CN118211495A - Unmanned mining electric shovel dynamic excavation resistance modeling method based on data-model combined driving - Google Patents
Unmanned mining electric shovel dynamic excavation resistance modeling method based on data-model combined driving Download PDFInfo
- Publication number
- CN118211495A CN118211495A CN202410627598.0A CN202410627598A CN118211495A CN 118211495 A CN118211495 A CN 118211495A CN 202410627598 A CN202410627598 A CN 202410627598A CN 118211495 A CN118211495 A CN 118211495A
- Authority
- CN
- China
- Prior art keywords
- resistance
- excavation
- excavating
- bucket
- mining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000009412 basement excavation Methods 0.000 title claims abstract description 101
- 238000005065 mining Methods 0.000 title claims abstract description 90
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000013499 data model Methods 0.000 title claims abstract description 15
- 239000000463 material Substances 0.000 claims abstract description 85
- 238000004364 calculation method Methods 0.000 claims abstract description 32
- 230000008569 process Effects 0.000 claims abstract description 27
- 238000005520 cutting process Methods 0.000 claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000003062 neural network model Methods 0.000 claims description 24
- 230000008859 change Effects 0.000 claims description 11
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000004134 energy conservation Methods 0.000 claims description 7
- 239000002689 soil Substances 0.000 claims description 6
- 230000001133 acceleration Effects 0.000 claims description 5
- 238000003825 pressing Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 4
- 230000005484 gravity Effects 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 2
- 238000001125 extrusion Methods 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000005381 potential energy Methods 0.000 claims description 2
- 238000013507 mapping Methods 0.000 abstract description 4
- 238000004458 analytical method Methods 0.000 description 5
- 238000004088 simulation Methods 0.000 description 5
- 230000009471 action Effects 0.000 description 4
- 238000005056 compaction Methods 0.000 description 3
- 238000011439 discrete element method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 239000011435 rock Substances 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000003971 tillage Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000005422 blasting Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000010276 construction 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
- 230000001186 cumulative effect Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005293 physical law Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Operation Control Of Excavators (AREA)
Abstract
The invention provides a modeling method for dynamic excavating resistance of an unmanned mining electric shovel based on data-model combined driving, and belongs to the field of dynamic excavating resistance of mining electric shovels. Firstly, basic parameters and historical mining data of the unmanned mining electric shovel are obtained. And secondly, calculating excavation depth information corresponding to different material levels. Third, the cutting resistance is calculated according to Balovnev resistance calculation models. Fourth, constructing Bi-LSTM neural network to predict radial resistance. Fifth, a mining energy loss function is constructed. Sixth, training the Bi-LSTM neural network based on the mining data to obtain a Bi-LSTM network model. And finally, the Bi-LSTM network model and the Balovnev resistance calculation model are combined to form an excavation resistance prediction model. The excavating resistance prediction model obtained by the invention is combined with a real resistance and an analytic formula to predict the excavating resistance, so that the excavating resistance prediction model can reveal the mapping relation between the material surface morphology, the state parameters of the electric shovel, the excavating track and the like and the excavating resistance, and can realize the rapid and accurate prediction of the external load in the excavating operation process of the mining large-scale electric shovel.
Description
Technical Field
The invention belongs to the field of dynamic excavating resistance of mining electric shovels, and relates to a modeling method for dynamic excavating resistance of an unmanned mining electric shovel based on data-model combined driving.
Background
With the continuous development of intelligent mine construction in China, the mining efficiency and safety requirements are continuously improved, wherein an unmanned mining electric shovel is used as key equipment for mining, is widely used for stripping and mining operation of solid resources (such as coal, ore materials and the like) of an open-pit mine, and can solve the safety problem caused by fatigue of operators due to large labor intensity of complex operation conditions (heavy load, strong impact, strong vibration and the like) and severe operation environments (extremely cold, high temperature, high altitude, high dust and the like). In the autonomous operation process of the mining electric shovel, the quality of the digging track directly influences the performance and is very important for realizing high-performance operation. Because the real rock has complex and changeable morphology and possibly buries the rock with larger body type, the mining electric shovel bucket is in direct contact with the excavated material and interacts with the excavated material in the excavating process, the excavating resistance acting on the bucket is changed severely and is transmitted to the whole system through the bucket rod-bucket, the lifting rope, the lifting wire rope and other parts of the front-end working device, and the driving mechanism is influenced. The dynamic mining resistance prediction model which accords with the actual situation is established and is the basis of subsequent mining safety evaluation, power consumption calculation and track planning.
