CN104598970A - Method for detecting work state of climbing frame group - Google Patents

Method for detecting work state of climbing frame group Download PDF

Info

Publication number
CN104598970A
CN104598970A CN201510010422.1A CN201510010422A CN104598970A CN 104598970 A CN104598970 A CN 104598970A CN 201510010422 A CN201510010422 A CN 201510010422A CN 104598970 A CN104598970 A CN 104598970A
Authority
CN
China
Prior art keywords
neural network
network model
layer
output
climbing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201510010422.1A
Other languages
Chinese (zh)
Other versions
CN104598970B (en
Inventor
秦建武
李宏
陈斌
陈东旭
施乾东
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.)
Ningbo University
Original Assignee
Ningbo University
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 Ningbo University filed Critical Ningbo University
Priority to CN201510010422.1A priority Critical patent/CN104598970B/en
Publication of CN104598970A publication Critical patent/CN104598970A/en
Application granted granted Critical
Publication of CN104598970B publication Critical patent/CN104598970B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Feedback Control In General (AREA)

Abstract

The invention discloses a method for detecting the work state of a climbing frame group. The method includes that a double-level neural network model is built, a height neural network model processes height difference information between adjacent climbing frames in the climbing frame group, an inclination neural network model processes inclination information of each climbing frame in the climbing frame group, a load neural network model processes load information of each climbing frame in the climbing frame group, outputs of the height neural network model, the inclination neural network model and the load neural network model serve as inputs of a second-level neural network model, and then outputs of the second-level neural network model can be obtained and the work state of each climbing frames in the climbing frame group can be judged. The method has the advantages that relevance of state data of the adjacent climbing frames in the climbing frame group is considered, data information of each climbing frame is integrated during processing, accordingly disturbing influence to the climbing frames brought by the outside is reduced, the comprehensive and accurate judgment result can be obtained, and safety detection for the climbing frames is improved.

Description

A kind of working state detecting method of climbing frame group
Technical field
The present invention relates to a kind of duty detection technique of climbing frame group, especially relate to a kind of working state detecting method of climbing frame group.
Background technology
Along with the fast development of national economy, the industry size of Building Trade in China constantly expands, and large-scale high-rise engineering is increasing, promotes developing rapidly of building scaffold technology.At present, the framing scaffold that high-building construction is conventional has console mode external scaffolding, overhanging scaffold and attachment type raise scaffold etc.Console mode external scaffolding because it is consuming time, the shortcoming such as work consuming and consumptive material and being eliminated gradually; Overhanging scaffold exists takes the shortcoming of tearing open repeatedly; And attachment type raise scaffold, i.e. climbing frame, only mounting or dismounting are needed once in building operation, successively can climb or decline simultaneously with construction speed multiple spot, overcome the shortcoming of overhanging scaffold preferably, thus save the expense such as manpower, material, increase work efficiency, have broad application prospects in high-building construction.
Building industry be a kind ofly have that work high above the ground is frequent, the industry of the feature such as mobility is strong, dangerous high and Frequent Accidents, be put into China's high risk industries.The safety problem of building scaffold is the difficult problem in China's building operation always, every year a lot of framing scaffold security incident all can occur, therefore provides the importance of certain Security Assurance Mechanism to be apparent for construction scaffolding.At present, climbing frame normally uses in groups, and each climbing frame group comprises multiple climbing frame.Specify in climbing frame safety standard, each climbing frame in climbing frame group must have the security mechanisms such as anti-dumping, anti-overload and synchronization lifting.In the climbing frame group course of work, need the duty of each climbing frame of Real-Time Monitoring, to avoid climbing frame security incident.
The working state detecting method of existing climbing frame group normally independently carries out detection to each climbing frame in climbing frame group and obtains detecting data, then adopt the process of threshold value manner of comparison to detect duty whether safety that data obtain this climbing frame.But there is following problem in the working state detecting method of existing climbing frame group: the Dynamic data exchange of each climbing frame in the method, uncorrelated mutually, do not consider the relevance of the status data of adjacent climbing frame in climbing frame group, do not consider the disturbing effect of external environment to each climbing frame in climbing frame group, cause the result of determination accuracy to whether the duty of climbing frame group is safe lower, there is larger potential safety hazard.
Summary of the invention
Technical matters to be solved by this invention is to provide a kind of working state detecting method of climbing frame group, and the method considers the relevance of the status data of each climbing frame in climbing frame group, and reduce external environment to the disturbing effect of each climbing frame, result of determination accuracy is higher.
The present invention solves the problems of the technologies described above adopted technical scheme: a kind of working state detecting method of climbing frame group, described climbing frame group comprises five climbing frames be arranged in order, and five climbing frames are followed successively by No. 1 climbing frame, No. 2 climbing frames, No. 3 climbing frames, No. 4 climbing frames and No. 5 climbing frames; Comprise the following steps:
1. double-level neural network model is built: first order neural network model comprises three neural network models, be respectively height neural network model, inclination angle neural network model and load neural network model, second level neural network model comprises a neural network model, each described neural network model comprises input layer, hidden layer and output layer, each described hidden layer of neural network model and the activation function of output layer are tansig function, and described tansig function is: e is the truth of a matter of natural logarithm;
The neuron node number of the input layer of described height neural network model is 4, and the neuron node number of the hidden layer of described height neural network model is 9, and the neuron node number of the output layer of described height neural network model is 1; The parameter of the input layer of described height neural network model is X 1=[x 1, x 2, x 3, x 4], the output function of hidden layer is y 1=f 1(iw 1x 1+ b 1), the output function of output layer is y 2=f 2(lw 2y 1+ b 2)=f 2(lw 2f 1(iw 1x 1+ b 1)+b 2), wherein iw 1for the weight matrix between input layer and hidden layer, b 1for the threshold matrix between input layer and hidden layer; Lw 2for the weight matrix between hidden layer and output layer, b 2for the threshold matrix between hidden layer and output layer; The output Y of described height neural network model h=y 2;
The neuron node number of the input layer of described inclination angle neural network model is 5, and the neuron node number of the hidden layer of described inclination angle neural network model is 11, and the neuron node number of the output layer of described inclination angle neural network model is 1; The parameter of the input layer of described inclination angle neural network model is X 2=[x 5, x 6, x 7, x 8, x 9], the output function of hidden layer is y 1'=f 1' (iw 1' X 2+ b 1'), the output function of output layer is y 2'=f 2' (lw 2' y 1'+b 2')=f 2' (lw 2' f 1' (iw 1' X 2+ b 1')+b 2'), wherein iw 1' be the weight matrix between input layer and hidden layer, b 1' be the threshold matrix between input layer and hidden layer; Lw 2' be the weight matrix between hidden layer and output layer, b 2' be the threshold matrix between hidden layer and output layer; The output Y of described inclination angle neural network model d=y 2';
The neuron node number of the input layer of described load neural network model is 5, and the neuron node number of the hidden layer of described load neural network model is 11, and the neuron node number of the output layer of described load neural network model is 1; The parameter of the input layer of described load neural network model is X 3=[x 10, x 11, x 12, x 13, x 14], the output function of hidden layer is y 1"=f 1" (iw 1" X 3+ b 1"), the output function of output layer is y 2"=f 2" (lw 2" y 1"+b 2")=f 2" (lw 2" f 1" (iw 1" X 3+ b 1")+b 2"), wherein iw 1" be the weight matrix between input layer and hidden layer, b 1" be the threshold matrix between input layer and hidden layer; Lw 2" be the weight matrix between hidden layer and output layer, b 2" be the threshold matrix between hidden layer and output layer; The output Y of described load neural network model f=y 2";
The neuron node number of the input layer of described second level neural network model is 3, and the neuron node number of the hidden layer of described second level neural network model is 7, and the neuron node number of the output layer of described second level neural network model is 1; The parameter of the input layer of described second level neural network model is X 4=[x 15, x 16, x 17], the output function of hidden layer is y 1" '=f 1" ' (iw 1" ' X 4+ b 1" '), the output function of output layer is y 2" '=f 2" ' (lw 2" ' y 1" '+b 2" ')=f 2" ' (lw 2" ' f 1" ' (iw 1" ' X 4+ b 1" ')+b 2" '), wherein iw 1" ' be the weight matrix between input layer and hidden layer, b 1" ' be the threshold matrix between input layer and hidden layer; Lw 2" ' be the weight matrix between hidden layer and output layer, b 2" ' be the threshold matrix between hidden layer and output layer; The output T=y of described second level neural network model 2" ';
2. adopt the sample data of climbing frame group to step 1. in each neural network model train, obtain iw 1, b 1, lw 2, b 2, iw 1', b 1', lw 2', b 2', iw 1", b 1", lw 2", b 2", iw 1" ', b 1" ', lw 2" ', b 2" ';
3. the duty of five climbing frames is periodically sampled: the sampled data in each cycle comprises the height of climbing frame, inclination angle and load data, and wherein, the height of No. 1 climbing frame is designated as h 1, inclination angle is designated as d 1, load is designated as F 1; The height of No. 2 climbing frames is designated as h 2, inclination angle is designated as d 2, load is designated as F 2; The height of No. 3 climbing frames is designated as h 3, inclination angle is designated as d 3, load is designated as F 3; The height of No. 4 climbing frames is designated as h 4, inclination angle is designated as d 4, load is designated as F 4; The height of No. 5 climbing frames is designated as h 5, inclination angle is designated as d 5, load is designated as F 5;
4. the height of two climbing frames adjacent in same period is subtracted each other, obtain Δ h 1, Δ h 2, Δ h 3with Δ h 4, wherein Δ h 1=h 1-h 2, Δ h 2=h 2-h 3, Δ h 3=h 3-h 4, Δ h 4=h 4-h 5;
5. to Δ h 1, Δ h 2, Δ h 3, Δ h 4, d 1, d 2, d 3, d 4, d 5, F 1, F 2, F 3, F 4and F 5substitute into formula respectively be normalized, in this formula, x represents the value before normalized, represent the value after x normalized, min represents the minimum value of the physical quantity represented by x, and max represents the maximum occurrences of the physical quantity represented by x; Δ h 1value after normalized is Δ h 2value after normalized is Δ h 3value after normalized is Δ h 4value after normalized is d 1value after normalized is d 2value after normalized is d 3value after normalized is d 4value after normalized is d 5value after normalized is f 1value after normalized is f 2value after normalized is f 3value after normalized is f 4value after normalized is f 5value after normalized is
6. process in the numerical value input double-level neural network model after step 5. normalized: will with as the parameter X of the input layer of height neural network model 1=[x 1, x 2, x 3, x 4] be input to height neural network model in process, obtain height neural network model output Y h; Will with as the parameter X of the input layer of inclination angle neural network model 2=[x 5, x 6, x 7, x 8, x 9] be input in the neural network model of inclination angle and process, obtain the output Y of inclination angle neural network model d; Will with as the parameter X of the input layer of load neural network model 3=[x 10, x 11, x 12, x 13, x 14] be input in load neural network model and process, obtain the output Y of load neural network model f;
7. by Y h, Y dand Y fas the parameter X of the input layer of second level neural network model 4=[x 15, x 16, x 17] be input in the neural network model of the second level, obtain the output T of second level neural network model;
8. the duty whether safety of climbing frame group is judged according to T: if 0≤T < 0.6, then judge that climbing frame group is in normal operating conditions, i.e. safe condition; If 0.6≤T < 0.85, then judge that climbing frame group is in an interim state, the duty of climbing frame group is between safe condition and precarious position; If 0.85≤T < 1, then judge that climbing frame group is in the hole.
The step 3. middle duty of sensor to five climbing frames that adopt periodically is sampled.
Step is 2. middle trains the iw obtained 1, b 1, lw 2, b 2, iw 1', b 1', lw 2', b 2', iw 1", b 1", lw 2", b 2", iw 1" ', b 1" ', lw 2" ' and b 2" ' value as follows with matrix representation respectively:
iw 1 = - 1.4294 2.1056 0.0984 0.2588 - 0.3989 1.0556 - 1.3734 - 1.4542 - 1.4519 0.6043 1.0364 - 1.0675 0.0777 - 1.1730 1.4206 1.7589 0.2263 - 0.7069 0.3344 2.5649 - 0.0953 2.2808 - 0.8899 0.7832 0.3359 2.2522 0.2358 - 0.7678 1.1341 - 2.1221 0.1014 - 1.7349 0.0833 - 0.9581 1.3257 1.1724 ;
b 1 = 2.2467 2.0805 1.5714 - 0.7583 1.0155 0.6044 1.2380 1.4571 2.8843 ;
lw 2＝[0.3018 0.9576 -0.9755 0.6558 -0.6398 -1.3462 0.0730 -1.0640 0.7539]；
b 2＝1.0666；
iw 1 &prime; = - 1.2045 - 0.3371 0.7560 0.7460 - 2.9123 1.8199 - 0.4599 1.2851 - 0.0337 0.7060 - 0.6696 - 1.2646 - 0.5386 - 0.4943 - 1.7590 0.9775 0.1653 0.6110 - 1.5494 1.0240 0.6144 1.3369 - 0.3300 - 0.4277 - 0.0052 - 0.7982 - 0.6136 - 1.5310 0.5218 - 1.3059 - 0.7351 - 0.9008 0.1288 0.9592 - 2.1402 0.7728 - 1.1101 0.7323 1.3522 - 0.7817 0.5835 - 0.1987 1.4816 - 1.1176 - 0.7901 0.4960 - 0.7126 0.8616 0.7017 - 0.9394 - 1.0355 - 0.8205 - 1.6641 0.2910 - 0.5537 ;
b 1 &prime; = 1.2327 - 2.3188 1.1721 - 1.1589 0.4361 0.4401 0.3611 1.2175 1.6637 2.4782 - 2.4590 ;
lw 2'＝[-0.4708 0.5644 0.0656 -0.2925 0.6840 -0.4306 0.3589 0.8449 -0.8681 -0.4029 -0.2514]；
b 2'＝[1.2941]；
iw 1 &prime; &prime; = 0.6975 0.7833 0.2469 1.3658 1.3978 0.4200 1.3684 - 0.2846 0.5294 0.0347 0.6389 - 1.9154 1.3066 - 1.4085 0.6155 - 1.3516 - 0.6848 0.8086 0.9298 - 0.8105 - 0.9144 - 1.1992 1.1232 0.5216 1.8870 - 0.3950 0.3728 1.6742 0.4410 3.3742 0.0646 0.8454 1.2649 0.9439 - 0.9336 0.2692 0.9738 0.9794 1.1372 0.4460 - 0.1712 1.3614 0.5462 0.8244 1.2522 - 1.5307 1.7310 - 0.0565 0.6100 0.8086 - 0.7942 1.9623 - 0.4815 - 1.5079 2.2581 ;
b 1 &prime; &prime; = - 2.3687 - 1.4547 - 0.7202 - 1.0085 - 0.6318 - 1.5964 - 3.3376 - 4.3990 2.3958 - 2.4027 - 3.7329 ;
lw 2″＝[-1.3493 2.7278 0.9061 0.8963 1.6039 -0.3664 -2.2133 0.3152 -0.2164 -0.4828 1.3420]；
b 2″＝[1.7548]；
iw 1 &prime; &prime; &prime; = 0.0595 - 1.5705 2.3612 - 0.9238 - 2.0974 - 0.6133 0.7470 2.4332 0.6421 0.5393 - 0.2921 2.2924 - 1.6518 1.3493 1.8521 1.8997 1.2728 0.5721 1.5665 1.2850 - 1.7419 ;
b 1 &prime; &prime; &prime; = - 2.4927 1.7161 - 0.9046 - 0.5676 - 1.0763 2.6683 2.6999 ;
lw 2″'＝[0.5816 -0.3338 0.3656 -0.1650 0.7490 0.8836 0.2446]；
b 2″'＝0.7275。
Compared with prior art, the invention has the advantages that by building double-level neural network model, first order neural network model comprises three neural network models, be respectively height neural network model, inclination angle neural network model and load neural network model, second level neural network model comprises a neural network model, height neural network model processes the difference in height information between the adjacent climbing frame in climbing frame group, the obliquity information of inclination angle neural network model to each climbing frame in climbing frame group processes, the load information of load neural network to each climbing frame in climbing frame group processes, height neural network model, the output of inclination angle neural network model and load neural network model is as the input of second level neural network model, obtain the output of second level neural network model thus and carry out the duty that judges to obtain climbing frame group accordingly, in method of the present invention, consider the relevance of the status data of adjacent climbing frame in climbing frame group, fusion treatment is carried out to the data message of each climbing frame in climbing frame group, reduce external environment to the disturbing effect of each climbing frame, comprehensively obtain comprehensively, result of determination accurately, improve the accuracy rate detecting climbing frame group safety case, enhance anti-interference, ensure that security of system is reliable.
Accompanying drawing explanation
Fig. 1 is theory diagram of the present invention;
Fig. 2 is the structural drawing of double-level neural network model of the present invention;
Fig. 3 is workflow diagram of the present invention.
Embodiment
Below in conjunction with accompanying drawing embodiment, the present invention is described in further detail.
Embodiment: a kind of working state detecting method of climbing frame group, climbing frame group comprises five climbing frames be arranged in order, and five climbing frames are followed successively by No. 1 climbing frame, No. 2 climbing frames, No. 3 climbing frames, No. 4 climbing frames and No. 5 climbing frames; Comprise the following steps:
1. double-level neural network model is built: first order neural network model comprises three neural network models, be respectively height neural network model 1, inclination angle neural network model 2 and load neural network model 3, second level neural network model comprises a neural network model, each neural network model comprises input layer, hidden layer and output layer, the hidden layer of each neural network model and the activation function of output layer are tansig function, and tansig function is: e is the truth of a matter of natural logarithm;
The neuron node number of the input layer 11 of height neural network model 1 is 4, and the neuron node number of the hidden layer 12 of height neural network model 1 is 9, and the neuron node number of the output layer 13 of height neural network model 1 is 1; The parameter of the input layer 11 of height neural network model 1 is X 1=[x 1, x 2, x 3, x 4], the output function of hidden layer 12 is y 1=f 1(iw 1x 1+ b 1), the output function of output layer 13 is y 2=f 2(lw 2y 1+ b 2)=f 2(lw 2f 1(iw 1x 1+ b 1)+b 2), wherein iw 1for the weight matrix between input layer 11 and hidden layer 12, b 1for the threshold matrix between input layer 11 and hidden layer 12; Lw 2for the weight matrix between hidden layer 12 and output layer 13, b 2for the threshold matrix between hidden layer 12 and output layer 13; The output Y of height neural network model 1 h=y 2;
The neuron node number of the input layer 21 of inclination angle neural network model 2 is 5, and the neuron node number of the hidden layer 22 of inclination angle neural network model 2 is 11, and the neuron node number of the output layer 23 of inclination angle neural network model 2 is 1; The parameter of the input layer 21 of inclination angle neural network model 2 is X 2=[x 5, x 6, x 7, x 8, x 9], the output function of hidden layer 22 is y 1'=f 1' (iw 1' X 2+ b 1'), the output function of output layer 23 is y 2'=f 2' (lw 2' y 1'+b 2')=f 2' (lw 2' f 1' (iw 1' X 2+ b 1')+b 2'), wherein iw 1' be the weight matrix between input layer 21 and hidden layer 22, b 1' be the threshold matrix between input layer 21 and hidden layer 22; Lw 2' be the weight matrix between hidden layer 22 and output layer 23, b 2' be the threshold matrix between hidden layer 22 and output layer 23; The output Y of inclination angle neural network model 2 d=y 2';
The neuron node number of the input layer 31 of load neural network model 3 is 5, and the neuron node number of the hidden layer 32 of load neural network model 3 is 11, and the neuron node number of the output layer 33 of load neural network model 3 is 1; The parameter of the input layer 31 of load neural network model 3 is X 3=[x 10, x 11, x 12, x 13, x 14], the output function of hidden layer 32 is y 1"=f 1" (iw 1" X 3+ b 1"), the output function of output layer 33 is y 2"=f 2" (lw 2" y 1"+b 2")=f 2" (lw 2" f 1" (iw 1" X 3+ b 1")+b 2"), wherein iw 1" be the weight matrix between input layer 31 and hidden layer 32, b 1" be the threshold matrix between input layer 31 hidden layer 32; Lw 2" be the weight matrix between hidden layer 32 and output layer 33, b 2" be the threshold matrix between hidden layer 32 and output layer 33; The output Y of load neural network model 3 f=y 2";
The neuron node number of the input layer 41 of second level neural network model 4 is 3, and the neuron node number of the hidden layer 42 of second level neural network model 4 is 7, and the neuron node number of the output layer 43 of second level neural network model 4 is 1; The parameter of the input layer 41 of second level neural network model 4 is X 4=[x 15, x 16, x 17], the output function of hidden layer 42 is y 1" '=f 1" ' (iw 1" ' X 4+ b 1" '), the output function of output layer 43 is y 2" '=f 2" ' (lw 2" ' y 1" '+b 2" ')=f 2" ' (lw 2" ' f 1" ' (iw 1" ' X 4+ b 1" ')+b 2" '), wherein iw 1" ' be the weight matrix between input layer 41 and hidden layer 42, b 1" ' be the threshold matrix between input layer 41 and hidden layer 42; Lw 2" ' be the weight matrix between hidden layer 42 and output layer 43, b 2" ' be the threshold matrix between hidden layer 42 and output layer 43; The output T=y of second level neural network model 4 2" ';
2. adopt the sample data of climbing frame group to step 1. in each neural network model train, obtain iw 1, b 1, lw 2, b 2, iw 1', b 1', lw 2', b 2', iw 1", b 1", lw 2", b 2", iw 1" ', b 1" ', lw 2" ', b 2" '; The sample data of climbing frame group can obtain from the statistics the climbing frame course of work according to the data demand of each neural network model process or climbing frame group be carried out to test of many times operation and obtain, and adopts the neural network training method of existing maturation to train;
3. the duty of five climbing frames is periodically sampled: the sampled data in each cycle comprises the height of climbing frame, inclination angle and load data, and wherein, the height of No. 1 climbing frame is designated as h 1, inclination angle is designated as d 1, load is designated as F 1; The height of No. 2 climbing frames is designated as h 2, inclination angle is designated as d 2, load is designated as F 2; The height of No. 3 climbing frames is designated as h 3, inclination angle is designated as d 3, load is designated as F 3; The height of No. 4 climbing frames is designated as h 4, inclination angle is designated as d 4, load is designated as F 4; The height of No. 5 climbing frames is designated as h 5, inclination angle is designated as d 5, load is designated as F 5; Sampling period can be determined according to climbing frame real work situation, and in the present embodiment, the sampling period is 100ms;
4. the height of two climbing frames adjacent in same period is subtracted each other, obtain Δ h 1, Δ h 2, Δ h 3with Δ h 4, wherein Δ h 1=h 1-h 2, Δ h 2=h 2-h 3, Δ h 3=h 3-h 4, Δ h 4=h 4-h 5;
5. to Δ h 1, Δ h 2, Δ h 3, Δ h 4, d 1, d 2, d 3, d 4, d 5, F 1, F 2, F 3, F 4and F 5substitute into formula respectively be normalized, in this formula, x represents the value before normalized, represent the value after x normalized, min represents the minimum value of the physical quantity represented by x, and max represents the maximum occurrences of the physical quantity represented by x, and the value of min and max can obtain from climbing frame keeps the safety in production specification; Δ h 1value after normalized is Δ h 2value after normalized is Δ h 3value after normalized is Δ h 4value after normalized is d 1value after normalized is d 2value after normalized is d 3value after normalized is d 4value after normalized is d 5value after normalized is f 1value after normalized is f 2value after normalized is f 3value after normalized is f 4value after normalized is f 5value after normalized is
6. process in the numerical value input double-level neural network model after step 5. normalized: will with as the parameter X of the input layer 11 of height neural network model 1 1=[x 1, x 2, x 3, x 4] be input to height neural network model 1 in process, obtain height neural network model 1 output Y h; Will with as the parameter X of the input layer 21 of inclination angle neural network model 2 2=[x 5, x 6, x 7, x 8, x 9] be input in inclination angle neural network model 2 and process, obtain the output Y of inclination angle neural network model 2 d; Will with as the parameter X of the input layer 31 of load neural network model 3 3=[x 10, x 11, x 12, x 13, x 14] be input in load neural network model 3 and process, obtain the output Y of load neural network model 3 f;
7. by Y h, Y dand Y fas the parameter X of the input layer 41 of second level neural network model 4 4=[x 15, x 16, x 17] be input in second level neural network model 4, obtain the output T of second level neural network model 4;
8. the duty whether safety of climbing frame group is judged according to T: if 0≤T < 0.6, then judge that climbing frame group is in normal operating conditions, i.e. safe condition; If 0.6≤T < 0.85, then judge that climbing frame group is in an interim state, the duty of climbing frame group between safe condition and precarious position, just by safe condition to precarious position transition; If 0.85≤T < 1, then judge that climbing frame group is in the hole.
In the present embodiment, the step 3. middle duty of sensor to five climbing frames that adopt periodically is sampled.
In the present embodiment, step is 2. middle trains the iw obtained 1, b 1, lw 2, b 2, iw 1', b 1', lw 2', b 2', iw 1", b 1", lw 2", b 2", iw 1" ', b 1" ', lw 2" ' and b 2" ' value as follows with matrix representation respectively:
iw 1 = - 1.4294 2.1056 0.0984 0.2588 - 0.3989 1.0556 - 1.3734 - 1.4542 - 1.4519 0.6043 1.0364 - 1.0675 0.0777 - 1.1730 1.4206 1.7589 0.2263 - 0.7069 0.3344 2.5649 - 0.0953 2.2808 - 0.8899 0.7832 0.3359 2.2522 0.2358 - 0.7678 1.1341 - 2.1221 0.1014 - 1.7349 0.0833 - 0.9581 1.3257 1.1724 ;
b 1 = 2.2467 2.0805 1.5714 - 0.7583 1.0155 0.6044 1.2380 1.4571 2.8843 ;
lw 2＝[0.3018 0.9576 -0.9755 0.6558 -0.6398 -1.3462 0.0730 -1.0640 0.7539]；
b 2＝1.0666；
iw 1 &prime; = - 1.2045 - 0.3371 0.7560 0.7460 - 2.9123 1.8199 - 0.4599 1.2851 - 0.0337 0.7060 - 0.6696 - 1.2646 - 0.5386 - 0.4943 - 1.7590 0.9775 0.1653 0.6110 - 1.5494 1.0240 0.6144 1.3369 - 0.3300 - 0.4277 - 0.0052 - 0.7982 - 0.6136 - 1.5310 0.5218 - 1.3059 - 0.7351 - 0.9008 0.1288 0.9592 - 2.1402 0.7728 - 1.1101 0.7323 1.3522 - 0.7817 0.5835 - 0.1987 1.4816 - 1.1176 - 0.7901 0.4960 - 0.7126 0.8616 0.7017 - 0.9394 - 1.0355 - 0.8205 - 1.6641 0.2910 - 0.5537 ;
b 1 &prime; = 1.2327 - 2.3188 1.1721 - 1.1589 0.4361 0.4401 0.3611 1.2175 1.6637 2.4782 - 2.4590 ;
lw 2'＝[-0.4708 0.5644 0.0656 -0.2925 0.6840 -0.4306 0.3589 0.8449 -0.8681 -0.4029 -0.2514]；
b 2'＝[1.2941]；
iw 1 &prime; &prime; = 0.6975 0.7833 0.2469 1.3658 1.3978 0.4200 1.3684 - 0.2846 0.5294 0.0347 0.6389 - 1.9154 1.3066 - 1.4085 0.6155 - 1.3516 - 0.6848 0.8086 0.9298 - 0.8105 - 0.9144 - 1.1992 1.1232 0.5216 1.8870 - 0.3950 0.3728 1.6742 0.4410 3.3742 0.0646 0.8454 1.2649 0.9439 - 0.9336 0.2692 0.9738 0.9794 1.1372 0.4460 - 0.1712 1.3614 0.5462 0.8244 1.2522 - 1.5307 1.7310 - 0.0565 0.6100 0.8086 - 0.7942 1.9623 - 0.4815 - 1.5079 2.2581 ;
b 1 &prime; &prime; = - 2.3687 - 1.4547 - 0.7202 - 1.0085 - 0.6318 - 1.5964 - 3.3376 - 4.3990 2.3958 - 2.4027 - 3.7329 ;
lw 2″＝[-1.3493 2.7278 0.9061 0.8963 1.6039 -0.3664 -2.2133 0.3152 -0.2164 -0.4828 1.3420]；
b 2″＝[1.7548]；
iw 1 &prime; &prime; &prime; = 0.0595 - 1.5705 2.3612 - 0.9238 - 2.0974 - 0.6133 0.7470 2.4332 0.6421 0.5393 - 0.2921 2.2924 - 1.6518 1.3493 1.8521 1.8997 1.2728 0.5721 1.5665 1.2850 - 1.7419 ;
b 1 &prime; &prime; &prime; = - 2.4927 1.7161 - 0.9046 - 0.5676 - 1.0763 2.6683 2.6999 ;
lw 2″'＝[0.5816 -0.3338 0.3656 -0.1650 0.7490 0.8836 0.2446]；
b 2″'＝0.7275。

Claims (3)

1. a working state detecting method for climbing frame group, described climbing frame group comprises five climbing frames be arranged in order, and five climbing frames are followed successively by No. 1 climbing frame, No. 2 climbing frames, No. 3 climbing frames, No. 4 climbing frames and No. 5 climbing frames; It is characterized in that comprising the following steps:
1. double-level neural network model is built: first order neural network model comprises three neural network models, be respectively height neural network model, inclination angle neural network model and load neural network model, second level neural network model comprises a neural network model, each described neural network model comprises input layer, hidden layer and output layer, each described hidden layer of neural network model and the activation function of output layer are tansig function, and described tansig function is: f ( x ) = 2 1 + e - 2 x - 1 , E is the truth of a matter of natural logarithm;
The neuron node number of the input layer of described height neural network model is 4, and the neuron node number of the hidden layer of described height neural network model is 9, and the neuron node number of the output layer of described height neural network model is 1; The parameter of the input layer of described height neural network model is X 1=[x 1, x 2, x 3, x 4], the output function of hidden layer is y 1=f 1(iw 1x 1+ b 1), the output function of output layer is y 2=f 2(lw 2y 1+ b 2)=f 2(lw 2f 1(iw 1x 1+ b 1)+b 2), wherein iw 1for the weight matrix between input layer and hidden layer, b 1for the threshold matrix between input layer and hidden layer; Lw 2for the weight matrix between hidden layer and output layer, b 2for the threshold matrix between hidden layer and output layer; The output Y of described height neural network model h=y 2;
The neuron node number of the input layer of described inclination angle neural network model is 5, and the neuron node number of the hidden layer of described inclination angle neural network model is 11, and the neuron node number of the output layer of described inclination angle neural network model is 1; The parameter of the input layer of described inclination angle neural network model is X 2=[x 5, x 6, x 7, x 8, x 9], the output function of hidden layer is y 1'=f 1' (iw 1' X 2+ b 1'), the output function of output layer is y 2'=f 2' (lw 2' y 1'+b 2')=f 2' (lw 2' f 1' (iw 1' X 2+ b 1')+b 2'), wherein iw 1' be the weight matrix between input layer and hidden layer, b 1' be the threshold matrix between input layer and hidden layer; Lw 2' be the weight matrix between hidden layer and output layer, b 2' be the threshold matrix between hidden layer and output layer; The output Y of described inclination angle neural network model d=y 2';
The neuron node number of the input layer of described load neural network model is 5, and the neuron node number of the hidden layer of described load neural network model is 11, and the neuron node number of the output layer of described load neural network model is 1; The parameter of the input layer of described load neural network model is X 3=[x 10, x 11, x 12, x 13, x 14], the output function of hidden layer is y 1"=f 1" (iw 1" X 3+ b 1"), the output function of output layer is y 2"=f 2" (lw 2" y 1"+b 2")=f 2" (lw 2" f 1" (iw 1" X 3+ b 1")+b 2"), wherein iw 1" be the weight matrix between input layer and hidden layer, b 1" be the threshold matrix between input layer and hidden layer; Lw 2" be the weight matrix between hidden layer and output layer, b 2" be the threshold matrix between hidden layer and output layer; The output Y of described load neural network model f=y 2";
The neuron node number of the input layer of described second level neural network model is 3, and the neuron node number of the hidden layer of described second level neural network model is 7, and the neuron node number of the output layer of described second level neural network model is 1; The parameter of the input layer of described second level neural network model is X 4=[x 15, x 16, x 17], the output function of hidden layer is y 1" '=f 1" ' (iw 1" ' X 4+ b 1" '), the output function of output layer is y 2" '=f 2" ' (lw 2" ' y 1" '+b 2" ')=f 2" ' (lw 2" ' f 1" ' (iw 1" ' X 4+ b 1" ')+b 2" '), wherein iw 1" ' be the weight matrix between input layer and hidden layer, b 1" ' be the threshold matrix between input layer and hidden layer; Lw 2" ' be the weight matrix between hidden layer and output layer, b 2" ' be the threshold matrix between hidden layer and output layer; The output T=y of described second level neural network model 2" ';
2. adopt the sample data of climbing frame group to step 1. in each neural network model train, obtain iw 1, b 1, lw 2, b 2, iw 1', b 1', lw 2', b 2', iw 1", b 1", lw 2", b 2", iw 1" ', b 1" ', lw 2" ', b 2" ';
3. the duty of five climbing frames is periodically sampled: the sampled data in each cycle comprises the height of climbing frame, inclination angle and load data, and wherein, the height of No. 1 climbing frame is designated as h 1, inclination angle is designated as d 1, load is designated as F 1; The height of No. 2 climbing frames is designated as h 2, inclination angle is designated as d 2, load is designated as F 2; The height of No. 3 climbing frames is designated as h 3, inclination angle is designated as d 3, load is designated as F 3; The height of No. 4 climbing frames is designated as h 4, inclination angle is designated as d 4, load is designated as F 4; The height of No. 5 climbing frames is designated as h 5, inclination angle is designated as d 5, load is designated as F 5;
4. the height of two climbing frames adjacent in same period is subtracted each other, obtain Δ h 1, Δ h 2, Δ h 3with Δ h 4, wherein Δ h 1=h 1-h 2, Δ h 2=h 2-h 3, Δ h 3=h 3-h 4, Δ h 4=h 4-h 5;
5. to Δ h 1, Δ h 2, Δ h 3, Δ h 4, d 1, d 2, d 3, d 4, d 5, F 1, F 2, F 3, F 4and F 5substitute into formula respectively be normalized, in this formula, x represents the value before normalized, represent the value after x normalized, min represents the minimum value of the physical quantity represented by x, and max represents the maximum occurrences of the physical quantity represented by x; Δ h 1value after normalized is Δ h 2value after normalized is Δ h 3value after normalized is Δ h 4value after normalized is d 1value after normalized is d 2value after normalized is d 3value after normalized is d 4value after normalized is d 5value after normalized is f 1value after normalized is f 2value after normalized is f 3value after normalized is f 4value after normalized is f 5value after normalized is
6. process in the numerical value input double-level neural network model after step 5. normalized: will with as the parameter X of the input layer of height neural network model 1=[x 1, x 2, x 3, x 4] be input to height neural network model in process, obtain height neural network model output Y h; Will with as the parameter X of the input layer of inclination angle neural network model 2=[x 5, x 6, x 7, x 8, x 9] be input in the neural network model of inclination angle and process, obtain the output Y of inclination angle neural network model d; Will with as the parameter X of the input layer of load neural network model 3=[x 10, x 11, x 12, x 13, x 14] be input in load neural network model and process, obtain the output Y of load neural network model f;
7. by Y h, Y dand Y fas the parameter X of the input layer of second level neural network model 4=[x 15, x 16, x 17] be input in the neural network model of the second level, obtain the output T of second level neural network model;
8. the duty whether safety of climbing frame group is judged according to T: if 0≤T < 0.6, then judge that climbing frame group is in normal operating conditions, i.e. safe condition; If 0.6≤T < 0.85, then judge that climbing frame group is in an interim state, the duty of climbing frame group is between safe condition and precarious position; If 0.85≤T < 1, then judge that climbing frame group is in the hole.
2. the working state detecting method of a kind of climbing frame group according to claim 1, is characterized in that adopting the duty of sensor to five climbing frames periodically to sample during step 3..
3. the working state detecting method of a kind of climbing frame group according to claim 1, is characterized in that training the iw obtained during step 2. 1, b 1, lw 2, b 2, iw 1', b 1', lw 2', b 2', iw 1", b 1", lw 2", b 2", iw 1" ', b 1" ', lw 2" ' and b 2" ' value as follows with matrix representation respectively:
iw 1 = - 1.4294 2.1056 0.0984 0.2588 - 0.3989 1.0556 - 1.3734 - 1.4542 - 1.4519 0.6043 1.0364 - 1.0675 0.0777 - 1.1730 1.4206 1.7589 0.2263 - 0.7069 0.3344 2.5649 - 0.0953 2.3808 - 0.8899 - 0 . 7832 0.3359 2.2522 0.2358 - 0.7678 1.1341 - 2.1221 0.1014 - 1.7349 0.0833 - 0.9581 1.3257 1.1724 ;
b 1 = 2.2467 2.0805 1.5714 - 0.7583 1.0155 0.6044 1.2380 1.4571 2.8843 ;
lw 2＝[0.3018 0.9576 -0.9755 0.6558 -0.6398 -1.3462 0.0730 -1.0640 0.7539]；
b 2＝1.0666；
iw 1 &prime; = 1.2045 - 0.3371 0.7560 0.7460 - 2.9123 1.8199 - 0.4599 1.2851 - 0.0337 0.7060 - 0.6696 - 1.2646 - 0.5386 - 0.4943 - 1.7590 0.9755 0.1653 0.6110 - 1.5494 1.0240 - 0.6144 1.3369 - 0.3300 - 0.4277 - 0.0052 - 0.7982 - 0.6136 - 1.5310 0.5218 - 1.3059 - 0.7351 - 0.9008 0.1288 0.9596 - 2.1402 0.7728 - 1.1101 0.7323 1.3522 - 0.7817 0.5853 - 0.1987 1.4816 - 1.1176 - 0.7901 0.4960 - 0.7126 0.8616 0.7017 - 0.9394 - 1.0355 - 0.8205 - 1.6641 0.2910 - 0.5537 ;
b 1 &prime; = 1.2327 - 2.3188 1.1721 - 1.1589 0.4361 0.4401 0.3611 1.2175 1.6637 2.4782 - 2.4590 ;
lw 2'＝[-0.4708 0.5644 0.0656 -0.2925 0.6840 -0.4306 0.3589 0.8449 -0.8681 -0.4029 -0.2514]；
b 2'＝[1.2941]；
iw 1 &prime; &prime; = 0.6975 0.7833 0.2469 1.3658 1.3978 0.4200 1.3684 - 0.2846 0.5294 0.0347 0.6389 - 1.9154 1.3066 - 1.4085 0.6155 - 1.3516 - 0.6848 0.8086 0.9298 - 0.8105 - 0.9144 - 0.1922 1.1232 0.5216 1.8870 - 0.3950 0.3728 1.6742 0.4410 3.3742 0.0646 0.8454 1.2649 0.9439 - 0.9336 0.2692 0.9738 0.9794 1.1372 0.4460 - 0.1712 1.3614 0.5462 0.8244 1.2522 - 1.5307 1.7310 - 0.0565 0.6100 0.8086 - 0.7942 1.9623 - 0.4815 - 1.5079 2.2581 ;
b 1 &prime; &prime; = - 2.3687 - 1.4547 - 0.7202 - 1.0085 - 0.6318 - 1.5964 - 3.3376 - 4.3990 2.3958 - 2.4027 - 3.7329 ;
lw 2″＝[-1.3493 2.7278 0.9061 0.8963 1.6039 -0.3664 -2.2133 0.3152 -0.2164 -0.4828 1.3420]；
b 2″＝[1.7548]；
iw 1 &prime; &prime; &prime; = 0.0595 - 1.5705 2.3612 - 0.9238 - 2.0974 - 0.6133 0.7470 2.4332 0.6421 0.5393 - 0.2921 2.2924 - 1.6518 1.3493 1.8521 1.8997 1.2728 0.5721 1.5665 1.2850 - 1.7419 ;
b 1 &prime; &prime; &prime; = - 2.4927 1.7161 - 0.9046 - 0.5676 - 1.0763 2.6683 2.6999 ;
lw 2″'＝[0.5816 -0.3338 0.3656 -0.1650 0.7490 0.8836 0.2446]；
b 2″'＝0.7275。
CN201510010422.1A 2015-01-09 2015-01-09 Method for detecting work state of climbing frame group Active CN104598970B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510010422.1A CN104598970B (en) 2015-01-09 2015-01-09 Method for detecting work state of climbing frame group

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510010422.1A CN104598970B (en) 2015-01-09 2015-01-09 Method for detecting work state of climbing frame group

Publications (2)

Publication Number Publication Date
CN104598970A true CN104598970A (en) 2015-05-06
CN104598970B CN104598970B (en) 2017-02-22

Family

ID=53124741

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510010422.1A Active CN104598970B (en) 2015-01-09 2015-01-09 Method for detecting work state of climbing frame group

Country Status (1)

Country Link
CN (1) CN104598970B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109124635A (en) * 2018-09-25 2019-01-04 上海联影医疗科技有限公司 Model generating method, MRI scan method and system
CN110132353A (en) * 2019-05-15 2019-08-16 上海荷福人工智能科技（集团）有限公司 High-supported formwork safety monitoring method and system based on multisensor and artificial intelligence
CN110736980A (en) * 2019-11-04 2020-01-31 广东博智林机器人有限公司 method, device and system for monitoring climbing rack state information
CN113483813A (en) * 2021-06-18 2021-10-08 广州云奇科技有限公司 Scaffold safety detection method and system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030055797A1 (en) * 2001-07-30 2003-03-20 Seiji Ishihara Neural network system, software and method of learning new patterns without storing existing learned patterns
US7406451B2 (en) * 2003-04-10 2008-07-29 Bayer Technology Services Gmbh Method for training neural networks
CN103760299A (en) * 2013-04-25 2014-04-30 孝感毅恒安全科技有限公司 Comprehensive air quality prediction method based on two-stage neural network
CN104166095A (en) * 2014-08-29 2014-11-26 东南大学 Fault information fusion diagnosis method based on double-edge linear motor

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030055797A1 (en) * 2001-07-30 2003-03-20 Seiji Ishihara Neural network system, software and method of learning new patterns without storing existing learned patterns
US7406451B2 (en) * 2003-04-10 2008-07-29 Bayer Technology Services Gmbh Method for training neural networks
CN103760299A (en) * 2013-04-25 2014-04-30 孝感毅恒安全科技有限公司 Comprehensive air quality prediction method based on two-stage neural network
CN104166095A (en) * 2014-08-29 2014-11-26 东南大学 Fault information fusion diagnosis method based on double-edge linear motor

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
PFISTER G.: ""Multisensor /multicriteria fire detection: a new trend rapidly becomes state of the art"", 《FIRE TECHNOLOGY》 *
WU WEI ETL.: ""Convergence analysis of online gradient method for BP neural networks"", 《NEURAL NETWORKS》 *
周斌: ""基于BP神经网络的内燃机排放性能建模与应用研究"", 《中国优秀博硕士学位论文全文数据库(博士) 工程科技Ⅱ辑》 *
杨光祥: ""附着升降脚手架同步升降控制方法研究"", 《工业建筑》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109124635A (en) * 2018-09-25 2019-01-04 上海联影医疗科技有限公司 Model generating method, MRI scan method and system
CN110132353A (en) * 2019-05-15 2019-08-16 上海荷福人工智能科技（集团）有限公司 High-supported formwork safety monitoring method and system based on multisensor and artificial intelligence
CN110736980A (en) * 2019-11-04 2020-01-31 广东博智林机器人有限公司 method, device and system for monitoring climbing rack state information
CN110736980B (en) * 2019-11-04 2021-09-21 广东博智林机器人有限公司 Method, device and system for monitoring state information of climbing frame
CN113483813A (en) * 2021-06-18 2021-10-08 广州云奇科技有限公司 Scaffold safety detection method and system

Also Published As

Publication number Publication date
CN104598970B (en) 2017-02-22

Similar Documents

Publication Publication Date Title
CN107941537B (en) A kind of mechanical equipment health state evaluation method
CN105067209B (en) The method of bridge structure stiffness variation is judged based on bridge health monitoring deformation data
WO2023197461A1 (en) Gearbox fault early warning method and system based on working condition similarity evaluation
CN104598970A (en) Method for detecting work state of climbing frame group
CN114559819B (en) Electric automobile battery safety early warning method based on signal processing
CN104914775B (en) Multi-modal procedure failure testing method and system based on the description of vine copula correlations
CN113255188B (en) Bridge safety early warning method and system based on accident tree
CN104976517A (en) Wharf water supply pipe network online supervision method
CN105139388A (en) Method and apparatus for building facade damage detection in oblique aerial image
CN112798042A (en) Intelligent diagnosis method for working state and supporting quality of hydraulic support
CN106429689A (en) Elevator maintenance system based on Internet-of-things big data support
CN109635879B (en) Coal mining machine fault diagnosis system with optimal parameters
CN111191832A (en) Typhoon disaster power distribution network tower fault prediction method and system
Jin et al. Structural damage recognition based on filtered feature selection and convolutional neural network
WO2019019429A1 (en) Anomaly detection method, device and apparatus for virtual machine, and storage medium
CN102951516B (en) Elevator operation monitoring system
CN104216397B (en) Failure recognition and detection method for intelligent drive axle system
CN116664846B (en) Method and system for realizing 3D printing bridge deck construction quality monitoring based on semantic segmentation
CN104985003A (en) Rolling process slip diagnosis and prediction method based on data driving
CN114582083B (en) Tunnel monitoring multi-sensor data fusion fire disaster early warning method
CN114897084A (en) Tower crane structure safety monitoring method based on graph convolution neural network
CN105095649B (en) The data analysing method and system of building inclination angle
CN114757365A (en) High-speed railway roadbed settlement prediction and early warning method based on deep learning
CN110954349B (en) Crane structure health state monitoring method based on residual stress distortion rate
CN107218180A (en) A kind of wind power generating set driving unit fault alarm method measured based on vibration acceleration

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant