Summary of the invention
Technical problem to be solved by the present invention lies in provide a kind of to be able to achieve complicated multi-aquifer water bursting source type
The complicated multi-aquifer water bursting in mine water source recognition methods quickly and effectively identified, that it is suitable for hydrogeologic conditions is more complex, fills
The more water bursting in mine water source identification in water water-bearing layer.
The present invention solves above-mentioned technical problem using following technical scheme: a kind of knowledge of complexity multi-aquifer water bursting in mine water source
Other method, method includes the following steps:
S1: according to the water chemistry data of known aqueous layer water sample, the water source database of mine is established;
S2: by data detection and exceptional value treated data as modeling sample;
S3: distinguishing indexes and its threshold value and having using the distinguishing indexes in the preliminary identification model of back substitution test and judge are determined
Effect property;
S4: according to effective distinguishing indexes and Fisher method of identification, " comprehensive-gradually method of identification " model is established;
S5: measuring water sample distinguishing indexes to be sentenced, and is successively determined by " comprehensive-gradually method of identification " model, identifies water source
Type.
As one of preferred embodiment of the invention, the water chemistry data of the water sample are each aquifer water of the affiliated coal mine of water sample
Various mass of ion concentration and each ion milliequivalent and ratio;Conventional ion can be Na++K+、 Ca2+、Mg2+、Cl-、
HCO3-、SO4 2-Deng;This water chemistry data builds library index as establish water source database.
As one of preferred embodiment of the invention, the method for the data detection is anion-cation balance inspection, error control
System is within ± 5%;The screening of exceptional sample and processing method are using each index histogram, case figure, Q~Q figure and clustering
The one or more of of figure combine, and the water sample data after filtering out exceptional sample and rejecting are as modeling sample
As one of preferred embodiment of the invention, the selection method of the distinguishing indexes is according to each water-bearing layer water sample of modeling
Water chemistry data, the index in each water-bearing layer can be distinguished as distinguishing indexes by therefrom selecting, and determine its threshold value, and identification water
Identification relationship when source;The standard of distinguishing indexes selection is changes of contents of the distinguishing indexes in the water sample of water-bearing layer than other indexs
Changes of contents it is big;Water sample distinguishing indexes to be measured are less than or greater than determining knowledge by the identification relationship of a certain identification when identifying
The water sample of other metrics-thresholds is classified as one kind, and distinguishing indexes can at least distinguish the water sample in two water-bearing layers;Wherein, distinguishing indexes are
The mass concentration of conventional ion and the milliequivalent ratio of conventional ion.
As one of preferred embodiment of the invention, the mass concentration of the conventional ion and the milliequivalent ratio of conventional ion
Value is called characteristic ion comparison and ion ratio coefficient respectively.
As one of preferred embodiment of the invention, using in the preliminary identification model of back substitution test and judge in the step S3
The validity of distinguishing indexes method particularly includes: using the water sample of known aqueous channel type as new samples, successively substitute into the first of foundation
It walks in identification model, if recognition result and reality are consistent, illustrates the preliminary identification model of selection and the foundation of distinguishing indexes
It is effective;If same water-bearing layer water sample recognition result is most inconsistent with reality, new distinguishing indexes should be reselected,
Establish new preliminary identification model.
As one of preferred embodiment of the invention, it is comprehensive in the step S4-gradually method of identification be it is comprehensive using feature from
Sub- method of comparison, ion ratio Y-factor method Y and Fisher method of identification use distinct methods to the identification of different water cut layer water source, first simple
It is complicated afterwards, gradually determine water source type;I.e. for a certain wait sentence single water-bearing layer water sample, by water sample characteristic ion to be measured when identification
Or ion ratio coefficient is classified as certain water source type less than the water sample of determining distinguishing indexes threshold value, or the index is greater than threshold
The water sample of value is classified as certain water source type, as can judge otherwise to be classified as new sample to be sentenced as a result, then stop identifying,
New distinguishing indexes and its threshold value are selected this, are so recycled;When to that can not determine water bursting source type with distinguishing indexes, then adopt
With Fisher method of identification.It can be realized the quick identification of complicated multi-aquifer mine water outlet water source type using the present invention.
As one of preferred embodiment of the invention, gradually specific step is as follows for method of identification for the synthesis-:
The first step judges that the water sample is ash difficult to understand using characteristic ion method of comparison for a certain wait sentence single water-bearing layer water sample
Water or two aqueous, as can judging otherwise to carry out second step identification as a result, then stop identifying;
Second step judges that the water sample is nappe water or coal measures water using ion ratio Y-factor method Y;
Third step, if the water sample is determined as nappe water, uses characteristic ion method of comparison on the basis of second step,
Determine that the water sample is nappe gneiss water and the cold buck of nappe;If the water sample is determined as coal measures water, Fisher is used
Method of identification determines that the water sample is Sandstone Water or too buck.
So far, by above-mentioned three step, should wait sentence single water-bearing layer water sample be judged as two aqueous, nappe gneiss water,
Nappe is trembled with fear buck, Sandstone Water, too one of buck, Ordovician karst water.
As one of preferred embodiment of the invention, the synthesis-gradually method of identification concrete operation method are as follows: passing through data
In water sample after verification and abnormity removing: firstly, working as TDS > 4300mg/L, Cl->2000mg/L、 Na++K+When > 1350mg/L,
The water sample is Ordovician karst water, end of identification;Otherwise, further identification, as TDS < 440mg/L, Cl-<50mg/L、Na++K+<
When 80mg/L, the water sample is two aqueous, end of identification;Otherwise, further identification, as γ Cl-/γCa2+When < 5.2, the water
Sample is nappe water, is further identified, as TDS < 1360mg/L, Cl-When < 474mg/L, the water sample is nappe water gneiss
Water, end of identification;Otherwise the water sample is the cold buck of nappe water, end of identification;As γ Cl-/γCa2+When > 5.2, the water
Sample is coal measures water, recycles Fisher method of identification, identifies Sandstone Water or too buck, at this time end of identification.
The present invention compared with prior art the advantages of be: it is an advantage of the invention that it is comprehensive using characteristic ion method of comparison, from
Sub- proportionality coefficient method and Fisher method of identification etc. use distinct methods to the identification of different water cut layer water source, complicated after first simple, i.e.,
It can gradually determine the Identification of Water Source in Mines type that hydrogeologic condition is more complex, filled water bearing strata is more.The present invention directly passes through detection
Distinguishing indexes used by identification process in unknown water sample can substitute into " comprehensive-gradually method of identification " model established, into
The identification of row water source.It is thus possible to improve the rapidity and accuracy of complicated multi-aquifer water bursting in mine water source identification.
Embodiment 1
As shown in Figure 1, a kind of complicated multi-aquifer water bursting in mine water source recognition methods of the present embodiment, this method include with
Lower step:
S1: according to the water chemistry data of known aqueous layer water sample, the water source database of mine is established;The aquation of the water sample
Learn the milliequivalent and ratio of various mass of ion concentration and each ion that data are each aquifer water of the affiliated coal mine of water sample;It is conventional
Ion can be Na++K+、Ca2+、Mg2+、Cl-、HCO3 -、SO4 2-Deng;This water chemistry data builds library as establish water source database
Index.
S2: by data detection and exceptional value treated data as modeling sample;The side of the data detection
Method is anion-cation balance inspection, and control errors are within ± 5%.In the anion-cation balance of the data detection is examined:
The screening of exceptional sample and processing method use the one or more of each index histogram, case figure, Q~Q figure and clustering figure
In conjunction with the water sample data after filtering out exceptional sample and rejecting are as modeling sample.
S3: distinguishing indexes and its threshold value and having using the distinguishing indexes in the preliminary identification model of back substitution test and judge are determined
Effect property;The selection method of the distinguishing indexes is according to the water chemistry data for modeling each water-bearing layer water sample, and therefrom selecting can distinguish
The index in each water-bearing layer determines its threshold value as distinguishing indexes, and the identification relationship at identification water source;Distinguishing indexes selection
Standard is that changes of contents of the distinguishing indexes in the water sample of water-bearing layer is bigger than the changes of contents of other indexs;The identification of a certain identification is closed
System, that is, the water sample that water sample distinguishing indexes to be measured are less than or greater than to determining distinguishing indexes threshold value when identifying are classified as one kind, identify
Index can at least distinguish the water sample in two water-bearing layers;Wherein, distinguishing indexes are the mass concentration and conventional ion of conventional ion
Milliequivalent ratio, the mass concentration of the conventional ion and the milliequivalent ratio of conventional ion are called characteristic ion respectively
Comparison and ion ratio coefficient.Using the effective of the distinguishing indexes in the preliminary identification model of back substitution test and judge in the step S3
Property method particularly includes: using the water sample of known aqueous channel type as new samples, successively substitute into the preliminary identification model of foundation, if
Recognition result and reality are consistent, then illustrate that the preliminary identification model of selection and the foundation of distinguishing indexes is effective;If same
Water-bearing layer water sample recognition result is most inconsistent with reality, then should reselect new distinguishing indexes, establish new preliminary knowledge
Other model.
S4: according to effective distinguishing indexes and Fisher method of identification, " comprehensive-gradually method of identification " model is established;It is described
Gradually method of identification is that synthesis uses characteristic ion method of comparison, ion ratio Y-factor method Y and Fisher method of identification to synthesis-in step S4,
Distinct methods are used to the identification of different water cut layer water source, it is complicated after first simple, gradually determine water source type;I.e. for a certain wait sentence
Water sample characteristic ion to be measured or ion ratio coefficient are less than determining distinguishing indexes threshold value when identification by single water-bearing layer water sample
Water sample be classified as certain water source type, or the water sample that the index is greater than threshold value is classified as certain water source type, as can judging to tie
Fruit, then stop identifying, is otherwise classified as new sample to be sentenced, and selects this new distinguishing indexes and its threshold value, so recycles;
When to that can not determine water bursting source type with distinguishing indexes, then Fisher method of identification is used.Fisher method of identification divides totality
Cloth does not have particular requirement, is a kind of linear identification method.Its feature is that high dimensional data point is projected to lower dimensional space is (such as one-dimensional
Straight line) on, such data point can become than comparatively dense, so as to overcome due to caused by dimension height " the dimension seed of trouble ".
The principle of projection is will to separate as far as possible between totality and totality, then the smallest according between class distance maximum, inter- object distance
Principle determines discriminance analysis function, and then new sample is carried out classification knowledge.It can be realized complicated multi-aquifer using the present invention
The quick identification of mine water outlet water source type.
Gradually specific step is as follows for method of identification for the synthesis-:
The first step judges that the water sample is ash difficult to understand using characteristic ion method of comparison for a certain wait sentence single water-bearing layer water sample
Water or two aqueous, as can judging otherwise to carry out second step identification as a result, then stop identifying;
Second step judges that the water sample is nappe water or coal measures water using ion ratio Y-factor method Y;
Third step, if the water sample is determined as nappe water, uses characteristic ion method of comparison on the basis of second step,
Determine that the water sample is nappe gneiss water and the cold buck of nappe;If the water sample is determined as coal measures water, Fisher is used
Method of identification determines that the water sample is Sandstone Water or too buck.
So far, by above-mentioned three step, should wait sentence single water-bearing layer water sample be judged as two aqueous, nappe gneiss water,
Nappe is trembled with fear buck, Sandstone Water, too one of buck, Ordovician karst water.
S5: measuring water sample distinguishing indexes to be sentenced, and is successively determined by " comprehensive-gradually method of identification " model, identifies water source
Type.
The synthesis-gradually method of identification concrete operation method are as follows: in the water sample after data check and abnormity removing:
Firstly, working as TDS > 4300mg/L, Cl->2000mg/L、Na++K+When > 1350mg/L, the water sample is Ordovician karst water, end of identification;
Otherwise, further identification, as TDS < 440mg/L, Cl-<50mg/L、 Na++K+When < 80mg/L, the water sample is two aqueous, is known
Do not terminate;Otherwise, further identification, as γ Cl-/ γCa2+When < 5.2, the water sample is nappe water, is further identified, when
TDS<1360mg/L、Cl-When < 474mg/L, the water sample is nappe water gneiss water, end of identification;Otherwise the water sample
For the cold buck of nappe water, end of identification;As γ Cl-/γCa2+When > 5.2, the water sample is coal measures water, recycles Fisher
Method of identification identifies Sandstone Water or too buck, at this time end of identification.
