CN104296801A - Hydrometallurgy thick washing process key variable detection method - Google Patents
Hydrometallurgy thick washing process key variable detection method Download PDFInfo
- Publication number
- CN104296801A CN104296801A CN201410258081.5A CN201410258081A CN104296801A CN 104296801 A CN104296801 A CN 104296801A CN 201410258081 A CN201410258081 A CN 201410258081A CN 104296801 A CN104296801 A CN 104296801A
- Authority
- CN
- China
- Prior art keywords
- model
- data
- concentration
- washing process
- dense washing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a hydrometallurgy thick washing process key variable real-time prediction method which comprises the steps of process data acquisition, auxiliary variable selection and standardization processing, hybrid model establishment and the like. The method is characterized by establishing a parallel structure hybrid model formed based on a mechanism model and a data driver model; and taking the model based on data driving as an error compensation model of the mechanism model. The invention further provides a software system for carrying out thick washing process key variable prediction. The software system comprises a main program, a database and a man-machine interaction interface. The software system takes a model computer of a hydrometallurgy process control system as a hardware platform. The method is applied to the thick washing process of some hydrometallurgy factory and can predicate overflow concentration and underflow concentration, and the predication results are within the predetermined error range. The advantages of the method are that the method is simple in model, high in explainability, good in extrapolation performance and high in prediction precision.
Description
Technical field
The invention belongs to technical field of wet metallurgy.A kind of hydrometallurgy dense washing process underflow density detection method is provided especially, namely a kind of method of real-time estimate underflow density is provided.
Background technology
Hydrometallurgical processes be one ancient and have the modern science and technology of tremendous expansion future, compared with traditional pyrometallurgical smelting, hydrometallurgical technology have be convenient to many metals resources separation and recovery, do not produce smoke pollution and environment amenable advantage.The particularly feature of complicated for Mineral Resources in China, difficult choosing and low-grade ore, hydrometallurgical technology has more superiority, should say that hydrometallurgy more can adapt to the requirement of current Sustainable Development of Mineral Resource.
China's nonferrous smelting equipment technological level was existing in recent years significantly promotes, but compared with external advanced enterprises, in throughput rate, production cost and energy consumption, environmental pollution and mineral recovery rate etc., also have larger gap, smelting process automation level is low is one of main reason causing this gap.The automated arm level of China's hydrometallurgy key equipment and process control level can not meet its industrialized needs far away, seriously hinder the further raising of hydrometallurgical processes level and production efficiency.
Dense washing process is one critical process of hydrometallurgy process.Concentrator is common equipment wherein, and its underflow density is the key index weighing dense washing process quality.The control of current thickener underflow concentration is still in manual operation state, and the fluctuation of its concentration, flow is correspondingly all larger, affects greatly the production target of subsequent production process.Concentrator is due to characteristics such as influence factor is many, large dead time, slow time-varying simultaneously, and make to realize its automatic control is a difficult problem always.Realize dense washing process optimal control, it is crucial that Establishment and optimization controls relevant dense washing process key variables (underflow density) forecast model.
Concentrator is the solid-liquid separating equipment based on gravity settling effect, by concentration be 10% ~ 20% ore pulp by gravity settling simmer down to concentration be 45% ~ 55% underflow ore pulp, by the effect of rake being installed on microrunning in concentrator, the underflow ore pulp of thickening is drawn off by underflow opening bottom concentrator.Concentrator top produces more peace and quiet clarified solution (overflow), is discharged by the annular chute at top.The working space of concentrator can be divided into free setting district, flocking settling district, hindered settling district and four, pressure negative area district according to concentration, as shown in Figure 1.
Settling process is as follows:
(1) need concentrated suspending liquid slurry first freedom of entry negative area, solid particle relies on the rapid sedimentation of deadweight, enters flocking settling district;
(2) in flocking settling district, oneself formation of the solid particle in suspending liquid floc sedimentation more closely, floc sedimentation still continues sedimentation, but its speed is slower;
(3) in hindered settling district, partial particulate is due to Gravitative Loads sedimentation, and partial particulate is then subject to the obstruction of compacted grains, is difficult to continue sedimentation;
(4) the sedimentation bulk concentration entering pressure negative area is larger, because being provided with rotor segment herein, sometimes the part in this district is shallow conical surface, water in concentrate can overflow again under the pressure effect of scraper plate, suspension concentration improves further, finally discharged by mouth at the bottom of concentrator, become the underflow product of concentrator, obtain the mine tailing slip of high concentration.
Because dense washing process is complicated, severe operational environment, carries out Analysis on Mechanism to it and sets up the larger difficulty of mathematical models existence.Since J.V.N.Dorr in 1905 has invented the thickener of continuous operations, a lot of scholar has carried out mathematical simulation to dense washing process.In recent decades, concentrator solid-liquid separation technique has a great development, but is nowhere near to the research of the mathematical model of dense washing process.The mathematical model of dense washing process contributes to describing and understanding settling process mechanism, provides theoretical guidance to system and lectotype selection; Contribute to simulating the dynamic change of settling process and the change of research ore pulp output, actual production run can be instructed.Along with the development of solid-liquid separation technique, the research of mathematical model must be carried out, thus more profoundly be familiar with the phenomenon and rule studied.
Hydrometallurgy dense washing process modeling method is actual can apply mainly contain following several frequently seen method:
(1) based on the method for mechanism model
After the comprehensively deep movement mechanism of solution preocess, the theorem using some known, law and principle, as energy-balance equation, heat and mass principle, principle of dynamics etc., the mathematical model of process of establishing, the process internal characteristics of motion contained with the internal mechanism expressing process.Usual mechanism model is made up of Algebraic Equation set or differential equation group.Because mechanism relation clearly can show immanent structure and the contact of process, be therefore called white box modeling.
(2) based on the method for data model
According to the inputoutput data of system, set up the method with the mathematical model of system external characteristics equivalence, be called data modeling.Data modeling regards system as black box, when not understanding internal system structure and mechanism, choose one group to have with leading variable and maintain close ties with and the secondary variable easily measured, according to certain optiaml ciriterion, utilize the mathematical model between statistical method structure secondary variable and leading variable.
(3) based on the method for mixture model
The mathematical model of multi-modeling method establishment object used in combination, can reach the effect that various method is learnt from other's strong points to offset one's weaknesses, become the focus of research at present.
If system has the physical knowledge of priori to utilize, then utilize as far as possible, so that blackbox model is changed into greybox model, thus mechanism method and data method are combined.Data method can extract mechanism method the complex information of unaccountable object inside, and mechanism model can improve the Generalization Ability of data model.Combination is generally divided into parallel and serial two kinds.
Serial combination as shown in Figure 2, first obtains the model structure of a band parameter, then determines those parameters with data method by mechanism method.Such as Psichogios and Ungar adopts which to fermentation reaction process model building exactly, estimates throughput rate constant by neural network, then delivers to the mechanism model represented by mass balance equation.
Parallel combination as shown in Figure 3, adopts data method to determine a compensator, compensates the result that mechanism model obtains.Such as Su etc. take parallel combination, and to continuous polymerization reaction modeling, process data is delivered in mechanism model and neural network model simultaneously, and the output of neural network model is added in the output of mechanism model, compensates it.The application of priori, compared with the blackbox model merely set up according to data, improves the precision of model, enhances the Generalization Ability of model, and decrease the data needed for parameter estimation, decrease calculated amount.
At present, there is not yet the report about hydrometallurgy dense washing process key variables Forecasting Methodology.Still there is no the accurate dense washing process model that can be applied to optimal control.Dense washing process Modeling Research is still in the exploratory stage.
Summary of the invention
Object of the present invention, is to provide a kind of hydrometallurgy dense washing process key variables detection method.
The present invention is one of Ministry of Science and Technology's approval " selecting smelting detection and Optimized-control Technique " brainstorm project main contents, national high-tech research development plan (863 Program) project of relying on Northeastern University to bear-hydrometallurgy whole process comprehensive intelligent one Optimized-control Technique, carrying out the research of hydrometallurgy dense washing process underflow density flexible measurement method and application aspect, controlling to lay the foundation for realizing dense washing process optimal control and hydrometallurgy overall process optimization.
Dense washing process key variables detection method provided by the present invention, its feature comprises (1) data acquisition; (2) the choosing and the steps such as data processing (3) mixture model foundation of auxiliary variable.
(1) data acquisition
The device used in data acquisition comprise dense washing process key variables detection system, host computer, PLC, on-the-spot sensing become send part.Wherein dense washing process key variables detection system comprises data receiver part and forecast model part, receiving unit receives the data of host computer collection and does corresponding process, and forecast model part is carried out the prediction of key variables according to the data received and result carried out preservation and chart display.Wherein on-the-spot sensing change send part to comprise the measuring instruments such as pressure, concentration, flow.At dense washing process in-site installation measuring instrument, the signal of collection is delivered to PLC by measuring instrument, collection signal is sent to host computer by Ethernet timing by PLC, and host computer passes to dense washing process key variables prognoses system the data received, and carries out the real-time estimate of key variables.
