured based on the beneficiation process variables in the standardized beneficiation process data set in the step S1;
s4: establishing a regression model for predicting the process variable to be detected selected in the step S3;
s5: predicting the process variables to be detected in all window sub-data sets by using the regression model in the S4, and subtracting the predicted value from the true value to construct a process variable prediction deviation;
s6: according to the process variable prediction deviation in each window obtained in the step S5, measuring the process variable prediction deviation distribution difference between adjacent windows by utilizing the least square density difference, and comparing the process variable prediction deviation distribution difference with a set threshold value to perform modal division to obtain a final modal identification result;
s7: obtaining characteristics of each mode according to the mode identification result, and respectively establishing a soft measurement model;
s8: and (3) acquiring an online beneficiation process variable, calculating a process variable prediction deviation by using the regression model constructed in the step (S4), comparing the process variable prediction deviation with the prediction deviation distribution in each mode in the step (S6), and selecting a model with the minimum distance to finish the prediction of overflow granularity.
In some embodiments, step S1 comprises: collecting m ore dressing process variables related to overflow granularity to obtain N samples, which are respectively recorded asFirst sample->Corresponding beneficiation process variable setSample N->Corresponding beneficiation process variable set ∈ ->These samples were normalized.
In some embodiments, step S2 comprises: sliding with window size HCutting the beneficiation process data set by using the window to obtain P windows, wherein the first windowkThe sub-data sets in the windows may be represented as，The method comprises the steps of carrying out a first treatment on the surface of the The process data set can then be expressed as +.>，/>The size of the window width H is 100-400 sampling data lengths.
In some embodiments, step S4 comprises: establishing beneficiation process variablesAs independent variable, process variable to be measured +.>Multiple-input multiple-output partial least squares regression model for dependent variables>。
In some embodiments, step S5 comprises:
inputting the beneficiation process variable in each window sub-data set in S2 into a regression modelObtaining a predicted value of the process variable to be measured +.>；
To the true value of the process variable to be measuredAnd predictive value->Making a difference to obtain a process variablePrediction bias，/>。
In some embodiments, step S6 comprises:
s61: setting a stable mode thresholdCalculating adjacent window data in process variable prediction bias EL 2 Distance whenL 2 Less than->When the two windows are in the same stable modeL 2 Is greater than->Step S62, dividing transition sub-modes;
s62: when calculating between the kth window and the kth+1th window dataL 2 The distance is greater than the stable mode thresholdAt this time, the two window data are combined with the (k-1) th window data to obtain a new data set +.>The method is used for identifying the transition sub-mode in the next step;
s63: sliding window pair datasetCutting, giving transition mode threshold +.>Thereby calculating the data between each window and the first windowL 2 Distance, when the data between the (q) th window and the (1) st window is obtainedL 2 Distance is greater than->Then the first data in the q-th window is the starting point of the first transition sub-mode, and then the space between the q-th window and the rest windows is calculatedL 2 Distance and->All modalities are identified by comparison and so on.
In some embodiments, in steps S61, S62, S63, a calculation is performed between adjacent model prediction bias windowsL 2 The distance process is as follows:
by the formulaAnd->Calculating to obtain h and->，
In the method, in the process of the invention,for nuclear width->And->Representation for computingL 2 The adjacent window dataset of distances, in the invention, the values that are brought in are adjacent windows in the process variable prediction bias E, used to calculate the distance between themL 2 Distance (L)>Represents the center of the Gaussian kernel, taken from +.>And->Is used as a reference to the value of (a),n represents the amount of data in the window, since the window size is H, n=h, d represents the dimension of the data when in use, and E is H-dimensional in the present invention, so d=h when in use.
After obtaining h andafter that, by the formula->Calculated->Then can pass throughObtaining between adjacent window dataL 2 A distance;
in the method, in the process of the invention,for regularization coefficient, ++>Representing a b-dimensional identity matrix.
In some embodiments, step S7 comprises: p modes are obtained according to S6, and PLSR regression models of overflow granularity in the ore separation process are respectively established in data of each mode:
modality 1:；
modality 2:；
modality p:。
in some embodiments, step S8 comprises: on-line to be acquiredThe model prediction deviation of part of the beneficiation process variable is obtained by inputting the least square regression model constructed by S4 into the beneficiation process variable, and the model prediction deviation in each mode in S7 are calculated respectivelyL 2 And selecting a model with the smallest distance to finish the prediction of overflow granularity.
