CN112906143B - Fully-mechanized coal mining face mine pressure prediction model establishment method considering data distribution domain adaptation - Google Patents
Fully-mechanized coal mining face mine pressure prediction model establishment method considering data distribution domain adaptation Download PDFInfo
- Publication number
- CN112906143B CN112906143B CN202010944130.6A CN202010944130A CN112906143B CN 112906143 B CN112906143 B CN 112906143B CN 202010944130 A CN202010944130 A CN 202010944130A CN 112906143 B CN112906143 B CN 112906143B
- Authority
- CN
- China
- Prior art keywords
- hydraulic support
- data
- source
- domain
- target
- 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
Links
- 238000009826 distribution Methods 0.000 title claims abstract description 83
- 238000005065 mining Methods 0.000 title claims abstract description 79
- 238000000034 method Methods 0.000 title claims abstract description 61
- 239000003245 coal Substances 0.000 title claims abstract description 32
- 230000006978 adaptation Effects 0.000 title claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 36
- 230000003044 adaptive effect Effects 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 45
- 238000012549 training Methods 0.000 claims description 23
- 238000012360 testing method Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 9
- 238000010801 machine learning Methods 0.000 claims description 8
- 239000004576 sand Substances 0.000 claims description 7
- 230000008859 change Effects 0.000 description 4
- 238000005520 cutting process Methods 0.000 description 4
- 239000011435 rock Substances 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005056 compaction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Geometry (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention discloses a method for establishing a fully-mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation, relates to the technical field of mine pressure prediction of coal mine fully-mechanized coal mining faces, and can improve the accuracy of the established mine pressure prediction model. The method comprises the following steps: collecting working resistance time sequence data of a hydraulic support of a fully mechanized mining face, and preprocessing the collected data; determining input and output indexes of a mine pressure prediction model of the fully mechanized mining face, and determining an input and output data set of the mine pressure prediction model according to the input and output indexes; determining a source hydraulic support and a target hydraulic support in the fully mechanized mining face; processing source domain and target domain data of a source hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing a source hydraulic support mine pressure prediction model; and processing the source domain and target domain data of the target hydraulic support mine pressure prediction model, and establishing the target hydraulic support mine pressure prediction model. The method is suitable for the mine pressure prediction modeling of the fully mechanized coal mining face.
Description
Technical Field
The invention relates to the technical field of mine pressure prediction of a coal mine fully-mechanized coal mining face, in particular to a method for establishing a fully-mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation.
Background
The change rule of the working resistance of the hydraulic support reflects the breaking characteristics of an overlying rock layer to a certain extent, the mine pressure of the fully mechanized mining face is sensed and predicted based on the monitoring data of the working resistance of the hydraulic support, the method is an effective means for realizing periodic advance early warning and advance response of the working face, and the method plays an important role in dynamically improving the adaptability of the support and optimizing the control quality of surrounding rocks.
The mining environment change of the underground fully mechanized mining face is complex, the mining dynamic stress is coupled with the compaction and bearing of the goaf, the mechanism model of the mining system is complex, and accurate modeling is difficult, so that the mining pressure time sequence prediction model of the fully mechanized mining face hydraulic support is established by adopting a data-driven modeling method to attract wide attention, and relevant students can carry out research on the mining pressure prediction by adopting methods such as a neural network and an expert system. However, the fully mechanized mining face is always a 'dynamic' process in the mining process, mining induced stress damages the original rock stress balance, the surrounding rock stress state of the coal wall is continuously redistributed, the mining working conditions are complex and changeable due to multi-factor coupling, and the phenomenon that the working resistance time sequence data of the same hydraulic support is inconsistent before and after the data distribution structure time sequence along with the advancing of the working face is directly shown, the whole working face is up to hundreds of hydraulic supports, the working resistance time sequence data distribution difference of the hydraulic supports at different positions is more obvious, and the accuracy of the built mine pressure prediction model cannot meet the industrial production requirements.
Disclosure of Invention
In view of this, the embodiment of the present invention provides a method for establishing a mine pressure prediction model of a fully mechanized mining face, which considers data distribution domain adaptation, and can improve the accuracy of the established mine pressure prediction model.
The embodiment of the invention provides a method for establishing a fully mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation, which comprises the following steps: collecting working resistance time sequence data of a hydraulic support of a fully mechanized mining face, and preprocessing the collected data; determining input and output indexes of a mine pressure prediction model of the fully mechanized mining face, and determining an input and output data set of the mine pressure prediction model according to the input and output indexes; determining a source hydraulic support and a target hydraulic support in the fully mechanized mining face; processing source domain and target domain data of a source hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing a source hydraulic support mine pressure prediction model; and processing the source domain and target domain data of the target hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing the target hydraulic support mine pressure prediction model.
