CN104239738A

CN104239738A - Method for predicting floor original water flowing fractured zone

Info

Publication number: CN104239738A
Application number: CN201410503595.2A
Authority: CN
Inventors: 于小鸽; 施龙青; 韩进
Original assignee: Shandong University of Science and Technology
Current assignee: Shandong University of Science and Technology
Priority date: 2014-09-28
Filing date: 2014-09-28
Publication date: 2014-12-24
Anticipated expiration: 2034-09-28
Also published as: CN104239738B

Abstract

The invention belongs to the technical field of coal mine safety production and relates to a method for predicting a floor original water flowing fractured zone. The method comprises the following steps: firstly establishing a BP (Back Propagation) neural network model, dividing geological structure complexity into four levels, i.e., simple, more simple, more complex and complex, then determining a water-rich area of an area to be predicted and water richness thereof according to geological and hydrological conditions of the area to be predicted, determining the floor original water flowing fractured zone of the area to be predicted according to the determined overlapped areas of more complex and complex geological structure areas and the water-rich area, and finally predicting the height of the determined floor original water flowing fractured zone by using the BP neural network prediction model established through MATLAB. The method for predicting the floor original water flowing fractured zone has the advantages of simple prediction principle, simple prediction method, intelligent and convenient operation, high prediction accuracy and prediction environmental-friendliness.

Description

The original Forecasting Methodology leading high-band of a kind of base plate

Technical field:

The invention belongs to Safety of Coal Mine Production technical field, relate to the original Forecasting Methodology leading high-band of the original Forecasting Methodology leading high-band, particularly a kind of base plate in the control of a kind of mine water disaster.

Background technology:

Original high-band of leading refers to that piestic water in water-bearing zone is under the effect of high hydraulic pressure, along the height that the splitting in hypomere water-resisting layer or fractured zones are risen, its distribution opposed bottom gushing water has primitiveness and is called original height of leading, and original high-band of leading has following characteristics: one is this band rock because water chemical action is in elastoplasticity or mecystasis, two is that its cranny development is uneven, and becomes water inrush channel, three is these band rock stratum poor continuity, four is that original high-band of leading loses water-resisting ability, and it is little that Floor water gives prominence to on-way resistance from this band, according to statistics, the colliery of China nearly 10% in various degree be subject to pressure-bearing water mitigation, the each coal producer of injured area and the order of severity Jun Ju world is the first, along with coal mining shifts to deep year by year, water damage problem is day by day serious, Chinese scholars is this has been a large amount of research work, a lot of theory and means is proposed, but few people study separately original high-band of leading, its basic reason is: original basis of leading high formation is tectonic fissure, development and the power raised are wedge action and the stress corrosion of water, originally lead the high impact by tectonic fissure and confined aquifer, a continuity is not had in the process of growing, that usually grows different scale leads height, and its upper bound is uneven, not of uniform size, some mining areas or position even do not have originally to lead height, these influence factors make original high-band of leading study difficult, its prediction is extremely difficult especially, there is not yet the original relevant record leading high-band Forecasting Methodology in prior art, but, original high-band of leading but plays vital effect in the prediction of Water Inrush, therefore, seek to design the original Forecasting Methodology leading high-band of a kind of base plate and there is important economic implications and great realistic meaning.

Summary of the invention:

The object of the invention is to the defect overcoming prior art existence, seeking to design provides a kind of base plate the original Forecasting Methodology leading high-band, and strengthen leading high-band to base plate and leading a liter machine-processed understanding, the prediction for Water Inrush provides important Technical Reference.

To achieve these goals, the Forecasting Methodology that the present invention relates to, comprises following concrete steps:

(1) the architectonic complexity of survey region is determined: according to the factor affecting geological structural complexity, set up BP (Back Propagation) neural network model, geological structural complexity is divided into simply, more simply, more complicated, complicated four ranks, its step is as follows:

1. the factor of geological structural complexity will be affected, comprise Fault density M, FAULT STRENGTH index F, tomography and coal seam angle index Q, seam inclination absolute value A, seat earth inclination angle coefficient of variation R, thickness of coal seam abnormal D, floor level luffing H and area of structure loss coefficient S are normalized, and are about to parameter quantitative qualitatively, be normalized further again, by parameter value with 0 ~ 1 numerical value represent;

2. utilize matrix labotstory (MATLAB) to set up BP neural network prediction model, by four of geological structural complexity ranks: simply, more simply, the corresponding output item of more complicated and complicated difference (1,0,0,0), (0,1,0,0), (0,0,1,0) and (0,0,0,1);

3. predict according to the BP neural network model set up, divide geological structural complexity region, and on picture, draw a circle to approve more complicated and complex region;

(2) watery of estimation range is determined: the geological hydrology condition residing for estimation range, utilize three-dimensional physical prospecting means, difference according to resistivity judges, the region watery that resistivity is high is poor, the region watery that resistivity is low is good, draws a circle to approve water enrichment area and the watery thereof of estimation range with this;

(3) determine that the base plate of estimation range is original and lead high-band: the base plate of the delineation estimation range, overlapping region determined according to step (1) and step (2) is original leads high-band, MATLAB is utilized to set up BP neural network prediction model, finally lead high-band to the base plate of delineation and carry out that base plate is original leads high prediction, its step is as follows:

The water enrichment area of the floor water-bearing rock that the more complicated and complex region of the tectonic structure 1. drawn a circle to approve out according to step (1) and step (2) are drawn a circle to approve out, determines that the two intersection is the original region of leading high-band and comparatively growing of base plate;

2. selected part mine bottom plate is original leads high data of testing as sample, utilizes MATLAB to set up BP neural network prediction model, then determines that the base plate of estimation range is original with the model set up and lead height;

The MATLAB of utilization of the present invention sets up BP neural network prediction model, its step is as follows: selected part mine leads high data of testing as sample data to base plate is original, utilize MATLAB to be normalized sample data, and sample data is divided into learning sample and test samples; Choose after suitable parameters is trained learning sample and set up BP neural network prediction model, test samples is utilized to test to BP neural network prediction, assay reaches training requirement, illustrate that the forecast model set up is feasible, when having new data again, direct input makes it run, and just can obtain the result wanted; When dividing geological structural complexity, the data affecting geological structural complexity factor be normalized, input runs, and just can mark off the geological structural complexity belonging to it; Measure original when leading high, originally lead high model, input data according to what set up, just can dope and originally lead height.

The present invention compared with prior art, its data related to all take from field measurement, ensure the objectivity of data, utilize MATLAB to set up BP neural network prediction model, improve speed and the precision of prediction, its prediction principle is reliable, Forecasting Methodology is simple, operative intelligence is convenient, and precision of prediction is high, prediction environmental friendliness.

Accompanying drawing illustrates:

Fig. 1 is the flow process schematic block diagram of Forecasting Methodology of the present invention.

Fig. 2 is that the BP neural network model three layers of experimental network target that the present invention relates to export the loose some schematic diagram with actual output error.

Fig. 3 is the BP neural network training process curve synoptic diagram that the present invention relates to.

Embodiment:

The present invention to be elaborated by embodiment below in conjunction with accompanying drawing.

Embodiment 1:

MATLAB is utilized to set up BP neural network prediction model described in the present embodiment, its step is as follows: selected part mine leads high data of testing as sample data to base plate is original, utilize MATLAB to be normalized sample data, and sample data is divided into learning sample and test samples; Choose after suitable parameters is trained learning sample and set up BP neural network prediction model, test samples is utilized to test to BP neural network prediction, assay reaches training requirement, illustrate that the forecast model set up is feasible, when having new data again, direct input makes it run, and just can obtain the result wanted; When dividing geological structural complexity, the data affecting geological structural complexity factor be normalized, input runs, and just can mark off the geological structural complexity belonging to it; Measure original when leading high, originally lead high model, input data according to what set up, just can dope and originally lead height.

Claims

1. the original Forecasting Methodology leading high-band of base plate, is characterized in that comprising following concrete steps:

(1) the architectonic complexity of survey region is determined: according to the factor affecting geological structural complexity, set up BP neural network model, geological structural complexity is divided into simply, more simply, more complicated, complicated four ranks, its partiting step is as follows:

2. utilize matrix labotstory to set up BP neural network prediction model, by four of geological structural complexity ranks: simply, more simply, the corresponding output item of more complicated and complicated difference (1,0,0,0), (0,1,0,0), (0,0,1,0) and (0,0,0,1);

(3) determine that the base plate of estimation range is original and lead high-band: the base plate of the delineation estimation range, overlapping region determined according to step (1) and step (2) is original leads high-band, MATLAB is utilized to set up BP neural network prediction model, finally to delineation base plate lead high-band carry out base plate original lead high-band prediction, its step is as follows:

2. selected part mine bottom plate is original leads high data of testing as sample, utilizes MATLAB to set up BP neural network prediction model, then determines that the base plate of estimation range is original with the model set up and lead high-band;

The described MATLAB of utilization sets up BP neural network prediction model, its step is as follows: selected part mine leads high data of testing as sample data to base plate is original, utilize MATLAB to be normalized sample data, and sample data is divided into learning sample and test samples; Choose after suitable parameters is trained learning sample and set up BP neural network prediction model, test samples is utilized to test to BP neural network prediction, assay reaches training requirement, illustrate that the forecast model set up is feasible, when having new data again, direct input makes it run, and just can obtain the result wanted; When dividing geological structural complexity, the data affecting geological structural complexity factor be normalized, input runs, and just can mark off the geological structural complexity belonging to it; Measure original when leading high, originally lead high model, input data according to what set up, just can dope and originally lead height.