Compared with the prior art, the invention discloses a soft measurement modeling method suitable for overflow granularity in a multi-station ore dressing process, which is used for solving the problem that the prediction performance of an established soft measurement model for predicting overflow granularity is reduced due to the distribution change of ore dressing process variables such as collected cyclone ore feeding pressure, ore feeding flow, ore feeding concentration and the like caused by the change of a set value, material quantity, environment and the like in the ore dressing process; cutting continuous process variables collected in the beneficiation process by utilizing a sliding window method, and selecting partial variables from measurable process variables as variables to be measured; establishing a multi-input multi-output partial least square regression model of the process variable to be measured and all the process variables in the first window data; predicting the process variable to be detected in all windows by using the regression model, and obtaining a process variable prediction deviation by solving the difference between the process variable and the true value; finally utilizing the least square density differenceL 2 Distance) measures the distribution difference of adjacent window deviation information, compares the distribution difference with a set threshold value to determine whether the mode is changed, so as to divide the modes and establish a soft measurement model of each mode; after the online data is acquired, calculating the prediction deviation of the process variable and the history modeL 2 Distance, selecting online mode and history modeL 2 The model with the minimum distance completes online prediction of overflow granularity; the invention, from the viewpoint of measuring the model prediction error, excavates the deterioration degree of the prediction model by detecting the prediction deviation change of the process variable, so as to perform modal identification, establish a soft measurement model and improve the prediction precision of overflow granularity in the actual industrial process.
Drawings
FIG. 1 is a flow chart of the present invention;
fig. 2 is a flow chart of the mode identification in S63 of the present invention.
Detailed Description
The invention is further described below with reference to examples, which should be construed as not limiting the scope of the invention as claimed.
As shown in fig. 1, a soft measurement modeling method for overflow granularity in a multi-working-condition beneficiation process comprises the following steps:
s1: and collecting and storing m ore dressing process variables related to overflow granularity as an ore dressing process data set, and carrying out standardized treatment on the ore dressing process data set.
Collecting m ore dressing process variables related to overflow granularity to obtain N samples, which are respectively recorded asFirst sample->Corresponding beneficiation process variable set +.>Sample NCorresponding beneficiation process variable set +.>These samples were normalized.
The normalization method is as follows:。
preprocessing the acquired N samples by using a normalization method to obtain a normalized process variable data setQuality variable dataset +.>。
In this embodiment, the m beneficiation process variables are cyclone feeding pressure, cyclone feeding flow, cyclone feeding concentration, pump pool liquid level, slurry pump frequency, overflow flow, overflow concentration, cyclone group number and sand settling amount, and these parameters are used as beneficiation process variables for modal classification, and overflow particle size obtained by laboratory analysis is collected for soft measurement model training after modal classification.
S2: and (3) based on the standard beneficiation process data set in the step (S1), cutting the beneficiation process data set according to time sequence by adopting a sliding window method to obtain a plurality of window sub-data sets with the same size.
Cutting the beneficiation process data set by utilizing a sliding window with window size H to obtain P windows, wherein the first windowkThe sub-data sets in the windows may be represented as，/>The method comprises the steps of carrying out a first treatment on the surface of the The beneficiation process data set can then be expressed as +.>，/>The size of the window width H is 100-400 sampling data lengths.
In this embodiment, the window size is selectedHSliding window data =200XAnd cutting to obtain n/H window data.
S3: and selecting part of the beneficiation process variables as process variables to be measured based on the beneficiation process variables in the standardized beneficiation process data set in the step S1.
Assuming that the first window data obtained in S2 is in a stable mode, h (0.7 m<h.ltoreq.m) measurable process variables，/>As a variable to be measured, a beneficiation process variable +.>Is an independent variable, the variable to be measured +>Multiple-input multiple-output partial least squares regression model for dependent variables>。
In this embodiment, cyclone feed pressure, cyclone feed flow, and cyclone feed concentration are selected as process variables to be measured in the first window.
S4: and (3) establishing a PLSR regression model for predicting the selected process variable to be detected in the step S3.