According to a specific implementation manner of the embodiment of the invention, the determining of the source hydraulic support and the target hydraulic support in the fully mechanized mining face comprises the following steps: dividing the whole working surface into n different areas; in different areas, the hydraulic support at the most middle position of each area is determined to be a source hydraulic support, and the hydraulic supports at other positions in each area are determined to be target hydraulic supports.
According to a specific implementation manner of the embodiment of the invention, after the source hydraulic support and the target hydraulic support in the fully mechanized mining face are determined, the method further comprises the following steps: determining source domain data and target domain data of a source hydraulic support, and determining source domain data and target domain data of a target hydraulic support; the source domain data of the source hydraulic support is training set data of a mine pressure model to be established by the source hydraulic support, and the target domain data of the source hydraulic support is test set data of the mine pressure model to be established by the source hydraulic support; the source domain data of the target hydraulic support is training set data of a mine pressure model to be established by the source hydraulic support, and the target domain data of the target hydraulic support is test set data of the mine pressure model to be established by the target hydraulic support; the method for establishing the mine pressure prediction model of the source hydraulic support by utilizing the data distribution domain adaptive algorithm to process the source domain and target domain data of the mine pressure prediction model of the source hydraulic support comprises the following steps: calculating a projection matrix by using a data distribution domain adaptive algorithm according to input data in source domain data and target domain data of the source hydraulic support; projecting the input data in the source domain data and the target domain data of the source hydraulic support to a unified public space according to the projection matrix to obtain the source domain input data of the source hydraulic support after projection and the target domain input data of the source hydraulic support after projection; and training the mine pressure prediction model of the hydraulic support by using a machine learning algorithm by using the source domain input data of the source hydraulic support after projection and the source domain output data of the source hydraulic support before projection to obtain the mine pressure prediction model of the source hydraulic support after training.
According to a specific implementation manner of the embodiment of the invention, the method for establishing the fully mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation further comprises the following steps: and predicting the future mine pressure value of the source hydraulic support by using the trained mine pressure prediction model of the hydraulic support and the input data in the projected target domain data of the source hydraulic support.
According to a specific implementation manner of the embodiment of the invention, the source domain and the target domain data of the target hydraulic support mine pressure prediction model are processed by using a data distribution domain adaptive algorithm, and the target hydraulic support mine pressure prediction model is established, wherein the method comprises the following steps: calculating a projection matrix by using input data in source domain data of the source hydraulic support and input data in target domain data of the target hydraulic support; according to the projection matrix, projecting input data in source domain data of the source hydraulic support and input data in a target domain of the target hydraulic support to a unified public space to obtain source domain input data of the source hydraulic support after projection and target domain input data of the target hydraulic support after projection; and training a target hydraulic support mine pressure prediction model by using a machine learning algorithm by using the source domain input data of the source hydraulic support after projection and the output data in the source domain data of the source hydraulic support before projection to obtain the trained target hydraulic support mine pressure prediction model.
According to a specific implementation manner of the embodiment of the invention, the method for establishing the fully mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation further comprises the following steps: and predicting the future mine pressure value of the target hydraulic support by using the trained mine pressure prediction model of the target hydraulic support and input data in the target domain data of the projected target hydraulic support.
According to a specific implementation manner of the embodiment of the invention, the calculation of the projection matrix by using the input data in the working resistance source domain data and the target domain data of the hydraulic support comprises the following steps: according to the working resistance input data of the hydraulic support in the source domain data and the target domain data, calculating a projection matrix by adopting a manifold regular domain adaptive algorithm, and performing data distribution consistency processing on the source domain data and the target domain data of the mine pressure time sequence data, wherein the data distribution consistency processing is specifically shown as the following formula:
(λ1XsLs(Xs)T+λ2XMcXT)-1(Xt(Xt)T)P=τP;
wherein, X issAnd XtInput data in the source domain data and input data in the target domain data, respectively, and X ═ Xs Xt]Said λ1,λ2The weighted value is P, the P is the projection matrix to be solved, and tau is taken as a characteristic value;
said LsAnd McThe matrix is constructed as follows:
the matrix McHas a dimension of (n)s+nt)×(ns+nt);
Wherein ε is the local neighbor radius, t is the sample population variance, and the matrix L issHas a dimension of ns×ns。
The invention discloses a method for establishing a fully-mechanized mining face mine pressure prediction model considering data distribution domain adaptation, which comprises the steps of acquiring working resistance time sequence data of a hydraulic support of a fully-mechanized mining face, preprocessing the acquired data, determining input and output indexes of the fully-mechanized mining face mine pressure prediction model, and determining an input and output data set of the mine pressure prediction model according to the input and output indexes; determining a source hydraulic support and a target hydraulic support in the fully mechanized mining face; processing source domain and target domain data of a source hydraulic support mine pressure prediction model through a data distribution domain adaptive processing algorithm, and establishing a source hydraulic support mine pressure prediction model; the source domain and the target domain data of the target hydraulic support mine pressure prediction model are processed, the target hydraulic support mine pressure prediction model is established, the distribution consistency of the hydraulic support working resistance time sequence data in each region can be greatly kept, the influence of the time sequence distribution difference of the multi-working-condition data on the model precision is reduced, and therefore the precision of the established mine pressure prediction model can be improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for establishing a fully mechanized coal mining face mine pressure prediction model in consideration of data distribution domain adaptation according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a method for establishing a fully mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation according to another embodiment of the present invention;
fig. 3 is a schematic flow chart of a method for establishing a fully mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation according to another embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention mainly aims to provide a method for establishing a fully-mechanized mining working face mine pressure prediction model considering data distribution domain adaptation, aiming at the problem of model misalignment caused by inconsistent distribution of time sequence data in the current working face mine pressure prediction based on hydraulic support working resistance data. The method comprises the steps that on the basis of observed hydraulic support working resistance time sequence data, divided hydraulic support source domain data sets and divided hydraulic support target domain data sets are subjected to data distribution consistency processing through a data distribution domain adaptive processing algorithm, specifically, data of the source domain data sets and the data of the target domain data sets are mapped to a unified public space through the data distribution domain adaptive processing algorithm, and the data distribution structures of the source domain data sets and the target domain data sets are consistent; and then, migrating a data distribution structure of a mine pressure prediction model built by the source hydraulic support, thereby completing a modeling task of the target hydraulic support, greatly keeping the data distribution consistency of the working resistance time sequence data of the hydraulic supports in the same region, and reducing the influence of the time sequence distribution difference of the multi-working-condition data on the model precision.
Fig. 1 is a schematic flow chart of a method for establishing a fully mechanized mining face mine pressure prediction model considering data distribution domain adaptation according to an embodiment of the present invention, as shown in fig. 1, the method of the embodiment may include the steps of:
s100, collecting working resistance time sequence data of the hydraulic support of the fully mechanized mining face, and preprocessing the collected data.
Preprocessing the acquired data includes rejecting outliers and invalid values.
S102, determining input and output indexes of the fully mechanized mining face mine pressure prediction model, and determining an input and output data set of the mine pressure prediction model according to the input and output indexes.
And determining input and output indexes of the fully mechanized mining face mine pressure prediction model, namely determining the working resistance values (namely the output indexes Y) of the following hydraulic supports according to the working resistance data (namely the input indexes X) of the hydraulic supports at the previous moments.
In one example, determining input and output indicators of the fully mechanized coal mining face mine pressure prediction model comprises: for a certain hydraulic support S in a working face, S pieces of working resistance data of the hydraulic support S in the k-2 cutting coal cutting process of a coal mining machine are selected as input of a model (namely, in the process that the coal mining machine cuts coal from the end of the working face to the tail of the machine for one cut, the number of the working resistance data monitored by the hydraulic support S is S), and l (l is more than or equal to 1) pieces of working resistance data of the hydraulic support S in the k cutting coal cutting process of the coal mining machine are output of the model, namely, a prediction model with S-dimensional input and l-dimensional output is determined.
And S104, determining a source hydraulic support and a target hydraulic support in the fully mechanized mining face.
In one example, the hydraulic support at the most middle position in the fully mechanized mining face can be a source hydraulic support, and the hydraulic supports at other positions can be target hydraulic supports.
In another example, for a plurality of hydraulic support mine pressure prediction models for the entire fully mechanized mining face, the determining a source hydraulic support and a target hydraulic support in the fully mechanized mining face (step S104) may include: dividing the whole working surface into n different areas; in different areas, the hydraulic support at the most middle position of each area is determined to be a source hydraulic support, and the hydraulic supports at other positions in each area are determined to be target hydraulic supports.
Preferably, the whole working surface can be divided into n areas such as a head area, a middle upper part, a middle lower part and a tail, the number of the hydraulic supports in the head area and the tail area is 11-21 (odd number), and the number of the hydraulic supports in other areas is required to be 30-40.
And S106, processing source domain and target domain data of the source hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing the source hydraulic support mine pressure prediction model.
And S108, processing source domain and target domain data of the target hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing the target hydraulic support mine pressure prediction model.