The present embodiment is comprehensive using characteristic ion method of comparison, ion ratio Y-factor method Y and Fisher method of identification etc., to difference
Aquifer water identifing source uses distinct methods, complicated after first simple, can gradually determine that hydrogeologic condition is more complex, water-filling contains
The more Identification of Water Source in Mines type of water layer.The present embodiment directly passes through identification used by identification process in the unknown water sample of detection and refers to
Mark can substitute into " comprehensive-gradually method of identification " model established, and carry out water source identification.It is thus possible to improve complicated contain more
The rapidity and accuracy of water layer water bursting in mine water source identification.
Selection below is new to collect aqueous Cenozoic two in two water quality accounts, nappe gneiss water, nappe cold buck, sandstone
The water-bearing layers such as water and too buck water sample data are as database, to be specifically described the design scheme and theoretical foundation of the present embodiment.
Data detection and exceptional sample processing
When carrying out single characteristics on Aquifer index and judging, it is desirable that the water sample Hydrochemical Composition analyzed need to be the water-bearing layer
Typical Representative.Therefore, it before carrying out characteristic index analysis and being established with water source identification model, need to test to each water-bearing layer water sample
With exceptional sample processing.
The method that sample data is examined is anion-cation balance inspection, due to Na++K+、Ca2+、Mg2+、Cl-、 HCO3 -、
SO4 2-Index is measured value, therefore it is required that control errors within ± 5%, i.e., test error is qualified water within this range
Sample carries out next step analysis as basic data.
Though it is qualified to examine through zwitterion, there are a small number of water samples durings water sampling, the judgement of layer position, test analysis etc.
A certain link there are some problems, cause the water sample that can not really reflect a certain water-bearing layer water quality characteristic, in water analysis
Outliers are shown as in the process, need to carry out screening rejecting processing.Below by taking Ordovician karst water sample as an example, the sieve to exceptional sample is illustrated
Choosing and processing method, such as Fig. 2-4.
To 17 Ordovician karst water samples TDS, Na of collection++K+、Ca2+、Cl-Ion compares and analyzes, as shown in Fig. 2.Label
It is obvious for the above-mentioned several indexs of tri- water samples of O-03, O-04, O-05 and remaining 14 water sample difference.Further to judge Ordovician karst water
O-04, O-05 and No. O-06 three water samples and remaining water sample relationship, have carried out hierarchial-cluster analysis to Ordovician karst water sample.Using TDS,
Na++K+、Ca2+、Mg2+、Cl-、HCO3 -、 SO4 2-Deng 7 indexs as variable, cluster result is as shown in Figure 3.As seen from Figure 3, it removes
Outside marked as O-03, O-04, O-05 sample, remaining sample room otherness is minimum, can most represent Ordovician karst water property, and marked as
O-03, O-04, O-05 and remaining sample interval finally just complete classification from maximum.And box traction substation, it is that one kind is used to describe number
According to the statistical graph of distribution, it can be used to show the descriptive statistics amounts such as median, 4 quantiles and the extreme value of observation data, from
The distribution situation of the angular observation variate-value of vision.Ordovician karst water 17 water samples TDS, Na++K+、Ca2+、Mg2+、Cl-Index case figure is such as
Shown in Fig. 4.As seen from the figure, marked as tri- water samples of O-3, O-4, O-5 in this 5 indexs, also display has with remaining water sample
Significant difference.
Therefore, O-03, O-04, No. O-05 three samples can be considered exceptional sample in 17 samples of Ordovician karst water, cannot represent
Ordovician karst water characteristic feature, will reject in subsequent water sample type identification model foundation, participate in modeling not as standard water sample.
, nappe gneiss water aqueous to Cenozoic two, nappe Cambrian system water, Sandstone Water and too buck etc. remaining 5
Single water-bearing layer water sample is all made of the synthesis screening and rejecting that above-mentioned or other a variety of methods carry out outliers.It is examined by data
Test and exceptional sample reject after data, the sample as modeling.
Distinguishing indexes and threshold value determine
According to Such analysis, two aqueous and Ordovician karst waters are in TDS, Cl-And Na++K+There is characteristic feature in content, and it is other
Water-bearing layer water sample can be distinguished effectively.Therefore, select above-mentioned 3 indexs as two is aqueous and the distinguishing indexes of Ordovician karst water,
Its content range is as shown in table 1.
Table 1 two is aqueous and Ordovician karst water characteristic index content range
(2) nappe water and coal measures water
Nappe gneiss water is close with the cold buck conventional ion of nappe and water quality type, and Sandstone Water is conventional with too buck
Ion and water quality type are close.Here nappe gneiss water and the cold buck of nappe are first classified as one kind, are referred to as nappe
Water;Sandstone Water and too buck is classified as one kind, is referred to as coal measures water.By modeling sample statistics indicate that, nappe water and coal
It is characteristic ion proportionality coefficient (the γ Cl of water-/γCa2+) otherness is obvious (table 2), can be used as the nappe water reclassified and
The diacritics of coal measures water.
2 nappe water of table and coal measures water characteristic ion proportionality coefficient scope
(3) nappe gneiss water and the cold buck of nappe
Nappe gneiss water and nappe Cambrian system water are in TDS and Cl-There is characteristic feature in content, can be used as differentiation
The mark of these two types of aquifer waters.
3 nappe water characteristic index content range of table
(4) Sandstone Water and too buck
The difference in feature example index and ion ratio coefficient is unobvious with too buck for Sandstone Water, only from above-mentioned two
Aspect conventional means are not possible to distinguish.For this purpose, comprehensive selection TDS, Ca2+、Mg2+、Na++K+、 HCO3 -、SO4 2-、Cl-Refer to Deng 7
It is denoted as the recognition factor for Sandstone Water and too classified between buck, judgement water source type is carried out using Fisher method of identification.
Fisher method of identification is a kind of linear identification method to the overall no particular requirement of distribution.Its feature be by
High dimensional data point projects on lower dimensional space (such as one-dimensional straight line), and such data point can become than comparatively dense, so as to
Overcome due to caused by dimension height " the dimension seed of trouble ".The principle of projection is will to separate as far as possible between totality and totality, then
According between class distance, maximum, the smallest principle of inter- object distance determines discriminance analysis function, and then new sample is carried out classification knowledge.
Recognition methods selection and identification step
This two mine complexity multi-aquifer water source type recognition methods of new collection is using " comprehensive-gradually method of identification ".
Specific step is as follows:
The first step judges that the water sample is ash difficult to understand using characteristic ion method of comparison for a certain wait sentence single water-bearing layer water sample
Water or two aqueous, as can judging otherwise to carry out second step identification as a result, then stop identifying.
Second step judges that the water sample is nappe water or coal measures water using ion ratio Y-factor method Y.
Third step, if the water sample is determined as nappe water, uses characteristic ion method of comparison on the basis of second step,
Determine that the water sample is nappe gneiss water and the cold buck of nappe;If the water sample is determined as coal measures water, Fisher is used
Method of identification determines that the water sample is Sandstone Water and too buck.
So far, by above-mentioned three step, should wait sentence single water-bearing layer water sample be judged as two aqueous, nappe gneiss water,
Nappe is trembled with fear buck, Sandstone Water, too one of buck, Ordovician karst water.
The foundation of " comprehensive-gradually method of identification " model
For a certain wait sentence single water-bearing layer water sample, data detection and exceptional sample processing, distinguishing indexes and boundary are being carried out
After limit value, recognition methods selection and identification step determine, the identification model and process of foundation are as shown in Figure 3.
The inspection of recognition effect
Using the validity of back substitution test and judge distinguishing indexes, using the water sample of known aqueous channel type as new samples, according to
It is secondary to substitute into the identification model established, if recognition result and practical consistent, illustrate distinguishing indexes selection and foundation it is " comprehensive
Closing-gradually method of identification " model is effective;If same water-bearing layer water sample recognition result is most inconsistent with reality, should weigh
New distinguishing indexes are newly selected, new water source identification model is established.
After determining " comprehensive-gradually method of identification " model, to unknown water-bearing layer water sample, pass through the content of test distinguishing indexes
Afterwards, established model is substituted into, the aqueous channel type of you can get it unknown water sample.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.