Establishing beneficiation process variablesAs independent variable, process variable to be measured +.>Multiple-input multiple-output partial least squares regression model for dependent variables>。
S5: and (3) using the regression model constructed in the step (S4) for predicting all window sub-data sets in the step (S2), and obtaining a process variable prediction deviation by making a difference with the true value of the selected part of beneficiation process variables in the step (S3).
The specific process is as follows:
inputting the beneficiation process variable in each window sub-data set in S2 into a regression modelObtaining a predicted value of the process variable to be measured +.>；
To the true value of the process variable to be measuredAnd predictive value->Making a difference to obtain a predicted deviation of the process variable，/>。
S6: and (3) according to the process variable prediction deviation in each window obtained in the step (S5), measuring the process variable prediction deviation distribution difference between adjacent windows by utilizing the least square density difference, and comparing the process variable prediction deviation distribution difference with a set threshold value to perform modal division to obtain a final modal identification result.
S61, calculating the prediction deviation window between adjacent models by using the least square density differenceL 2 Distance ofBy->To measure the distribution difference of the prediction bias of different window process variables when it is greater than a given threshold +.>When the mode is changed, the threshold value +.>The value is 0.7%>。
Specifically, a stable mode threshold is setCan be adjusted according to actual conditions, and adjacent window data in the data set E is calculatedL 2 Distance and->Compared with the division of stable modes.
S62, assume S61 identifies that the process variable prediction bias is between window k and k+1L 2 Distance is greater than a given thresholdTaking into account that the modal change point positions may be in window k or k+1, integrating the window k-1, k, k+1 window data together forms a new dataset +.>And using window size L, L<H sliding window pair->Cutting to obtain Q small window data sets, wherein Q=3H/L,/I>，/>Wherein the size of the small window width L is selected to be 1/5H of the sample data length,/L->For the next step of transition sub-mode identification.
S63, selecting a sliding window pair dataset with window size l=40Cutting, giving a transition mode threshold valueCan be adjusted according to actual conditions. Thereby calculating the data between each window and the first windowL 2 Distance, when the data between the (q) th window and the (1) st window is obtainedL 2 Distance is greater than->Then the first data in the q-th window is the starting point of the first transition sub-mode, and then the space between the q-th window and the rest windows is calculatedL 2 Distance ofAnd->All modalities are identified by comparison and so on.
In particular, since the starting point of the transitional mode is identified as being located in window k or k+1, the data setThe first window data in the list belongs to window k-1, so that the first window is in stable mode, and each window are calculated backwards in turnL 2 Distance until the calculatedL 2 Distance greater than a given threshold->Threshold->The value of (2) is +.>The method comprises the steps of carrying out a first treatment on the surface of the Let the q-th window and the first windowL 2 Distance is greater than->I.e. +.>Then the first sample point in the q-th window is the starting point of the transition sub-mode, and then the previous operation is repeated by taking the q-th window as the reference until all data sets are +.>The last modality is considered to be the new stable modality and the detailed procedure is shown in fig. 2.
In steps S61, S62, S63, the prediction bias window between adjacent models is calculatedL 2 The distance process is as follows:
the least squares density difference can be expressed as:，/>，/>is a b-dimensional basis function vector, the basis function uses the Gaussian kernel model +.>，/>Represents the center of the Gaussian kernel, i.e. c is taken from +.>And->，/>，/>，/>For regularization coefficient, ++>For nuclear width->Represents a b-dimensional identity matrix, H is +.>Dimension matrix->。
Obtaining adjacent windowsAnd->After that, c is taken from the window data +.>And->Then it can pass through the formulaAnd->Calculating to obtain h and->In obtaining h and->After that, by the formula->Calculated->Then can pass->Obtaining between adjacent window dataL 2 Distance.
S7: according to the p modes obtained in S6, a PLSR regression model of overflow granularity in the ore separation process is respectively established in each mode data, so that the water feeding amount and the ore feeding amount can be conveniently adjusted by workers.
Modality 1:；
modality 2:；
modality p:。
s8: obtaining on-line beneficiation process variablesCalculating to obtain predicted values of cyclone ore feeding pressure, cyclone ore feeding flow and cyclone ore feeding concentration by using a PLSR regression model constructed by S4, and subtracting the predicted values from the true values to obtain a model predicted deviation data set. Calculating +.>From model predictive bias datasets in each modalityL 2 Distance according toL 2 And selecting a model with the smallest distance from the distance to finish the prediction of the online overflow granularity.