The invention discloses a method for establishing a fully-mechanized mining face mine pressure prediction model considering data distribution domain adaptation, which comprises the steps of acquiring working resistance time sequence data of a hydraulic support of a fully-mechanized mining face, preprocessing the acquired data, determining input and output indexes of the fully-mechanized mining face mine pressure prediction model, and determining an input and output data set of the mine pressure prediction model according to the input and output indexes; determining a source hydraulic support and a target hydraulic support in the fully mechanized mining face; the method comprises the steps of processing source domain and target domain data of a source hydraulic support mine pressure prediction model through a data distribution domain adaptive processing algorithm, establishing the source hydraulic support mine pressure prediction model, processing source domain and target domain data of a target hydraulic support mine pressure prediction model, establishing the target hydraulic support mine pressure prediction model, greatly keeping the distribution consistency of hydraulic support working resistance time sequence data, reducing the influence of time sequence distribution difference of multi-working condition data on model precision, and improving the precision of the established mine pressure prediction model.
Referring to fig. 2, in one embodiment, after determining a source hydraulic support and a target hydraulic support in a fully mechanized coal face (step S104), the method further includes the steps of:
s105, determining source domain data and target domain data of the source hydraulic support, and determining source domain data and target domain data of the target hydraulic support.
The source domain data of the source hydraulic support is training set data of a mine pressure model to be established by the source hydraulic support, and the target domain data of the source hydraulic support is test set data of the mine pressure model to be established by the source hydraulic support.
The source domain data of the target hydraulic support is training set data of a mine pressure model to be established by the source hydraulic support, and the target domain data of the target hydraulic support is test set data of the mine pressure model to be established by the target hydraulic support.
Wherein the source domain data is recorded as DsdAnd target Domain Source Domain data as Dtd. The source domain data and the target domain source domain data respectively comprise multiple sets of input and output data, such as source domain data Dsd=[Xsd,Ysd]Target domain data Dtd=[Xtd,Ytd]。
Establishing a mine pressure prediction model aiming at a source hydraulic support in a certain area, wherein the source area data are training set data of the source hydraulic support to-be-established mine pressure model and are expressed asThe target domain data is test set data of a mine pressure model to be built of the source hydraulic support and is expressed asFor the target within a certain areaThe method comprises the steps that a mine pressure prediction model is built on a hydraulic support, and source domain data are training set data of the mine pressure model to be built on the hydraulic support and are expressed asThe target domain data is test set data of a to-be-built ore pressure model of the target hydraulic support and is expressed as
The method for establishing the mine pressure prediction model of the source hydraulic support by utilizing the data distribution domain adaptive algorithm includes the following steps of:
s1061, calculating a projection matrix by using input data in source domain data and target domain data of the source hydraulic support;
s1062, projecting the input data in the source domain data and the target domain data of the source hydraulic support to a unified public space according to the projection matrix to obtain the source domain input data of the source hydraulic support after projection and the target domain input data of the source hydraulic support after projection;
and S1063, training the mine pressure prediction model of the hydraulic support by using a machine learning algorithm by using the source domain input data of the source hydraulic support after projection and the output data in the source domain data of the source hydraulic support before projection to obtain the mine pressure prediction model of the source hydraulic support after training.
Thereafter, future mine pressure values of the source hydraulic support can be predicted by using the trained hydraulic support mine pressure prediction model and input data in the target domain data of the projected source hydraulic support.
Under the condition that the whole working surface is divided into n different areas, in one example, the input and output data of the source hydraulic support mine pressure prediction model are processed by using a data distribution domain adaptive algorithm, and the source hydraulic support mine pressure prediction model of the nth area is establishedThe method can comprise the following steps:
s1061a input data in source domain data and target domain data using source hydraulic supportAndcalculating a projection matrix P1j(j=1,2,...,n);
S1062a, according to the projection matrix P1jInput data in the source domain data and the target domain dataAndprojection into a unified common space, i.e. projected source domain input data asThe input data of the target domain after projection is
S1063a, inputting data by using the projected source domainAnd source domain data output data before said projectionTraining the source hydraulic support mine pressure prediction model by using a machine learning algorithm to obtain the trained source hydraulic support mine pressure prediction model
Thereafter, the model can be utilizedAnd the projected target field input dataAnd predicting the future mine pressure value of the source hydraulic support.
The method for establishing the mine pressure prediction model of the target hydraulic support by utilizing the data distribution domain adaptive algorithm comprises the following steps of processing source domain data and target domain data of the mine pressure prediction model of the target hydraulic support (step S108), wherein the method comprises the following steps:
s1081, calculating a projection matrix by using input data in source domain data of the source hydraulic support and input data in target domain data of the target hydraulic support;
s1082, projecting input data in source domain data of the source hydraulic support and input data of target domain data of the target hydraulic support to a unified public space according to the projection matrix to obtain source domain input data of the source hydraulic support after projection and target domain input data of the target hydraulic support after projection;
s1083, training a target hydraulic support mine pressure prediction model by using a machine learning algorithm by using the source domain input data of the source hydraulic support after projection and the output data in the source domain data of the source hydraulic support before projection to obtain the trained target hydraulic support mine pressure prediction model.
And then, predicting the future mine pressure value of the target hydraulic support by using the trained mine pressure prediction model of the target hydraulic support and the input data in the target domain data of the projected target hydraulic support.
Under the condition that the whole working face is divided into n different areas, in one example, input and output data of a source hydraulic support mine pressure prediction model and a target hydraulic support mine pressure prediction model are processed by using a data distribution domain adaptive algorithm, and a target hydraulic support mine pressure prediction model of the nth area is established(i is the number of the target hydraulic supports in a certain area), the method can comprise the following steps:
s1081a, selecting a suitable data distribution domain adaptive processing algorithm, and utilizing input data of the source domain of the source hydraulic support and the target domain of the target hydraulic supportAndcalculating a projection matrix P2i(i=1,2,..,k);
S1082a, according to the projection matrix P2iInputting data of the source domain of the source hydraulic support and the target domain of the target hydraulic supportAndprojection into a unified common space, i.e. projected source domain input data asThe input data of the target domain after projection is
S1083a, inputting data by using the projected source domainAnd output data of the source domain dataTraining the target hydraulic support mine pressure prediction model by using a machine learning algorithm to obtain the trained target hydraulic support mine pressure prediction model
Thereafter, the model can be utilizedAnd input data of the target domain of the target hydraulic supportAnd predicting the future mine pressure value of the target hydraulic support.
And repeating the step S106 to the step S108, and establishing a mine pressure prediction model of each hydraulic support in other areas of the working face.
In step S106 or step S108, calculating a projection matrix using the input data in the hydraulic support working resistance source domain data and the target domain data may include: according to the working resistance input data of the hydraulic support in the source domain data and the target domain data, calculating a projection matrix by adopting a manifold regular domain adaptive algorithm, and performing data distribution consistency processing on the source domain data and the target domain data of the mine pressure time sequence data, wherein the data distribution consistency processing is specifically shown as the following formula:
(λ1XsLs(Xs)T+λ2XMcXT)-1(Xt(Xt)T)P=τP;
wherein, X issAnd XtInput data in the source domain data and input data in the target domain data, respectively, and X ═ Xs Xt]Said λ1,λ2The weighted value is P, the P is the projection matrix to be solved, and tau is taken as a characteristic value;
said LsAnd McThe matrix is constructed as follows:
the matrix McHas a dimension of (n)s+nt)×(ns+nt);
Wherein ε is the local neighbor radius, t is the sample population variance, and the matrix L issHas a dimension of ns×ns。
The invention relates to a method for establishing a mine pressure prediction model of a fully mechanized mining face in consideration of data distribution domain adaptation.A divided source domain data set and a divided target domain data set of a hydraulic support are subjected to data distribution structure uniformization processing through a data distribution domain adaptation processing algorithm based on observed working resistance time sequence data of the hydraulic support, and specifically, the source domain data and the target domain data are uniformly mapped to a common space through the domain adaptation processing algorithm, so that the data distribution structures of the source domain data set and the target domain data set are consistent; and then, migrating a data distribution structure of a mine pressure prediction model built by the source hydraulic support, thereby completing a modeling task of the target hydraulic support, greatly keeping the distribution consistency of the working resistance time sequence data of the hydraulic support, and reducing the influence of time sequence distribution difference of multi-working-condition data on the precision of the model.
The method for establishing the mine pressure prediction model of the fully mechanized mining face considering the data distribution domain adaptation according to the embodiment of the invention is described below with reference to a specific example.
Referring to fig. 3, taking a fully mechanized coal mining face with 174 hydraulic supports as an example, the method for establishing the mine pressure prediction model of the fully mechanized coal mining face considering data distribution domain adaptation of the invention comprises the following steps:
step S1: collecting working resistance time sequence data of N groups of fully mechanized coal mining face hydraulic supportsNormalizing the collected original data and eliminating zero value, and expressing the processed data as
Step S2: according to determined input/output indexDividing the data set; assuming that the current moment is t, the working resistance of the hydraulic support at the current moment isFor a mine pressure prediction model of a hydraulic support, the output isAnd input isTherefore, a 12-dimensional input and 1-dimensional output mine pressure prediction model can be established;
step S3: in this embodiment, because the selected working surface is too long, the whole working surface is divided into 6 regions including a head region (including 21 hydraulic supports), a tail region (including 21 hydraulic supports), 2 regions (including 33 hydraulic supports), 3 regions (including 33 hydraulic supports), 4 regions (including 33 hydraulic supports) and 5 regions (including 33 hydraulic supports); the source hydraulic supports of the 6 areas are No. 11, 38, 71, 104, 137 and 164 hydraulic supports respectively;
step S4: determining source domain data DsdAnd target domain data DtdThe source domain data is denoted as Dsd=[Xsd,Ysd]The target domain data is represented as Dtd=[Xtd,Ytd](ii) a Wherein, the source domain data of the source hydraulic support is the training set data of the ore pressure model to be established of the source hydraulic support and is expressed asData size is top 0.8N1The target domain data is its test set data, denoted asData size of last 0.2N1(ii) a And for the whole working face, the source domain data is the training set data of the mine pressure model to be established of the source hydraulic support and is expressed asThe data volume is 0.8N before the source hydraulic support data1And the target domain data is test set data of a mine pressure model to be established of the target hydraulic support and is expressed asThe data volume is 0.2N after the target hydraulic support data1;
Step S501: adopting a data distribution domain adaptive processing algorithm, and utilizing input data of a source domain and a target domain of a source hydraulic supportAndcalculating a projection matrix P11;
Step S502: according to projection matrix P11Inputting the source domain and the target domain of the source hydraulic supportAndprojection into a unified common space, i.e. projected source domain input data asThe input data of the target domain after projection is
Step S503: inputting data by using projected source domainAnd source domain output data before projectionTraining the source hydraulic support mine pressure prediction model, preferably, training the hydraulic support mine pressure prediction model by using a least square algorithm to obtain the trained source hydraulic support mine pressure prediction model
Step S504: according to the obtained modelAnd projected target field input dataPredicting a future mine pressure value of the source hydraulic support;
step S6: establishing a mine pressure prediction model of the 1 st target hydraulic support on the left side of the source hydraulic support of the end head region
Step S601: using a data distribution domain adaptive processing algorithm, using input data of a source domain of a source hydraulic support and a target domain of a target hydraulic supportAndcalculating a projection matrix P21;
Step S602: according to projection matrix P21Inputting data of the source region of the source hydraulic mountAnd input data of a target field of a target hydraulic mountProjection into a unified common space, i.e. projected source domain input data asThe input data of the target domain after projection is
Step S603: inputting data by using projected source domainAnd source domain output data before projectionPreferably, the target hydraulic support mine pressure prediction model can be trained by using a least square algorithm to obtain the trained target hydraulic support mine pressure prediction model
Step S604: according to the obtained modelAnd input data of a target field of a target hydraulic mountPredicting a future mine pressure value of the target hydraulic support;
step S7: and repeating the steps from S5 to S6, and establishing a mine pressure prediction model of other target hydraulic supports in the head area and a mine pressure prediction model of hydraulic supports in other areas.
The invention relates to a mining pressure prediction model building method of a fully mechanized mining face considering data distribution domain adaptation, which considers the inconsistency of actual production multi-working condition data distribution and adopts a data distribution domain adaptation processing algorithm to perform data distribution processing on input and output data of a model, preferably adopts a manifold regular domain adaptation algorithm to calculate a projection matrix and perform data distribution consistency processing on a source domain and a target domain of mining pressure time sequence data, and is shown as the following formula:
(λ1XsLs(Xs)T+λ2XMcXT)-1(Xt(Xt)T)P=τP;
in the above formula, P is the projection matrix. Mixing XsIs replaced byMixing XtIs replaced by Considering P as a feature vector and τ as a feature value, preferably λ1=λ2The projection matrix P is solved using eigenvalue decomposition algorithm, 0.5. L issAnd McThe matrix is constructed as follows:
the matrix McDimension of (A) is N1×N1。
In the above formula, preferably, epsilon is 10. The matrix LsHas a dimension of 0.8N1×0.8N1。
The invention relates to a method for establishing a mine pressure prediction model of a fully mechanized mining face in consideration of data distribution domain adaptation.A divided source domain data set and a divided target domain data set of a hydraulic support are subjected to data distribution structure uniformization processing through a data distribution domain adaptation processing algorithm based on observed working resistance time sequence data of the hydraulic support, and specifically, the source domain data and the target domain data are mapped to a uniform public space through the domain adaptation processing algorithm, so that the data distribution structures of the source domain data set and the target domain data set are consistent; and then, migrating a data distribution structure of a mine pressure prediction model built by the source hydraulic support, thereby completing a modeling task of the target hydraulic support, greatly keeping the distribution consistency of the working resistance time sequence data of the hydraulic support, and reducing the influence of time sequence distribution difference of multi-working-condition data on the precision of the model.
According to the method for establishing the mine pressure prediction model of the fully mechanized coal mining face, which is adaptive to the data distribution domain, the problem of inaccurate prediction model caused by influence on the working resistance data distribution structure of the hydraulic support due to frequent change of working conditions in the supporting process of the hydraulic support of the fully mechanized coal mining face can be effectively solved, the mine pressure advanced prediction of the fully mechanized coal mining face based on the working resistance data of the hydraulic support is realized, and a basis is provided for subsequent analysis of the mine pressure display rule of the fully mechanized coal mining face, advanced adaptation to the environment change of a stope and guidance of normal mining of the working face.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (4)
1. A method for establishing a fully mechanized coal mining face mine pressure prediction model considering data distribution domain adaptation is characterized by comprising the following steps:
collecting working resistance time sequence data of a hydraulic support of a fully mechanized mining face, and preprocessing the collected data;
determining input and output indexes of a mine pressure prediction model of the fully mechanized mining face, and determining an input and output data set of the mine pressure prediction model according to the input and output indexes;
determining a source hydraulic support and a target hydraulic support in the fully mechanized mining face;
processing source domain and target domain data of a source hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing a source hydraulic support mine pressure prediction model;
processing source domain and target domain data of a target hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing the target hydraulic support mine pressure prediction model;
the method comprises the following steps of processing source domain and target domain data of a target hydraulic support mine pressure prediction model by using a data distribution domain adaptive algorithm, and establishing the target hydraulic support mine pressure prediction model, wherein the method comprises the following steps: calculating a projection matrix by using input data in source domain data of the source hydraulic support and input data in target domain data of the target hydraulic support; according to the projection matrix, projecting input data in source domain data of the source hydraulic support and input data in a target domain of the target hydraulic support to a unified public space to obtain source domain input data of the source hydraulic support after projection and target domain input data of the target hydraulic support after projection; training a target hydraulic support mine pressure prediction model by using a machine learning algorithm by using the source domain input data of the source hydraulic support after projection and the output data in the source domain data of the source hydraulic support before projection to obtain the trained target hydraulic support mine pressure prediction model;
after determining the source hydraulic support and the target hydraulic support in the fully mechanized mining face, the method further comprises:
determining source domain data and target domain data of a source hydraulic support, and determining source domain data and target domain data of a target hydraulic support; the source domain data of the source hydraulic support is training set data of a mine pressure model to be established by the source hydraulic support, and the target domain data of the source hydraulic support is test set data of the mine pressure model to be established by the source hydraulic support; the source domain data of the target hydraulic support is training set data of a mine pressure model to be established by the source hydraulic support, and the target domain data of the target hydraulic support is test set data of the mine pressure model to be established by the target hydraulic support;
the method for establishing the mine pressure prediction model of the source hydraulic support by utilizing the data distribution domain adaptive algorithm to process the source domain and target domain data of the mine pressure prediction model of the source hydraulic support comprises the following steps: calculating a projection matrix by using input data in source domain data and target domain data of the source hydraulic support; projecting the input data in the source domain data and the target domain data of the source hydraulic support to a unified public space according to the projection matrix to obtain the source domain input data of the source hydraulic support after projection and the target domain input data of the source hydraulic support after projection; training the mine pressure prediction model of the hydraulic support by using a machine learning algorithm by using the source domain input data of the source hydraulic support after projection and the source domain output data of the source hydraulic support before projection to obtain the mine pressure prediction model of the source hydraulic support after training;
the method comprises the following steps of calculating a projection matrix by utilizing input data in source domain data and target domain data of the working resistance of the hydraulic support, wherein the projection matrix comprises the following steps:
according to input data in source domain data and target domain data of the working resistance of the hydraulic support, calculating a projection matrix by adopting a manifold regular domain adaptive algorithm, and performing source domain and target domain data distribution consistency processing on mine pressure time series data, wherein the data distribution consistency processing is specifically shown as the following formula:
(λ1XsLs(Xs)T+λ2XMcXT)-1(Xt(Xt)T)P=τP;
wherein, X issAnd XtInput data in the source domain data and input data in the target domain data, respectively, and X ═ XsXt]Said λ1,λ2The weighted value is P, the P is the projection matrix to be solved, and tau is taken as a characteristic value;
said LsAnd McThe matrix is constructed as follows:
the matrix McHas a dimension of (n)s+nt)×(ns+nt);
Wherein ε is the local neighbor radius, t is the sample population variance, and the matrix L issHas a dimension of ns×ns。
2. The method for building the mine pressure prediction model of the fully mechanized mining face with consideration of data distribution domain adaptation according to claim 1, wherein the determining of a source hydraulic support and a target hydraulic support in the fully mechanized mining face comprises: dividing the whole fully mechanized mining face into n different areas; in different areas, the hydraulic support at the most middle position of each area is determined to be a source hydraulic support, and the hydraulic supports at other positions in each area are determined to be target hydraulic supports.
3. The method for building the mine pressure prediction model of the fully mechanized mining face with consideration of data distribution domain adaptation according to claim 1, further comprising: and predicting the future mine pressure value of the source hydraulic support by using the trained mine pressure prediction model of the source hydraulic support and the input data in the target domain data of the projected source hydraulic support.
4. The method for building the mine pressure prediction model of the fully mechanized mining face with consideration of data distribution domain adaptation according to claim 1, further comprising: and predicting the future mine pressure value of the target hydraulic support by using the trained mine pressure prediction model of the target hydraulic support and input data in the target domain data of the projected target hydraulic support.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010944130.6A CN112906143B (en) | 2020-09-09 | 2020-09-09 | Fully-mechanized coal mining face mine pressure prediction model establishment method considering data distribution domain adaptation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010944130.6A CN112906143B (en) | 2020-09-09 | 2020-09-09 | Fully-mechanized coal mining face mine pressure prediction model establishment method considering data distribution domain adaptation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112906143A CN112906143A (en) | 2021-06-04 |
CN112906143B true CN112906143B (en) | 2022-03-25 |
Family
ID=76110911
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010944130.6A Active CN112906143B (en) | 2020-09-09 | 2020-09-09 | Fully-mechanized coal mining face mine pressure prediction model establishment method considering data distribution domain adaptation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112906143B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114792177B (en) * | 2022-06-24 | 2022-09-23 | 中煤科工开采研究院有限公司 | Multi-hydraulic support load prediction method and device for fully mechanized mining face and electronic equipment |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180024968A1 (en) * | 2016-07-22 | 2018-01-25 | Xerox Corporation | System and method for domain adaptation using marginalized stacked denoising autoencoders with domain prediction regularization |
CN109798138B (en) * | 2019-01-09 | 2020-05-22 | 天地科技股份有限公司 | Circulation end resistance prediction method of hydraulic support suitable for fully mechanized mining face |
CN109653779B (en) * | 2019-01-09 | 2020-05-22 | 天地科技股份有限公司 | Dynamic prediction method for hydraulic support load in fully mechanized mining face coal mining cycle |
CN109798149B (en) * | 2019-01-28 | 2020-06-09 | 辽宁工程技术大学 | Method for predicting inflow intensity of mining field of thick coal seam with hard roof in grading manner |
CN110929384A (en) * | 2019-10-29 | 2020-03-27 | 青岛本末岩控技术有限公司 | Mine pressure big data real-time analysis system and method based on fully mechanized coal mining face |
-
2020
- 2020-09-09 CN CN202010944130.6A patent/CN112906143B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN112906143A (en) | 2021-06-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Qi et al. | Prediction of open stope hangingwall stability using random forests | |
CN108487916B (en) | Tunnel stability protection prediction system and tunnel rock mass deformation prediction method | |
Shahin | Use of evolutionary computing for modelling some complex problems in geotechnical engineering | |
Kalantary et al. | An investigation on the Su–NSPT correlation using GMDH type neural networks and genetic algorithms | |
Gholami et al. | Improved RMR rock mass classification using artificial intelligence algorithms | |
Qi et al. | Back-analysis method for stope displacements using gradient-boosted regression tree and firefly algorithm | |
CN113297272B (en) | Bridge monitoring data association rule mining and health early warning method and system | |
CN103700255A (en) | Time and space related data mining-based traffic flow prediction method | |
Ozturk et al. | An artificial neural network model for virtual Superpave asphalt mixture design | |
Oral et al. | Predicting construction crew productivity by using Self Organizing Maps | |
Najafzadeh et al. | New expression-based models to estimate scour depth at clear water conditions in rectangular channels | |
Leng et al. | A hybrid data mining method for tunnel engineering based on real-time monitoring data from tunnel boring machines | |
CN106408120B (en) | Local area landslide prediction device and method | |
CN112906143B (en) | Fully-mechanized coal mining face mine pressure prediction model establishment method considering data distribution domain adaptation | |
CN111691679A (en) | Digital twinning-based intelligent tensioning method for prestressed steel structure | |
CN111828095B (en) | Dynamic prediction method for top plate incoming pressure | |
Pal | On application of machine learning method for history matching and forecasting of times series data from hydrocarbon recovery process using water flooding | |
Sabamehr et al. | System identification and model updating of highway bridges using ambient vibration tests | |
CN109063269B (en) | Hydraulic support group top plate supporting control method based on graph model and storage medium | |
CN113887637A (en) | Coal and gas outburst accident prediction and identification method based on deep cycle neural network | |
JP2010128779A (en) | Method for extracting multiple regression equation | |
Elkatatny et al. | An artificial intelligent approach to predict static Poisson's ratio | |
Li et al. | Stability evaluations of three-layered soil slopes based on extreme learning neural network | |
Cai et al. | Identification of soil strata based on general regression neural network model from CPTU data | |
CN117037432B (en) | Risk evaluation geological disaster early warning method based on multi-method cooperation |
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 |