The functions of apparatus of the present invention:
A () on-the-spot sensing becomes send part: comprise the measuring instruments such as pressure, concentration, flow, be made up of sensor, is responsible for collection and the transmission of process data;
(b) PLC: be responsible for the signal A/D conversion gathering, and by Ethernet, signal sent to host computer;
C () host computer: collect local plc data, sends dense washing process key variables prognoses system to, real-time estimate process key variable.
(2) the choosing and data processing of auxiliary variable
The selection of auxiliary variable is the first step of process of establishing data model, and this step determines the input information matrix of model, thus directly determines structure and the output of process model, most crucial to the success or not of modeling.The selection of auxiliary variable comprises the selection of types of variables and the selection of check point position.By the analysis to process mechanism, larger variable is affected on underflow density and effluent concentration and comprises feed thickness, pan feeding flow, underflow flow, excess flow, these four variablees can be surveyed in industry spot simultaneously, therefore select them as input variable, select effluent concentration, underflow density as output variable, carry out the research of next step mixture model and model tuning.
In order to prevent each detection variable from having an impact to data model due to unit difference, first the sensor measurement data collected is carried out standardization:
X in formula
idata after-process;
X
i-sample data;
X
max-sample data maximal value;
X
min-sample data minimum value.
(3) mixture model is set up
(a) mechanism model
The essence of the process of dense washing is the process of solid-liquid by sedimentation separation, first mathematician Cauchy really setting up the Ren Shi Birmingham, GBR university of sedimentation theory.Nineteen fifty-two he delivered famous paper " sedimentation theory ", this section of article establishes dynamic settling theory according to the diffusion of sedimentation ripple in suspending liquid, and Cauchy's theory thinks that suspending liquid is non-individual body, and settling process can be described by the continuity equation of solid phase:
The basic assumption of Cauchy's theory can be summed up as following 3 points:
(1) in any flat seam in suspended solid district, the concentration of suspended solid is uniform, and total solids is with same velocity sedimentation.Particle shape, size and composition do not affect settling process;
(2) settling velocity of solid particle is only the function of local concentration near particle;
(3) within the scope of whole precipitate height, initial concentration is uniform, or increases gradually along depth of precipitation.
Cauchy's sedimentation formula describes the settling process mechanism in thickening pond, considers dispersion and the dense washing course of work (pan feeding, underflow discharge, overflow discharge), obtains the mechanism model of dense washing process according to law of conservation of mass:
Q in formula
f-unit area pan feeding flow;
C
f-input concentration;
D-dispersion coefficient;
δ-dirac momentum;
The height of z-concentrator;
Z
f-entry level;
V
s-settling velocity;
C-pulp density;
The t-time;
Unit area flow in q-thickening pond.
Adopt the thought of layering, as shown in Figure 4.Thickening pond is divided into the dense layer of some even concentration by layering, and every one deck of layering meets Cauchy's sedimentation theory hypothesis.
After layering, above-mentioned mass balance equation is converted into first order differential equation system, the partial differential equation being about to solve complexity is converted into and solves first order differential equation system, such as formula 4.
Q in formula
e-unit area excess flow;
Q
u-unit area underflow flow;
Q
f-unit area pan feeding flow;
C
f-input concentration;
D-dispersion coefficient;
The number of plies of n-layering;
The height of z-concentrator;
Z
f-entry level;
V
s-settling velocity;
The concentration of C-ore pulp.
Can find out by observing settling process mechanism model, the key of modeling selects suitable terminal velocity model, the present invention regards as unified precipitation process various precipitated form, set up unified mathematical model, when the concentration of ore pulp is very low, settling process depends on the character of particle, but when the concentration of ore pulp is higher, settling process depends on the concentration of ore pulp, so low concentration settling process had both depended on that particle properties also depended on the concentration of ore pulp.Adopt unified terminal velocity model, Tak á cs terminal velocity model is:
The concentration of C-sedimentation body in formula;
V
0the solid particle of '-in fact maximum settlement speed;
V
0-theoretical solids particle maximum settlement speed;
R
hthe performance parameter that-decision speed diminishes, because the interaction hourly velocity between particle diminishes;
R
pwhen-solution concentration is less, decision speed becomes large performance parameter;
C
minwhen-settling velocity is 0, minimum concentration.
(b) data model
Under many circumstances, utilize merely mechanism model to be not enough to all characteristics of description process, the measurable variable in some processes, due to complicated mechanism, is difficult to all be included among mechanism model; In addition, some X factors understand the precision of prediction of Influencing Mechanism model equally, at this moment can compensate the Unmarried pregnancy in mechanism model with utilizing data model, to improve the precision of prediction of model.
PLS is a kind of dimensionality reduction technology utilizing the useful information process of establishing model of Statistics leaching process data.It not only can complete Data Dimensionality Reduction and feature extraction, also contemplates the regression relation between inputoutput data.
PLS proposes the method adopting constituents extraction, in principal component analysis (PCA), in order to find the generalized variable summarizing former data message best, extracts first principal component F respectively for argument data Table X and dependent variable tables of data Y
1, G
1, make F
1, G
1middle comprised former data variation information can reach maximum, i.e. var (F
1) → max, var (G
1) → max, ensures F simultaneously
1, G
1the degree of correlation is maximum.
The data matrix of independent variable X after standardization is designated as E
0=(E
01..., Ε
0p)
n × p, because the data matrix of scalar Y after standardization is designated as F
0=(F
01..., F
0p)
n × p, t
1e
0first principal component t
1=E
0w
1, w
1e
0the first axle, it is a vector of unit length, || w
1||=1; u
1f
0first principal component u
1=F
0c
1, c
1f
0the first axle, it is a vector of unit length, || c
1||=1.Var (t
1) → max, var (u
1) → max, requires t
1to u
1there is maximum interpretability r (t
1, u
1) → max.
PLS algorithm nonlinear iteration partial least square algorithm realization step is as follows:
Step one: get E
0=X, F
0=Y, h=1;
Step 2: get u
h=y
j, y
jf
h-1in arbitrary column vector, obtain variance maximum column vector;
Step 3: calculate input weight vector w
h: w
h t=u
h te
h-1/ u
h tu
h, by w
hnormalization: w
h=w
h/ || w
h||;
Step 4: calculate input score vector: t
h=E
h-1w
h/ (w
h tw
h);
Step 5: calculate output load vector q
h: q
h t=t
h tf
h-1/ (t
h tt
h) by q
hnormalization: q
h=q
h/ || q
h||, calculate and export score vector: u
h=F
h-1q
h/ q
h tq
h;
Step 6: repeat step 3 to step 6, until convergence.Check the way of convergence be see " with previous difference whether within the range of permission, be usually no more than 10 iteration and will reach convergence (if Y only contains a variable, then step 5 can be omitted to step 7, directly puts q=1);
Step 7: calculate input u
h=F
h-1q
h/ q
h tq
hload vector: p
h=t
he
h-1/ (t
h tt
h);
Step 8: calculate internal model regression coefficient b
h: b
h=u
h tt
h/ (t
h tt
h);
Step 9: to residual matrix E
h-1and F
h-1reduce:
Step 10: make h=h+1, goes to step two, until calculate all proper vectors.
After calculating PLS model, first calculate the PLS model parameter of former data, then calculate the PLS model regression coefficient Matrix C for predicting
pLS.C
pLSfollowing formula can be adopted to calculate
[33]:
C
PLS＝W(P
TW)
-1BQ
T (6)
When obtaining new samples X
newtime, utilize C
pLSdirectly prediction deviation is calculated by following formula:
(c) mixture model
Dense washing process key variables detection method based on mixture model is carried out according to following steps:
Step one, mechanism model parameter calculate: according to the parameter in historical data identification mechanism model;
Step 2, mechanism model are predicted: utilize mechanism model to predict effluent concentration and underflow density, and record predicts the outcome;
Step 3, image data: collect the effluent concentration of off-line chemical examination and the process operation parameter of underflow density value and the sensor measurement corresponding to off-line laboratory values;
Step 4, will predict the outcome compares with true testing result, the difference between computational prediction result and actual value;
Step 5, data model are trained: by the sensor measurement data collected and above-mentionedly to predict the outcome and difference between actual value forms inputoutput data pair, utilize PLS method to train, obtain the parameter in data model;
The prediction of step 6, mixture model: mechanism model and data model are composed in parallel mixture model, realize carrying out real-time estimate to effluent concentration and underflow density, the structure of mixture model as shown in Figure 5.
The present invention has following advantage:
1) the present invention is by the research of this gordian technique of hard measurement, realizes monitoring complex process;
2) method with mixed model of the present invention had both considered the advantage of mechanism model, considered again the feature of data model, and thus achievement in research of the present invention can also be applied to the industrial circles such as other chemical industry.