Claims (9)

1. A soft measurement modeling method for overflow granularity in a multi-station mineral separation process is characterized by comprising the following steps:
s1: collecting and storing m ore dressing process variables related to overflow granularity as an ore dressing process data set, and carrying out standardized treatment on the ore dressing process data set;
s2: cutting the standard beneficiation process data set according to time sequence by adopting a sliding window method based on the standard beneficiation process data set in the S1 to obtain a plurality of window sub-data sets with the same size;
s3: selecting part of beneficiation process variables as process variables to be measured based on the beneficiation process variables in the standardized beneficiation process data set in the step S1;
s4: establishing a regression model for predicting the process variable to be detected selected in the step S3;
s5: predicting the process variables to be detected in all window sub-data sets by using the regression model in the S4, and subtracting the predicted value from the true value to construct a process variable prediction deviation;
s6: according to the process variable prediction deviation in each window obtained in the step S5, measuring the process variable prediction deviation distribution difference between adjacent windows by utilizing the least square density difference, and comparing the process variable prediction deviation distribution difference with a set threshold value to perform modal division to obtain a final modal identification result;
s7: obtaining characteristics of each mode according to the mode identification result, and respectively establishing a soft measurement model;
s8: and (3) acquiring an online beneficiation process variable, calculating a process variable prediction deviation by using the regression model constructed in the step (S4), comparing the process variable prediction deviation with the prediction deviation distribution in each mode in the step (S6), and selecting a model with the minimum distance to finish the prediction of overflow granularity.
2. The soft measurement modeling method of overflow granularity of multi-condition beneficiation process according to claim 1, wherein the step S1 comprises: collecting m ore dressing process variables related to overflow granularity to obtain N samples, which are respectively recorded asFirst sample->Corresponding beneficiation process variable set ∈ ->Sample N->Corresponding beneficiation process variable set ∈ ->These samples were normalized.
3. The soft measurement modeling method of overflow granularity of multi-condition beneficiation process according to claim 1, wherein the step S2 comprises: cutting the beneficiation process data set by utilizing a sliding window with window size H to obtain P windows, wherein the first windowkNumber of children in a windowThe dataset may be represented as，/>The method comprises the steps of carrying out a first treatment on the surface of the The process data set can then be expressed as +.>，/>The size of the window width H is 100-400 sampling data lengths.
4. A soft measurement modeling method of overflow granularity of a multi-condition beneficiation process according to claim 3, wherein the step S4 comprises: establishing beneficiation process variablesAs independent variable, process variable to be measured +.>Multiple-input multiple-output partial least squares regression model for dependent variables>。
5. The soft measurement modeling method of overflow granularity of multi-condition beneficiation process according to claim 4, wherein the step S5 comprises:
inputting the beneficiation process variable in each window sub-data set in S2 into a regression modelObtaining a predicted value of the process variable to be measured +.>；
To the true value of the process variable to be measuredAnd predictive value->Making a difference to obtain a predicted deviation of the process variable，/>。
6. The method for modeling the soft measurement of overflow granularity of a multi-condition beneficiation process according to claim 5, wherein the step S6 comprises:
s61: setting a stable mode thresholdCalculating adjacent window data in process variable prediction bias EL 2 Distance whenL 2 Less than->When the two windows are in the same stable modeL 2 Is greater than->Step S62, dividing transition sub-modes;
s62: when calculating between the kth window and the kth+1th window dataL 2 The distance is greater than the stable mode thresholdAt this time, the two window data are combined with the (k-1) th window data to obtain a new data set +.>The method is used for identifying the transition sub-mode in the next step;
s63: sliding window pair datasetCutting, giving transition mode threshold +.>Thereby calculating the data between each window and the first windowL 2 Distance, when the data between the (q) th window and the (1) st window is obtainedL 2 Distance is greater than->Then the first data in the q-th window is the starting point of the first transition sub-mode, and then the space between the q-th window and the rest windows is calculatedL 2 Distance and->All modalities are identified by comparison and so on.
7. The method for modeling the overflow granularity of a multi-condition beneficiation process according to claim 6, wherein in the steps S61, S62 and S63, the prediction deviation windows of adjacent models are calculatedL 2 The distance process is as follows:
by the formulaAnd->Calculating to obtain h and->，
In the method, in the process of the invention,for nuclear width->And->Representation for computingL 2 The adjacent window dataset of distances, in the invention, the values that are brought in are adjacent windows in the process variable prediction bias E, used to calculate the distance between themL 2 Distance (L)>Represents the center of the Gaussian kernel, taken from +.>And->N represents the amount of data in the window, n=h when in use, d represents the dimension of the data, in the present invention E is H-dimensional, so d=h when in use;
after obtaining h andafter that, by the formula->Calculated->Then can pass->Obtaining between adjacent window dataL 2 A distance;
in the method, in the process of the invention,for regularization coefficient, ++>Representing a b-dimensional identity matrix.
8. The soft measurement modeling method of overflow granularity of multi-condition beneficiation process according to claim 6, wherein the step S7 comprises: p modes are obtained according to S6, and PLSR regression models of overflow granularity in the ore separation process are respectively established in data of each mode:
modality 1:；
modality 2:；
modality p:。
9. the soft measurement modeling method of overflow granularity of multi-condition beneficiation process according to claim 7, wherein the step S8 comprises: inputting the obtained online beneficiation process variable into a least square regression model constructed in S4 to obtain model prediction deviation of part of beneficiation process variable, and respectively calculating model prediction deviation and model prediction deviation in each mode in S7L 2 And selecting a model with the smallest distance to finish the prediction of overflow granularity.
CN202310984717.3A 2023-08-07 2023-08-07 Soft measurement modeling method for overflow granularity in multi-station ore dressing process Pending CN117034601A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310984717.3A CN117034601A (en) 2023-08-07 2023-08-07 Soft measurement modeling method for overflow granularity in multi-station ore dressing process

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310984717.3A CN117034601A (en) 2023-08-07 2023-08-07 Soft measurement modeling method for overflow granularity in multi-station ore dressing process

Publications (1)

Publication Number Publication Date
CN117034601A true CN117034601A (en) 2023-11-10

Family

ID=88629305

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310984717.3A Pending CN117034601A (en) 2023-08-07 2023-08-07 Soft measurement modeling method for overflow granularity in multi-station ore dressing process

Country Status (1)

Country Link
CN (1) CN117034601A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117776413A (en) * 2023-12-28 2024-03-29 武汉飞博乐环保工程有限公司 Method for treating high-hardness wastewater by using carbon dioxide waste gas

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117776413A (en) * 2023-12-28 2024-03-29 武汉飞博乐环保工程有限公司 Method for treating high-hardness wastewater by using carbon dioxide waste gas

Similar Documents

Publication Publication Date Title
CN109062196B (en) Blast furnace process monitoring and fault diagnosis method integrated with PCA-ICA
TWI521360B (en) Metrology sampling method and computer program product thereof
CN110033141B (en) Method for establishing desulfurization system operation condition database
CN110561191B (en) Numerical control machine tool cutter abrasion data processing method based on PCA and self-encoder
CN108241901B (en) Transformer early warning evaluation method and device based on prediction data
CN110929347A (en) Hot continuous rolling strip steel convexity prediction method based on gradient lifting tree model
CN102693452A (en) Multiple-model soft-measuring method based on semi-supervised regression learning
CN113177358B (en) Soft measurement method for cement quality based on fuzzy fine-grained feature extraction
CN110968618B (en) Method for mining quantitative association rule of welding parameters and application
CN114789364B (en) Multi-index drilling quality control method and device and equipment thereof
Xu et al. Multi-model soft measurement method of the froth layer thickness based on visual features
CN116244625A (en) Overflow type mill load indirect forecasting method based on multi-feature fusion neural network
CN112305441A (en) Power battery health state assessment method under integrated clustering
CN112418522B (en) Industrial heating furnace steel temperature prediction method based on three-branch integrated prediction model
Yan et al. Estimation of copper concentrate grade for copper flotation
CN117034601A (en) Soft measurement modeling method for overflow granularity in multi-station ore dressing process
CN116304948A (en) Unsupervised electricity consumption anomaly detection method integrating multi-scale fuzzy information particles
CN109648397B (en) Broach life prediction method based on correlation between width of cutting edge strip and broaching load
CN110673470B (en) Industrial non-stationary process soft measurement modeling method based on local weighting factor model
WO2022126197A1 (en) Method and system for automated rock recognition
Jeyaraj et al. A data-driven model to predict dressing interval during a multi-flute end mill grooving process using a multilayered diamond grinding wheel
CN114384870B (en) Complex industrial process running state evaluation method based on nuclear local linear embedded PLS
CN115755608B (en) Energy consumption optimization decision method for high-pressure roller mill
Jing et al. Prediction of concentrate grade and recovery rate of tailings in the process of production based on chaotic ant colony algorithm
CN112475904B (en) Numerical control milling and boring machine machining precision prediction method based on thermal analysis

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