The mining electric shovel has the advantages that the mining resistance of the mining electric shovel is accurately predicted, the evolution rule of the mining electric shovel is established, the mining electric shovel is very important for subsequent work such as structural performance response analysis and track decision, and the mining electric shovel is important in load input for carrying out the subsequent work. In the existing research, the mining electric shovel excavation resistance modeling mainly comprises a numerical method, an analytic method and an experimental test method, and the method comprises the following steps of:
In the numerical method, for the characteristics of the materials to be excavated, a Finite Element Method (FEM), a Discrete Element Method (DEM), a gridless method and the like are often selected for analysis. For example, kimy S et al in papers "Kim Y S, Siddique M A A, Kim W S, et al. DEM simulation for draft force prediction of moldboard plow according to the tillage depth in cohesive soil[J]. Computers and Electronics in Agriculture, 2021, 189: 106368." and "Kim Y S, Lee S D, Baek S M, et al. Development of DEM-MBD coupling model for draft force prediction of agricultural tractor with plowing depth[J]. Computers and Electronics in Agriculture, 2022, 202: 107405." model soil based on discrete element simulation methods and predict the change in the excavation resistance of a blade cutting soil at different excavation depths. However, discrete element simulation is huge in calculation amount, long in prediction period, and needs simulation calculation for different mining conditions, so that the discrete element simulation is inconvenient to use in track planning.
Compared with a numerical method, the analysis method has the characteristics of high calculation efficiency, convenience in parameterization expression, strong physical interpretability and the like. For example, gill et al consider the influence of the digging speed when calculating the digging resistance in paper "Gill W R, Berg G E V. Soil dynamics in tillage and traction[M]. Agricultural Research Service, US Department of Agriculture, 1967." , and more approach to the real digging process, thereby effectively improving the model precision. However, since the bucket squeezes materials to generate radial compaction force perpendicular to the movement direction of the bucket teeth in the process of digging, the size of the radial compaction force mainly depends on the digging track and the hardness of the dug medium, and the radial compaction force is difficult to be explicitly expressed by an analytical expression.
Experimental testing methods are typically combined with data processing methods, with experimental testing being performed first to obtain relevant data, followed by data processing and building a relevant resistance prediction model. For example Ilaria Palomba et al in paper "[43] Palomba I, Richiedei D, Trevisani A, et al. Estimation of the digging and payload forces in excavators by means of state observers[J]. Mechanical systems and signal processing, 2019, 134: 106356." solve for forces between the earth-bucket and cumulative loads in the bucket during operation of the excavator using a state estimation method, the forces and external loads can be determined effectively without consideration of the earth-tool interaction model. But unlike analytical methods, the method lacks interpretable physical meaning and is prone to introducing modeling variances in the solution process.
In summary, most of the methods in the current stage of research cannot be combined efficiently, accurately and reliably, and cannot play a role in trajectory planning. The accurate excavation resistance can be obtained in a short time, so that key information can be provided for excavation safety evaluation, power consumption calculation and track planning.
Disclosure of Invention
Aiming at the problem that effective information cannot be obtained quickly in the prior art, the invention provides a data-model combined driving-based unmanned mining electric shovel dynamic excavation resistance modeling method.
The technical scheme adopted by the invention is as follows:
A modeling method for dynamic excavation resistance of an unmanned mining electric shovel based on data-model combined driving comprises the steps of firstly, obtaining historical excavation data of the unmanned mining electric shovel. And secondly, calculating excavation depth information corresponding to different material levels. Third, the cutting resistance is calculated according to Balovnev resistance calculation models. Fourth, constructing Bi-LSTM neural network model to predict radial resistance. Fifth, a mining energy loss function is constructed. And sixthly, training the Bi-LSTM neural network based on the mining data until the mining energy loss function reaches the minimum value, so as to obtain the Bi-LSTM network model. Finally, the Bi-LSTM network model and Balovnev resistance calculation model together form an excavation resistance prediction model. The method specifically comprises the following steps:
step1: historical excavation data of the unmanned mining electric shovel is acquired, and the method specifically comprises the following steps of:
In the process of excavating, excavating resistance generated by pressing the bucket into the material is influenced by excavating operation and a mining electric shovel structure besides the properties of the material, such as cohesive force, viscosity coefficient, material repose angle and the like, so that historical excavating data to be acquired comprise: parameters related to the excavation strategy include an excavation trajectory, an excavation speed, an acceleration, and an excavation time, and mechanism parameters related to the electric shovel include a crown wheel radius, a bucket width, an arm length, and a boom weight.
Step 2: the mining depth information corresponding to different material levels is calculated, and the mining depth information is specifically as follows:
In the modeling process of the excavating resistance, dynamic excavating depth information is introduced according to the shape of a piling surface and an excavating track, and an additional resistance item brought by the weight item and the excavating speed of the excavated material is added into the bucket, so that the action mechanism from the bucket to the material of the mining electric shovel in a three-dimensional space is more met. It is therefore necessary to calculate the excavation depth from the surface information of the material level and the excavation track, and the weight of the material in the bucket at each moment. In the digging process, digging depth information at different moments Can be changed along with the change of the shape of the material surface and the excavation track. When the bucket is inserted into the material, the value is the vertical distance between the tooth tip of the bucket and the surface of the material; when the material is separated, the value is. Depth of excavationAccording to formula (1):
Wherein, Representing a charge level function; representing an excavation trajectory function; Represents a distance in the horizontal direction; And Indicating the horizontal distance the bucket begins to insert and the horizontal distance from the material, respectively.
The weight of the excavated material in the bucketFor the product of the excavated volume and the density, as shown in formula (2):
Wherein, Representing the density of the material; representing a bucket width; Representing the excavation volume, which is the excavation depth With respect to horizontal distanceIs a function of the integral of (a).
Step 3: the cutting resistance is calculated according to Balovnev resistance calculation model, concretely as follows:
when the mining electric shovel excavates the inclined plane stacker, the excavated material in front of the bucket is regarded as an approximate wedge body according to the material plane failure surface hypothesis theory. According to the mechanical characteristics of wedge-shaped soil bodies, the basic material information, the mining electric shovel mechanism parameters and the dynamic excavation depth information in the step 1 and the step 2 are introduced, and the excavation resistance of the bucket is built In the form of:
Wherein, The gravity of the bucket is a constant value in the process of excavating, and the gravity of the material is gradually increased and is obtained by a formula (2); for cutting resistance, the tangential movement of the bucket is mainly blocked, and is obtained by a Balovnev resistance calculation model; radial resistance generated by extruding the matrix material by the bucket in the excavating process is provided; in order to cause friction between the bottom surface of the bucket and the base material due to normal resistance, the direction being parallel to the direction of movement of the bucket, wherein the normal resistance comprises the radial resistance and the normal component of the cutting resistance generated by the extrusion of the material by the bucket, according to AndIs calculated.
Step 4: constructing a Bi-LSTM neural network model to predict radial resistance, wherein the method comprises the following steps of:
The radial resistance is perpendicular to the bucket movement direction, and is related to the digging operation, and is related to the hardness and the characteristics of the base material, so that an analytical expression is difficult to establish. And the relevant variables in the excavation resistance model are related not only to the current moment state information, but also to the historical excavation trajectories, so the radial resistance predictive modeling is dynamic. According to the method, the Bi-LSTM neural network model is built by combining the mining information and the cutting force calculation result in the step 3, and radial resistance parameterization expression generated by pressing the bucket into the material is built by means of the neural network model, so that dynamic radial resistance is predicted. The Bi-LSTM neural network model is characterized in that the input of the Bi-LSTM neural network model comprises parameters related to an excavation strategy, such as excavation track, excavation speed, acceleration, excavation depth and the like in the step 1 and the step 2, and the output is radial resistance caused by the fact that a bucket is pressed into materials.
Wherein,Representation ofThe input of the time neural network unit is the parameter related to the mining strategy; Representation of Radial resistance predicted at time; Representation of Time pairAndIs a preliminary feature value of (a); Representation of Time pairAndIs a characteristic value of (2); Representing an activation function; Representing the Hadamard product; ，，， respectively representing forgetting gate, input gate, output gate and characteristic extraction process Weight coefficient of (2);，，， respectively representing forgetting gate, input gate, output gate and characteristic extraction process Weight coefficient of (2); Is that Forget the value of the door at the moment; Is that Inputting the value of the gate at the moment; Is that Outputting the value of the door at the moment; Is that The radial resistance predicted at the moment.
Step 5: the mining energy loss function is constructed as follows:
A mining energy loss function is constructed, which consists of a lagrangian mechanical loss function, an energy conservation loss function, and a regularized loss, as shown in equation (5). The Lagrange mechanical loss function is built according to the minimized Lagrange mechanics and is used for describing excavation force loss in the excavation process; the energy conservation loss function is built by solving the kinetic energy and the time change rate of the potential energy through a central differential filter, is used for describing the energy change of the mining electric shovel system in the excavating operation, ensures that the total energy of the mining electric shovel system is equal at each moment of operation, and realizes the minimum energy loss in the prediction process by using the loss function; regularization loss, which reduces the likelihood of finding an overfitting solution by reducing the solution space, uses And (5) regularization.
Wherein,Indicating a bucket digging angle;、 And Is a super parameter, and determines the weight occupied by each item in the optimized objective function; Representing a lagrangian mechanical loss function; representing an energy conservation loss function; representing regularization loss, preventing model overfitting.
Step 6: training the Bi-LSTM neural network based on the mined data, specifically as follows:
And (3) calculating the excavation energy loss function according to the cutting resistance calculated by the Balovnev resistance calculation model in the step (3) and the radial resistance predicted by the neural network in the step (4), and solving the optimization problem shown in the formula (6) by a gradient descent method and combining the calculation result of the loss function to obtain the Bi-LSTM neural network model parameters. And the termination condition is that the output result of the loss function is minimum, and when the termination condition is met, training is completed, and the Bi-LSTM neural network model parameters are obtained.
Step 7: the Bi-LSTM network and Balovnev resistance calculation model together form an excavation resistance prediction model, and the method is specifically as follows:
And (3) the Bi-LSTM neural network model and the Balovnev resistance calculation model which are trained in the step (6) jointly form an excavation resistance prediction model. The excavation resistance prediction model takes basic information such as mining structure information, excavation track, material stacking surface, material viscosity and the like as input, and can rapidly and accurately predict the excavation resistance at different moments in the excavation process.
Compared with the prior art, the invention has the following beneficial effects:
(1) In order to deal with the problems of unavoidable introduction of modeling deviation and long calculation period due to incomplete knowledge of the shape and the accumulation structure of the materials in the known geotechnical equations and empirical formulas from the practical point of view, the invention adopts a method of combining the traditional mechanical equilibrium equation and the artificial intelligence technology, so that the excavation resistance can be rapidly predicted on the premise of ensuring the maximum reliability, reliable data can be obtained, and the calculation time can be greatly reduced.
(2) According to the invention, the minimized Lagrange mechanics description physical law is utilized to construct the excavation resistance prediction model loss function to carry out model training, and finally modeling solution of hidden variable excavation resistance is realized.
(3) The invention provides a mining resistance prediction model formed by a Bi-LSTM neural network and a Balovnev resistance calculation model, wherein the model is combined with a real resistance and an analytic formula to predict the mining resistance, so that the mining resistance prediction model reveals the mapping relation between the material surface morphology, the state parameters of the electric shovel, the mining track and the like and the mining resistance, and realizes the accurate prediction of the external load in the mining large-scale electric shovel mining operation process.
Drawings
FIG. 1 is a flow chart of modeling and predicting the mining resistance by data-model combination driving in accordance with an embodiment of the present invention.
Fig. 2 is a graph of lift moment contrast for validating the predictive model of excavation resistance in accordance with the present invention.
Fig. 3 is a graph showing a comparison of pushing moment for verifying the predictive model of excavating resistance according to the present invention.
Detailed Description
The invention will be further described with reference to the accompanying drawings and specific examples.
As shown in FIG. 1, the invention provides a data-model combined driving-based unmanned mining electric shovel dynamic excavation resistance modeling method, which comprises the following specific implementation measures:
Step 1: the excavating resistance prediction model predicts on the basis of the structure and material information of the mining electric shovel, so that basic structure information such as the distance between the center of a crown wheel and a pushing shaft, the distance between the far end of a movable arm and the pushing shaft, the height between the center of the pushing gear and the ground and the like, and material information such as blasting material density, cohesion, viscosity coefficient, material static angle and the like are required to be obtained before model prediction. In order to obtain enough training data, the mining electric shovel performs multiple excavation experiments on rock materials in different tracks, and covers different excavation working conditions, including the shape of a material surface, the full bucket condition of a bucket and the distance between the electric shovel and the material surface. Corresponding working mechanism posture information, lifting and pushing motor torque information and point cloud data describing the shape of the material surface are collected and used as a sample set.
Step 2: the material surface shape corresponding to the excavation track is determined according to modeling of material surface point cloud data obtained through laser radar scanning. When the mining electric shovel excavates the inclined plane stacker, the excavated material in front of the bucket can be regarded as an approximate wedge body according to the material plane failure surface hypothesis theory. In the actual excavating operation process of the mining electric shovel, the upper carriage and the lower carriage are kept fixed, the bucket motion is realized by virtue of the pushing and lifting mechanism, so that the transverse axis coordinates of the excavating track at different moments are kept unchanged, and the dynamic excavating depth in the excavating process is calculated under the condition. In the excavating process, materials in the bucket are continuously accumulated, and the mass of the materials in the bucket under a given excavating track is calculated by adopting a material level equation and a difference integral of the excavating track.
Step 3: the dynamic excavation depth information in the step 2, the weight items of the materials which are excavated and the additional resistance items brought by the excavation speed are introduced in the excavation resistance modeling process, so that the method is more in line with the action mechanism of the mining electric shovel bucket and materials in the three-dimensional space, and the expression of the dynamic excavation resistance applied to the mining electric shovel in the complex stacking morphology during the excavation operation is further deduced. In the expression, the cutting resistance refers to an existing bucket-material action model, three parts are split, and the bucket tooth cutting resistance, the resistance caused by the action of the excavated material on two side walls and the additional resistance caused by the influence of the excavation speed are obtained by a Balovnev resistance calculation model.
Step 4: the normal resistance of the excavation resistance is partly the normal component of the cutting resistance and partly the radial resistance due to the bucket pressing into the material, wherein the latter part of resistance is related to the excavation operation and is difficult to be explicitly expressed by Balovnev resistance calculation models. During the digging operation of the electric shovel, materials are continuously loaded into the bucket, and the materials in the bucket are continuously accumulated. The related variables in the mining resistance model are related to not only the state information at the current moment but also the historical mining track, so that the dynamic radial resistance predictive modeling is regarded as a time sequence modeling problem, and the Bi-LSTM neural network is constructed to predict the radial resistance. In the mining resistance modeling strategy, the neural network is used for establishing track information and functional mapping of the material surface morphology and radial resistance, has a relatively simple functional mapping relation, and can capture the relation between input and output. The input of the neural network model comprises parameters related to the excavation strategy, such as excavation track, excavation speed, acceleration, excavation depth and the like, and the output is radial resistance caused by the fact that the bucket is pressed into the material.
Step 5: the excavation resistance is the reaction force exerted by the excavated material on the bucket, and is difficult to directly measure based on the existing measuring equipment and measuring technology, so that the traditional end-to-end supervised model training method cannot be directly adopted. In order to realize end-to-end supervised neural network model training, the radial resistance value output by the neural network model needs to be projected into the generalized coordinates of the front-end working device, and the radial resistance value is used as a part of the generalized force in the Lagrangian mechanical equation to participate in dynamic calculation. The mining resistance prediction model provided by the invention takes energy consumption as an optimization target, and can realize end-to-end supervised neural network model training through a mining energy loss function constructed by a Lagrange mechanical loss function, an energy conservation loss function and regularization loss. And (3) calculating a function value according to the results of the step (3) and the step (4), and optimizing the Bi-LSTM neural network by taking the minimum value of the loss function as an optimization target in the training of the neural network model.
Step 6: and calculating the mining resistance by combining the Bi-LSTM neural network and the Balovnev resistance calculation model with the sample data, and obtaining a mining resistance prediction model loss function according to the resistance calculation result. And solving the minimum value of the loss function of the excavation resistance prediction model by a gradient descent method, wherein the Bi-LSTM neural network obtains the optimal model parameters. Model training time and selectionThe samples are used as a training set and,As a set of verification samples,The test set is set as 200, and an Adam optimizer is selected for model training to obtain an optimized neural network model.
Step 7: and (3) the Bi-LSTM neural network model and the Balovnev resistance calculation model which are trained in the step (6) jointly form an excavation resistance prediction model. The model takes basic information such as mining electric shovel structure information, excavation track, material pile surface, material viscosity and the like as input, and predicts the excavation resistance of the mining electric shovel at different moments.
Because the excavating resistance has complex composition, the specific resistance value is difficult to measure, the actual excavating resistance is difficult to compare and verify, and the actual lifting force and the actual pushing force of the electric shovel are convenient to obtain, the invention adopts an indirect comparison mode. The method comprises the steps of obtaining corresponding lifting moment and pushing moment by actual excavation of an electric shovel as comparison data, introducing predicted excavation resistance into an electric shovel dynamics model to calculate corresponding lifting moment and pushing moment as experimental data, and comparing the lifting moment and the pushing moment, if the change trend of the lifting moment and the pushing moment is similar, indicating that the excavation resistance encountered by the electric shovel in the excavation process is similar, and proving the effectiveness of the excavation resistance prediction model. Figures 2 and 3 show the excavating resistance predicted by the method of the invention as input, and the excavating resistance predicted by the method is similar to the actual excavating resistance according to the excavating force change curve and the actual excavating force change curve calculated by the dynamic model of the front-end working mechanism, thus proving the accuracy.
The examples described above represent only embodiments of the invention and are not to be understood as limiting the scope of the patent of the invention, it being pointed out that several variants and modifications may be made by those skilled in the art without departing from the concept of the invention, which fall within the scope of protection of the invention.
Claims (7)
1. The modeling method for the dynamic mining resistance of the unmanned mining electric shovel based on the data-model combined driving is characterized by comprising the following steps of:
step 1, acquiring historical excavation data of an unmanned mining electric shovel;
step 2, calculating excavation depth information corresponding to different material levels;
step 3, calculating cutting resistance according to Balovnev resistance calculation models;
step 4, constructing a Bi-LSTM neural network model to predict radial resistance; the step 4 specifically comprises the following steps:
Constructing a Bi-LSTM neural network model by combining the excavation information and the cutting force calculation result in the step 3, and establishing radial resistance parameterized expression generated by pressing the bucket into the material by means of the neural network model to realize prediction of dynamic radial resistance; the input of the Bi-LSTM neural network model comprises parameters related to an excavation strategy, such as an excavation track, an excavation speed, an acceleration, an excavation depth and the like in the step 1 and the step 2, and the output is radial resistance caused by pressing the bucket into the material;
/>
Wherein, Representation/>The input of the time neural network unit is the parameter related to the mining strategy; /(I)Representation/>Radial resistance predicted at time; /(I)Representation/>Time pair/>And/>Is a preliminary feature value of (a); /(I)Representation/>Time pair/>And/>Is a characteristic value of (2); Representing an activation function; /(I) Representing the Hadamard product; /(I)，/>，/>，/>Respectively represent forgetting gate, input gate, output gate and characteristic extraction process/>Weight coefficient of (2); /(I)，/>，/>，/>Respectively represent forgetting gate, input gate, output gate and characteristic extraction process/>Weight coefficient of (2); /(I)For/>Forget the value of the door at the moment; /(I)For/>Inputting the value of the gate at the moment; /(I)For/>Outputting the value of the door at the moment; /(I)For/>Radial resistance predicted at time;
Step 5, constructing a mining energy loss function;
Step 6, training the Bi-LSTM neural network based on the mining data until the mining energy loss function reaches the minimum value, so as to obtain a Bi-LSTM network model;
and 7, forming an excavation resistance prediction model by the Bi-LSTM network model and the Balovnev resistance calculation model.
2. The method for modeling the dynamic excavating resistance of the unmanned mining electric shovel based on the data-model combined driving according to claim 1, wherein the historical excavating data in the step 1 comprises the following steps: parameters related to the excavation strategy include an excavation trajectory, an excavation speed, an acceleration, and an excavation time, and mechanism parameters related to the electric shovel include a crown wheel radius, a bucket width, an arm length, and a boom weight.
3. The method for modeling the dynamic excavating resistance of the unmanned mining electric shovel based on the data-model combined driving according to claim 1, wherein the step 2 is specifically as follows:
In the modeling process of excavating resistance, introducing dynamic excavating depth information according to the shape of a piling surface and an excavating track, and adding an excavated material weight item and an additional resistance item brought by excavating speed in a bucket; calculating the excavation depth according to the surface information of the material surface and the excavation track, and the weight of the material in the bucket at each moment;
In the digging process, digging depth information at different moments Can change along with the change of the shape of the material surface and the excavation track; when the bucket is inserted into the material, the value is the vertical distance between the tooth tip of the bucket and the surface of the material; when the material is separated, the value is/>; Dig depth/>According to formula (1):
Wherein, Representing a charge level function; /(I)Representing an excavation trajectory function; /(I)Represents a distance in the horizontal direction; /(I)AndIndicating the horizontal distance the bucket begins to insert into the material and the horizontal distance away from the material, respectively;
The weight of the excavated material in the bucket For the product of the excavated volume and the density, as shown in formula (2):
Wherein, Representing the density of the material; /(I)Representing a bucket width; /(I)Represents the excavation volume, which is the excavation depth/>With respect to horizontal distance/>Is a function of the integral of (a).
4. The method for modeling the dynamic excavating resistance of the unmanned mining electric shovel based on the data-model combined driving according to claim 1, wherein the step 3 is specifically as follows:
when the mining electric shovel excavates the inclined plane stacker, according to the material plane failure surface hypothesis theory, the excavated material in front of the shovel is regarded as an approximate wedge; according to the mechanical characteristics of wedge-shaped soil bodies, the basic material information, mining electric shovel mechanism parameters and dynamic excavation depth information in the step 1 and the step 2 are introduced, and the excavation resistance of the bucket is established In the form of:
Wherein, The gravity of the bucket is a constant value in the process of excavating, and the gravity of the material is gradually increased and is obtained by a formula (2); /(I)For cutting resistance, the tangential movement of the bucket is mainly blocked, and is obtained by a Balovnev resistance calculation model; /(I)Radial resistance generated by extruding the matrix material by the bucket in the excavating process is provided; /(I)The direction of friction between the bottom surface of the bucket and the base material caused by normal resistance is parallel to the direction of motion of the bucket, wherein the normal resistance comprises the radial resistance and the normal component of cutting resistance generated by the extrusion of the bucket by the material, according to/>And/>Is calculated.
5. The method for modeling the dynamic excavating resistance of the unmanned mining electric shovel based on the data-model combined driving according to claim 1, wherein the step 5 is specifically as follows:
Constructing a mining energy loss function, wherein the loss function consists of a Lagrangian mechanical loss function, an energy conservation loss function and a regularization loss function, as shown in a formula (5); the Lagrange mechanical loss function is built according to the minimized Lagrange mechanics and is used for describing excavation force loss in the excavation process; the energy conservation loss function is built by solving kinetic energy and potential energy time change rate through a central differential filter, is used for describing energy change of the mining electric shovel system in the excavating operation, ensures that the total energy of the mining electric shovel system is equal at each moment of operation, and realizes the minimum energy loss in the prediction process by using the loss function; regularized loss function adoption Regularization;
Wherein, Indicating a bucket digging angle; /(I)、/>And/>Is a super parameter, and determines the weight occupied by each item in the optimized objective function; /(I)Representing a lagrangian mechanical loss function; /(I)Representing an energy conservation loss function; /(I)Representing regularization loss.
6. The method for modeling the dynamic excavating resistance of the unmanned mining electric shovel based on the data-model combined driving according to claim 1, wherein the step 6 is specifically as follows:
Calculating the excavation energy loss function according to the cutting resistance calculated by the Balovnev resistance calculation model in the step 3 and the radial resistance predicted by the neural network in the step 4, and solving the optimization problem shown in the formula (6) by a gradient descent method and combining the calculation result of the loss function to obtain Bi-LSTM neural network model parameters; the termination condition is that the output result of the loss function is minimum, and when the termination condition is met, training is completed, bi-LSTM neural network model parameters are obtained, and then the Bi-LSTM neural network model is obtained;
。
7. the method for modeling the dynamic excavating resistance of the unmanned mining electric shovel based on the data-model combined driving according to claim 1, wherein the step 7 is specifically as follows:
The trained Bi-LSTM neural network model and Balovnev resistance calculation model jointly form an excavation resistance prediction model; the excavation resistance prediction model takes basic information such as mining structure information, excavation track, material stacking surface, material viscosity and the like as input, and can rapidly and accurately predict the excavation resistance at different moments in the excavation process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410627598.0A CN118211495B (en) | 2024-05-21 | 2024-05-21 | Unmanned mining electric shovel dynamic excavation resistance modeling method based on data-model combined driving |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410627598.0A CN118211495B (en) | 2024-05-21 | 2024-05-21 | Unmanned mining electric shovel dynamic excavation resistance modeling method based on data-model combined driving |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118211495A true CN118211495A (en) | 2024-06-18 |
| CN118211495B CN118211495B (en) | 2024-07-30 |
Family
ID=91452756
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410627598.0A Active CN118211495B (en) | 2024-05-21 | 2024-05-21 | Unmanned mining electric shovel dynamic excavation resistance modeling method based on data-model combined driving |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118211495B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106836364A (en) * | 2017-01-17 | 2017-06-13 | 大连理工大学 | The automatic control system and optimal trajectory planning method of intelligent excavator |
| CN112836404A (en) * | 2021-01-07 | 2021-05-25 | 大连理工大学 | Method for constructing digital twin body of structural performance of intelligent excavator |
| CN113849910A (en) * | 2021-09-23 | 2021-12-28 | 四川大学 | Dropout-based BiLSTM network wing resistance coefficient prediction method |
| CN116933441A (en) * | 2023-08-15 | 2023-10-24 | 厦门大学 | Digital twin data driving model modeling method for autonomous shoveling process of loader |
| WO2024078003A1 (en) * | 2022-10-10 | 2024-04-18 | 华为云计算技术有限公司 | Simulation training method and apparatus, and computing device cluster |
-
2024
- 2024-05-21 CN CN202410627598.0A patent/CN118211495B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106836364A (en) * | 2017-01-17 | 2017-06-13 | 大连理工大学 | The automatic control system and optimal trajectory planning method of intelligent excavator |
| CN112836404A (en) * | 2021-01-07 | 2021-05-25 | 大连理工大学 | Method for constructing digital twin body of structural performance of intelligent excavator |
| CN113849910A (en) * | 2021-09-23 | 2021-12-28 | 四川大学 | Dropout-based BiLSTM network wing resistance coefficient prediction method |
| WO2024078003A1 (en) * | 2022-10-10 | 2024-04-18 | 华为云计算技术有限公司 | Simulation training method and apparatus, and computing device cluster |
| CN116933441A (en) * | 2023-08-15 | 2023-10-24 | 厦门大学 | Digital twin data driving model modeling method for autonomous shoveling process of loader |
Non-Patent Citations (3)
| Title |
|---|
| DAMIANA CATANOSO 等: "Analysis of Sample Acquisition Dynamics Using Discrete Element Method", 《IEEE2020》, 31 December 2020 (2020-12-31) * |
| YUNHAO CUI 等: "Memory-Augmented Point Cloud Registration Network for Bucket Pose Estimation of the Intelligent Mining Excavator", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》, 31 December 2022 (2022-12-31) * |
| 冯一飞 等: "东水泉煤矿围岩变形智能感知技术研究", 《绿色科技》, 31 December 2023 (2023-12-31) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118211495B (en) | 2024-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jong et al. | State-of-the-art review of geotechnical-driven artificial intelligence techniques in underground soil-structure interaction | |
| CN104769396B (en) | The method and system of the load weight of exploitation excavating equipment is determined for trying to be the first | |
| Ebid | 35 Years of (AI) in geotechnical engineering: state of the art | |
| Patel et al. | Soil-tool interaction as a review for digging operation of mini hydraulic excavator | |
| Awuah-Offei et al. | Cable shovel digging optimization for energy efficiency | |
| CN109978226A (en) | Shield construction ground settlement prediction method based on Recognition with Recurrent Neural Network | |
| WO2005103396A1 (en) | Method of estimating parameters of a medium to be moved by a digging apparatus | |
| Huo et al. | Intelligent prediction for digging load of hydraulic excavators based on RBF neural network | |
| CN116360425A (en) | Unmanned excavator autonomous mining excavation track planning method based on digital twin | |
| CN114722586A (en) | Intelligent electric shovel optimal automatic excavation operation method based on stockpile face prediction model | |
| Zhao et al. | A novel method of soil parameter identification and force prediction for automatic excavation | |
| CN111119902A (en) | Tunnel dynamic construction method based on BP neural network | |
| Ur Rehman et al. | Understanding how speed, tractive effort, digging height, and rake angle affect bucket penetration and resistive forces for rubber tire loaders | |
| CN118211495B (en) | Unmanned mining electric shovel dynamic excavation resistance modeling method based on data-model combined driving | |
| Huo et al. | Chaos-based support vector regression for load power forecasting of excavators | |
| Skonieczny | Lightweight robotic excavation | |
| Reddy | A machine learning framework for predicting displacements due to deep excavations and tunnels | |
| Chen et al. | Shovel parameter sensitivity analysis and online optimization method for unmanned loaders | |
| Rasimarzabadi et al. | Particle flow mechanism into cable shovel dippers | |
| Rehman | Using similitude theory and discrete element modeling to understand the effects of digging parameters on excavation performance for rubber tire loaders | |
| CN115030250B (en) | Resistance prediction method and device for shovel loading operation of loader | |
| Iakupov et al. | Working parts of means of minerals excavation | |
| Ur Rehman et al. | Effectiveness of similitude theory for bucket design and analysis for rubber tire loaders | |
| Somua-Gyimah | Dragline excavation simulation, real-time terrain recognition and object detection | |
| Kashizadeh | Theoretical and experimental analysis of the cutting process in sand |
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 |