CN113307436B - HMM model-based evaporative crystallization solid-liquid separation system and method - Google Patents

HMM model-based evaporative crystallization solid-liquid separation system and method Download PDF

Info

Publication number
CN113307436B
CN113307436B CN202110574119.XA CN202110574119A CN113307436B CN 113307436 B CN113307436 B CN 113307436B CN 202110574119 A CN202110574119 A CN 202110574119A CN 113307436 B CN113307436 B CN 113307436B
Authority
CN
China
Prior art keywords
solid
hmm model
liquid
centrifugal separator
cyclone
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.)
Active
Application number
CN202110574119.XA
Other languages
Chinese (zh)
Other versions
CN113307436A (en
Inventor
赵泽盟
刘清宝
史元腾
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.)
China Coal Energy Research Institute Co Ltd
Original Assignee
China Coal Energy Research Institute Co Ltd
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 China Coal Energy Research Institute Co Ltd filed Critical China Coal Energy Research Institute Co Ltd
Priority to CN202110574119.XA priority Critical patent/CN113307436B/en
Publication of CN113307436A publication Critical patent/CN113307436A/en
Application granted granted Critical
Publication of CN113307436B publication Critical patent/CN113307436B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F1/00—Treatment of water, waste water, or sewage
    • C02F1/02—Treatment of water, waste water, or sewage by heating
    • C02F1/04—Treatment of water, waste water, or sewage by heating by distillation or evaporation
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F1/00—Treatment of water, waste water, or sewage
    • C02F1/38—Treatment of water, waste water, or sewage by centrifugal separation
    • C—CHEMISTRY; METALLURGY
    • C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
    • C02F2303/00—Specific treatment goals
    • C02F2303/14—Maintenance of water treatment installations

Landscapes

  • Chemical & Material Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Hydrology & Water Resources (AREA)
  • Environmental & Geological Engineering (AREA)
  • Water Supply & Treatment (AREA)
  • Organic Chemistry (AREA)
  • Analytical Chemistry (AREA)
  • Mechanical Engineering (AREA)
  • Centrifugal Separators (AREA)
  • Cyclones (AREA)

Abstract

The invention provides an HMM model-based evaporative crystallization solid-liquid separation system and method, which comprises the following steps: the image preprocessing step, wherein the preprocessing technology comprises Gaussian filtering, mean filtering, median filtering, minimum mean square error filtering and Gabor filtering; an HMM construction step, namely training a hidden Markov model to obtain relevant parameters of the HMM; a step of predicting materials to be subjected to solid-liquid separation, which is to input information of the materials to be subjected to solid-liquid separation into a trained HMM to obtain a predicted state, wherein an HMM model is obtained by training image information in a swirler of the same evaporative crystallization system and the corresponding solid content of the materials; and a material separation step, namely scheduling the material forward direction according to the predicted material state to be separated. The invention can automatically control the solid-liquid separation system of the evaporative crystallization process, and prevent the inseparable material from entering the subsequent process section to cause the problems of agglomeration, blockage and the like of the subsequent process section. The cloud computing technology is adopted, so that the reliability and the compatibility of data acquisition are improved, and the computing cost is reduced.

Description

HMM model-based evaporative crystallization solid-liquid separation system and method
Technical Field
The invention relates to the field of solid-liquid separation in an evaporative crystallization process, in particular to an evaporative crystallization solid-liquid separation system and method based on an HMM model.
Background
One key point in 'zero discharge' of mine water and coal chemical wastewater and resource engineering is the solid-liquid separation process of the saturated solution of the crystallized salt after evaporation and crystallization. Whether this process is unobstructed directly influences the steady operation of whole engineering to and front end reuse water volume and treatment water volume.
At the present stage, zero discharge of mine water and industrial wastewater and resource engineering evaporative crystallization processes often occur, and solid content of materials entering a centrifugal separator is too thin, so that the centrifugal separator cannot separate solid and liquid phases, subsequent drying equipment is agglomerated, the whole crystallization process is forced to stop, and the drying equipment is cleaned. The continuous and stable operation of the whole 'zero emission' and resource engineering is seriously influenced, and the workload is increased. In the 'zero discharge' project of partial mine water and industrial wastewater, the scale of a front cyclone entering the centrifugal separator is enlarged, and the solid content entering the centrifugal separator is ensured, so that the long-period stable operation of the centrifugal separator is ensured. However, the technical route can cause the pipeline between the cyclone and the centrifugal separator to be blocked, the pipeline needs to be cleaned frequently, the whole zero emission and the stable operation of a recycling system are also not facilitated, and the problem of solid-liquid separation in the evaporative crystallization process is not solved fundamentally.
In addition, the conventional HMM model does not assume that the dependent variable follows a normal distribution, and the solid content in the actual solid-liquid material to be separated is not normally distributed, so that the conventional HMM model cannot predict accurately.
Disclosure of Invention
Aiming at the problems that the existing mine water and industrial wastewater zero emission and evaporative crystallization process in the resource technology can not effectively separate solid and liquid, and the conventional HMM model can not accurately describe the distribution condition of the solid content of the actual material, the invention provides an evaporative crystallization solid-liquid separation system and method based on the HMM model; the system has a simple structure, effectively prevents materials with low solid content from entering the drying equipment and agglomerating, thereby playing a role in ensuring continuous and stable operation of the evaporative crystallization process.
The invention is realized by the following technical scheme:
an HMM model-based evaporative crystallization solid-liquid separation system comprises a controller and a solid-liquid separation unit;
the solid-liquid separation unit comprises a to-be-cyclone separator, a centrifugal separator, a to-be-separated material device, evaporation crystallization process equipment and drying equipment; the output end of the material device to be separated is connected with the input end of the swirler; the output end of the cyclone is respectively connected with a centrifugal separator and evaporative crystallization process equipment; the output end of the centrifugal separator is respectively connected with evaporative crystallization process equipment and drying equipment;
the input end of the controller is respectively connected with the image acquisition module and the signal input module, and the output end of the controller is respectively connected with the image preprocessing module and the data processing module; the input end of the image acquisition module is respectively connected with the first high-definition camera and the second high-definition camera; the input end of the signal input module is connected with the sample collector;
the first high-definition camera is assembled on the cyclone, the second high-definition camera is assembled at a material outlet of the centrifugal separator, and the sample collector is assembled on the material device to be separated.
Preferably, a human-computer interaction module is arranged on the controller, and the input end of the human-computer interaction module is connected with the output end of the controller and used for displaying images acquired by the first high-definition camera and the second high-definition camera and displaying signal information of the sample collector, the image preprocessing module and the data processing module.
Preferably, the output end of the controller is connected with a monitor through a monitoring module, and the monitors are respectively assembled on the material outlet output ends of the cyclone and the centrifugal separator.
Preferably, the output end of the cyclone is provided with an automatic valve for the materials with unqualified cyclone outlets and an automatic valve for the materials with qualified cyclone outlets; the automatic valve of the unqualified material at the outlet of the cyclone is connected with evaporation crystallization process equipment; the automatic valve for qualified materials at the outlet of the cyclone is connected with the input end of the centrifugal separator.
Preferably, the output end of the centrifugal separator is provided with an automatic valve for qualified materials at the outlet of the centrifugal separator and an automatic valve for unqualified materials at the outlet of the centrifugal separator; the automatic valve of the unqualified material at the outlet of the centrifugal separator is connected with evaporative crystallization process equipment; the automatic valve for qualified materials at the outlet of the centrifugal separator is connected with drying equipment.
An HMM model-based evaporative crystallization solid-liquid separation method is based on the evaporative crystallization solid-liquid separation system and comprises the following steps,
step 1, collecting a training sample set by using a sample collector, wherein the training sample set acquires image information of a material to be separated in a cyclone;
step 2, carrying out image preprocessing on the acquired image information through an image preprocessing module;
step 3, screening the materials in the cyclone through a data processing module after image preprocessing, feeding the normal solid-liquid separable materials into a centrifugal separator to execute step 4, and feeding the solid-liquid inseparable materials back to the evaporation crystallization process equipment to execute step 1 again;
step 4, acquiring image information in the centrifugal separator, and carrying out image preprocessing on the acquired image information through an image preprocessing module;
step 5, screening the materials in the centrifugal separator through a data processing module after image preprocessing, feeding the normal solid-liquid separable materials into drying equipment, and finishing the separation work; returning the solid-liquid inseparable material to the evaporation crystallization process equipment to continue evaporation and concentration, and then re-executing the steps.
Preferably, the image preprocessing employs noise reduction preprocessing on the image information, wherein the image preprocessing techniques include gaussian filtering, mean filtering, median filtering, minimum mean square error filtering, and Gabor filtering.
Preferably, the data processing module adopts an HMM model technology to sieve the materials in the cyclone and at the outlet of the centrifugal separator, and the specific method is as follows:
constructing a first HMM model and a second HMM model;
inputting the image information of the material in the cyclone after the pretreatment of the material to be separated into a first HMM model to output a probability value P (O-lambda) 1 (ii) a Probability value P (O-lambda) 1 Respectively corresponding to normal solid-liquid separable state S 11 And a solid-liquid inseparable state S 12 ；
When the image information of the material in the cyclone is output by the first HMM model, the probability value P (O-lambda) is output 1 In a normal solid-liquid separable state S 11 In the range of (1), the material enters a subsequent centrifugal separator; when the image information of the material in the cyclone is output with a first HMM model, a probability value P (O-lambda) 1 In a solid-liquid inseparable state S 12 Returning the material to the evaporative crystallization process equipment 10;
inputting the image information of the centrifuge outlet material after the pretreatment of the material to be separated into a second HMM model to output a probability value P (O-lambda) 2 (ii) a Probability value P (O-lambda) 2 Respectively corresponding to normal solid-liquid separable state S 21 And a solid-liquid inseparable state S 22 ；
When the image information of the material at the outlet of the centrifugal separator is output with the second HMM model, the probability value P (O-lambda) is output 2 In a normal solid-liquid separable state S 21 The material enters a drying device within the range of (1); when the image information of the material at the outlet of the centrifugal separator is output with the second HMM model, the probability value P (O-lambda) is output 2 In a solid-liquid inseparable stateS 22 And returning the material to the evaporation crystallization process equipment.
Further, the first HMM model and the second HMM model are obtained through supervised learning of EM, wherein the probability density distribution form of the first HMM model in the EM algorithm is obtained according to the solid content of the materials in the cyclone and the K-S test of the same evaporative crystallization process equipment. The second HMM model is obtained according to the solid content of the material at the outlet of a centrifugal separator of the same evaporative crystallization process equipment and K-S test in a probability density distribution form in an EM algorithm.
Further, a normal solid-liquid separable state S 11 The solid content of the material is more than or equal to 30wt%; solid-liquid inseparable state S 12 ＜30wt％；
The normal solid-liquid separable state S 21 The solid content of the material is more than or equal to 90wt%; solid-liquid inseparable state S 12 ＜90wt％。
Compared with the prior art, the invention has the following beneficial technical effects:
the invention provides an evaporative crystallization solid-liquid separation system based on an HMM model, which effectively monitors a solid-liquid separation unit through a controller, the controller respectively collects image information of a cyclone and a centrifugal separator through an image collection module and a signal input module, and effectively screens the cyclone and the centrifugal separator through an image preprocessing module and a data processing module, so that the accuracy of solid-liquid separation of materials in the solid-liquid separation unit is improved, unqualified materials can be returned to evaporative crystallization process equipment, the continuous and stable operation of the evaporative crystallization process is ensured, the materials with low solid content are prevented from entering drying equipment and agglomerating, the reliability and compatibility of data acquisition are improved by adopting the controller, and meanwhile, the calculation cost is also reduced.
Furthermore, a human-computer interaction module is arranged on the controller, and the input end of the human-computer interaction module is connected with the output end of the controller and used for displaying images acquired by the first high-definition camera and the second high-definition camera and displaying signal information of the sample collector, the image preprocessing module and the data processing module, so that an operator can effectively observe the image information condition in the equipment when performing solid-liquid separation operation.
Furthermore, the output end of the controller is connected with a monitor through a monitoring module, the monitor is respectively assembled on the output ends of the cyclone and the centrifugal separator, and preprocessed material image information in the cyclone and preprocessed material image information at the outlet of the centrifugal separator can be input into the data processing module for effective monitoring.
Furthermore, the output end of the cyclone is provided with an automatic valve for the materials with unqualified outlet of the cyclone and an automatic valve for the materials with qualified outlet of the cyclone, and the automatic valve for the materials with unqualified outlet of the cyclone is connected with evaporative crystallization process equipment to ensure the continuous and stable operation of the evaporative crystallization process; the automatic valve for qualified materials at the outlet of the cyclone is connected with the input end of the centrifugal separator, so that the separation operation between the cyclone and the centrifugal separator is facilitated.
Furthermore, the output end of the centrifugal separator is provided with an automatic valve for qualified materials at the outlet of the centrifugal separator and an automatic valve for unqualified materials at the outlet of the centrifugal separator, and the automatic valve for unqualified materials at the outlet of the centrifugal separator is connected with evaporative crystallization process equipment to ensure the continuous and stable operation of the evaporative crystallization process; the automatic valve for qualified materials at the outlet of the centrifugal separator is connected with the drying equipment, so that the separation operation between the centrifugal separator and the drying equipment is facilitated.
An evaporation crystallization solid-liquid separation method based on HMM model, the supplies image information in the cyclone judges whether the supplies to be separated enter the subsequent centrifuge through the data processing module, the supplies image information of the centrifuge exit judges whether the supplies to be separated enter the subsequent drying equipment through the data processing module, the invention avoids the supplies with lower solid content entering the drying equipment and agglomerating, thus play a role in guaranteeing the continuous, steady operation of evaporation crystallization process; the solid-liquid separation operation is more accurately carried out on the solid-liquid separation unit through the controller, and the operation stability of the evaporative crystallization process is improved.
Further, the data processing module adopts an HMM model technology to respectively sieve the materials in the cyclone and at the outlet of the centrifugal separator, and the probability density distribution form of the HMM model is obtained through the solid content of the materials in the cyclone of the evaporative crystallization process, the solid content of the materials at the outlet of the centrifugal separator and K-S test. And performing model parameter estimation by adopting normal distribution, logarithmic normal distribution and generalized extreme value distribution according to the actual distribution condition, so that the output result is fit with the actual condition, the monitoring control is more accurate and adaptive, and the operation stability of the evaporative crystallization process is improved.
Drawings
FIG. 1 is a schematic diagram of an HMM model-based evaporative crystallization solid-liquid separation system according to the present invention;
FIG. 2 is a schematic flow chart of an HMM model-based evaporative crystallization solid-liquid separation method according to the present invention;
fig. 3 is a schematic structural diagram of a controller according to the present invention.
In the figure: 1-a swirler; 2-a first high definition camera; 3-an automatic valve for unqualified materials at the outlet of the cyclone; 4-automatic valve for qualified materials at the outlet of the cyclone; 5-centrifugal separator; 6-a second high-definition camera; 7-automatic valve for qualified materials at the outlet of the centrifugal separator; 8-automatic valve for unqualified materials at the outlet of the centrifugal separator; 9-a controller; 10-evaporative crystallization process equipment; 11-drying equipment; 12-a material to be separated device.
Detailed Description
The present invention will now be described in further detail with reference to specific examples, which are intended to be illustrative, but not limiting, of the invention.
The invention provides an HMM model-based evaporative crystallization solid-liquid separation system, which comprises a controller 9 and a solid-liquid separation unit, as shown in FIG. 1;
the solid-liquid separation unit comprises a to-be-cyclone 1, a centrifugal separator 5, a to-be-separated material device 12, evaporative crystallization process equipment 10 and drying equipment 11; the output end of the material device 12 to be separated is connected with the input end of the swirler 1; the output end of the swirler 1 is respectively connected with a centrifugal separator 5 and evaporation crystallization process equipment 10; the output end of the centrifugal separator 5 is respectively connected with evaporative crystallization process equipment 10 and drying equipment 11; the first high-definition camera 2 is assembled on the cyclone 1, the second high-definition camera 6 is assembled on a material outlet of the centrifugal separator 5, and the sample collector is assembled on a material device 12 to be separated.
As shown in fig. 3, the input end of the controller 9 is connected to the image acquisition module and the signal input module, respectively, and the output end of the controller 9 is connected to the image preprocessing module and the data processing module, respectively; the input end of the image acquisition module is respectively connected with the first high-definition camera 2 and the second high-definition camera 6; the input end of the signal input module is connected with the sample collector; the controller 9 is provided with a human-computer interaction module, the input end of the human-computer interaction module is connected with the output end of the controller and used for displaying images acquired by the first high-definition camera 2 and the second high-definition camera 6 and displaying signal information of the sample collector, the image preprocessing module and the data processing module, and an operator can effectively observe the image information condition in the equipment when performing solid-liquid separation operation.
The output end of the controller is connected with the monitor through the monitoring module, the monitor is respectively assembled on the material outlets of the cyclone 1 and the centrifugal separator 5, and preprocessed material image information in the material cyclone to be separated and preprocessed material image information at the outlet of the centrifugal separator can be input into the data processing module for effective monitoring.
The output end of the cyclone 1 is provided with an automatic material valve 3 with an unqualified cyclone outlet and an automatic material valve 4 with a qualified cyclone outlet; the automatic valve 3 for the material with unqualified cyclone outlet is connected with evaporative crystallization process equipment 10; and the qualified material automatic valve 4 at the outlet of the cyclone is connected with the input end of a centrifugal separator 5.
The output end of the centrifugal separator is provided with an automatic valve 7 for qualified materials at the outlet of the centrifugal separator and an automatic valve 8 for unqualified materials at the outlet of the centrifugal separator; the automatic valve 8 for unqualified materials at the outlet of the centrifugal separator is connected with evaporative crystallization process equipment 10; and the automatic valve 7 for qualified materials at the outlet of the centrifugal separator is connected with a drying device 11.
Referring to fig. 2, an HMM model based evaporative crystallization solid-liquid separation method, based on the above mentioned evaporative crystallization solid-liquid separation system, includes the following steps,
step 1, collecting a training sample set by using a sample collector, wherein the training sample set acquires image information of a material to be separated in a cyclone 1;
step 2, carrying out image preprocessing on the acquired image information through an image preprocessing module;
step 3, screening the materials in the cyclone 1 through a data processing module after image preprocessing, feeding the normal solid-liquid separable materials into a centrifugal separator 5 to execute step 4, and feeding the solid-liquid inseparable materials back to the evaporative crystallization process equipment 10 to execute step 1 again;
step 4, acquiring image information of an outlet of the centrifugal separator 5, and performing image preprocessing on the acquired image information through an image preprocessing module;
step 5, screening the materials in the centrifugal separator 5 through a data processing module after image preprocessing, enabling the normal solid-liquid separable materials to enter a drying device 11, and finishing the separation work; returning the solid-liquid inseparable material to the evaporative crystallization process equipment 10 to continue evaporation and concentration, and then re-executing the step 1.
Image preprocessing employs noise reduction preprocessing on image information, wherein the image preprocessing techniques include gaussian filtering, mean filtering, median filtering, minimum mean square error filtering and Gabor filtering.
The data processing module adopts HMM model technology to sieve the materials in the cyclone 1 and at the outlet of the centrifugal separator 5 respectively, and the specific method is as follows:
constructing a first HMM model and a second HMM model;
inputting the image information of the material in the cyclone 1 after the pretreatment of the material to be separated into a first HMM model to output a probability value P (O-lambda) 1 (ii) a Probability value P (O-lambda) 1 Respectively corresponding to normal solid-liquid separable state S 11 And a solid-liquid inseparable state S 12 ；
When the material image information in the cyclone 1 is output with a first HMM model, a probability value P (O-lambda) 1 In a normal solid-liquid separable state S 11 In the range of (2), the material enters a subsequent centrifugal separator 5; when the material image information in the cyclone 1 is output with a first HMM model, a probability value P (O-lambda) 1 In a solid-liquid stateSeparation state S 12 Returning the material to the evaporative crystallization process equipment 10;
inputting the image information of the centrifuge outlet material after the pretreatment of the material to be separated into a second HMM model to output a probability value P (O-lambda) 2 (ii) a Probability value P (O-lambda) 2 Respectively corresponding to normal solid-liquid separable state S 21 And a solid-liquid inseparable state S 22 ；
When the image information of the material at the outlet of the centrifugal separator is output by the second HMM model, the probability value P (O-lambda) is output 2 In a normal solid-liquid separable state S 21 In the range of (1), the material enters drying equipment; when the image information of the material at the outlet of the centrifugal separator is output with the second HMM model, the probability value P (O-lambda) is output 2 In a solid-liquid inseparable state S 22 And the material returns to the evaporative crystallization process equipment 10.
The first HMM model and the second HMM model are both obtained by supervised learning by EM, wherein the first HMM model is obtained from the solid content of the material in the cyclone 1 and the K-S test in the same evaporative crystallization process equipment 10 in the form of probability density distribution in the EM algorithm. The second HMM model is obtained from the solid content of the material at the outlet of the centrifuge 5 and the K-S test of the same evaporative crystallization process equipment 10 using the probability density distribution form in the EM algorithm.
Wherein a normal solid-liquid separable state S 11 The solid content of the material is more than or equal to 30wt%; solid-liquid inseparable state S 12 ＜30wt％；
The normal solid-liquid separable state S 21 The solid content of the material is more than or equal to 90wt%; solid-liquid inseparable state S 12 ＜90wt％。
The probability density distribution form includes, but is not limited to, normal distribution, log-normal distribution, and generalized extreme distribution.
Examples
This example provides a solid-liquid separation method based on HMM model by evaporative crystallization, which comprises the following steps,
s101, obtaining image information of materials in a cyclone 1, wherein the image information of the materials in the cyclone 1 is obtained by shooting through a first high-definition camera 2; the image information comprises image information of solid-liquid separable materials and image information of solid-liquid inseparable materials;
s102: carrying out noise reduction on image information through a preprocessing technology to obtain an HMM training sequence O, wherein the image preprocessing technology comprises Gaussian filtering, mean filtering, median filtering, minimum mean square error filtering and Gabor filtering;
s103: training the preprocessed image information by adopting an EM algorithm to obtain a first HMM parameter combination lambda = (pi 1, A1, B1) with the maximum probability;
under the premise of training the sequence O and the model lambda, the probability xi (i, j) of transferring the material to be separated from the Si state to the Sj state at the moment t +1 is as follows:
ξ(i，j)＝P(q t ＝S i ,q t+1 ＝S j |O,λ)
wherein q is t Is the material at time t, q t+1 The material at the moment t + 1; s i And S j Is in the material state
According to a forward-backward algorithm xi t (i, j) may in turn be represented as:
Figure BDA0003083653500000101
wherein alpha is t (i) Indicating that the training time sequence is in state S at time t i The probability of the first half of (d); a is ij Indicating a hidden state S i Towards a hidden state S j The probability of a transition; b j (O t+1 ) Indicating that at the time t +1, the hidden state S is in the observation sequence O in the relation matrix B of the hidden state and the observation state j The corresponding observed probability value; beta is a beta t+1 (j) Indicates at time t +1 that it is in the hidden state S j The probability of the remaining portion of (a).
γ(i)＝P(q t ＝S i |0,λ)
Figure BDA0003083653500000102
Wherein γ (i) represents the time at tProbability of being in state i; p (O | λ) represents the probability of the observed sequence O observed in the HMM model; beta is a t (i) Indicates being in state S at time t i The probability of the remaining portion of (a).
Because the materials in the cyclone are continuous, the training sequence probability matrix is replaced by the training data probability density function.
Figure BDA0003083653500000103
Here, c jm Indicating that the mth Gaussian function is in an implicit state S j The mixing coefficient of (a); II indicates an implicit state S j With the mean vector mu jm Covariance matrix U jm The density of the mth gaussian function of (1). The parameters of the gaussian mixture density function can be calculated by the following three formulas:
Figure BDA0003083653500000111
Figure BDA0003083653500000112
Figure BDA0003083653500000113
wherein the content of the first and second substances,
Figure BDA0003083653500000114
and
Figure BDA0003083653500000115
HMM model parameters based on normal distribution;
wherein T represents the time series time number, gamma t (j, m) represents a hidden state probability density distribution function at time t, j and m being two parameters of gamma; o is t Representing the observation sequence at time t.
The two parameters μ and σ of the lognormal distribution are obtained by the above calculation formula:
Figure BDA0003083653500000116
Figure BDA0003083653500000117
the two formulas are HMM model parameters based on lognormal distribution; gamma ray t (i) Representing the probability of being in state i at time t; x is a radical of a fluorine atom t Representing the observed value at time t.
The three parameters δ, μ and ε of the generalized extremum distribution are obtained by the following equations:
Figure BDA0003083653500000121
Figure BDA0003083653500000122
Figure BDA0003083653500000123
the three equations above are based on HMM model parameters of generalized extremum distributions.
Wherein the content of the first and second substances,
Figure BDA0003083653500000124
Figure BDA0003083653500000125
the kth Gaussian component at time t is in state S j The probability of (d) can be expressed as:
Figure BDA0003083653500000126
Figure BDA0003083653500000127
wherein pi represents the hidden state S j With the mean vector mu jm Covariance matrix U jm The density of the mth gaussian function of (1).
In E step of the EM algorithm, γ t (i) And ξ (i, j) may be obtained by a forward-backward algorithm.
S104: based on the E-step calculation, HMM parameters can be obtained by M steps:
Figure BDA0003083653500000128
Figure BDA0003083653500000129
s105: iterations S103 and S104 are repeated until convergence.
S106: the image information in the cyclone of the solid-liquid material to be separated is brought into the first HMM model, and P (O-lambda) is calculated 1 And outputting the corresponding material state. If output P (O- λ) 1 The corresponding state is S 11 Then sending to a centrifugal separator, and otherwise, returning to the evaporative crystallization process.
S107: the calculation process of the second HMM model is the same as the calculation process of S101 to S106, and the description of the present invention is omitted.
According to the above technical solution, in the solid-liquid separation method for evaporative crystallization process based on HMM model provided by the present invention, the first HMM model outputs P (O- λ) 1 If output P (O- λ) 1 The corresponding state is S 11 When the separation is carried out, the materials to be separated can be continuously separated, otherwise, the materials cannot be separated and return to the front end. Second HMM model output P (O- λ) 2 If output P (O- λ) 2 Corresponding state is S 21 When the materials to be separated can be continuously separated, otherwise, the materials can not be separated and return to the front end. Also, HMM models are based on trainingSolid content information and K-S inspection results of the materials are modeled in a distribution form more fitting the reality, so that monitoring and control are more accurate and adaptive, and stable operation of the evaporative crystallization process is ensured.
The controller provided by the invention comprises an image acquisition module, a signal input module, an image preprocessing module, a data processing module and a monitoring module;
the image acquisition module and the signal input module are used for acquiring material image information of the material to be separated in the cyclone and material image information of an outlet of the centrifugal separator, preprocessing the material image information, and uploading the preprocessed material image information to the module for storage.
And the image preprocessing module and the data processing module are used for constructing the first HMM model and the second HMM model. Inputting the image information of the material in the cyclone after the pretreatment of the material to be separated into a first HMM model to output P (O-lambda) 1 . Inputting the image information of the centrifuge outlet material after the pretreatment of the material to be separated into a second HMM model to output P (O-lambda) 2 。
The monitoring module is used for inputting preprocessed material image information in the material cyclone to be separated into a first HMM model, and the first HMM model outputs P (O-lambda) 1 If P (O-lambda) of the material to be separated is present 1 Corresponding state S 11 And (4) feeding the material at the outlet of the cyclone into a centrifugal separator, and otherwise, returning the material at the outlet of the cyclone to the front-end evaporative crystallization process. Inputting the image information of the material pretreated at the outlet of the centrifugal separator into a second HMM model, wherein the second HMM model outputs P (O-lambda) 2 P (O-lambda) of the material to be separated 2 Corresponding state S 12 Then the mixture is sent to a drying device, and otherwise, the mixture returns to the evaporative crystallization process.
According to P (O-lambda) of the material to be separated 1 And P (O- λ) 2 And obtaining a control strategy.
In the first HMM model, if P (O- λ) 1 Corresponding to a normal solid-liquid separable state S 11 Opening an automatic valve 4 for qualified materials at the outlet of the cyclone, and closing an automatic valve 3 for unqualified materials at the outlet of the cyclone; p (O- λ) 1 Corresponding to the solid-liquid inseparable state S 12 Opening the outlet of the cycloneAnd (3) closing the automatic valve 4 for qualified materials at the outlet of the cyclone.
In the second HMM model, if P (O- λ) 2 Corresponding to a normal solid-liquid separable state S 21 Opening an automatic valve 7 for qualified materials at the outlet of the centrifugal separator, and closing an automatic valve 8 for unqualified materials at the outlet of the centrifugal separator; if P (O- λ) 2 Corresponding to the solid-liquid inseparable state S 22 And opening the automatic valve 8 for the unqualified materials at the outlet of the centrifugal separator and closing the automatic valve 7 for the qualified materials at the outlet of the centrifugal separator.
In a preferred embodiment, the HMM model is built by building an HMM module through a data processing module.
In a preferred embodiment, the parameter profile of the HMM model is determined from the solids content of the training material and the K-S test.

Claims (10)

1. An HMM model-based evaporative crystallization solid-liquid separation system is characterized by comprising a controller (9) and a solid-liquid separation unit;
the solid-liquid separation unit comprises a cyclone (1), a centrifugal separator (5), a material device to be separated (12), evaporative crystallization process equipment (10) and drying equipment (11); the output end of the material device (12) to be separated is connected with the input end of the swirler (1); the output end of the cyclone (1) is respectively connected with a centrifugal separator (5) and evaporative crystallization process equipment (10); the output end of the centrifugal separator (5) is respectively connected with evaporative crystallization process equipment (10) and drying equipment (11);
the input end of the controller (9) is respectively connected with the image acquisition module and the signal input module, and the output end of the controller (9) is respectively connected with the image preprocessing module and the data processing module; the input end of the image acquisition module is respectively connected with a first high-definition camera (2) and a second high-definition camera (6); the input end of the signal input module is connected with the sample collector; the data processing module adopts an HMM model technology to sieve the materials in the cyclone and at the outlet of the centrifugal separator respectively; the data processing module is used for establishing a first HMM model and a second HMM model;
the first HMM model is used for inputting image information of materials in the cyclone after pretreatment of the materials to be separated and outputting a probability value P (O-lambda) 1 (ii) a Probability value P (O-lambda) 1 Respectively corresponding to normal solid-liquid separable state S 11 And a solid-liquid inseparable state S 12 (ii) a Wherein, when the image information of the material in the cyclone is output with a first HMM model, a probability value P (O-lambda) 1 In a normal solid-liquid separable state S 11 In the range of (1), the material enters a subsequent centrifugal separator; when the image information of the material in the cyclone is output with a first HMM model, a probability value P (O-lambda) 1 In a solid-liquid inseparable state S 12 The material returns to the evaporation crystallization process equipment (10);
the second HMM model is used for inputting image information of the centrifuge outlet material after pretreatment of the material to be separated and outputting a probability value P (O-lambda) 2 (ii) a Probability value P (O-lambda) 2 Respectively corresponding to normal solid-liquid separable state S 21 And a solid-liquid inseparable state S 22 (ii) a Wherein, when the image information of the material at the outlet of the centrifugal separator outputs a probability value P (O-lambda) through the second HMM model 2 In a normal solid-liquid separable state S 21 The material enters a drying device within the range of (1); when the image information of the material at the outlet of the centrifugal separator is output with the second HMM model, the probability value P (O-lambda) is output 2 In a solid-liquid inseparable state S 22 Returning the material to the evaporation crystallization process equipment;
the first high-definition camera (2) is assembled on the cyclone (1), the second high-definition camera (6) is assembled at a material outlet of the centrifugal separator (5), and the sample collector is assembled on a material device (12) to be separated.
2. The HMM model-based evaporative crystallization solid-liquid separation system according to claim 1, wherein a human-computer interaction module is disposed on the controller (9), and an input end of the human-computer interaction module is connected to an output end of the controller, and is configured to display images acquired by the first high-definition camera (2) and the second high-definition camera (6) and display signal information of the sample collector, the image preprocessing module and the data processing module.
3. The HMM model based evaporative crystallization solid-liquid separation system according to claim 1, wherein the output of the controller (9) is connected to monitors via monitoring modules, the monitors being respectively mounted on the material outlet outputs of the cyclone (1) and the centrifuge (5).
4. The HMM model-based evaporative crystallization solid-liquid separation system according to claim 1, wherein the output end of the cyclone (1) is provided with an automatic cyclone outlet unqualified material valve (3) and an automatic cyclone outlet qualified material valve (4); the automatic valve (3) for the unqualified material at the outlet of the cyclone is connected with evaporation crystallization process equipment (10); and the qualified material automatic valve (4) at the outlet of the cyclone is connected with the input end of the centrifugal separator (5).
5. The HMM model-based evaporative crystallization solid-liquid separation system according to claim 1, wherein the output end of the centrifugal separator is provided with an automatic valve (7) for qualified materials at the outlet of the centrifugal separator and an automatic valve (8) for unqualified materials at the outlet of the centrifugal separator; an automatic valve (8) for the material with unqualified outlet of the centrifugal separator is connected with evaporative crystallization process equipment (10); the automatic valve (7) for qualified materials at the outlet of the centrifugal separator is connected with a drying device (11).
6. An HMM model-based evaporative crystallization solid-liquid separation method, characterized in that the evaporative crystallization solid-liquid separation system according to any one of claims 1 to 5 comprises the steps of,
step 1, collecting a training sample set by using a sample collector, wherein the training sample set acquires image information of a material to be separated in a cyclone (1);
step 2, carrying out image preprocessing on the acquired image information through an image preprocessing module;
step 3, screening the materials in the cyclone (1) through a data processing module after image preprocessing, enabling the materials with normal solid-liquid separation to enter a centrifugal separator (5) to execute step 4, and returning the materials with solid-liquid separation to the evaporation crystallization process equipment (10) to execute step 1 again;
step 4, acquiring image information in the centrifugal separator (5), and carrying out image preprocessing on the acquired image information through an image preprocessing module;
5, screening the materials in the centrifugal separator (5) through a data processing module after image preprocessing, feeding the normal solid-liquid separable materials into a drying device (11), and finishing the separation work; returning the solid-liquid inseparable material to the evaporation crystallization process equipment (10) for continuous evaporation and concentration, and then re-executing the step 1.
7. The HMM model-based evaporative crystallization solid-liquid separation method according to claim 6, wherein the image preprocessing is performed by noise reduction preprocessing on the image information, and the image preprocessing techniques include Gaussian filtering, mean filtering, median filtering, minimum mean square error filtering and Gabor filtering.
8. The HMM model-based evaporative crystallization solid-liquid separation method according to claim 6, wherein the data processing module adopts HMM model technology to sieve the materials in the cyclone (1) and at the outlet of the centrifugal separator (5), and the method comprises the following steps:
constructing a first HMM model and a second HMM model;
inputting the image information of the material in the cyclone (1) after the pretreatment of the material to be separated into a first HMM model output probability value P (O-lambda) 1 (ii) a Probability value P (O-lambda) 1 Respectively corresponding to normal solid-liquid separable state S 11 And a solid-liquid inseparable state S 12 ；
When the material image information in the cyclone (1) is output with a first HMM model to obtain a probability value P (O-lambda) 1 In a normal solid-liquid separable state S 11 In the range of (5), the material enters a subsequent centrifugal separator (5); when the cyclone(1) Outputting probability value P (O-lambda) from the image information of the inner material through the first HMM model 1 In a solid-liquid inseparable state S 12 The material returns to the evaporation crystallization process equipment (10);
inputting the image information of the centrifuge outlet material after the pretreatment of the material to be separated into a second HMM model to output a probability value P (O-lambda) 2 (ii) a Probability value P (O-lambda) 2 Respectively corresponding to normal solid-liquid separable state S 21 And a solid-liquid inseparable state S 22 ；
When the image information of the material at the outlet of the centrifugal separator is output with the second HMM model, the probability value P (O-lambda) is output 2 In a normal solid-liquid separable state S 21 In the range of (1), the material enters drying equipment; when the image information of the material at the outlet of the centrifugal separator is output by the second HMM model, the probability value P (O-lambda) is output 2 In a solid-liquid inseparable state S 22 And the material returns to the evaporative crystallization process equipment (10).
9. The HMM model based evaporative crystallization solid-liquid separation method according to claim 8, wherein the first HMM model and the second HMM model are both obtained by supervised learning of EM, wherein the probability density distribution form of the first HMM model in EM algorithm is obtained by solid content of material in the cyclone (1) and K-S test in the same evaporative crystallization process equipment (10); the second HMM model is obtained by adopting the probability density distribution form in the EM algorithm according to the solid content of the material at the outlet of the centrifugal separator (5) of the same evaporative crystallization process equipment (10) and K-S test.
10. The HMM model-based evaporative crystallization solid-liquid separation method of claim 8, wherein the normal solid-liquid separable state S is 11 The solid content of the material is more than or equal to 30wt%; solid-liquid inseparable state S 12 ＜30wt%；
The normal solid-liquid separable state S 21 The solid content of the material is more than or equal to 90wt%; solid-liquid inseparable state S 22 ＜90wt%。
CN202110574119.XA 2021-05-25 2021-05-25 HMM model-based evaporative crystallization solid-liquid separation system and method Active CN113307436B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110574119.XA CN113307436B (en) 2021-05-25 2021-05-25 HMM model-based evaporative crystallization solid-liquid separation system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110574119.XA CN113307436B (en) 2021-05-25 2021-05-25 HMM model-based evaporative crystallization solid-liquid separation system and method

Publications (2)

Publication Number Publication Date
CN113307436A CN113307436A (en) 2021-08-27
CN113307436B true CN113307436B (en) 2022-12-13

Family

ID=77374716

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110574119.XA Active CN113307436B (en) 2021-05-25 2021-05-25 HMM model-based evaporative crystallization solid-liquid separation system and method

Country Status (1)

Country Link
CN (1) CN113307436B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023123365A1 (en) * 2021-12-31 2023-07-06 东江环保股份有限公司 System for determining end point of evaporation by means of image recognition during ammonia-nitrogen wastewater evaporative crystallization process

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106599902A (en) * 2016-11-04 2017-04-26 浙江大学 Sugar crystal classification and recognition and crystallization quality control method based on image
CN106800507A (en) * 2015-11-25 2017-06-06 衡阳屹顺化工有限公司 A kind of solid-liquid separating method produced during sodium formate
CN207899610U (en) * 2017-12-20 2018-09-25 上海亭博分离技术有限公司 A kind of intelligent disconnecting link for horizontal centrifuge solid outlet
CN209522605U (en) * 2018-12-06 2019-10-22 中煤西安设计工程有限责任公司 A kind of evaporation and crystallization system of the high organic matter sewage of high saliferous

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10766782B2 (en) * 2015-09-23 2020-09-08 Solvay Sa Production of crystalline sodium bicarbonate

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106800507A (en) * 2015-11-25 2017-06-06 衡阳屹顺化工有限公司 A kind of solid-liquid separating method produced during sodium formate
CN106599902A (en) * 2016-11-04 2017-04-26 浙江大学 Sugar crystal classification and recognition and crystallization quality control method based on image
CN207899610U (en) * 2017-12-20 2018-09-25 上海亭博分离技术有限公司 A kind of intelligent disconnecting link for horizontal centrifuge solid outlet
CN209522605U (en) * 2018-12-06 2019-10-22 中煤西安设计工程有限责任公司 A kind of evaporation and crystallization system of the high organic matter sewage of high saliferous

Also Published As

Publication number Publication date
CN113307436A (en) 2021-08-27

Similar Documents

Publication Publication Date Title
CN113307436B (en) HMM model-based evaporative crystallization solid-liquid separation system and method
Zhu et al. Probability density function of bubble size based reagent dosage predictive control for copper roughing flotation
CN109492265B (en) Wastewater effluent index prediction method based on dynamic nonlinear PLS soft measurement method
EP2850522B1 (en) Anomalous system state identification
CN109284705B (en) Anomaly detection method for traffic monitoring video
CN108594790B (en) Fault detection and separation method based on structured sparse principal component analysis
CN109675935A (en) A kind of IPCA operation of rolling on-line fault diagnosis method becoming control limit
CN115937251A (en) Multi-target tracking method for shrimps
CN112685910A (en) Complex equipment power pack fault prediction method based on hybrid prediction model
CN114819054A (en) Power electronic system state monitoring method based on physical information neural network
Nieto et al. A new predictive model for the outlet turbidity in micro-irrigation sand filters fed with effluents using Gaussian process regression
CN115546899A (en) Examination room abnormal behavior analysis method, system and terminal based on deep learning
Shardt et al. Automated system identification in mineral processing industries: A case study using the zinc flotation cell
CN1462884A (en) Method of recognizing image of lung cancer cells with high accuracy and low rate of false negative
Fei et al. Online process monitoring for complex systems with dynamic weighted principal component analysis
Mahajan et al. Prediction of Fault in Gas Chromatograph using Convolutional Neural Network
CN113307435B (en) Dynamic PCA (principal component analysis) -based evaporative crystallization solid-liquid separation system and method
CN110175602B (en) Feature selection method based on feature reconstruction and heterogeneous data combined modeling
CN114596625A (en) Fish school track monitoring system based on edge calculation
Rienmüller et al. Hybrid estimation through synergic mode-set focusing
Sugawara et al. Calving prediction from video: Exploiting behavioural information relevant to calving signs in japanese black beef cows
Afzal et al. Forecasting in industrial process control: A hidden Markov model approach
CN110189360A (en) A kind of recognition and tracking method of pair of specific objective
CN115342817B (en) Method, system, equipment and medium for monitoring track tracking state of unmanned tracked vehicle
CN114528760B (en) Aero-engine gas circuit health state prediction method considering monitoring error

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant