CN110577091B - Method, system and medium for stabilizing quality of blended ore based on artificial intelligence - Google Patents

Method, system and medium for stabilizing quality of blended ore based on artificial intelligence Download PDF

Info

Publication number
CN110577091B
CN110577091B CN201910267375.7A CN201910267375A CN110577091B CN 110577091 B CN110577091 B CN 110577091B CN 201910267375 A CN201910267375 A CN 201910267375A CN 110577091 B CN110577091 B CN 110577091B
Authority
CN
China
Prior art keywords
content
representing
discharging
component content
neuron
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910267375.7A
Other languages
Chinese (zh)
Other versions
CN110577091A (en
Inventor
董宁
顾奕华
李刚
张慧峰
马力
尹川
林玉华
倪伟菊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Baoshan Iron and Steel Co Ltd
Shanghai Baosight Software Co Ltd
Original Assignee
Baoshan Iron and Steel Co Ltd
Shanghai Baosight Software Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Baoshan Iron and Steel Co Ltd, Shanghai Baosight Software Co Ltd filed Critical Baoshan Iron and Steel Co Ltd
Priority to CN201910267375.7A priority Critical patent/CN110577091B/en
Publication of CN110577091A publication Critical patent/CN110577091A/en
Application granted granted Critical
Publication of CN110577091B publication Critical patent/CN110577091B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
    • B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
    • B65G43/00—Control devices, e.g. for safety, warning or fault-correcting
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
    • B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
    • B65G65/00—Loading or unloading
    • B65G65/30—Methods or devices for filling or emptying bunkers, hoppers, tanks, or like containers, of interest apart from their use in particular chemical or physical processes or their application in particular machines, e.g. not covered by a single other subclass
    • B65G65/32—Filling devices
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
    • B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
    • B65G69/00—Auxiliary measures taken, or devices used, in connection with loading or unloading
    • B65G69/04—Spreading out the materials conveyed over the whole surface to be loaded; Trimming heaps of loose materials
    • C—CHEMISTRY; METALLURGY
    • C22—METALLURGY; FERROUS OR NON-FERROUS ALLOYS; TREATMENT OF ALLOYS OR NON-FERROUS METALS
    • C22B—PRODUCTION AND REFINING OF METALS; PRETREATMENT OF RAW MATERIALS
    • C22B1/00—Preliminary treatment of ores or scrap
    • 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

Abstract

The invention provides a method, a system and a medium for stabilizing quality of a uniformly mixed ore based on artificial intelligence, which comprises the following steps: a target rate obtaining step: acquiring a current blending stacking plan, calculating and acquiring a total target speed, a target component content of the material and a discharging speed of a discharging groove according to the acquired blending stacking plan, and calculating the current component content of the material according to the discharging speed of the discharging groove; material content judging step: judging whether the current component content of the material meets the target component content of the material or not according to the obtained target component content of the material and the current component content of the material: if yes, returning to the target rate obtaining step to continue execution; otherwise, entering a slot rate optimization step to continue execution. According to the method for adjusting the cutting rate of each groove on line through the neural network, the stability of the cut material content is intelligently controlled, the cost of frequent trial and error of manpower is reduced, and the system is more intelligent.

Description

Method, system and medium for stabilizing quality of blended ore based on artificial intelligence
Technical Field
The invention relates to the technical field of steel manufacturing, in particular to a method, a system and a medium for stabilizing quality of uniformly mixed ore based on artificial intelligence.
Background
In the field of steel manufacturing industry, the blending technology of a raw material factory is always a core technology, qualified finished product blended ore is blended, the material is mainly prepared in a mode of manually combining computer-aided calculation, the cutting rate of a disk feeder is adjusted, the output finished product blended ore and the content of TFe reach qualified indexes by a multi-trial method, although the method is implemented for many years, actual needs can be basically met, a great amount of labor is consumed to try to adjust the rate, and the qualified finished product blended ore can be blended more quickly by depending on years of experience. Because most raw material plants are open-air, the contents of finished product blending ore, TFe and the like are deviated due to the influence of external factors such as climate, temperature or misoperation of operators, so that operators are required to readjust the speed to maintain the quality of the finished product blending ore, and the quality deviation is large and even the loss of the finished product blending ore is caused sometimes.
The qualified finished product blended ore is mixed in a raw material factory, the material is mainly mixed in a manner of manually combining with computer-aided calculation, the cutting rate of a disc feeder is adjusted, and the contents of the output finished product blended ore and TFe reach the qualified indexes by a multi-trial method.
An intelligent mixing system is developed, and the cutting rate of the disc feeder is adjusted on line in real time by using an artificial intelligence method to stabilize the material quality, so that the cost is reduced, and the mixing effect and efficiency are improved
Patent document CN104649036B (application number: 201410811600.6) discloses a stacking method for improving the stability of blending material, which makes a stacking plan according to the relationship between the total number M of blending varieties and the number N of blending bins of blending ore, and the blending capacity V of the blending bins: the method comprises the steps of setting m BLOCKs of grouped ingredients according to total stockpile amount A and participated ingredients N, when a certain BLOCK has more ingredients and less ingredient bins, selecting the ingredients according to a near-Si principle to realize that two or more ingredients are fed in the same bin in a single BLOCK, namely the SiO2 contents of the two or more selected ingredients are similar, dividing the single BLOCK into a plurality of sections, sequentially feeding the ingredients into the bins according to the planned amount of pre-proportioning, and sequentially cutting the ingredients out to realize that two or more ingredients are fed in the same bin in the single BLOCK to ensure the ingredient stability of the blended ore ingredients.
Patent document CN1299051A (application No. 00100426.3) discloses an intelligent stacking control method for a blend mine, which includes the steps of making a rough plan according to a stacking plan table and an expected component table, calculating the optimum values of the variety and quantity of each raw material according to the experience and the fuzzy comprehensive judgment principle to obtain the rough plan table, dynamically allocating the cut quantity of each ingredient tank, that is, calculating the contents of specific components such as SiO2 and TFe in the stacked blend mine according to the components and quantity of the raw material output from a quantitative feeder, and adjusting the cutting speed of the quantitative feeder according to the values
Patent document CN104561411A (application number: 201510034618.4) discloses a blending method capable of effectively improving the quality of blended ore, which is characterized by comprising a raw material preparation step of pre-blending and stacking small ore species and peripheral blended ore; the number of the prepared material mixing bins is equal to the number of material stacking layers/stacking layers + X, wherein X is more than or equal to 0; according to the priority of the component difference: SiO2 > TFe > moisture > granularity, the raw material loading step of selecting the batching ore species with big component difference to load in two adjacent bins; under the condition of meeting production requirements, the stacking and feeding amount approaches to a feeding amount lower limit value, wherein: and (4) stacking the materials with the lower limit value of the feeding amount being equal to the lower limit value of the measuring range of the proportioning bins multiplied by the number of the proportioning bins.
The above patents are all groove and ingredient methods, and the invention relates to an artificial intelligence speed regulation method.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a method, a system and a medium for stabilizing the quality of a uniformly mixed ore based on artificial intelligence.
The method for stabilizing the quality of the blended ore based on artificial intelligence provided by the invention comprises the following steps:
a target rate obtaining step: acquiring a current blending stacking plan, calculating and acquiring a total target speed, a target component content of the material and a discharging speed of a discharging groove according to the acquired blending stacking plan, and calculating the current component content of the material according to the discharging speed of the discharging groove;
material content judging step: judging whether the current component content of the material meets the target component content of the material or not according to the obtained target component content of the material and the current component content of the material: if yes, returning to the target rate obtaining step to continue execution; otherwise, entering a groove rate optimization step to continue execution;
a groove speed optimizing step: calculating the optimized discharging rate of each discharging groove by using a neural network algorithm according to the obtained target component content of the material and the current component content of the material;
a groove speed adjusting step: and adjusting the discharging speed of each discharging groove according to the obtained optimized discharging speed and the total target speed of each discharging groove.
Preferably, the target rate acquiring step:
the current blending windrow plan comprises: the method comprises the following steps of (1) stacking planning time table, raw material distribution usage table, raw material composition table, dry weight, water content percentage, total blending stacking time, residual discharging time of a discharging groove, sum of residual moisture contents of all kinds of materials in the discharging groove at each moment and component content of each kind of material;
according to the current blending and stacking plan, obtaining the dry weight and the water content percentage of each variety of material, and calculating to obtain the moisture content of each variety of material, wherein the calculation formula is as follows:
Figure GDA0003046644480000031
obtaining total blending and stacking time according to the current blending and stacking plan, and calculating to obtain a total target speed according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000032
obtaining the target component content of each variety of materials according to the current blending and stacking plan, and then calculating the target component content of the obtained materials according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000033
moisture content of variety material
According to the current blending and stacking plan, obtaining the residual moisture content of all kinds of materials in the discharge chute and the residual time of the discharge chute, and calculating the discharge rate of the discharge chute, wherein the calculation formula is as follows:
Figure GDA0003046644480000034
calculating to obtain the current component content of the material according to the obtained discharge rate of the discharge chute, wherein the calculation formula is as follows:
current ingredient content of the material
The ingredient content of the variety of material being cut out in the discharge chute 1 is multiplied by the discharge rate of the discharge chute 1
+ the contents of the ingredients of the variety material being cut out by the discharge chute 2 x the discharge rate of the discharge chute 2 + …
+ the component content of the material of the variety being cut by the discharge chute n x the discharge rate of the discharge chute n
Wherein the content of the first and second substances,
n represents the number of discharge chutes.
Preferably, the tank rate optimizing step:
acquiring a speed change value of the discharge chute: and calculating the optimized discharging rate of each discharging groove by taking the obtained target component content of the material, the current component content of the material and the deviation of the preset component content as the input of a neural network algorithm, wherein the calculation process is as follows:
the input of the jth neuron of the hidden layer of the artificial neural network is as follows:
Figure GDA0003046644480000041
Figure GDA0003046644480000042
wherein the content of the first and second substances,
m is the number of neurons of the input layer;
Figure GDA0003046644480000043
representing the jth neuron of the input layer of the pth training sample, and respectively inputting the obtained target component content of the material, the current component content of the material and the deviation of the preset component content into the neuron
Figure GDA0003046644480000044
Figure GDA0003046644480000045
Representing the input of the ith neuron of the hidden layer under the action of the p training sample;
ωjirepresenting the weight between the input layer and the hidden layer;
Figure GDA0003046644480000046
representing the output of the jth neuron of the hidden layer under the action of the pth training sample;
Figure GDA0003046644480000047
representing the output of the ith neuron of the hidden layer under the action of the p training sample;
θia threshold representing hidden layer neuron i;
g (x) is a non-linear mapping function for hidden layer neurons, comprising: a Sigmoid function;
the input of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000048
wherein the content of the first and second substances,
Figure GDA0003046644480000049
representing the input of the kth neuron of the output layer under the action of the pth training sample;
ωkirepresenting a weight coefficient between the output layer and the hidden layer;
θka threshold value representing output layer neuron k;
q is the number of neurons in the hidden layer;
the output of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA00030466444800000410
wherein the content of the first and second substances,
Figure GDA00030466444800000411
representing the output of the kth neuron of the output layer, namely the speed change value of the discharge chute;
output layer activation function
Figure GDA00030466444800000412
The derivative function of (d) is:
Figure GDA0003046644480000051
the quadratic error function of the input pattern pair for the p-th training sample, i.e. the performance index, is:
Figure GDA0003046644480000052
wherein the content of the first and second substances,
Jprepresenting a performance index;
Figure GDA0003046644480000053
representing a preset target output value;
determination of Performance index JpWhether the standard meets the preset standard or not: if yes, entering a groove speed adjusting step to continue execution; otherwise, entering the performance index adjusting step to continue execution.
Preferably, the performance index adjusting step:
press error function JpAnd reducing the fastest direction adjustment weighting coefficient, and utilizing a gradient descent method until a satisfactory weighting coefficient set is obtained, wherein the adjustment process is as follows:
Figure GDA0003046644480000054
wherein the content of the first and second substances,
η represents the learning rate;
Δωkirepresents an adjustment increment of the weight coefficient;
Figure GDA0003046644480000055
expressing a performance index derivation;
Figure GDA0003046644480000056
the derivation of the weight coefficients representing the hidden layer and the output layer;
Figure GDA0003046644480000057
representing the output derivative to the output layer;
therefore, the weighting coefficient modification formula of any neuron k of the output layer is:
Figure GDA0003046644480000058
Figure GDA0003046644480000059
wherein the content of the first and second substances,
Figure GDA00030466444800000510
an intermediate variable representing an output layer;
the weight variable of the available hidden layer is adjusted as follows:
Figure GDA0003046644480000061
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
Figure GDA0003046644480000062
an intermediate variable representing an input layer;
the weighting coefficient improvement formula of any neuron k of the output layer when the p training sample acts is as follows:
Figure GDA0003046644480000063
Δωkirepresenting the weight coefficient increment between the adjusted output layer and the hidden layer;
k represents a neuron number of the output layer;
the weighting coefficient improvement formula of any neuron k of the hidden layer when the p training sample acts is as follows:
Figure GDA0003046644480000064
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
the learning process adjusts the weighting coefficient according to the direction which enables the error function J to reduce the fastest, and the weighting coefficient increment formula when all samples of any neuron k and i of the output layer and the hidden layer act can be obtained:
Figure GDA0003046644480000065
Figure GDA0003046644480000066
according to the obtained weighting coefficient increment, the weighting coefficient omega is subjected tokiAnd ωjiThe return slot rate optimization step continues with the modification.
Preferably, the tank rate adjusting step:
according to the obtained total target rate and the output of the k-th neuron of the output layer
Figure GDA0003046644480000067
Adjusting the speed of each discharge chute, adding the adjusted speeds of the discharge chutes to obtain the sum of the speeds of the discharge chutes after the neural network optimization, and calculating to obtain the final target speed of the discharge chute k, wherein the calculation mode is as follows:
Figure GDA0003046644480000068
the final target speed of the discharge chute k is equal to the target speed of the chute k after the optimization of the Muxneural network, and k is equal to 1, 2, … …, n;
and adjusting the speed of each discharge chute according to the obtained final target speed of the discharge chute k.
The system for stabilizing the quality of the blended ore based on artificial intelligence can be realized through the steps and the flows of the method for stabilizing the quality of the blended ore based on artificial intelligence. The method for stabilizing the quality of the blended ore based on artificial intelligence can be understood as a preferred example of the system for stabilizing the quality of the blended ore based on artificial intelligence by those skilled in the art.
The system for stabilizing the quality of the blended ore based on artificial intelligence provided by the invention comprises:
a target rate acquisition module: acquiring a current blending stacking plan, calculating and acquiring a total target speed, a target component content of the material and a discharging speed of a discharging groove according to the acquired blending stacking plan, and calculating the current component content of the material according to the discharging speed of the discharging groove;
material content judging module: judging whether the current component content of the material meets the target component content of the material or not according to the obtained target component content of the material and the current component content of the material: if yes, calling a target rate acquisition module; otherwise, calling a slot rate optimization module;
a tank rate optimization module: calculating the optimized discharging rate of each discharging groove by using a neural network algorithm according to the obtained target component content of the material and the current component content of the material;
a tank rate adjustment module: and adjusting the discharging speed of each discharging groove according to the obtained optimized discharging speed and the total target speed of each discharging groove.
Preferably, the target rate obtaining module:
the current blending windrow plan comprises: the method comprises the following steps of (1) stacking planning time table, raw material distribution usage table, raw material composition table, dry weight, water content percentage, total blending stacking time, residual discharging time of a discharging groove, sum of residual moisture contents of all kinds of materials in the discharging groove at each moment and component content of each kind of material;
according to the current blending and stacking plan, obtaining the dry weight and the water content percentage of each variety of material, and calculating to obtain the moisture content of each variety of material, wherein the calculation formula is as follows:
Figure GDA0003046644480000071
obtaining total blending and stacking time according to the current blending and stacking plan, and calculating to obtain a total target speed according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000072
obtaining the target component content of each variety of materials according to the current blending and stacking plan, and then calculating the target component content of the obtained materials according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000073
moisture content of variety material
According to the current blending and stacking plan, obtaining the residual moisture content of all kinds of materials in the discharge chute and the residual time of the discharge chute, and calculating the discharge rate of the discharge chute, wherein the calculation formula is as follows:
Figure GDA0003046644480000081
calculating to obtain the current component content of the material according to the obtained discharge rate of the discharge chute, wherein the calculation formula is as follows:
current ingredient content of the material
The ingredient content of the variety of material being cut out in the discharge chute 1 is multiplied by the discharge rate of the discharge chute 1
+ the contents of the ingredients of the variety material being cut out by the discharge chute 2 x the discharge rate of the discharge chute 2 + …
+ the component content of the material of the variety being cut by the discharge chute n x the discharge rate of the discharge chute n
Wherein the content of the first and second substances,
n represents the number of discharge chutes.
Preferably, the slot rate optimization module:
the speed change value of the discharge chute obtains the module: and calculating the optimized discharging rate of each discharging groove by taking the obtained target component content of the material, the current component content of the material and the deviation of the preset component content as the input of a neural network algorithm, wherein the calculation process is as follows:
the input of the jth neuron of the hidden layer of the artificial neural network is as follows:
Figure GDA0003046644480000082
Figure GDA0003046644480000083
wherein the content of the first and second substances,
m is the number of neurons of the input layer;
Figure GDA0003046644480000084
representing the jth neuron of the input layer of the pth training sample, and respectively inputting the obtained target component content of the material, the current component content of the material and the deviation of the preset component content into the neuron
Figure GDA0003046644480000085
Figure GDA0003046644480000086
Representing the input of the ith neuron of the hidden layer under the action of the p training sample;
ωjirepresenting between input and hidden layersA weight value;
Figure GDA0003046644480000087
representing the output of the jth neuron of the hidden layer under the action of the pth training sample;
Figure GDA0003046644480000088
representing the output of the ith neuron of the hidden layer under the action of the p training sample;
θia threshold representing hidden layer neuron i;
g (x) is a non-linear mapping function for hidden layer neurons, comprising: a Sigmoid function;
the input of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000091
wherein the content of the first and second substances,
Figure GDA0003046644480000092
representing the input of the kth neuron of the output layer under the action of the pth training sample;
ωkirepresenting a weight coefficient between the output layer and the hidden layer;
θka threshold value representing output layer neuron k;
q is the number of neurons in the hidden layer;
the output of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000093
wherein the content of the first and second substances,
Figure GDA0003046644480000094
representing the output of the kth neuron of the output layer, namely the speed change value of the discharge chute;
output layer activation function
Figure GDA0003046644480000095
The derivative function of (d) is:
Figure GDA0003046644480000096
the quadratic error function of the input pattern pair for the p-th training sample, i.e. the performance index, is:
Figure GDA0003046644480000097
wherein the content of the first and second substances,
Jprepresenting a performance index;
Figure GDA0003046644480000098
representing a preset target output value;
determination of Performance index JpWhether the standard meets the preset standard or not: if yes, calling a slot rate adjusting module; otherwise, the performance index adjusting module is called.
Preferably, the performance index adjustment module:
press error function JpAnd reducing the fastest direction adjustment weighting coefficient, and utilizing a gradient descent method until a satisfactory weighting coefficient set is obtained, wherein the adjustment process is as follows:
Figure GDA0003046644480000099
wherein the content of the first and second substances,
η represents the learning rate;
Δωkirepresents an adjustment increment of the weight coefficient;
Figure GDA0003046644480000101
expressing a performance index derivation;
Figure GDA0003046644480000102
the derivation of the weight coefficients representing the hidden layer and the output layer;
Figure GDA0003046644480000103
representing the output derivative to the output layer;
therefore, the weighting coefficient modification formula of any neuron k of the output layer is:
Figure GDA0003046644480000104
Figure GDA0003046644480000105
wherein the content of the first and second substances,
Figure GDA0003046644480000106
an intermediate variable representing an output layer;
the weight variable of the available hidden layer is adjusted as follows:
Figure GDA0003046644480000107
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
Figure GDA0003046644480000108
an intermediate variable representing an input layer;
the weighting coefficient improvement formula of any neuron k of the output layer when the p training sample acts is as follows:
Figure GDA0003046644480000109
Δωkirepresenting the weight coefficient increment between the adjusted output layer and the hidden layer;
k represents a neuron number of the output layer;
the weighting coefficient improvement formula of any neuron k of the hidden layer when the p training sample acts is as follows:
Figure GDA00030466444800001010
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
the learning process adjusts the weighting coefficient according to the direction which enables the error function J to reduce the fastest, and the weighting coefficient increment formula when all samples of any neuron k and i of the output layer and the hidden layer act can be obtained:
Figure GDA00030466444800001011
Figure GDA00030466444800001012
according to the obtained weighting coefficient increment, the weighting coefficient omega is subjected tokiAnd ωjiModifying, and calling a slot rate optimization module;
the slot rate adjustment module:
according to the obtained total target rate and the output of the k-th neuron of the output layer
Figure GDA0003046644480000111
Adjusting the speed of each discharge chute, adding the adjusted speeds of the discharge chutes to obtain the sum of the speeds of the discharge chutes after the neural network optimization, and calculating to obtain the final target speed of the discharge chute k, wherein the calculation mode is as follows:
Figure GDA0003046644480000112
the final target speed of the discharge chute k is equal to the target speed of the chute k after the optimization of the Muxneural network, and k is equal to 1, 2, … …, n;
and adjusting the speed of each discharge chute according to the obtained final target speed of the discharge chute k.
According to the invention, the computer-readable storage medium is stored with a computer program, and the computer program is characterized in that when being executed by a processor, the computer program realizes the steps of any one of the above-mentioned methods for stabilizing the quality of the blended ore based on artificial intelligence.
Compared with the prior art, the invention has the following beneficial effects:
according to the method for adjusting the cutting rate of each groove on line through the neural network, the stability of the cut material content is intelligently controlled, the cost of frequent trial and error of manpower is reduced, and the system is more intelligent. In the field operation process, the method for intelligently adjusting the cutting rate of the disk feeder is effectively utilized, the cut material quality is stable through one-year monitoring, the waste rate is reduced, the effect of intelligently adjusting the cutting rate of the disk feeder is achieved, and the cost is effectively saved for a raw material factory.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
fig. 1 is a schematic diagram of a neural network model according to preferred embodiment 1 of the present invention.
Fig. 2 is a schematic flow chart of the intelligent blending system according to preferred embodiment 1 of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
The method for stabilizing the quality of the blended ore based on artificial intelligence provided by the invention comprises the following steps:
a target rate obtaining step: acquiring a current blending stacking plan, calculating and acquiring a total target speed, a target component content of the material and a discharging speed of a discharging groove according to the acquired blending stacking plan, and calculating the current component content of the material according to the discharging speed of the discharging groove;
material content judging step: judging whether the current component content of the material meets the target component content of the material or not according to the obtained target component content of the material and the current component content of the material: if yes, returning to the target rate obtaining step to continue execution; otherwise, entering a groove rate optimization step to continue execution;
a groove speed optimizing step: calculating the optimized discharging rate of each discharging groove by using a neural network algorithm according to the obtained target component content of the material and the current component content of the material;
a groove speed adjusting step: and adjusting the discharging speed of each discharging groove according to the obtained optimized discharging speed and the total target speed of each discharging groove.
Specifically, the target rate obtaining step:
the current blending windrow plan comprises: the method comprises the following steps of (1) stacking planning time table, raw material distribution usage table, raw material composition table, dry weight, water content percentage, total blending stacking time, residual discharging time of a discharging groove, sum of residual moisture contents of all kinds of materials in the discharging groove at each moment and component content of each kind of material;
according to the current blending and stacking plan, obtaining the dry weight and the water content percentage of each variety of material, and calculating to obtain the moisture content of each variety of material, wherein the calculation formula is as follows:
Figure GDA0003046644480000121
obtaining total blending and stacking time according to the current blending and stacking plan, and calculating to obtain a total target speed according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000122
obtaining the target component content of each variety of materials according to the current blending and stacking plan, and then calculating the target component content of the obtained materials according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000123
moisture content of variety material
According to the current blending and stacking plan, obtaining the residual moisture content of all kinds of materials in the discharge chute and the residual time of the discharge chute, and calculating the discharge rate of the discharge chute, wherein the calculation formula is as follows:
Figure GDA0003046644480000124
calculating to obtain the current component content of the material according to the obtained discharge rate of the discharge chute, wherein the calculation formula is as follows:
current ingredient content of the material
The ingredient content of the variety of material being cut out in the discharge chute 1 is multiplied by the discharge rate of the discharge chute 1
+ the contents of the ingredients of the variety material being cut out by the discharge chute 2 x the discharge rate of the discharge chute 2 + …
+ the component content of the material of the variety being cut by the discharge chute n x the discharge rate of the discharge chute n
Wherein the content of the first and second substances,
n represents the number of discharge chutes.
Specifically, the tank rate optimization step:
acquiring a speed change value of the discharge chute: and calculating the optimized discharging rate of each discharging groove by taking the obtained target component content of the material, the current component content of the material and the deviation of the preset component content as the input of a neural network algorithm, wherein the calculation process is as follows:
the input of the jth neuron of the hidden layer of the artificial neural network is as follows:
Figure GDA0003046644480000131
Figure GDA0003046644480000132
wherein the content of the first and second substances,
m is the number of neurons of the input layer;
Figure GDA0003046644480000133
representing the jth neuron of the input layer of the pth training sample, and respectively inputting the obtained target component content of the material, the current component content of the material and the deviation of the preset component content into the neuron
Figure GDA0003046644480000134
Figure GDA0003046644480000135
Representing the input of the ith neuron of the hidden layer under the action of the p training sample;
ωjirepresenting the weight between the input layer and the hidden layer;
Figure GDA0003046644480000136
representing the output of the jth neuron of the hidden layer under the action of the pth training sample;
Figure GDA0003046644480000137
representing the output of the ith neuron of the hidden layer under the action of the p training sample;
θia threshold representing hidden layer neuron i;
g (x) is a non-linear mapping function for hidden layer neurons, comprising: a Sigmoid function;
the input of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000138
wherein the content of the first and second substances,
Figure GDA0003046644480000141
representing the input of the kth neuron of the output layer under the action of the pth training sample;
ωkirepresenting a weight coefficient between the output layer and the hidden layer;
θka threshold value representing output layer neuron k;
q is the number of neurons in the hidden layer;
the output of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000142
wherein the content of the first and second substances,
Figure GDA0003046644480000143
representing the output of the kth neuron of the output layer, namely the speed change value of the discharge chute;
output layer activation function
Figure GDA0003046644480000144
The derivative function of (d) is:
Figure GDA0003046644480000145
the quadratic error function of the input pattern pair for the p-th training sample, i.e. the performance index, is:
Figure GDA0003046644480000146
wherein the content of the first and second substances,
Jprepresenting a performance index;
Figure GDA0003046644480000147
representing a preset target output value;
determination of Performance index JpWhether the standard meets the preset standard or not: if yes, entering a groove speed adjusting step to continue execution; otherwise, entering the performance index adjusting step to continue execution.
Specifically, the performance index adjusting step:
press error function JpAnd reducing the fastest direction adjustment weighting coefficient, and utilizing a gradient descent method until a satisfactory weighting coefficient set is obtained, wherein the adjustment process is as follows:
Figure GDA0003046644480000148
wherein the content of the first and second substances,
η represents the learning rate;
Δωkirepresents an adjustment increment of the weight coefficient;
Figure GDA0003046644480000149
expressing a performance index derivation;
Figure GDA0003046644480000151
representing a hidden layer andderivation of the weight coefficient of the output layer;
Figure GDA0003046644480000152
representing the output derivative to the output layer;
therefore, the weighting coefficient modification formula of any neuron k of the output layer is:
Figure GDA0003046644480000153
Figure GDA0003046644480000154
wherein the content of the first and second substances,
Figure GDA0003046644480000155
an intermediate variable representing an output layer;
the weight variable of the available hidden layer is adjusted as follows:
Figure GDA0003046644480000156
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
Figure GDA0003046644480000157
an intermediate variable representing an input layer;
the weighting coefficient improvement formula of any neuron k of the output layer when the p training sample acts is as follows:
Figure GDA0003046644480000158
Δωkirepresenting the weight coefficient increment between the adjusted output layer and the hidden layer;
k represents a neuron number of the output layer;
the weighting coefficient improvement formula of any neuron k of the hidden layer when the p training sample acts is as follows:
Figure GDA0003046644480000159
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
the learning process adjusts the weighting coefficient according to the direction which enables the error function J to reduce the fastest, and the weighting coefficient increment formula when all samples of any neuron k and i of the output layer and the hidden layer act can be obtained:
Figure GDA00030466444800001510
Figure GDA00030466444800001511
according to the obtained weighting coefficient increment, the weighting coefficient omega is subjected tokiAnd ωjiThe return slot rate optimization step continues with the modification.
Specifically, the tank rate adjustment step:
according to the obtained total target rate and the output of the k-th neuron of the output layer
Figure GDA00030466444800001512
Adjusting the speed of each discharge chute, adding the adjusted speeds of the discharge chutes to obtain the sum of the speeds of the discharge chutes after the neural network optimization, and calculating to obtain the final target speed of the discharge chute k, wherein the calculation mode is as follows:
Figure GDA0003046644480000161
the final target speed of the discharge chute k is equal to the target speed of the chute k after the optimization of the Muxneural network, and k is equal to 1, 2, … …, n;
and adjusting the speed of each discharge chute according to the obtained final target speed of the discharge chute k.
The system for stabilizing the quality of the blended ore based on artificial intelligence can be realized through the steps and the flows of the method for stabilizing the quality of the blended ore based on artificial intelligence. The method for stabilizing the quality of the blended ore based on artificial intelligence can be understood as a preferred example of the system for stabilizing the quality of the blended ore based on artificial intelligence by those skilled in the art.
The system for stabilizing the quality of the blended ore based on artificial intelligence provided by the invention comprises:
a target rate acquisition module: acquiring a current blending stacking plan, calculating and acquiring a total target speed, a target component content of the material and a discharging speed of a discharging groove according to the acquired blending stacking plan, and calculating the current component content of the material according to the discharging speed of the discharging groove;
material content judging module: judging whether the current component content of the material meets the target component content of the material or not according to the obtained target component content of the material and the current component content of the material: if yes, calling a target rate acquisition module; otherwise, calling a slot rate optimization module;
a tank rate optimization module: calculating the optimized discharging rate of each discharging groove by using a neural network algorithm according to the obtained target component content of the material and the current component content of the material;
a tank rate adjustment module: and adjusting the discharging speed of each discharging groove according to the obtained optimized discharging speed and the total target speed of each discharging groove.
Specifically, the target rate obtaining module:
the current blending windrow plan comprises: the method comprises the following steps of (1) stacking planning time table, raw material distribution usage table, raw material composition table, dry weight, water content percentage, total blending stacking time, residual discharging time of a discharging groove, sum of residual moisture contents of all kinds of materials in the discharging groove at each moment and component content of each kind of material;
according to the current blending and stacking plan, obtaining the dry weight and the water content percentage of each variety of material, and calculating to obtain the moisture content of each variety of material, wherein the calculation formula is as follows:
Figure GDA0003046644480000162
obtaining total blending and stacking time according to the current blending and stacking plan, and calculating to obtain a total target speed according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000171
obtaining the target component content of each variety of materials according to the current blending and stacking plan, and then calculating the target component content of the obtained materials according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure GDA0003046644480000172
moisture content of variety material
According to the current blending and stacking plan, obtaining the residual moisture content of all kinds of materials in the discharge chute and the residual time of the discharge chute, and calculating the discharge rate of the discharge chute, wherein the calculation formula is as follows:
Figure GDA0003046644480000173
calculating to obtain the current component content of the material according to the obtained discharge rate of the discharge chute, wherein the calculation formula is as follows:
current ingredient content of the material
The ingredient content of the variety of material being cut out in the discharge chute 1 is multiplied by the discharge rate of the discharge chute 1
+ the contents of the ingredients of the variety material being cut out by the discharge chute 2 x the discharge rate of the discharge chute 2 + …
+ the component content of the material of the variety being cut by the discharge chute n x the discharge rate of the discharge chute n
Wherein the content of the first and second substances,
n represents the number of discharge chutes.
Specifically, the slot rate optimization module:
the speed change value of the discharge chute obtains the module: and calculating the optimized discharging rate of each discharging groove by taking the obtained target component content of the material, the current component content of the material and the deviation of the preset component content as the input of a neural network algorithm, wherein the calculation process is as follows:
the input of the jth neuron of the hidden layer of the artificial neural network is as follows:
Figure GDA0003046644480000174
Figure GDA0003046644480000175
wherein the content of the first and second substances,
m is the number of neurons of the input layer;
Figure GDA0003046644480000176
representing the jth neuron of the input layer of the pth training sample, and respectively inputting the obtained target component content of the material, the current component content of the material and the deviation of the preset component content into the neuron
Figure GDA0003046644480000181
Figure GDA0003046644480000182
Representing the input of the ith neuron of the hidden layer under the action of the p training sample;
ωjirepresenting the weight between the input layer and the hidden layer;
Figure GDA0003046644480000183
representing the output of the jth neuron of the hidden layer under the action of the pth training sample;
Figure GDA0003046644480000184
representing the output of the ith neuron of the hidden layer under the action of the p training sample;
θia threshold representing hidden layer neuron i;
g (x) is a non-linear mapping function for hidden layer neurons, comprising: a Sigmoid function;
the input of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000185
wherein the content of the first and second substances,
Figure GDA0003046644480000186
representing the input of the kth neuron of the output layer under the action of the pth training sample;
ωkirepresenting a weight coefficient between the output layer and the hidden layer;
θka threshold value representing output layer neuron k;
q is the number of neurons in the hidden layer;
the output of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000187
wherein the content of the first and second substances,
Figure GDA0003046644480000188
representing the output of the kth neuron of the output layer, namely the speed change value of the discharge chute;
output layer activation function
Figure GDA0003046644480000189
The derivative function of (d) is:
Figure GDA00030466444800001810
the quadratic error function of the input pattern pair for the p-th training sample, i.e. the performance index, is:
Figure GDA00030466444800001811
wherein the content of the first and second substances,
Jprepresenting a performance index;
Figure GDA00030466444800001812
representing a preset target output value;
determination of Performance index JpWhether the standard meets the preset standard or not: if yes, calling a slot rate adjusting module; otherwise, the performance index adjusting module is called.
Specifically, the performance index adjustment module:
press error function JpAnd reducing the fastest direction adjustment weighting coefficient, and utilizing a gradient descent method until a satisfactory weighting coefficient set is obtained, wherein the adjustment process is as follows:
Figure GDA0003046644480000191
wherein the content of the first and second substances,
η represents the learning rate;
Δωkirepresents an adjustment increment of the weight coefficient;
Figure GDA0003046644480000192
expressing a performance index derivation;
Figure GDA0003046644480000193
the derivation of the weight coefficients representing the hidden layer and the output layer;
Figure GDA0003046644480000194
representing the output derivative to the output layer;
therefore, the weighting coefficient modification formula of any neuron k of the output layer is:
Figure GDA0003046644480000195
Figure GDA0003046644480000196
wherein the content of the first and second substances,
Figure GDA0003046644480000197
an intermediate variable representing an output layer;
the weight variable of the available hidden layer is adjusted as follows:
Figure GDA0003046644480000198
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
Figure GDA0003046644480000199
an intermediate variable representing an input layer;
the weighting coefficient improvement formula of any neuron k of the output layer when the p training sample acts is as follows:
Figure GDA00030466444800001910
Δωkirepresenting the weight coefficient increment between the adjusted output layer and the hidden layer;
k represents a neuron number of the output layer;
the weighting coefficient improvement formula of any neuron k of the hidden layer when the p training sample acts is as follows:
Figure GDA0003046644480000201
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
the learning process adjusts the weighting coefficient according to the direction which enables the error function J to reduce the fastest, and the weighting coefficient increment formula when all samples of any neuron k and i of the output layer and the hidden layer act can be obtained:
Figure GDA0003046644480000202
Figure GDA0003046644480000203
according to the obtained weighting coefficient increment, the weighting coefficient omega is subjected tokiAnd ωjiModifying, and calling a slot rate optimization module;
the slot rate adjustment module:
according to the obtained total target rate and the output of the k-th neuron of the output layer
Figure GDA0003046644480000204
Adjusting the speed of each discharge chute, adding the adjusted speeds of the discharge chutes to obtain the sum of the speeds of the discharge chutes after the neural network optimization, and calculating to obtain the dischargeThe final target velocity for slot k is calculated as follows:
Figure GDA0003046644480000205
the final target speed of the discharge chute k is equal to the target speed of the chute k after the optimization of the Muxneural network, and k is equal to 1, 2, … …, n;
and adjusting the speed of each discharge chute according to the obtained final target speed of the discharge chute k.
According to the invention, the computer-readable storage medium is stored with a computer program, and the computer program is characterized in that when being executed by a processor, the computer program realizes the steps of any one of the above-mentioned methods for stabilizing the quality of the blended ore based on artificial intelligence.
The present invention will be described more specifically below with reference to preferred examples.
Preferred example 1:
the invention mainly solves the technical problem of providing a method for stabilizing the material quality by adjusting the cutting-out rate of a disc feeder of a blending system based on artificial intelligence, wherein SiO in the blending process2When the content of TFe fluctuates, the method can adjust the cutting-out speed of the disc feeder on line in real time to ensure the quality of the mixed material to be stable, the adopted algorithm is a neural network algorithm, the characteristics of feedforward operation and error back propagation of a BP neural network and the approximation to any nonlinear function are utilized to ensure that the cutting-out speed of each groove reaches reasonable requirements, and SiO2When the TFe content is within the acceptable range, the cell velocities (y1 to y10) in the following graph are adjusted so that the following formula is finally established:
SiO2target content is SiO of material being cut out by groove 12Content X groove 1 Rate (y1)
+ groove 2 cutting out SiO of the material2Content x groove 2 Rate (y2) + …
+ groove 10 cutting out SiO of the material2Content x groove 10 Rate (y10)
The specific embodiment is as follows:
the blending system first loads the heap planThe large heap plan includes a heap plan time table, a raw material distribution amount table, a raw material component table, etc., and SiO to be mixed is calculated based on the plan table2Target content of TFe (in SiO)2For example, the calculation formula of TFe is the same), the total target rate of the large heap, and the like.
Figure GDA0003046644480000211
Figure GDA0003046644480000212
Figure GDA0003046644480000213
Figure GDA0003046644480000214
SiO2The current content being the SiO of the material being cut out of the groove 12Content x current rate of cell 1
+ groove 2 cutting out SiO of the material2Content x slot 2 Current Rate + …
+ groove 10 cutting out SiO of the material2Content x current rate of cell 10
And then selecting a large pile, namely starting the large pile, starting all disk feeders to run, selecting varieties to be cut for each groove in a non-full-automatic control mode, generally more than half of the grooves, clicking 'optimization', adjusting the cutting rate of each groove on line through a neural network, or selecting a 'full-automatic mode' button in an automatic mode, cutting out the uniform mixing system according to a default sequence, and automatically starting the neural network to optimize the rate of each groove when the deviation between the current SiO2 content and the target SiO2 content is greater than an allowable deviation.
The proposed neural network is realized in a computer by C + +, and the specific method is as follows:
1. SiO22Target content、SiO2Current content, target content of TFe, current content of TFe and SiO2TFe content deviation as input end of neural network, SiO2Target content, SiO2The current content, TFe target content and TFe current content are all calculated by the formula and are calculated according to the hidden layer neuron formula
Figure GDA0003046644480000215
In the course of numerous sample training, increasing or decreasing the number of neurons in the hidden layer, when 12 neurons in the hidden layer are found, the effect is optimal, and the optimal rate of 10 slots is output at the output end of the artificial neural network control module, and the model of the neural network is shown in fig. 1.
2. Through the feedback principle, the feedback value is used as the input of control, which is the core of the automatic control theory, the intelligent blending system uses the principle, and the flow chart is as shown in the attached figure 2:
3. after the model is built, the neural network algorithm is realized:
1) carrying out normalization processing on training sample data obtained by sampling;
2) initialization weight omegajiAnd ωki，ωjiIs the weight, omega, between the input layer and the hidden layer of the artificial neural networkkiIs the weight between the hidden layer and the output layer of the artificial neural network;
the forward calculation mode is as follows:
the input of the jth neuron of the hidden layer of the artificial neural network is as follows:
Figure GDA0003046644480000221
Figure GDA0003046644480000222
wherein the content of the first and second substances,
m is the number of neurons of the input layer;
Figure GDA0003046644480000223
representing the jth neuron of the input layer of the pth training sample;
Figure GDA0003046644480000224
representing the input of the ith neuron of the hidden layer under the action of the sample P
ωjiRepresenting weights between input and hidden layers
Figure GDA0003046644480000225
Representing the output of the jth neuron of the hidden layer under the action of the sample P;
θithreshold representing hidden layer neuron i
Where g (x) is a non-linear mapping function for hidden layer neurons, such as Sigmoid function:
Figure GDA0003046644480000226
the input of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000227
wherein the content of the first and second substances,
Figure GDA0003046644480000228
represents the input of the k-th neuron of the output layer under the action of the sample P
ωkiRepresenting weights between a hidden layer and an output layer
θkRepresenting threshold values for output layer neurons i
q is the number of neurons in the hidden layer,
Figure GDA0003046644480000231
the output of the ith neuron of the hidden layer.
The output of the kth neuron of the artificial neural network output layer is as follows:
Figure GDA0003046644480000232
namely:
Figure GDA0003046644480000233
Figure GDA0003046644480000234
...
Figure GDA0003046644480000235
in the equation, because y1, y2 … … y10 are rate increment values that may be increasing or decreasing, the output layer neuron activation function takes the form of a symmetric Sigmoid function:
Figure GDA0003046644480000236
Figure GDA0003046644480000237
representing the output of the kth neuron of the output layer
Figure GDA0003046644480000238
Representing the output of the kth neuron of the output layer
Output layer activation function
Figure GDA0003046644480000239
The derivative function of (d) is:
Figure GDA00030466444800002310
the quadratic error function of the input pattern pair for each sample p, i.e. the performance index J, is:
Figure GDA00030466444800002311
JPindicating performance index
Figure GDA00030466444800002312
Representing target output
Learning process pressing error function JpAnd reducing the fastest direction adjustment weighting coefficient, and utilizing a gradient descent method until a satisfactory weighting coefficient set is obtained.
Figure GDA0003046644480000241
Wherein the content of the first and second substances,
η represents the learning rate;
Δωkirepresents an adjustment increment of the weight coefficient;
Figure GDA0003046644480000242
expressing a performance index derivation;
Figure GDA0003046644480000243
the derivation of the weight coefficients representing the hidden layer and the output layer;
Figure GDA0003046644480000244
representing the output derivative to the output layer;
ωkirepresents: weight coefficients between output layer and hidden layer
Figure GDA0003046644480000245
Represents: output of hidden layer
θkRepresents: threshold of output layer
Figure GDA0003046644480000246
Wherein the content of the first and second substances,
Figure GDA0003046644480000247
represents: to simplify the calculation, intermediate variables of the output layer are set.
Therefore, the weighting coefficient correction formula of any neuron k of the output layer is
Then finally
Figure GDA0003046644480000248
Similarly, the weight variable of the available hidden layer is adjusted as follows:
Figure GDA0003046644480000249
Δωjirepresents: adjusted weight coefficient delta between input layer and hidden layer
Figure GDA00030466444800002410
Represents: to simplify the calculation, intermediate variables of the input layer are set.
Figure GDA00030466444800002411
Represents: output of input layer
The weight coefficient improvement formula of any neuron k of the output layer when the sample p acts is as follows:
Figure GDA00030466444800002412
Δωkirepresents: weight coefficient delta between adjusted output layer and hidden layer
k represents: a certain neuron of the output layer
The weight coefficient improvement formula of any neuron k of the hidden layer when the sample p acts is as follows:
Figure GDA0003046644480000251
Δωjirepresents: adjusted weight coefficient delta between input layer and hidden layer
If the learning process adjusts the weighting coefficients in the direction that makes the error function J decrease the fastest, a similar derivation process can be used to obtain the weighting coefficient increment formula when all samples of any neuron k and i of the output layer and the hidden layer act:
Figure GDA0003046644480000252
Figure GDA0003046644480000253
4. after the neural network outputs the velocity of each slot each time, the velocity of each slot needs to be added, and the total target velocity is used for amplification or reduction, and the applied formula is as follows:
Figure GDA0003046644480000254
final target rate of slot k (k 1, 2 … … 10) optimized slot k target rate of muxneural network
5. The neural network algorithm is used for regulating the cutting-out rate of the disc feeder and is compiled by adopting a C + + development language.
Preferred example 2:
according to the algorithm of the neural network, a C + + compiling program is adopted, a C # compiling foreground picture is adopted, when the operation is carried out, a large pile is selected on the picture, namely the large pile is started, all disc feeders start to operate, and a selectable full-automatic mode is arranged on the picture. The variety to be cut out is selected for each groove in a non-full-automatic control mode, generally more than half of the groove is selected, optimization is clicked, the cut-out speed of each groove can be adjusted on line through a neural network, or in an automatic mode, namely, a full-automatic mode button is selected, the blending system can cut out the groove according to a default sequence, and when the deviation between the current content and the target content is found to be more than the allowable deviation, the neural network is automatically started to optimize the speed of each groove.
Those skilled in the art will appreciate that, in addition to implementing the systems, apparatus, and various modules thereof provided by the present invention in purely computer readable program code, the same procedures can be implemented entirely by logically programming method steps such that the systems, apparatus, and various modules thereof are provided in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system, the device and the modules thereof provided by the present invention can be considered as a hardware component, and the modules included in the system, the device and the modules thereof for implementing various programs can also be considered as structures in the hardware component; modules for performing various functions may also be considered to be both software programs for performing the methods and structures within hardware components.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.

Claims (10)

1. A method for stabilizing quality of a blended ore based on artificial intelligence is characterized by comprising the following steps:
a target rate obtaining step: acquiring a current blending stacking plan, calculating and acquiring a total target speed, a target component content of the material and a discharging speed of a discharging groove according to the acquired blending stacking plan, and calculating the current component content of the material according to the discharging speed of the discharging groove;
material content judging step: judging whether the current component content of the material meets the target component content of the material or not according to the obtained target component content of the material and the current component content of the material: if yes, returning to the target rate obtaining step to continue execution; otherwise, entering a groove rate optimization step to continue execution;
a groove speed optimizing step: calculating the optimized discharging rate of each discharging groove by using a neural network algorithm according to the obtained target component content of the material and the current component content of the material;
a groove speed adjusting step: and adjusting the discharging speed of each discharging groove according to the obtained optimized discharging speed and the total target speed of each discharging groove.
2. The artificial intelligence based method for stabilizing the quality of the blended ore according to claim 1, wherein the target rate obtaining step comprises:
the current blending windrow plan comprises: the method comprises the following steps of (1) stacking planning time table, raw material distribution usage table, raw material composition table, dry weight and water content percentage of various materials, total blending stacking time, residual discharging time of a discharging groove, sum of residual moisture contents of all the various materials in the discharging groove at each moment and component content of various materials;
according to the current blending and stacking plan, obtaining the dry weight and the water content percentage of each variety of material, and calculating to obtain the moisture content of each variety of material, wherein the calculation formula is as follows:
Figure FDA0003046644470000011
obtaining total blending and stacking time according to the current blending and stacking plan, and calculating to obtain a total target speed according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure FDA0003046644470000012
obtaining the target component content of each variety of materials according to the current blending and stacking plan, and then calculating the target component content of the obtained materials according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure FDA0003046644470000013
moisture content of variety material
According to the current blending and stacking plan, obtaining the residual moisture content of all kinds of materials in the discharge chute and the residual time of the discharge chute, and calculating the discharge rate of the discharge chute, wherein the calculation formula is as follows:
Figure FDA0003046644470000021
calculating to obtain the current component content of the material according to the obtained discharge rate of the discharge chute, wherein the calculation formula is as follows: current ingredient content of the material
The component content of the variety material being cut out by the discharging chute 1 × the discharging rate of the discharging chute 1 + the component content of the variety material being cut out by the discharging chute 2 × the discharging rate of the discharging chute 2 + … + the component content of the variety material being cut out by the discharging chute n × the discharging rate of the discharging chute n
Wherein the content of the first and second substances,
n represents the number of discharge chutes.
3. The artificial intelligence-based method for stabilizing the quality of the blended ore according to claim 2, wherein the slot velocity optimization step comprises:
acquiring a speed change value of the discharge chute: and calculating the optimized discharging rate of each discharging groove by taking the obtained target component content of the material, the current component content of the material and the deviation of the preset component content as the input of a neural network algorithm, wherein the calculation process is as follows:
the input of the jth neuron of the hidden layer of the artificial neural network is as follows:
Figure FDA0003046644470000022
Figure FDA0003046644470000023
wherein the content of the first and second substances,
m is the number of neurons of the input layer;
Figure FDA0003046644470000024
representing the jth neuron of the input layer of the pth training sample, and respectively inputting the obtained target component content of the material, the current component content of the material and the deviation of the preset component content into the neuron
Figure FDA0003046644470000025
Figure FDA0003046644470000026
Representing the input of the ith neuron of the hidden layer under the action of the p training sample;
ωjirepresenting the weight between the input layer and the hidden layer;
Figure FDA0003046644470000027
representing the output of the jth neuron of the hidden layer under the action of the pth training sample;
Figure FDA0003046644470000031
representing the output of the ith neuron of the hidden layer under the action of the p training sample;
θia threshold representing hidden layer neuron i;
g (x) is a non-linear mapping function for hidden layer neurons, comprising: a Sigmoid function;
the input of the kth neuron of the artificial neural network output layer is as follows:
Figure FDA0003046644470000032
wherein the content of the first and second substances,
Figure FDA0003046644470000033
representing the input of the kth neuron of the output layer under the action of the pth training sample;
ωkirepresenting a weight coefficient between the output layer and the hidden layer;
θka threshold value representing output layer neuron k;
q is the number of neurons in the hidden layer;
the output of the kth neuron of the artificial neural network output layer is as follows:
Figure FDA0003046644470000034
wherein the content of the first and second substances,
Figure FDA0003046644470000035
representing the output of the kth neuron of the output layer, namely the speed change value of the discharge chute;
output layer activation function
Figure FDA0003046644470000036
The derivative function of (d) is:
Figure FDA0003046644470000037
the quadratic error function of the input pattern pair for the p-th training sample, i.e. the performance index, is:
Figure FDA0003046644470000038
wherein the content of the first and second substances,
Jprepresenting a performance index;
Figure FDA0003046644470000039
representing a preset target output value;
determination of Performance index JpWhether the standard meets the preset standard or not: if yes, entering a groove speed adjusting step to continue execution; otherwise, entering the performance index adjusting step to continue execution.
4. The method for stabilizing the quality of the blended ore based on artificial intelligence according to claim 3, wherein the performance index adjusting step comprises:
press error function JpAnd reducing the fastest direction adjustment weighting coefficient, and utilizing a gradient descent method until a satisfactory weighting coefficient set is obtained, wherein the adjustment process is as follows:
Figure FDA0003046644470000041
wherein the content of the first and second substances,
η represents the learning rate;
Δωkirepresents an adjustment increment of the weight coefficient;
Figure FDA0003046644470000042
expressing a performance index derivation;
Figure FDA0003046644470000043
the derivation of the weight coefficients representing the hidden layer and the output layer;
Figure FDA0003046644470000044
representing the output derivative to the output layer;
therefore, the weighting coefficient modification formula of any neuron k of the output layer is:
Figure FDA0003046644470000045
Figure FDA0003046644470000046
wherein the content of the first and second substances,
Figure FDA0003046644470000047
an intermediate variable representing an output layer;
the weight variable of the available hidden layer is adjusted as follows:
Figure FDA0003046644470000048
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
Figure FDA0003046644470000049
an intermediate variable representing an input layer;
the weighting coefficient improvement formula of any neuron k of the output layer when the p training sample acts is as follows:
Figure FDA00030466444700000410
Δωkirepresenting the weight coefficient increment between the adjusted output layer and the hidden layer;
k represents a neuron number of the output layer;
the weighting coefficient improvement formula of any neuron k of the hidden layer when the p training sample acts is as follows:
Figure FDA00030466444700000411
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
the learning process adjusts the weighting coefficient according to the direction which enables the error function J to reduce the fastest, and the weighting coefficient increment formula when all samples of any neuron k and i of the output layer and the hidden layer act can be obtained:
Figure FDA0003046644470000051
Figure FDA0003046644470000052
according to the obtained weighting coefficient increment, the weighting coefficient omega is subjected tokiAnd ωjiThe return slot rate optimization step continues with the modification.
5. The artificial intelligence based method for stabilizing the quality of the blended ore according to claim 4, wherein the tank rate adjusting step comprises:
according to the obtained total target rate and the output of the k-th neuron of the output layer
Figure FDA0003046644470000053
Adjusting the speed of each discharge chute to adjust each discharge chuteAdding the velocities of the discharge chutes to obtain the sum of the velocities of the chutes after the neural network optimization, and calculating to obtain the final target velocity of the discharge chute k, wherein the calculation mode is as follows:
Figure FDA0003046644470000054
the final target speed of the discharge chute k is equal to the target speed of the discharge chute k after the optimization of the Muxneural network, and k is equal to 1, 2, … …, n;
and adjusting the speed of each discharge chute according to the obtained final target speed of the discharge chute k.
6. The utility model provides a system for stabilize blending ore quality based on artificial intelligence which characterized in that includes:
a target rate acquisition module: acquiring a current blending stacking plan, calculating and acquiring a total target speed, a target component content of the material and a discharging speed of a discharging groove according to the acquired blending stacking plan, and calculating the current component content of the material according to the discharging speed of the discharging groove;
material content judging module: judging whether the current component content of the material meets the target component content of the material or not according to the obtained target component content of the material and the current component content of the material: if yes, calling a target rate acquisition module; otherwise, calling a slot rate optimization module;
a tank rate optimization module: calculating the optimized discharging rate of each discharging groove by using a neural network algorithm according to the obtained target component content of the material and the current component content of the material;
a tank rate adjustment module: and adjusting the discharging speed of each discharging groove according to the obtained optimized discharging speed and the total target speed of each discharging groove.
7. The system for stabilizing quality of blended ore based on artificial intelligence of claim 6, wherein the target rate obtaining module:
the current blending windrow plan comprises: the method comprises the following steps of (1) stacking planning time table, raw material distribution usage table, raw material composition table, dry weight and water content percentage of various materials, total blending stacking time, residual discharging time of a discharging groove, sum of residual moisture contents of all the various materials in the discharging groove at each moment and component content of various materials;
according to the current blending and stacking plan, obtaining the dry weight and the water content percentage of each variety of material, and calculating to obtain the moisture content of each variety of material, wherein the calculation formula is as follows:
Figure FDA0003046644470000061
obtaining total blending and stacking time according to the current blending and stacking plan, and calculating to obtain a total target speed according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure FDA0003046644470000062
obtaining the target component content of each variety of materials according to the current blending and stacking plan, and then calculating the target component content of the obtained materials according to the obtained moisture content of each variety of materials, wherein the calculation formula is as follows:
Figure FDA0003046644470000063
moisture content of variety material
According to the current blending and stacking plan, obtaining the residual moisture content of all kinds of materials in the discharge chute and the residual time of the discharge chute, and calculating the discharge rate of the discharge chute, wherein the calculation formula is as follows:
Figure FDA0003046644470000064
calculating to obtain the current component content of the material according to the obtained discharge rate of the discharge chute, wherein the calculation formula is as follows: current ingredient content of the material
The component content of the variety material being cut out by the discharging chute 1 × the discharging rate of the discharging chute 1 + the component content of the variety material being cut out by the discharging chute 2 × the discharging rate of the discharging chute 2 + … + the component content of the variety material being cut out by the discharging chute n × the discharging rate of the discharging chute n
Wherein the content of the first and second substances,
n represents the number of discharge chutes.
8. The system for stabilizing blending ore quality based on artificial intelligence of claim 7, wherein the tank rate optimization module:
the speed change value of the discharge chute obtains the module: and calculating the optimized discharging rate of each discharging groove by taking the obtained target component content of the material, the current component content of the material and the deviation of the preset component content as the input of a neural network algorithm, wherein the calculation process is as follows:
the input of the jth neuron of the hidden layer of the artificial neural network is as follows:
Figure FDA0003046644470000071
Figure FDA0003046644470000072
wherein the content of the first and second substances,
m is the number of neurons of the input layer;
Figure FDA0003046644470000073
representing the jth neuron of the input layer of the pth training sample, and respectively inputting the obtained target component content of the material, the current component content of the material and the deviation of the preset component content into the neuron
Figure FDA0003046644470000074
Figure FDA0003046644470000075
Representing the input of the ith neuron of the hidden layer under the action of the p training sample;
ωjirepresenting the weight between the input layer and the hidden layer;
Figure FDA0003046644470000076
representing the output of the jth neuron of the hidden layer under the action of the pth training sample;
Figure FDA0003046644470000077
representing the output of the ith neuron of the hidden layer under the action of the p training sample;
θia threshold representing hidden layer neuron i;
g (x) is a non-linear mapping function for hidden layer neurons, comprising: a Sigmoid function;
the input of the kth neuron of the artificial neural network output layer is as follows:
Figure FDA0003046644470000078
wherein the content of the first and second substances,
Figure FDA0003046644470000079
representing the input of the kth neuron of the output layer under the action of the pth training sample;
ωkirepresenting a weight coefficient between the output layer and the hidden layer;
θka threshold value representing output layer neuron k;
q is the number of neurons in the hidden layer;
the output of the kth neuron of the artificial neural network output layer is as follows:
Figure FDA00030466444700000710
wherein the content of the first and second substances,
Figure FDA00030466444700000711
representing the output of the kth neuron of the output layer, namely the speed change value of the discharge chute;
output layer activation function
Figure FDA00030466444700000712
The derivative function of (d) is:
Figure FDA00030466444700000713
the quadratic error function of the input pattern pair for the p-th training sample, i.e. the performance index, is:
Figure FDA0003046644470000081
wherein the content of the first and second substances,
Jprepresenting a performance index;
Figure FDA0003046644470000082
representing a preset target output value;
determination of Performance index JpWhether the standard meets the preset standard or not: if yes, calling a slot rate adjusting module; otherwise, the performance index adjusting module is called.
9. The system for stabilizing quality of blended ore based on artificial intelligence of claim 8, wherein the performance index adjustment module:
press error function JpAnd reducing the fastest direction adjustment weighting coefficient, and utilizing a gradient descent method until a satisfactory weighting coefficient set is obtained, wherein the adjustment process is as follows:
Figure FDA0003046644470000083
wherein the content of the first and second substances,
η represents the learning rate;
Δωkirepresents an adjustment increment of the weight coefficient;
Figure FDA0003046644470000084
expressing a performance index derivation;
Figure FDA0003046644470000085
the derivation of the weight coefficients representing the hidden layer and the output layer;
Figure FDA0003046644470000086
representing the output derivative to the output layer;
therefore, the weighting coefficient modification formula of any neuron k of the output layer is:
Figure FDA0003046644470000087
Figure FDA0003046644470000088
wherein the content of the first and second substances,
Figure FDA0003046644470000089
an intermediate variable representing an output layer;
the weight variable of the available hidden layer is adjusted as follows:
Figure FDA0003046644470000091
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
Figure FDA0003046644470000092
an intermediate variable representing an input layer;
the weighting coefficient improvement formula of any neuron k of the output layer when the p training sample acts is as follows:
Figure FDA0003046644470000093
Δωkirepresenting the weight coefficient increment between the adjusted output layer and the hidden layer;
k represents a neuron number of the output layer;
the weighting coefficient improvement formula of any neuron k of the hidden layer when the p training sample acts is as follows:
Figure FDA0003046644470000094
Δωjirepresenting the weight coefficient increment between the adjusted input layer and the hidden layer;
the learning process adjusts the weighting coefficient according to the direction which enables the error function J to reduce the fastest, and the weighting coefficient increment formula when all samples of any neuron k and i of the output layer and the hidden layer act can be obtained:
Figure FDA0003046644470000095
Figure FDA0003046644470000096
according to the obtained weighting coefficient increment, the weighting coefficient omega is subjected tokiAnd ωjiModifying, and calling a slot rate optimization module;
the slot rate adjustment module:
according to the obtained total target rate and the output of the k-th neuron of the output layer
Figure FDA0003046644470000097
Adjusting the speed of each discharge chute, adding the adjusted speeds of the discharge chutes to obtain the sum of the speeds of the discharge chutes after the neural network optimization, and calculating to obtain the final target speed of the discharge chute k, wherein the calculation mode is as follows:
Figure FDA0003046644470000098
the final target speed of the discharge chute k is equal to the target speed of the discharge chute k after the optimization of the Muxneural network, and k is equal to 1, 2, … …, n;
and adjusting the speed of each discharge chute according to the obtained final target speed of the discharge chute k.
10. A computer-readable storage medium storing a computer program which, when executed by a processor, performs the steps of the method for artificial intelligence based stabilization of blending ore quality of any of claims 1 to 5.
CN201910267375.7A 2019-04-03 2019-04-03 Method, system and medium for stabilizing quality of blended ore based on artificial intelligence Active CN110577091B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910267375.7A CN110577091B (en) 2019-04-03 2019-04-03 Method, system and medium for stabilizing quality of blended ore based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910267375.7A CN110577091B (en) 2019-04-03 2019-04-03 Method, system and medium for stabilizing quality of blended ore based on artificial intelligence

Publications (2)

Publication Number Publication Date
CN110577091A CN110577091A (en) 2019-12-17
CN110577091B true CN110577091B (en) 2021-07-06

Family

ID=68810442

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910267375.7A Active CN110577091B (en) 2019-04-03 2019-04-03 Method, system and medium for stabilizing quality of blended ore based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN110577091B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114516553A (en) * 2022-03-31 2022-05-20 宝钢资源控股(上海)有限公司 Uniform mixing production system and method for multi-variety iron ores at port

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1299051A (en) * 2000-01-31 2001-06-13 上海宝钢集团公司 Intelligent stack control method for uniformly mixed ores
CN102059071A (en) * 2010-11-16 2011-05-18 吕斌 Automatic blending material control system for sintering production
US9810065B2 (en) * 2015-05-29 2017-11-07 Joy Mm Delaware, Inc. Controlling an output of a mining system
JP6668643B2 (en) * 2015-08-31 2020-03-18 宇部興産機械株式会社 Control method of belt conveyor system
CN205739532U (en) * 2016-05-13 2016-11-30 缪磊 A kind of ceramic powder raw material intelligent proportioning system
CN110697438B (en) * 2017-09-19 2021-07-02 中国计量大学 Controller of direct falling weight loss type material blanking machine based on neural network
CN107697660B (en) * 2017-09-19 2019-11-29 中国计量大学 Screw material disperser control method based on machine learning
CN207786513U (en) * 2017-12-26 2018-08-31 广州发展集团股份有限公司 Automatic Allocating Coal System
CN108946205A (en) * 2018-08-28 2018-12-07 上海慧安科技股份有限公司 A kind of iron ore ingredient blending system based on harbour

Also Published As

Publication number Publication date
CN110577091A (en) 2019-12-17

Similar Documents

Publication Publication Date Title
Ramasamy et al. Control of ball mill grinding circuit using model predictive control scheme
CN101226377B (en) Robust control method for asphalt mixing plant batching error
CN103968924A (en) Multistage-control-based batching weighing control method
CN101907867B (en) Self-tuning expert control method of burning trough point parameter based on working condition recognition
CN100546791C (en) Asphalt mixture stirring equipment cold aggregate feeding control method
Harja et al. Improvements in dissolved oxygen control of an activated sludge wastewater treatment process
CN110577091B (en) Method, system and medium for stabilizing quality of blended ore based on artificial intelligence
EP3859287A1 (en) Intelligent system control and dosing device
CN106406099B (en) Cement batching system and method based on fuzzy matching sum value feedback
CN103129942A (en) Granule or powder feeding and weighing controlling system
CN102054125B (en) Method for stabilizing chemical constituents of charging agglomerate
CN106022496A (en) Raw material sintering burdening optimization method and system
CN111668851A (en) Electric heating load adjustment optimization method and device
CN114807596A (en) Batching control method and device for ore heap
CN114637206B (en) Weight control method based on dual-cluster self-adaptive fuzzy neural network
CN115496403A (en) Workshop logistics system scheduling method and device, electronic equipment and storage medium
Mutturi Dynamic optimization of fed-batch bioprocesses using flower pollination algorithm
CN110703713B (en) Method for improving switching efficiency of single-device multi-product processing scheme
CN114118898A (en) Improved blending ingredient Block plan generation method and device and storage medium
CN113935233A (en) Feed formula optimization method, system, computer and storage medium
CN102540888B (en) Online guiding automatic control method
Duran-Villalobos et al. Iterative learning modelling and control of batch fermentation processes
CN110825045B (en) Method for improving switching efficiency of multi-device multi-product processing schemes
CN111086268B (en) Integral reinforcement learning control method of calcium carbide furnace lime powder ball pressing control system
CN117660707A (en) On-line control method, device and storage medium for mixing and piling BLOCK

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