The system software of enforcement provided by the present invention dense washing process key variables Forecasting Methodology, it comprises primary module, algoritic module, database and interface.Primary module mainly carries out initialization to program, reads input data, starts clock, periodically by field measurement data write into Databasce needed for software, close database file; Data acquisition, data processing, model parameter amendment, prediction algorithm calculating is mainly comprised in algoritic module.Present system is provided with 3 interfaces: comprise switching between interface of main interface-realize and main program module runs; Prediction interface-realize the display of dense washing process key variables real-time estimate; The correlation parameter of parameters input and real-time display interface-input and correction model and realize the real-time display of Site Detection data and the input of offline metrology data.
Dense washing process key variables Forecasting Methodology provided by the present invention and system are supporting with the basic automation systems at scene, in use it is loaded in the computing machine of a prototype, namely using prototype as hardware platform of the present invention, become with the host computer of basic automatization part, PLC, on-the-spot sensing and send part collaborative work.Wherein on-the-spot sensing change send part to comprise the measuring instruments such as pressure, concentration, flow.At dense washing process in-site installation measuring instrument, the signal of collection is delivered to slave computer by measuring instrument, by the timing of Ethernet slave computer, collection signal is sent to host computer, host computer passes to the dense washing process key variables prognoses system of prototype the data received, carry out effluent concentration and underflow density real-time estimate, and show in host computer configuration interface.
Accompanying drawing explanation
Fig. 1 is concentrator inside processing spatial distribution map;
Fig. 2 is mixture model serial structure figure;
Fig. 3 is mixture model parallel organization figure;
Fig. 4 is mechanism model hierarchical diagram;
Fig. 5 is dense washing process mixture model structural drawing;
Fig. 6 is the contrast of effluent concentration prediction effect;
Fig. 7 is the contrast of underflow density prediction effect;
Fig. 8 is dense washing process prediction interface figure.
Embodiment
Be described in detail below in conjunction with the embodiment of accompanying drawing to content of the present invention.
Adopt the validity of Forecasting Methodology that the process units of the dense washing process of certain factory illustrates specific embodiment of the invention step and applies.
(1) embodiment 1
The Forecasting Methodology of Key of Implementation variable on concentrator production line.
Specific implementation process is as follows:
1) choosing auxiliary variables: the selection of auxiliary variable is the first step setting up soft-sensing model, this step determines the input information matrix of hard measurement, thus directly determines structure and the output of soft-sensing model, most crucial to the success or not of hard measurement.The selection of auxiliary variable comprises the selection of the selection of types of variables, the selection of variable quantity and check point position.
In dense washing process, we choose feed thickness, pan feeding flow, underflow flow, excess flow, are auxiliary variable.
2) data acquisition and processing (DAP): carried out the collection of data at the scene of dense washing process.Concrete measurement instrument is the corresponding instrument of introduction above, collects on-the-spot corresponding working condition data.
Due in practical problems, the measuring unit of each variable is inconsistent, if without certain process, can exaggerate the effect of wherein large dimension data, and ignore its dependent variable, thus truly can not reflect the situation of change of data itself.Therefore, the dimensional effect of variable be eliminated, make each variable have equal expressive ability.Prediction due to model can only provide the expectation value of response variable to the independent variable in given range, namely the predictive ability of model has interpolate value characteristic, and does not have extrapolated value characteristic, and the input exceeding maximal value will produce larger output error.Therefore, in order to make soft-sensing model have good predictive ability, the modeling training sample that we choose should comprise the minimum of each variable and maximal value as far as possible, thus makes checking data all within the scope of sample data, reduces predicated error.Its mathematic(al) representation is:
X in formula
idata after-process;
X
i-sample data;
X
max-sample data maximal value;
X
min-sample data minimum value.
The collection of data gathers mainly through PLC system and on-the-spot Sensor section.These calculating can call primary module in dense washing process prognoses system software by the computing machine in prototype and corresponding algoritic module-data processing completes automatically.
3) reduced kinetic mechanism prediction and calculation: the data after process are brought in the mechanism model of simplification and carry out mechanism model calculating.These calculate and call primary module in dense washing process key variables prognoses system software by the computing machine in prototype and corresponding algoritic module completes automatically.
4) mixture model prediction and calculation: being predicted the outcome by mechanism model compares with true testing result, calculating mechanism predicts the outcome and difference between actual value; By the sensor measurement data collected and above-mentionedly to predict the outcome and difference between actual value forms inputoutput data pair, utilize above-mentioned PLS method to train, obtain the parameter in data model; Mechanism model and data model are composed in parallel mixture model, and real-time estimate is carried out to dense washing process key variables.These calculating can call primary module in dense process key variable prediction system software by the computing machine in prototype and corresponding algoritic module-mixture model completes automatically.
The dense washing process key variables detection model set up according to this modeling method is in the dense washing process units of certain factory.In order to the validity of model is described, calculating data and laboratory assay value is adopted to compare.The curve of model output effluent concentration value and assay value compares sees accompanying drawing 6.Partial results is in table 1.
Table 1 effluent concentration predicts the outcome
Table 2 underflow density predicts the outcome
When predicting effluent concentration, the square error (MSE) of mixture model is 0.8054, and maximum absolute error (MAE) 0.0425 the results show that the precision comparison of effluent concentration forecast model is high; When predicting underflow density, the square error (MSE) of mixture model is 0.0512, and maximum absolute error (MAE) 0.0127 the results show that the precision comparison of underflow density forecast model is high; Can find out that mixture model predicted value and laboratory values trend comparison coincide from table 1, table 2 and accompanying drawing 6, accompanying drawing 7.This shows, mixture model modeling method has good estimated performance to dense washing process key variables (effluent concentration and underflow density), is applicable to the online computing application of industry.
Hybrid modeling method of the present invention is when detecting dense washing process key variables, and friendly human-computer interaction interface is absolutely necessary.Invention contemplates this requirement, prediction interface is combined with process control interface harmonious, be illustrated in figure 8 dense washing process key variables prediction interface.
The function that human-computer interaction interface realizes:
(1) process control tab realizes the real-time control of underflow density and the historical query of data;
(2) model prediction tab realizes the setting of dense washing process model parameter, the Real-time Collection of mode input amount, the prediction of underflow density and effluent concentration and Dynamic Announce, and can enquiry of historical data;
(3) IP of setting options card fulfillment database server is arranged.
Claims (3)
1. hydrometallurgy dense washing process key variables detection method, adopt the dense washing process technique of known hydrometallurgy, be made up of hardware support platform and software systems, it is characterized in that the feed rate, input concentration, excess flow and the underflow flow that directly control during the course to enter in concentrator, by mixture model modeling, real-time estimate effluent concentration and underflow density, comprise data acquisition, choosing auxiliary variables and data processing, mixture model establishment step:
1) data acquisition
Data acquisition device hardware used comprises dense washing process key variables detection system, host computer, PLC and on-the-spot sensing and becomes and send part, and wherein on-the-spot sensing becomes and send part to comprise concentration, flow and pressure instrumentation; Measuring instrument is responsible for collection and the transmission of process data, at dense process-field, measuring instrument is installed, the pressure distribution information of the concentration value of collection, flow and differing heights is delivered to PLC by data bus by measuring instrument in real time, collection signal is sent to host computer by Ethernet timing by PLC, host computer passes to dense washing process key variables prognoses system the data received, and carries out the real-time detection of effluent concentration and underflow density;
PLC function: be responsible for the signal A/D conversion gathering, and by Ethernet, signal sent to host computer;
Host computer function: collect local plc data, sends dense washing process key variables detection system to, detects effluent concentration and underflow density in real time;
2) choosing auxiliary variables and data processing
The auxiliary variable selected comprises:
(a) input concentration x
1;
(b) feed rate x
2;
(c) underflow flow x
3;
(d) excess flow x
4;
First the sensor measurement data collected is carried out standardization:
①
X in formula
idata after-process;
X
i-sample data;
X
max-sample data maximal value;
X
min-sample data minimum value;
3) mixture model is set up:
Mechanism model and data model are composed in parallel mixture model, realizes detecting in real time dense washing process key variables;
The mechanism model of dense washing process is set up according to solids flux theory and mass conservation theory;
②
Q in formula
f-unit area pan feeding flow;
C
f-input concentration;
D-dispersion coefficient;
δ-dirac momentum;
The height of z-concentrator;
Z
f-entry level;
V
s-settling velocity;
C-pulp density;
The t-time;
Unit area flow in q-thickening pond;
Adopt the thought of layering, thickening pond is divided into the dense layer of some even concentration, every one deck all meets Cauchy's sedimentation theory hypothesis;
After layering, above-mentioned mass balance equation is converted into first order differential equation system, is about to solve complicated partial differential equation and is converted into and solves first order differential equation system, such as formula 3.;
③
Q in formula
e-unit area excess flow;
Q
u-unit area underflow flow;
Q
f-unit area pan feeding flow;
C
f-input concentration;
D-dispersion coefficient;
The number of plies of n-layering;
The height of z-concentrator;
Z
f-entry level;
V
s-settling velocity;
The concentration of C-ore pulp;
Settling velocity function selects Tak á cs terminal velocity model to be:
④
The concentration of C-sedimentation body in formula;
V
0the solid particle of '-in fact maximum settlement speed;
V
0-theoretical solids particle maximum settlement speed;
R
hthe performance parameter that-decision speed diminishes, because the interaction hourly velocity between particle diminishes;
R
pwhen-solution concentration is less, decision speed becomes large performance parameter;
C
minwhen-settling velocity is 0, minimum concentration;
Adopt PLS offset minimum binary as the Unmarried pregnancy of Data Modeling Method matching mechanism model in the present invention;
Dense washing process key variables Forecasting Methodology based on mixture model is carried out according to following steps:
Step one, mechanism model parameter calculate: according to the parameter in historical data identification mechanism model;
Step 2, mechanism model are predicted: utilize mechanism model to predict effluent concentration and underflow density, and record predicts the outcome;
Step 3, image data: collect the effluent concentration of off-line chemical examination and the process operation parameter of underflow density and the sensor measurement corresponding to off-line laboratory values;
Step 4, will predict the outcome compares with true testing result, the difference between computational prediction result and actual value;
Step 5, data model are trained: by the sensor measurement data collected and above-mentionedly to predict the outcome and difference between actual value forms inputoutput data pair, utilize above-mentioned PLS method to train, obtain the parameter in data model;
The prediction of step 6, mixture model: mechanism model and data model are composed in parallel mixture model, realizes carrying out real-time estimate to effluent concentration and underflow density;
2. hydrometallurgy according to claim 1 dense washing process key variables Forecasting Methodology, is characterized in that: software systems, and it comprises primary module, algoritic module, database and interface, it is characterized in that:
1) primary module carries out initialization to program, reads input data, starts clock, periodically by field measurement data write into Databasce needed for software, closes database file;
2) algoritic module comprises data acquisition, data processing, model parameter amendment, prediction algorithm calculating; For primary module recursive call;
3) system interface comprises: main interface---realize the switching between interface and main program module operation; Prediction interface---realize the real-time estimate display of effluent concentration and underflow density; Parameters input and real-time display interface---input and correction model correlation parameter and realize the real-time display of Site Detection data and the input of offline metrology data.
3. hydrometallurgy according to claim 2 dense washing process key variables prognoses system, is characterized in that hardware platform made by the prototype of primary module, algoritic module, database and the interface system of metallurgical dense washing process key variables Forecasting Methodology in a wet process;
The above effluent concentration and underflow density detection method and system are supporting with the basic automation systems at scene, it loads in the computing machine of a prototype in use, namely using prototype as hardware platform of the present invention, become with the host computer of basic automatization part, PLC, on-the-spot sensing and send part collaborative work; Wherein on-the-spot sensing change send part to comprise concentration, flow and pressure instrumentation; At dense washing process in-site installation measuring instrument, the signal of collection is delivered to slave computer by measuring instrument, by the timing of Ethernet slave computer, collection signal is sent to host computer, host computer passes to the dense washing process key variables prognoses system of prototype the data received, carry out effluent concentration and underflow density real-time estimate, and show in host computer configuration interface.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410258081.5A CN104296801B (en) | 2014-06-12 | 2014-06-12 | Hydrometallurgy thick washing process key variable detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410258081.5A CN104296801B (en) | 2014-06-12 | 2014-06-12 | Hydrometallurgy thick washing process key variable detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104296801A true CN104296801A (en) | 2015-01-21 |
| CN104296801B CN104296801B (en) | 2017-01-18 |
Family
ID=52316632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410258081.5A Active CN104296801B (en) | 2014-06-12 | 2014-06-12 | Hydrometallurgy thick washing process key variable detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104296801B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108536979A (en) * | 2018-04-18 | 2018-09-14 | 东北大学 | A kind of underflow density prediction technique based on concentrator mechanism model |
| CN110276128A (en) * | 2019-06-21 | 2019-09-24 | 东北大学 | A kind of thickener underflow concentration prediction method based on DAJYPLS algorithm |
| WO2021203912A1 (en) * | 2020-04-10 | 2021-10-14 | 中国恩菲工程技术有限公司 | Online prediction method for parameters in copper converting process based on oxygen bottom blowing furnace |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008099457A1 (en) * | 2007-02-09 | 2008-08-21 | Toshiba Mitsubishi-Electric Industrial Systems Corporation | Process line control unit and method of controlling the line |
| CN101526814A (en) * | 2009-02-03 | 2009-09-09 | 东北大学 | Leaching rate prediction and optimization operation method in wet metallurgical leaching process |
| CN101813932A (en) * | 2009-02-03 | 2010-08-25 | 东北大学 | Method for component content prediction and optimization operation in wet-process metallurgic extraction process |
| CN102033523A (en) * | 2009-09-25 | 2011-04-27 | 上海宝钢工业检测公司 | Strip steel quality forecasting, furnace condition early-warning and fault diagnosis method based on partial least square |
-
2014
- 2014-06-12 CN CN201410258081.5A patent/CN104296801B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008099457A1 (en) * | 2007-02-09 | 2008-08-21 | Toshiba Mitsubishi-Electric Industrial Systems Corporation | Process line control unit and method of controlling the line |
| CN101526814A (en) * | 2009-02-03 | 2009-09-09 | 东北大学 | Leaching rate prediction and optimization operation method in wet metallurgical leaching process |
| CN101813932A (en) * | 2009-02-03 | 2010-08-25 | 东北大学 | Method for component content prediction and optimization operation in wet-process metallurgic extraction process |
| CN102033523A (en) * | 2009-09-25 | 2011-04-27 | 上海宝钢工业检测公司 | Strip steel quality forecasting, furnace condition early-warning and fault diagnosis method based on partial least square |
Non-Patent Citations (4)
| Title |
|---|
| 周天驰: "湿法冶金浸出过程机理模型", 《有色冶金设计与研究》 * |
| 张俊 等: "湿法冶金浸出过程金浓度的混合预报模型", 《仪器仪表学报》 * |
| 胡广浩 等: "湿法冶金浸出过程建模与仿真研究", 《湿法冶金浸出过程建模与仿真研究》 * |
| 袁青云 等: "湿法冶金中预测金产量的混合建模方法", 《东北大学学报（自然科学版）》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108536979A (en) * | 2018-04-18 | 2018-09-14 | 东北大学 | A kind of underflow density prediction technique based on concentrator mechanism model |
| CN108536979B (en) * | 2018-04-18 | 2021-08-03 | 东北大学 | Underflow concentration prediction method based on thickener mechanism model |
| CN110276128A (en) * | 2019-06-21 | 2019-09-24 | 东北大学 | A kind of thickener underflow concentration prediction method based on DAJYPLS algorithm |
| WO2021203912A1 (en) * | 2020-04-10 | 2021-10-14 | 中国恩菲工程技术有限公司 | Online prediction method for parameters in copper converting process based on oxygen bottom blowing furnace |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104296801B (en) | 2017-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101526814B (en) | Leaching rate prediction and optimization operation method in wet metallurgical leaching process | |
| CN104062953B (en) | Golden hydrometallurgy process running status on-line evaluation method | |
| CN104460721B (en) | Method for optimally controlling underflow concentration in hydrometallurgy dense washing process | |
| CN101813932B (en) | Method for component content prediction and optimization operation in wet-process metallurgic extraction process | |
| CN109935280B (en) | Blast furnace molten iron quality prediction system and method based on ensemble learning | |
| CN101482750B (en) | Cobalt oxalate granularity prediction method in hydrometallurgical synthesis process | |
| CN101382556A (en) | Soft measurement method for floating fine coal ash from slurry based on data drive | |
| CN104597755A (en) | Hydrometallurgical gold cyanide leaching optimization method | |
| CN103412489A (en) | Ore grinding granularity online prediction system and method | |
| CN105278520A (en) | Complex industrial process running state evaluation method and application based on T-KPRM | |
| CN105974953A (en) | Reaction vessel negative pressure rectification fuzzy control method | |
| CN105574297B (en) | Self adaptation blast furnace molten iron silicon content trend prediction method | |
| CN104296801A (en) | Hydrometallurgy thick washing process key variable detection method | |
| CN108536979B (en) | Underflow concentration prediction method based on thickener mechanism model | |
| CN109885012A (en) | A kind of gold hydrometallurgy whole process real-time optimization compensation method | |
| CN108037747B (en) | Golden hydrometallurgy whole process hierarchical block operating status on-line evaluation method | |
| Hodouin | Process observers and data reconciliation using mass and energy balance equations | |
| CN209027943U (en) | Single-well crude oil moisture content online weighing measuring system | |
| CN104263960B (en) | On-line prediction method for replacement rate and gold mud grade in gold hydrometallurgy replacement process | |
| CN110276128B (en) | Underflow concentration prediction method of thickener based on DAJYPLS algorithm | |
| CN104199285A (en) | Leaching rate prediction method for wet metallurgy gold cyaniding leaching process | |
| CN107832880B (en) | Blast furnace state variable prediction method based on material distribution parameters | |
| CN106886154B (en) | Hydrometallurgy whole process modeling method and optimization method based on interval number | |
| CN107583755A (en) | A kind of predictor method on semi-autogenous mill power | |
| Zhang et al. | Research on soft-sensor based on support vector regression for particle size of grinding and classification process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |