CN117669373A - Energy consumption prediction method and system for hydraulic system of forging forming equipment - Google Patents
Energy consumption prediction method and system for hydraulic system of forging forming equipment Download PDFInfo
- Publication number
- CN117669373A CN117669373A CN202311624189.7A CN202311624189A CN117669373A CN 117669373 A CN117669373 A CN 117669373A CN 202311624189 A CN202311624189 A CN 202311624189A CN 117669373 A CN117669373 A CN 117669373A
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- data set
- data
- characteristic
- prediction
- 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.)
- Pending
Links
- 238000005265 energy consumption Methods 0.000 title claims abstract description 177
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000005242 forging Methods 0.000 title claims abstract description 35
- 230000003068 static effect Effects 0.000 claims abstract description 53
- 230000007246 mechanism Effects 0.000 claims abstract description 38
- 230000009466 transformation Effects 0.000 claims abstract description 15
- 238000004088 simulation Methods 0.000 claims abstract description 14
- 238000012544 monitoring process Methods 0.000 claims abstract description 13
- 238000012952 Resampling Methods 0.000 claims abstract description 12
- 238000011156 evaluation Methods 0.000 claims abstract description 11
- 238000004140 cleaning Methods 0.000 claims abstract description 8
- 230000004927 fusion Effects 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 15
- 238000004458 analytical method Methods 0.000 claims description 12
- 230000002787 reinforcement Effects 0.000 claims description 12
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000010586 diagram Methods 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 10
- 238000010801 machine learning Methods 0.000 claims description 10
- 238000012417 linear regression Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 5
- 230000001502 supplementing effect Effects 0.000 claims description 5
- 238000012706 support-vector machine Methods 0.000 claims description 5
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 claims description 4
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000003066 decision tree Methods 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the disclosure provides a method and a system for predicting energy consumption of a hydraulic system of forging forming equipment, which belong to the technical field of data processing and specifically comprise the following steps: obtaining a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set; removing the characteristics of which the energy consumption correlation degree is lower than a correlation threshold; for time sequence data in the energy consumption data set, a sliding time window step length is obtained by adopting Fourier transformation, for static data in the energy consumption data set, characteristic state clustering is carried out after resampling, and classification evaluation is carried out on the operation working condition of the current equipment; analyzing power energy flow, and establishing a mechanism simulation model to obtain a first energy consumption predicted value; and respectively inputting the time sequence data set and the static data set into different prediction models, fusing the first energy consumption prediction, constructing an energy consumption prediction model with mechanism and data driving fusion, and obtaining and monitoring target energy consumption prediction. By the scheme, the energy consumption prediction efficiency, accuracy and stability are improved.
Description
Technical Field
The embodiment of the disclosure relates to the technical field of data processing, in particular to a method and a system for predicting energy consumption of a hydraulic system of forging forming equipment.
Background
With the development of intelligent factories and manufacturing technologies, the physical structure system of intelligent manufacturing equipment is gradually complicated, the production requirements of manufacturing industries on the intelligent manufacturing equipment are improved, and the equipment with high performance, high precision, low cost and low power consumption can ensure better processing quality. The hydraulic system has the characteristic of high power density, but also has the defects of high pollution and low energy efficiency, so the hydraulic system is very important for energy consumption prediction and early warning of the hydraulic system in intelligent manufacturing equipment. However, some large intelligent manufacturing equipment has a complex physical structure system, and monitoring energy consumption according to priori knowledge or predicting energy consumption by using a mechanism simulation model alone can cause waste of time and energy. Moreover, for some new intelligent manufacturing equipment, a priori knowledge and a mechanism simulation model thereof need to be reconstructed.
With the development of digital technology and artificial intelligence, the popularization of the multi-source sensor technology enables collected intelligent manufacturing equipment data to be more and more rich, and the method plays a basic supporting role for a data-driven energy consumption prediction method. The traditional data driving method is mostly single-model prediction energy consumption, and lacks corresponding feature selection and reconstruction processes, so that the problems of state space explosion of data and poor final prediction effect are caused. Meanwhile, the hydraulic systems under different scene working conditions are not distinguished, so that the effect of energy consumption prediction is poor under certain working conditions, and the flexibility of energy consumption prediction is lacking.
Therefore, a method for predicting the energy consumption of the hydraulic system of the forging forming equipment with high prediction efficiency and high precision is needed.
Disclosure of Invention
In view of the above, embodiments of the present disclosure provide a method and a system for predicting energy consumption of a hydraulic system of forging forming equipment, which at least partially solve the problem in the prior art that the prediction efficiency and the accuracy are poor.
In a first aspect, an embodiment of the present disclosure provides a method for predicting energy consumption of a hydraulic system of a forging forming apparatus, including:
step 1, collecting state characteristics of forging forming equipment, obtaining a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set;
step 2, removing the features with the degree of correlation with the energy consumption lower than a correlation threshold value in the standard feature data set through mechanism analysis and a filter to obtain an energy consumption data set;
step 3, for time sequence data in the energy consumption data set, a sliding time window step length is obtained by adopting Fourier transformation to form a time sequence data set, for static data in the energy consumption data set, characteristic state clustering is carried out after resampling, and classification evaluation is carried out on the operation working condition of the current equipment to form a static data set;
step 4, analyzing the power energy flow, and establishing a mechanism simulation model to obtain a first energy consumption predicted value;
and 5, respectively inputting the time sequence data set and the static data set into different prediction models, fusing the first energy consumption prediction, constructing an energy consumption prediction model fusing mechanism and data driving, and obtaining and monitoring target energy consumption prediction.
According to a specific implementation manner of the embodiment of the present disclosure, the step 1 specifically includes:
step 1.1, numbering hydraulic system components such as a hydraulic pump, a servo valve, a hydraulic actuator, an energy accumulator, a pipeline and the like in a state to be monitored based on a physical structure of intelligent equipment of a forging program, and designing a data acquisition device to acquire different characteristic data of the hydraulic system to form an original characteristic data set Wherein X is i Representing a set of sample characteristic inputs of different hydraulic system components, t moment sample characteristic input set +.> Sample characteristic input variables, y, representing hydraulic system components i Representing the sum of the energy consumption acquisition values of the hydraulic system, < + >>Representing a real space, n representing the number of samples, and m representing the number of sample input characteristic variables;
and 1.2, supplementing missing values in the original characteristic data set based on a KNN algorithm, and removing noise points and abnormal values in the original characteristic data set based on a density clustering method to obtain a standard characteristic data set.
According to a specific implementation manner of the embodiment of the present disclosure, the step 1.2 specifically includes:
step 1.2.1, calculating and complementing Euclidean distance between two adjacent characteristic data, wherein the Euclidean distance is expressed as
Wherein, l represents a time index of the sample data;
and 1.2.2, calculating the centroid distance of each characteristic data from the cluster, and rejecting the characteristic data exceeding the distance threshold value theta as an abnormal value and a standard characteristic data set.
According to a specific implementation manner of the embodiment of the present disclosure, the step 2 specifically includes:
step 2.1, carrying out mechanism analysis on the energy consumption related state characteristics of the hydraulic system of the forging forming equipment, and primarily selecting related characteristics;
and 2.2, filtering the related features to obtain the maximum information coefficient of each related feature, and removing the related features with the maximum information coefficients lower than the correlation threshold value to obtain the energy consumption data set.
According to a specific implementation manner of the embodiment of the present disclosure, the step 3 specifically includes:
step 3.1, for the time sequence data in the energy consumption data set, obtaining a sliding time window step by adopting Fourier transformation, and obtaining the time sequence data set by taking the length of one period as the sliding time window step
Where N represents a time-series range and k represents a time-series number.
Step 3.2, for static data in the energy consumption data set, clustering the characteristic states after resampling, classifying and evaluating the operation condition of the current equipment, and calculating Euclidean distances { d between every two n samples by adopting a hierarchical clustering method ij Distance matrix d= [ D ] ij ]Hierarchical clustering is combined according to two types which are nearest to each other, and the minimum distance expression is as follows
mindist＝min dist(X i ,Xj)＝min[d ij ]
And marking different cluster marks after classification evaluation is completed as a static sub-data set, and forming all the static sub-data sets into a static data set.
According to a specific implementation manner of the embodiment of the present disclosure, the step 4 specifically includes:
step 4.1, analyzing the power of each sub-component of the hydraulic system to obtain a power bonding diagram;
step 4.2, establishing a mechanism simulation model according to the power bonding diagram to obtain a first energy consumption predicted value M 1 。
According to a specific implementation manner of the embodiment of the present disclosure, the step 5 specifically includes:
step 5.1, for each static sub-data set in the static data set, transforming the data sequence by adopting a CNN one-dimensional convolution kernel feature extraction mode, wherein the one-dimensional convolution calculation formula is as follows:
where N is the length of f (m), S (N) is the data sequence of the characteristic transformation after convolution, f represents the amplitude of the input signal, and g represents the amplitude of the output signal.
Step 5.2, selecting three basic learners to construct an enhanced machine learning model, wherein the enhanced machine learning model comprises a multi-layer perceptron, a support vector machine and a gradient lifting decision tree which are sequentially stacked;
step 5.3, inputting the data sequence into the enhanced machine learning model for training, and inputting the data sequence which is not in accordance with the preset condition in the previous layer training as a new data set into the next layer for training, wherein the specific process is as follows:
Dataset (i+1) ＝Ψ(Dataset (i) )
wherein, dataset (i) To input the training samples of the i-th layer, the expression form of the conversion rule is as follows:
MAPE is a relative absolute value average error, and the calculation formula is as follows:
wherein y is i ' is the predicted value of energy consumption, y i Is the actual value of energy consumption;
step 5.4, performing energy consumption prediction on the three trained basic learners, and performing linear regression element learning on the respective output results to obtain a second energy consumption predicted value M 2 ；
Step 5.5, for the time sequence data set, establishing a GRU-reinforcement learning frame to extract time sequence characteristics, wherein the forward propagation formula of the GRU-reinforcement learning frame is as follows
r t ＝σ(W r ·[h t ,x t ]+b r )
z t ＝σ(W z ·[h t ,x t ]+b z )
Wherein W is r ，W z ，W h And b r ，b z ，b h Weight matrix and bias, r, are reset gate, update gate and calculate hidden state, respectively t And z t To reset the gate and update the gate outputs, h' t For candidate hidden state of cell, h t A cell hidden state;
step 5.6, after the time sequence feature extraction is completed, outputting the energy consumption value through the full connection layer at the time t-1Calculating the loss by a relative error expressed as
Meanwhile, calculating an error gradient at the time t-1, obtaining a learning parameter by adopting a reinforcement learning pi function, multiplying the learning parameter by the gradient, and updating the parameter by gradient descent, wherein the pi function is defined as follows:
π(a t |s t )＝μ(s t |θ μ )
the energy consumption at time t+1 is then predicted by the new parameters and the relative error is calculatedCalculating a prize function value r t+1 If the GRU prediction effect after reinforcement learning updating is better than that of GRU prediction effect without reinforcement learning updating, rewarding is obtained, and if not, punishment is obtained, wherein the rewarding function formula is defined as follows:
finally, obtaining a third energy consumption predicted value from the time sequence data set
Step 5.7, predicting output M for mechanical energy consumption 1 And static dataset energy consumption prediction output M 2 Obtaining energy consumption predicted value M from time sequence data set 3 Outputting by a linear regression element learning method to obtain a target energy consumption predicted value E p And monitoring whether the target energy consumption predicted value exceeds an energy consumption threshold in real time, wherein the expression of the target energy consumption predicted value is
E p ＝ω 1 M 1 +ω 2 M 2 +ω 3 M 3 +b
Wherein omega 1 Representation omega 2 Representation omega 3 The expression "b" means.
In a second aspect, embodiments of the present disclosure provide a forge forming apparatus hydraulic system energy consumption prediction system comprising:
the acquisition module is used for acquiring the state characteristics of the forging forming equipment, acquiring a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set;
the selection module is used for removing the characteristics of which the degree of the correlation with the energy consumption is lower than a correlation threshold value in the standard characteristic data set through mechanism analysis and a filter to obtain an energy consumption data set;
the reconstruction module is used for obtaining sliding time window step length by adopting Fourier transformation on time sequence data in the energy consumption data set to form the time sequence data set, clustering characteristic states after resampling on static data in the energy consumption data set, and classifying and evaluating the operation working condition of the current equipment to form the static data set;
the first prediction module is used for analyzing the power energy flow and establishing a mechanism simulation model to obtain a first energy consumption predicted value;
the second prediction module is used for respectively inputting the time sequence data set and the static data set into different prediction models, fusing the first energy consumption prediction, constructing an energy consumption prediction model with mechanism and data driving fusion, and obtaining and monitoring target energy consumption prediction.
The hydraulic system energy consumption prediction scheme of the forging forming equipment in the embodiment of the disclosure comprises the following steps: step 1, collecting state characteristics of forging forming equipment, obtaining a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set; step 2, removing the features with the degree of correlation with the energy consumption lower than a correlation threshold value in the standard feature data set through mechanism analysis and a filter to obtain an energy consumption data set; step 3, for time sequence data in the energy consumption data set, a sliding time window step length is obtained by adopting Fourier transformation to form a time sequence data set, for static data in the energy consumption data set, characteristic state clustering is carried out after resampling, and classification evaluation is carried out on the operation working condition of the current equipment to form a static data set; step 4, analyzing the power energy flow, and establishing a mechanism simulation model to obtain a first energy consumption predicted value; and 5, respectively inputting the time sequence data set and the static data set into different prediction models, fusing the first energy consumption prediction, constructing an energy consumption prediction model fusing mechanism and data driving, and obtaining and monitoring target energy consumption prediction.
The beneficial effects of the embodiment of the disclosure are that: according to the scheme, a multi-model integrated learning framework is constructed, an integration mechanism is integrated in the existing intelligent prediction technology, thinking and collective decision of a human brain are referenced, and data information is subjected to fusion processing through complementation, enhancement, stacking and fusion of models and combined with a mechanism method, so that the efficiency, precision and stability of energy consumption prediction are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and other drawings may be obtained according to these drawings without inventive effort to a person of ordinary skill in the art.
Fig. 1 is a schematic flow chart of a method for predicting energy consumption of a hydraulic system of forging forming equipment according to an embodiment of the disclosure;
FIG. 2 is a schematic diagram of a specific implementation flow of a hydraulic system energy consumption prediction method for forging forming equipment according to an embodiment of the disclosure;
FIG. 3 is a corresponding system module distribution diagram of a hydraulic system energy consumption prediction method for a forging forming apparatus according to an embodiment of the present disclosure;
fig. 4 is a schematic structural diagram of an energy consumption prediction system of a hydraulic system of forging forming equipment according to an embodiment of the present disclosure.
Detailed Description
Embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
Other advantages and effects of the present disclosure will become readily apparent to those skilled in the art from the following disclosure, which describes embodiments of the present disclosure by way of specific examples. It will be apparent that the described embodiments are merely some, but not all embodiments of the present disclosure. The disclosure may be embodied or practiced in other different specific embodiments, and details within the subject specification may be modified or changed from various points of view and applications without departing from the spirit of the disclosure. It should be noted that the following embodiments and features in the embodiments may be combined with each other without conflict. All other embodiments, which can be made by one of ordinary skill in the art without inventive effort, based on the embodiments in this disclosure are intended to be within the scope of this disclosure.
It is noted that various aspects of the embodiments are described below within the scope of the following claims. It should be apparent that the aspects described herein may be embodied in a wide variety of forms and that any specific structure and/or function described herein is merely illustrative. Based on the present disclosure, one skilled in the art will appreciate that one aspect described herein may be implemented independently of any other aspect, and that two or more of these aspects may be combined in various ways. For example, an apparatus may be implemented and/or a method practiced using any number of the aspects set forth herein. In addition, such apparatus may be implemented and/or such methods practiced using other structure and/or functionality in addition to one or more of the aspects set forth herein.
It should also be noted that the illustrations provided in the following embodiments merely illustrate the basic concepts of the disclosure by way of illustration, and only the components related to the disclosure are shown in the drawings and are not drawn according to the number, shape and size of the components in actual implementation, and the form, number and proportion of the components in actual implementation may be arbitrarily changed, and the layout of the components may be more complicated.
In addition, in the following description, specific details are provided in order to provide a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects may be practiced without these specific details.
The embodiment of the disclosure provides a hydraulic system energy consumption prediction method for forging forming equipment, which can be applied to equipment energy consumption prediction process in industrial manufacturing scenes.
Referring to fig. 1, a flow chart of a hydraulic system energy consumption prediction method for forging forming equipment is provided in an embodiment of the disclosure. As shown in fig. 1 and 2, the method mainly comprises the following steps:
step 1, collecting state characteristics of forging forming equipment, obtaining a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set;
further, the step 1 specifically includes:
step 1.1, numbering hydraulic system components such as a hydraulic pump, a servo valve, a hydraulic actuator, an energy accumulator, a pipeline and the like in a state to be monitored based on a physical structure of intelligent equipment of a forging program, and designing a data acquisition device to acquire different characteristic data of the hydraulic system to form an original characteristic data set Wherein X is i Representing a set of sample characteristic inputs of different hydraulic system components, t moment sample characteristic input set +.> Sample characteristic input variables, y, representing hydraulic system components i Representing the sum of the energy consumption acquisition values of the hydraulic system, < + >>Representing a real space, n representing the number of samples, and m representing the number of sample input characteristic variables;
and 1.2, supplementing missing values in the original characteristic data set based on a KNN algorithm, and removing noise points and abnormal values in the original characteristic data set based on a density clustering method to obtain a standard characteristic data set.
Further, the step 1.2 specifically includes:
step 1.2.1, calculating and complementing Euclidean distance between two adjacent characteristic data, wherein the Euclidean distance is expressed as
Wherein, l represents a time index of the sample data;
and 1.2.2, calculating the centroid distance of each characteristic data from the cluster, and rejecting the characteristic data exceeding the distance threshold value theta as an abnormal value and a standard characteristic data set.
In specific implementation, the distribution diagram of the entity system module corresponding to the energy consumption prediction method of the hydraulic system of the forging forming equipment can be shown as in fig. 3, the components such as the hydraulic pump, the servo valve, the hydraulic actuator, the energy accumulator, the pipeline and the like in the state to be monitored can be numbered based on the entity structure of the intelligent equipment of the forging program, and a data acquisition device is designed, wherein the data acquisition device comprises a temperature sensor, a current sensor, a flowmeter and the like, different acquisition devices can acquire different characteristics of the hydraulic system, and raw characteristic data is acquired as follows Wherein X is i Representing a set of sample characteristic inputs of different hydraulic system components, t moment sample characteristic input set +.> Sample characteristic input variables representing hydraulic system components, such as hydraulic pump pressure, flow, electrical signals and other data variables; y is i Representing the sum of the energy consumption acquisition values of the hydraulic system, < + >>Representing real space, n representing the number of samples, and m representing the number of sample input characteristic variables.
And then supplementing missing values in the original characteristic data set based on a KNN algorithm, removing noise points and abnormal values in the original characteristic data set based on a density clustering method to obtain a standard characteristic data set, and specifically supplementing the missing values based on the KNN algorithm, and removing the noise points and the abnormal values based on the density clustering method.
The distance measurement mode in the KNN method is defined by Euclidean distance, and the formula is as follows:
the method for detecting the noise points and the abnormal values by clustering is as follows, and the abnormal points are obtained by measuring the distance between the centers of mass of the clusters and exceeding a certain distance threshold value theta:
||X j -μ j ||>θ。
step 2, removing the features with the degree of correlation with the energy consumption lower than a correlation threshold value in the standard feature data set through mechanism analysis and a filter to obtain an energy consumption data set;
on the basis of the above embodiment, the step 2 specifically includes:
step 2.1, carrying out mechanism analysis on the energy consumption related state characteristics of the hydraulic system of the forging forming equipment, and primarily selecting related characteristics;
and 2.2, filtering the related features to obtain the maximum information coefficient of each related feature, and removing the related features with the maximum information coefficients lower than the correlation threshold value to obtain the energy consumption data set.
In specific implementation, the specific process of selecting the feature data in the standard feature data set may be as follows:
s21, carrying out mechanism analysis on the energy consumption related state characteristics of the hydraulic system of the forging forming equipment (initially selecting related characteristics such as oil outlet pressure of the plunger pump, output flow of the plunger pump, motor rotation speed, temperature and the like).
S22, filtering the relevant state characteristics of the energy consumption of the hydraulic system of the forging forming equipment, further filtering the pre-characteristics of the S21, measuring the importance of the characteristics by adopting a Maximum Information Coefficient (MIC) as the filter, wherein the MIC is closer to 1, the maximum degree of correlation between the variables X and Y is represented, and the calculation expression is as follows:
and then comparing the MIC value with a preset correlation threshold value, and eliminating the correlation characteristic with the maximum information coefficient lower than the correlation threshold value to obtain an energy consumption data set.
Step 3, for time sequence data in the energy consumption data set, a sliding time window step length is obtained by adopting Fourier transformation to form a time sequence data set, for static data in the energy consumption data set, characteristic state clustering is carried out after resampling, and classification evaluation is carried out on the operation working condition of the current equipment to form a static data set;
on the basis of the above embodiment, the step 3 specifically includes:
step 3.1, for the time sequence data in the energy consumption data set, obtaining a sliding time window step by adopting Fourier transformation, and obtaining the time sequence data set by taking the length of one period as the sliding time window step
Where N represents a time-series range and k represents a time-series number.
Step 3.2, for static data in the energy consumption data set, clustering the characteristic states after resampling, classifying and evaluating the operation condition of the current equipment, and calculating Euclidean distances { d between every two n samples by adopting a hierarchical clustering method ij Distance matrix d= [ D ] ij ]Hierarchical clustering is combined according to two types which are nearest to each other, and the minimum distance expression is as follows
mindist＝min dist(X i ,X j )＝min[d ij ]
And marking different cluster marks after classification evaluation is completed as a static sub-data set, and forming all the static sub-data sets into a static data set.
In specific implementation, the specific steps of reconstructing the energy consumption data set may be as follows:
s31, dividing the original data set into a time sequence data set and a static data set, obtaining a sliding time window step length by adopting Fourier transformation for the time sequence data set, taking the length of one period as the sliding time window step length,
s32, for a static data set, carrying out resampling, then carrying out feature state clustering, carrying out classification evaluation on the operation condition of the current equipment, adopting hierarchical clustering in a clustering mode, and calculating Euclidean distance { d) between every two samples ij Distance matrix d= [ D ] ij ]Hierarchical clustering is combined according to two types which are nearest to each other, and the minimum distance expression is as follows
mindist＝min dist(X i ,X j )＝min[d ij ]
Different clusters after classification evaluation are marked as static sub-data sets, and each cluster is a data set with similar working condition characteristics of the hydraulic system and is marked as G 1 ,G 2 ,...,G k K is the number of clusters.
Step 4, analyzing the power energy flow, and establishing a mechanism simulation model to obtain a first energy consumption predicted value;
further, the step 4 specifically includes:
step 4.1, analyzing the power of each sub-component of the hydraulic system to obtain a power bonding diagram;
step 4.2, establishing a mechanism simulation model according to the power bonding diagram to obtain a first energy consumption predicted value M 1 。
And 5, respectively inputting the time sequence data set and the static data set into different prediction models, fusing the first energy consumption prediction, constructing an energy consumption prediction model fusing mechanism and data driving, and obtaining and monitoring target energy consumption prediction.
On the basis of the above embodiment, the step 5 specifically includes:
step 5.1, for each static sub-data set in the static data set, transforming the data sequence by adopting a CNN one-dimensional convolution kernel feature extraction mode, wherein the one-dimensional convolution calculation formula is as follows:
wherein N is the length of f (m), S (N) is the data sequence of the characteristic transformation after convolution, f represents the amplitude of an input signal, and g represents the amplitude of an output signal;
step 5.2, selecting three basic learners to construct an enhanced machine learning model, wherein the enhanced machine learning model comprises a multi-layer perceptron, a support vector machine and a gradient lifting decision tree which are sequentially stacked;
step 5.3, inputting the data sequence into the enhanced machine learning model for training, and inputting the data sequence which is not in accordance with the preset condition in the previous layer training as a new data set into the next layer for training, wherein the specific process is as follows:
Dataset (i+1) ＝Ψ(Dataset (i) )
wherein, dataset (i) To input the training samples of the i-th layer, the expression form of the conversion rule is as follows:
MAPE is a relative absolute value average error, and the calculation formula is as follows:
wherein y is i ' is the predicted value of energy consumption, y i Is the actual value of energy consumption;
step 5.4, the three trained basic learners are subjected to energy consumption prediction, and the output results are subjected to linear returnReturning to the original state and learning to obtain a second energy consumption predicted value M 2 ；
Step 5.5, for the time sequence data set, establishing a GRU-reinforcement learning frame to extract time sequence characteristics, wherein the forward propagation formula of the GRU-reinforcement learning frame is as follows
r t ＝σ(W r ·[h t ,x t ]+b r )
z t ＝σ(W z ·[h t ,x t ]+b z )
Wherein W is r ，W z ，W h And b r ，b z ，b h Weight matrix and bias, r, are reset gate, update gate and calculate hidden state, respectively t And z t To reset the gate and update the gate outputs, h' t For candidate hidden state of cell, h t A cell hidden state;
step 5.6, after the time sequence feature extraction is completed, outputting the energy consumption value through the full connection layer at the time t-1Calculating the loss by a relative error expressed as
Meanwhile, calculating an error gradient at the time t-1, obtaining a learning parameter by adopting a reinforcement learning pi function, multiplying the learning parameter by the gradient, and updating the parameter by gradient descent, wherein the pi function is defined as follows:
π(a t |s t )＝μ(s t |θ μ )
then predicting the energy consumption at the time t+1 by the new parameters, andand calculate the relative errorCalculating a prize function value r t+1 If the GRU prediction effect after reinforcement learning updating is better than that of GRU prediction effect without reinforcement learning updating, rewarding is obtained, and if not, punishment is obtained, wherein the rewarding function formula is defined as follows:
finally, obtaining a third energy consumption predicted value from the time sequence data set
Step 5.7, predicting output M for mechanical energy consumption 1 And static dataset energy consumption prediction output M 2 Obtaining energy consumption predicted value M from time sequence data set 3 Outputting by a linear regression element learning method to obtain a target energy consumption predicted value E p And monitoring whether the target energy consumption predicted value exceeds an energy consumption threshold in real time, wherein the expression of the target energy consumption predicted value is
E p ＝ω 1 M 1 +ω 2 M 2 +ω 3 M 3 +b
Wherein omega 1 Representation omega 2 Representation omega 3 The expression "b" means.
In specific implementation, the specific process of respectively inputting the time sequence data set and the static data set into different prediction models to perform prediction can be as follows:
s51, for static data set G completing working condition classification 1 ,G 2 ,...,G k For each sub-data set G l (l=1, 2,., k), then the one-dimensional convolution calculation formula is as follows, using the manner of CNN one-dimensional convolution kernel feature extraction:
where N is the length of f (m) and S (N) is the data sequence of the post-convolution feature transformation.
S52, an enhanced machine learning framework is established, three basic learners, namely a multi-layer perceptron (MLP), a Support Vector Machine (SVM) and a Gradient Boosting Decision Tree (GBDT) are selected. The enhanced principle is as follows, a stacked structure is adopted, three basic learners are respectively a first layer, a second layer and a third layer of a frame, the data with poor training of the front layer becomes a new data set and is combined to the rear layer for training,
Dataset (i+1) ＝Ψ(Dataset (i) )
wherein, dataset (i) To input the training samples of the i-th layer, the expression form of the conversion rule is as follows:
MAPE is the average error of relative absolute value, let y i ' is the predicted value of energy consumption, y i The calculation formula is as follows for the actual value of energy consumption:
finally, carrying out energy consumption prediction on the three trained basic learners, and obtaining an energy consumption prediction result M by respectively outputting results and then carrying out linear regression element learning 2 。
S53, for the energy time sequence data set, a GRU-reinforcement learning framework is established, time sequence characteristics are extracted through a GRU network, compared with LSTM, a reset gate is arranged, a forgetting gate and an input gate are combined into an updating gate, an output gate is removed, and a cell state and a cell hidden state are fused. The forward propagation formula is as follows:
r t ＝σ(W r ·[h t ,x t ]+b r )
z t ＝σ(W z ·[h t ,x t ]+b z )
wherein W is r ，W z ，W h And b r ，b z ，b h Weight matrix and bias, r, are reset gate, update gate and calculate hidden state, respectively t And z t To reset the gate and update the gate outputs, h' t For candidate hidden state of cell, h t Cell hidden state.
After the time sequence feature is extracted, the energy consumption value is output through the full connection layer at the time t-1The loss is calculated by the relative error, which is calculated as follows:
meanwhile, calculating an error gradient at the time t-1, obtaining a learning parameter by adopting a reinforcement learning pi function, multiplying the learning parameter by the gradient, and updating the parameter by gradient descent, wherein the pi function is defined as follows:
π(a t |s t )＝μ(s t |θ μ )
then, the energy consumption at time t+1 is predicted by the new parameters, and the relative error is calculatedCalculating a prize function value r t+1 If the GRU prediction effect updated through reinforcement learning is better, rewards are obtained, and if worse, penalties are obtained. The bonus function formula is defined as follows:
finally, obtaining the energy consumption predicted value from the time sequence data set
S54, predicting output M for mechanical energy consumption 1 And static dataset energy consumption prediction output M 2 Obtaining energy consumption predicted value M from time sequence data set 3 Outputting by a linear regression element learning method to obtain a final energy consumption predicted value E p ，
E p ＝ω 1 M 1 +ω 2 M 2 +ω 3 M 3 +b
Based on the final energy consumption predicted value sequence, energy consumption monitoring and early warning are carried out, for example, the forging forming equipment is controlled to be stopped when the energy consumption is high, and the work load and the pressure are reduced.
According to the energy consumption prediction method for the hydraulic system of the forging forming equipment, the data acquisition system of the forging forming equipment is involved, and the data are cleaned to obtain a required data set; secondly, screening to obtain forging forming equipment characteristic data of an input model by means of mechanism analysis and a filter, wherein the filter adopts maximum information coefficient calculation; again, the data set is split into two parts, one part containing the data set of static discontinuous operating characteristics and energy consumption values and one part containing the time-series energy consumption data set. For the static data set, classifying working conditions by adopting a hierarchical clustering method to obtain a sub data set; for a time sequence data set, obtaining the step length of a time window by adopting Fourier transform; further, based on mechanism energy consumption prediction of the forging forming equipment, analyzing power energy flow, and establishing a mechanism simulation model to obtain an output result; then, extracting features of a static data set subjected to working condition classification by adopting a CNN one-dimensional convolution mode, further designing a multi-layer model to increase an integrated learning framework, continuously learning samples with poor learning of a former layer model to a next layer model, and finally outputting through linear regression element learning; and extracting time sequence characteristics from the time sequence data set through the GRU network and carrying out energy consumption prediction through the full connection layer, wherein in the process, the LSTM network is optimized by reinforcement learning, so that the prediction error is reduced. And finally, outputting a method for carrying out linear regression element learning by fusing the mechanism and the data prediction result, so that energy consumption monitoring and early warning are realized, and the prediction efficiency and accuracy are improved.
In correspondence with the above method embodiment, referring to fig. 4, the disclosed embodiment also provides a forging forming equipment hydraulic system energy consumption prediction system 40, comprising:
the acquisition module 401 is used for acquiring the state characteristics of the forging forming equipment, acquiring a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set;
a selection module 402, configured to reject, through mechanism analysis and a filter, features in the standard feature data set that have a degree of correlation with energy consumption lower than a correlation threshold, to obtain an energy consumption data set;
a reconstruction module 403, configured to apply the time series data in the energy consumption data set, obtain a sliding time window step size by using fourier transform, form a time series data set, and perform classification evaluation on the operation condition of the current device after resampling on the static data in the energy consumption data set to form a static data set;
the first prediction module 404 is configured to analyze the power energy flow, and build a mechanism simulation model to obtain a first energy consumption prediction value;
the second prediction module 405 is configured to input the time-series dataset and the static dataset into different prediction models, fuse the first energy consumption prediction, construct an energy consumption prediction model with mechanism and data driven fusion, and obtain and monitor a target energy consumption prediction.
The system shown in fig. 4 may correspondingly execute the content in the foregoing method embodiment, and the portions not described in detail in this embodiment refer to the content described in the foregoing method embodiment, which is not described herein again.
It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.
The foregoing is merely specific embodiments of the disclosure, but the protection scope of the disclosure is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the disclosure are intended to be covered by the protection scope of the disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.
Claims (8)
1. The method for predicting the energy consumption of the hydraulic system of the forging forming equipment is characterized by comprising the following steps of:
step 1, collecting state characteristics of forging forming equipment, obtaining a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set;
step 2, removing the features with the degree of correlation with the energy consumption lower than a correlation threshold value in the standard feature data set through mechanism analysis and a filter to obtain an energy consumption data set;
step 3, for time sequence data in the energy consumption data set, a sliding time window step length is obtained by adopting Fourier transformation to form a time sequence data set, for static data in the energy consumption data set, characteristic state clustering is carried out after resampling, and classification evaluation is carried out on the operation working condition of the current equipment to form a static data set;
step 4, analyzing the power energy flow, and establishing a mechanism simulation model to obtain a first energy consumption predicted value;
and 5, respectively inputting the time sequence data set and the static data set into different prediction models, fusing the first energy consumption prediction, constructing an energy consumption prediction model fusing mechanism and data driving, and obtaining and monitoring target energy consumption prediction.
2. The method according to claim 1, wherein the step 1 specifically comprises:
step 1.1, numbering hydraulic system components such as a hydraulic pump, a servo valve, a hydraulic actuator, an energy accumulator, a pipeline and the like in a state to be monitored based on a physical structure of intelligent equipment of a forging program, and designing a data acquisition device to acquire different characteristic data of the hydraulic system to formRaw feature data set Wherein X is i Representing a set of sample characteristic inputs of different hydraulic system components, t moment sample characteristic input set +.> Sample characteristic input variables, y, representing hydraulic system components i Representing the sum of the energy consumption acquisition values of the hydraulic system, < + >>Representing a real space, n representing the number of samples, and m representing the number of sample input characteristic variables;
and 1.2, supplementing missing values in the original characteristic data set based on a KNN algorithm, and removing noise points and abnormal values in the original characteristic data set based on a density clustering method to obtain a standard characteristic data set.
3. The method according to claim 2, wherein the step 1.2 specifically comprises:
step 1.2.1, calculating and complementing Euclidean distance between two adjacent characteristic data, wherein the Euclidean distance is expressed as
Wherein, l represents a time index of the sample data;
and 1.2.2, calculating the centroid distance of each characteristic data from the cluster, and rejecting the characteristic data exceeding the distance threshold value theta as an abnormal value and a standard characteristic data set.
4. A method according to claim 3, wherein said step 2 comprises:
step 2.1, carrying out mechanism analysis on the energy consumption related state characteristics of the hydraulic system of the forging forming equipment, and primarily selecting related characteristics;
and 2.2, filtering the related features to obtain the maximum information coefficient of each related feature, and removing the related features with the maximum information coefficients lower than the correlation threshold value to obtain the energy consumption data set.
5. The method according to claim 4, wherein the step 3 specifically includes:
step 3.1, for the time sequence data in the energy consumption data set, obtaining a sliding time window step by adopting Fourier transformation, and obtaining the time sequence data set by taking the length of one period as the sliding time window step
Wherein N represents a time sequence range, i.e. the dimension of the sample vector, and k represents a time sequence number;
step 3.2, for static data in the energy consumption data set, clustering the characteristic states after resampling, classifying and evaluating the operation condition of the current equipment, and calculating Euclidean distances { d between every two n samples by adopting a hierarchical clustering method ij Distance matrix d= [ D ] ij ]Hierarchical clustering is combined according to two types which are nearest to each other, and the minimum distance expression is as follows
mindist＝min dist(X i ,X j )＝min[d ij ]
And marking different cluster marks after classification evaluation is completed as a static sub-data set, and forming all the static sub-data sets into a static data set.
6. The method according to claim 5, wherein the step 4 specifically includes:
step 4.1, analyzing the power of each sub-component of the hydraulic system to obtain a power bonding diagram;
step 4.2, establishing a mechanism simulation model according to the power bonding diagram to obtain a first energy consumption predicted value M 1 。
7. The method according to claim 6, wherein the step 5 specifically comprises:
step 5.1, for each static sub-data set in the static data set, transforming the data sequence by adopting a CNN one-dimensional convolution kernel feature extraction mode, wherein the one-dimensional convolution calculation formula is as follows:
wherein N is the length of f (m), S (N) is the data sequence of the characteristic transformation after convolution, f represents the amplitude of an input signal, and g represents the amplitude of an output signal;
step 5.2, selecting three basic learners to construct an enhanced machine learning model, wherein the enhanced machine learning model comprises a multi-layer perceptron, a support vector machine and a gradient lifting decision tree which are sequentially stacked;
step 5.3, inputting the data sequence into the enhanced machine learning model for training, and inputting the data sequence which is not in accordance with the preset condition in the previous layer training as a new data set into the next layer for training, wherein the specific process is as follows:
Dataset (i+1) ＝Ψ(Dataset (i) )
wherein, dataset (i) To input the training samples of the i-th layer, the expression form of the conversion rule is as follows:
MAPE is a relative absolute value average error, and the calculation formula is as follows:
wherein y is i ' is the predicted value of energy consumption, y i Is the actual value of energy consumption;
step 5.4, performing energy consumption prediction on the three trained basic learners, and performing linear regression element learning on the respective output results to obtain a second energy consumption predicted value M 2 ；
Step 5.5, for the time sequence data set, establishing a GRU-reinforcement learning frame to extract time sequence characteristics, wherein the forward propagation formula of the GRU-reinforcement learning frame is as follows
r t ＝σ(W r ·[h t ,x t ]+b r )
z t ＝σ(W z ·[h t ,x t ]+b z )
Wherein W is r ，W z ，W h And b r ，b z ，b h Weight matrix and bias, r, are reset gate, update gate and calculate hidden state, respectively t And z t To reset the gate and update the gate outputs, h' t For candidate hidden state of cell, h t A cell hidden state;
step 5.6, extracting the time sequence characteristicsAfter the energy consumption value is taken out, at the time t-1, the energy consumption value is output through the full connection layerCalculating the loss by a relative error expressed as
Meanwhile, calculating an error gradient at the time t-1, obtaining a learning parameter by adopting a reinforcement learning pi function, multiplying the learning parameter by the gradient, and updating the parameter by gradient descent, wherein the pi function is defined as follows:
π(a t |s t )＝μ(s t |θ μ )
the energy consumption at time t+1 is then predicted by the new parameters and the relative error is calculatedCalculating a prize function value r t+1 If the GRU prediction effect after reinforcement learning updating is better than that of GRU prediction effect without reinforcement learning updating, rewarding is obtained, and if not, punishment is obtained, wherein the rewarding function formula is defined as follows:
finally, obtaining a third energy consumption predicted value from the time sequence data set
Step 5.7, predicting output M for mechanical energy consumption 1 And static dataset energy consumption prediction output M 2 Obtaining energy consumption predicted value M from time sequence data set 3 Outputting by a linear regression element learning method to obtain a target energy consumption predicted value E p And monitoring in real time whether it exceeds an energy consumption threshold, wherein the targetThe expression of the energy consumption predicted value is
E p ＝ω 1 M 1 +ω 2 M 2 +ω 3 M 3 +b
Wherein omega 1 Representation omega 2 Representation omega 3 The expression "b" means.
8. A hydraulic system energy consumption prediction system for a forging forming apparatus, comprising:
the acquisition module is used for acquiring the state characteristics of the forging forming equipment, acquiring a plurality of characteristic data through a multi-sensor network and cleaning the characteristic data to obtain a standard characteristic data set;
the selection module is used for removing the characteristics of which the degree of the correlation with the energy consumption is lower than a correlation threshold value in the standard characteristic data set through mechanism analysis and a filter to obtain an energy consumption data set;
the reconstruction module is used for obtaining sliding time window step length by adopting Fourier transformation on time sequence data in the energy consumption data set to form the time sequence data set, clustering characteristic states after resampling on static data in the energy consumption data set, and classifying and evaluating the operation working condition of the current equipment to form the static data set;
the first prediction module is used for analyzing the power energy flow and establishing a mechanism simulation model to obtain a first energy consumption predicted value;
the second prediction module is used for respectively inputting the time sequence data set and the static data set into different prediction models, fusing the first energy consumption prediction, constructing an energy consumption prediction model with mechanism and data driving fusion, and obtaining and monitoring target energy consumption prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311624189.7A CN117669373A (en) | 2023-11-30 | 2023-11-30 | Energy consumption prediction method and system for hydraulic system of forging forming equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311624189.7A CN117669373A (en) | 2023-11-30 | 2023-11-30 | Energy consumption prediction method and system for hydraulic system of forging forming equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117669373A true CN117669373A (en) | 2024-03-08 |
Family
ID=90070709
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311624189.7A Pending CN117669373A (en) | 2023-11-30 | 2023-11-30 | Energy consumption prediction method and system for hydraulic system of forging forming equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117669373A (en) |
-
2023
- 2023-11-30 CN CN202311624189.7A patent/CN117669373A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ren et al. | Multi-bearing remaining useful life collaborative prediction: A deep learning approach | |
| CN114048600A (en) | Digital twin-driven multi-model fusion industrial system anomaly detection method | |
| Li et al. | A new lightweight deep neural network for surface scratch detection | |
| CN107273924B (en) | Multi-data fusion power plant fault diagnosis method based on fuzzy clustering analysis | |
| CN114282443B (en) | Residual service life prediction method based on MLP-LSTM supervised joint model | |
| CN113837000A (en) | Small sample fault diagnosis method based on task sequencing meta-learning | |
| CN111649779B (en) | Oil well oil content and total flow rate measuring method based on dense neural network and application | |
| CN114218872B (en) | DBN-LSTM semi-supervised joint model-based residual service life prediction method | |
| CN113325721B (en) | Model-free adaptive control method and system for industrial system | |
| CN114015825B (en) | Method for monitoring abnormal state of blast furnace heat load based on attention mechanism | |
| CN109634233A (en) | Industrial big data intellectual analysis decision-making technique, readable storage medium storing program for executing and terminal | |
| Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
| CN110757510A (en) | Method and system for predicting remaining life of robot | |
| CN115859077A (en) | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions | |
| CN115096627A (en) | Method and system for fault diagnosis and operation and maintenance in manufacturing process of hydraulic forming intelligent equipment | |
| CN109614451A (en) | Industrial big data intellectual analysis decision making device | |
| Liu et al. | Deep & attention: A self-attention based neural network for remaining useful lifetime predictions | |
| Maged et al. | Recognition of abnormal patterns in industrial processes with variable window size via convolutional neural networks and AdaBoost | |
| CN114563671A (en) | High-voltage cable partial discharge diagnosis method based on CNN-LSTM-Attention neural network | |
| Yu et al. | Deep unLSTM network: Features with memory information extracted from unlabeled data and their application on industrial unsupervised industrial fault detection | |
| CN116958553B (en) | Lightweight plant point cloud segmentation method based on non-parametric attention and point-level convolution | |
| KR102366787B1 (en) | Real-time sliding window based anomaly detection system for multivariate data generated by manufacturing equipment | |
| CN112001115A (en) | Soft measurement modeling method of semi-supervised dynamic soft measurement network | |
| CN117060353A (en) | Fault diagnosis method and system for high-voltage direct-current transmission system based on feedforward neural network | |
| Techane et al. | Rotating machinery prognostics and application of machine learning algorithms: Use of deep learning with similarity index measure for health status prediction |
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 |