CN109894495B - Extruder anomaly detection method and system based on energy consumption data and Bayesian network - Google Patents
Extruder anomaly detection method and system based on energy consumption data and Bayesian network Download PDFInfo
- Publication number
- CN109894495B CN109894495B CN201910028040.XA CN201910028040A CN109894495B CN 109894495 B CN109894495 B CN 109894495B CN 201910028040 A CN201910028040 A CN 201910028040A CN 109894495 B CN109894495 B CN 109894495B
- Authority
- CN
- China
- Prior art keywords
- hydraulic pump
- energy consumption
- bayesian network
- evidence
- probability
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005265 energy consumption Methods 0.000 title claims abstract description 54
- 238000001514 detection method Methods 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 22
- 238000004458 analytical method Methods 0.000 claims abstract description 14
- 238000010276 construction Methods 0.000 claims abstract description 7
- 238000001125 extrusion Methods 0.000 claims description 63
- 238000012545 processing Methods 0.000 claims description 24
- 238000005070 sampling Methods 0.000 claims description 24
- 230000005856 abnormality Effects 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 19
- 230000007246 mechanism Effects 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 9
- 230000001149 cognitive effect Effects 0.000 claims description 3
- 238000012854 evaluation process Methods 0.000 abstract description 6
- 238000004364 calculation method Methods 0.000 abstract description 5
- 238000011156 evaluation Methods 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 description 1
- 229910052782 aluminium Inorganic materials 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Images
Landscapes
- Extrusion Moulding Of Plastics Or The Like (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses an extruder anomaly detection method and system based on energy consumption data and a Bayesian network, wherein the anomaly detection method introduces an energy consumption model analysis method, obtains key parameters influencing energy consumption by analyzing the energy consumption data of an extruder so as to determine the dependency relationship between nodes, and constructs an anomaly detection Bayesian network structure according to the nodes and the dependency relationship, and the Bayesian network structure is applied to anomaly detection of the extruder; meanwhile, compared with the existing Bayesian network construction method which needs repeated training and evaluation and has high complexity of a network model, the method has the advantages that a large number of redundant calculation and evaluation processes are reduced, the complexity of Bayesian network construction is reduced, the robustness and the high efficiency of the Bayesian network are improved, and the problems that a large number of redundant calculation and evaluation processes, high complexity of the network model, low detection precision and the like exist in the existing anomaly detection method of the extruder are solved.
Description
Technical Field
The invention relates to the technical field of extruder energy consumption detection, in particular to an extruder abnormity detection method and system based on energy consumption data and a Bayesian network.
Background
The extruder is used as a core device in the production process of the aluminum profile, the development of the extruder is gradually large, complex and automatic, and once abnormality or fault occurs in the production process, the extruder can cause greater production stop loss, more maintenance cost and even more serious safety accidents. In order to reduce the probability of the extruder breaking down, improve the stability and reliability of extruder equipment, improve the production efficiency and pay attention to the problem of extruder abnormity detection.
The existing abnormal detection technology of the extruder only detects whether the extruder is abnormal, can not detect the specific reasons of the abnormal occurrence of the extruder, and has low accuracy of the detection result;
in addition, in the anomaly detection of the mechanical system, an anomaly detection method based on the bayesian network is used, but since the structure learning method based on the evaluation function and the search algorithm and the structure learning method based on the node dependency relationship are mainly adopted, the following disadvantages exist: 1. the establishment process of the causal relationship among the parameters is not clear enough, the optimal Bayesian network structure can be found through multiple learning and evaluation processes, so that the complexity of the network model is increased continuously for improving the model accuracy, and the accuracy of the final experiment result of the network model with proper complexity is not high enough; 2. the conditional independence among the nodes needs to be calculated, and for the nodes with more determining factors, a great amount of redundant work is generated because the conditional probability among the nodes needs to be calculated to obtain the strong and weak dependence relationship, so that the workload is huge, and a great amount of storage space is consumed.
Disclosure of Invention
The invention provides an extruder anomaly detection method and system based on energy consumption data and a Bayesian network, and aims to solve the problems that a large number of redundant calculation and evaluation processes exist in the existing extruder anomaly detection method, the complexity of a network model is high, the detection accuracy is not high enough, and the like.
In order to achieve the above purpose, the technical means adopted is as follows:
an extruder anomaly detection method based on energy consumption data and a Bayesian network comprises the following steps:
s1, analyzing an energy flow mechanism of an extruder from the energy consumption angle in the using process of the extruder to obtain factors influencing energy consumption; the factors comprise piston diameter D, piston rod diameter D and theoretical flow Q of the hydraulic pumpmFlow Q of input extrusion cylinder, moving speed V of extrusion rod, and energy loss Delta P of hydraulic pumpmhEnergy loss delta P of hydraulic valve groupfhMechanical efficiency eta of hydraulic pumpmVolumetric efficiency eta of hydraulic pumpvVolumetric efficiency of the extrusion cylinder, total power P for the extrusion cycle, master cylinder pressure PaPump port pressure P of hydraulic pump, extrusion cylinder thrust F and extrusion cylinder input power PciRotating speed n of hydraulic pumpvAnd torque Tn；
S2, finding out the dependency relationship among the nodes of the Bayes network by combining the Bayes algorithm and the factors in the step S1, constructing a Bayes network structure and setting an observable variable set therein as an evidence variable set;
s3, collecting energy consumption data of the extruder as a sample data set, dividing the sample data set into a training data set and a testing data set, and then processing the energy consumption data;
s4, carrying out nonparametric estimation processing on the nodes of the evidence variable set to obtain probability density functions of all the nodes; obtaining a conditional probability table of all nodes by using the data of the training data set;
s5, carrying out Bayesian network reasoning by utilizing a Gibbs sampling algorithm to obtain a posterior probability about the occurrence of an anomaly under the currently input evidence variable;
s6, calculating the actual probability of the occurrence of the abnormality under the evidence variable currently input in the test data set, calculating the logarithmic ratio of the actual probability to the posterior probability in the step S5, setting a confidence coefficient according to the ratio, and judging whether the abnormality occurs according to the confidence coefficient; and if the abnormity occurs, calculating the posterior probability of the abnormity under different evidence variables, and defining the evidence variable corresponding to the maximum value of the posterior probability as the evidence variable causing the abnormity to occur.
Preferably, step S1 specifically includes:
the total power used by each extrusion process of the hydraulic system is expressed as:
P＝ΔPmh+ΔPfh+Pci＝UI
wherein, Δ PmhRepresenting the lost energy of the hydraulic pump, Δ PfhRepresenting the lost energy of the valve group of the hydraulic system, PciThe input power of the extrusion oil cylinder is represented, U represents the current voltage value, and I represents the current value;
input power P of hydraulic pumpmiIs represented by Pmi＝2·π·nv·Tn
Wherein n isvRepresenting the rotational speed of the pump shaft of the hydraulic pump, TnRepresenting the input torque of the hydraulic pump;
p for output power of hydraulic pumpmoIs represented by Pmo＝p·Qm
Wherein p represents the pump port pressure of the hydraulic pump, QmRepresenting the actual flow rate of the hydraulic pump, can be calculated by:
wherein, VmIs the traveling speed of the hydraulic pump, DmIs the diameter of the hydraulic pump;
the energy loss Δ P of the hydraulic pumpmhExpressed as:
ΔPmh＝Pmi-Pmo＝2·π·nv·Tn-p·Qm
the loss of energy consumption of the hydraulic valve group is expressed as Δ Pfh＝n·Δpf·ΔQf
Wherein n represents the number of valves, Δ pfRepresenting the valve port pressure difference, Δ QfRepresenting the valve port flow;
p for input power of extrusion cylinderciIs represented by Pci＝F·V
Wherein F represents the force to which the extrusion rod is subjected, and V represents the speed at which the extrusion rod moves;
Q＝Qm
wherein, in the formula: a represents the effective area of the extrusion oil cylinder; etamThe mechanical efficiency of the extrusion oil cylinder is shown;
ηvthe volumetric efficiency of the extrusion cylinder is shown; d represents the piston diameter; d represents the piston rod diameter; q represents the flow rate of the input extrusion oil cylinder; p is a radical ofaIndicates master cylinder pressure; indicating the volumetric efficiency of the squeeze cylinders.
Preferably, in the bayesian network structure in step S2, the piston diameter D, the piston rod diameter D and the theoretical flow rate Q of the hydraulic pumpmIs a fixed parameter; flow Q of input extrusion cylinder, moving speed V of extrusion rod, and energy loss Delta P of hydraulic pumpmhEnergy loss delta P of hydraulic valve groupfhHydraulic pumpMechanical efficiency etamVolumetric efficiency eta of hydraulic pumpvVolumetric efficiency of the extrusion cylinder, total power P for the extrusion cycle, master cylinder pressure PaPump port pressure P of hydraulic pump, extrusion cylinder thrust F and extrusion cylinder input power PciIs a variable with cognitive uncertainty, in which the master cylinder pressure paPump port pressure p of hydraulic pump, and rotation speed n of hydraulic pumpvAnd torque TnIs an observable variable.
Preferably, the data processing in step S3 includes:
s31, setting D as the set of all nodes, pi (D)i) Is DiOf parent node of (2), pi (D)i)＝{Dai}；
And S32, discretizing and symbolizing the value of the energy consumption data attribute.
Preferably, step S4 specifically includes:
s41, obtaining a set D { D ] by adopting a parzen window estimation method1，D2，D3，D4...DnIn each node DiWherein the kernel function is of the form:
pn(x) Represents the superposition of the respective sample-centered normal probability density functions:
where n denotes the size of the sample, hnRepresents the length of the window;
s42, dividing the training data into samples with different sizes for training, wherein n represents the size of the sample, hnIndicating the length of the window, adjusting n and hnObtaining the probability density function of all nodes;
and S43, obtaining a conditional probability table of all nodes with dependency relations by using the training data.
Preferably, step S5 specifically includes:
s51, utilizing Gibbs sampling algorithm to carry out Bayesian network inference, randomly selecting an evidence variable according to the result to be inferred, defining A', and then randomly generating a starting sample S1The starting sample S1Contains an evidence variable A';
s52, copying the last initial sample data to obtain new sample data S2Setting a sampling sequence of the non-evidence variables, and sampling the non-evidence variables according to the sampling sequence; wherein non-observable variables in the bayesian network structure are non-evidence variables;
s53, updating the current sample data according to the sampling result;
s54, obtaining the distribution probability of the sampling variable according to the formula and the conditional probability table obtained in the step S4, where the formula is as follows:
if D is1And P (D)1·D2)＝P(D1)·P(D2|D1) Independently of one another, then P (D)1·D2)＝P(D1)·P(D2)
If D is1And D2If there is a dependency, P (D)1·D2)＝P(D1)·P(D2|D1)
Bayesian formula:
chain rule:
s55, circularly executing m steps S52-S54 to generate m samples of distribution probability, and calculating the posterior probability of different values of the query variable in the m samples of distribution probability, namely the posterior probability of abnormity under the currently input evidence variable. Where the unknown variable currently desired to be computed is the query variable.
Preferably, the logarithmic ratio in step S6 is: the test data is centralizedThe actual probability of occurrence of an anomaly under the previously input evidence variable is defined as B, the posterior probability in step S5 is defined as C, and the log ratio is loga B/loga C。
The invention also provides a system applying the method, which comprises the following steps:
the equipment energy flow mechanism analysis and energy consumption modeling module: the method is used for analyzing the energy flow mechanism of the extruder from the energy consumption angle in the using process of the extruder to obtain factors influencing energy consumption;
a Bayesian network construction module: the energy flow mechanism analysis and energy consumption modeling module is used for analyzing the factors obtained by the energy flow mechanism analysis and energy consumption modeling module of the equipment, finding the dependency relationship among the nodes of the Bayesian network, constructing a Bayesian network structure and setting an observable variable set in the Bayesian network structure as an evidence variable set;
a data processing module: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring energy consumption data of an extruder as a sample data set, dividing the sample data set into a training data set and a testing data set, and then processing the energy consumption data;
a parameter learning module: the probability density function is used for carrying out nonparametric estimation processing on the nodes of the evidence variable set to obtain reasonable probability density functions of all the nodes; obtaining a conditional probability table of all nodes by using the data of the training data set;
bayesian network inference module: utilizing a Gibbs sampling algorithm to carry out Bayesian network reasoning to obtain the posterior probability of the occurrence of the abnormality under the currently input evidence variable;
an anomaly detection and analysis module: the system is used for calculating the actual probability of abnormality under the current input evidence in the test data set, calculating the logarithmic ratio of the actual probability to the posterior probability in the Bayesian network inference module, setting the confidence level according to the ratio, and judging whether the abnormality occurs according to the confidence level; and if the abnormity occurs, calculating the posterior probability of the abnormity under different evidence variables, and defining the evidence variable corresponding to the maximum value of the posterior probability as the evidence variable causing the abnormity to occur.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the method introduces an energy consumption model analysis method, obtains parameters influencing energy consumption by analyzing energy consumption data of the extruder so as to determine the dependency relationship among nodes of the Bayesian network, and constructs an anomaly detection Bayesian network structure according to the nodes and the dependency relationship, wherein the Bayesian network structure is applied to anomaly detection of the extruder; meanwhile, compared with the existing Bayesian network construction method which needs repeated training and evaluation and has high complexity of a network model, the method has the advantages that a large number of redundant calculation and evaluation processes are reduced, the complexity of Bayesian network construction is reduced, the robustness and the high efficiency of the Bayesian network are improved, and the problems that a large number of redundant calculation and evaluation processes, high complexity of the network model, low detection precision and the like exist in the existing anomaly detection method of the extruder are solved.
Drawings
FIG. 1 is a general flow diagram of the process of the present invention.
FIG. 2 is a block diagram of the system of the present invention.
Fig. 3 is a diagram showing the structure of the bayesian network in example 1.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1
As shown in fig. 1, a method for detecting extruder anomaly based on energy consumption data and bayesian network includes the following steps:
s1, analyzing an energy flow mechanism of an extruder from the energy consumption angle in the using process of the extruder to obtain factors influencing energy consumption; the factors comprise piston diameter D, piston rod diameter D and theoretical flow Q of the hydraulic pumpmFlow Q of input extrusion cylinder, moving speed V of extrusion rod, and energy loss Delta P of hydraulic pumpmhEnergy loss delta P of hydraulic valve groupfhMechanical efficiency eta of hydraulic pumpmVolumetric efficiency eta of hydraulic pumpvVolumetric efficiency of the extrusion cylinder, total power P for the extrusion cycle, master cylinder pressure PaPump port pressure P of hydraulic pump, extrusion cylinder thrust F and extrusion cylinder input power PciRotating speed n of hydraulic pumpvAnd torque Tn；
S2, finding out the dependency relationship among the nodes of the Bayes network by combining the Bayes algorithm and the factors in the step S1, constructing a Bayes network structure and setting an observable variable set therein as an evidence variable set; the bayesian network structure constructed in this embodiment is shown in fig. 3;
s3, collecting energy consumption data of the extruder as a sample data set, dividing the sample data set into a training data set and a testing data set, and then processing the energy consumption data;
s4, carrying out nonparametric estimation processing on the nodes of the evidence variable set to obtain probability density functions of all the nodes; obtaining a conditional probability table of all nodes by using the data of the training data set;
s5, carrying out Bayesian network reasoning by utilizing a Gibbs sampling algorithm to obtain a posterior probability about the occurrence of an anomaly under the currently input evidence variable;
s6, calculating the actual probability of the occurrence of the abnormality under the evidence variable currently input in the test data set, calculating the logarithmic ratio of the actual probability to the posterior probability in the step S5, setting a confidence coefficient according to the ratio, and judging whether the abnormality occurs according to the confidence coefficient; and if the abnormity occurs, calculating the posterior probability of the abnormity under different evidence variables, and defining the evidence variable corresponding to the maximum value of the posterior probability as the evidence variable causing the abnormity to occur.
Wherein, step S1 specifically includes:
the total power used by each extrusion process of the hydraulic system is expressed as:
P＝ΔPmh+ΔPfh+Pci＝UI
wherein, Δ PmhRepresenting the lost energy of the hydraulic pump, Δ PfhIndicating hydraulic system valveLoss of energy of the group, PciThe input power of the extrusion oil cylinder is represented, U represents the current voltage value, and I represents the current value;
input power P of hydraulic pumpmiIs represented by Pmi＝2·π·nv·Tn
Wherein n isvRepresenting the rotational speed of the pump shaft of the hydraulic pump, TnRepresenting the input torque of the hydraulic pump;
p for output power of hydraulic pumpmoIs represented by Pmo＝p·Qm
Wherein p represents the pump port pressure of the hydraulic pump, QmRepresenting the actual flow rate of the hydraulic pump, can be calculated by:
wherein, VmIs the traveling speed of the hydraulic pump, DmIs the diameter of the hydraulic pump;
the energy loss Δ P of the hydraulic pumpmhExpressed as:
ΔPmh＝Pmi-Pmo＝2·π·nv·Tn-p·Qm
the loss of energy consumption of the hydraulic valve group is expressed as Δ Pfh＝n·Δpf·ΔQf
Wherein n represents the number of valves, Δ pfRepresenting the valve port pressure difference, Δ QfRepresenting the valve port flow;
p for input power of extrusion cylinderciIs represented by Pci＝F·V
Wherein F represents the force to which the extrusion rod is subjected, and V represents the speed at which the extrusion rod moves;
Q＝Qm
wherein, in the formula: a represents the effective area of the extrusion oil cylinder; etamThe mechanical efficiency of the extrusion oil cylinder is shown; etavThe volumetric efficiency of the extrusion cylinder is shown; d represents the piston diameter; d represents the piston rod diameter; q represents the flow rate of the input extrusion oil cylinder; p is a radical ofaIndicates master cylinder pressure; indicating the volumetric efficiency of the squeeze cylinders.
In the bayesian network structure in step S2, the piston diameter D, the piston rod diameter D, and the theoretical flow rate Q of the hydraulic pumpmIs a fixed parameter; flow Q of input extrusion cylinder, moving speed V of extrusion rod, and energy loss Delta P of hydraulic pumpmhEnergy loss delta P of hydraulic valve groupfhMechanical efficiency eta of hydraulic pumpmVolumetric efficiency eta of hydraulic pumpvVolumetric efficiency of the extrusion cylinder, total power P for the extrusion cycle, master cylinder pressure PaPump port pressure P of hydraulic pump, extrusion cylinder thrust F and extrusion cylinder input power PciIs a variable with cognitive uncertainty, in which the master cylinder pressure paPump port pressure p of hydraulic pump, and rotation speed n of hydraulic pumpvAnd torque TnIs an observable variable.
Wherein the data processing in step S3 includes:
s31, setting D as the set of all nodes, pi (D)i) Is DiOf parent node of (2), pi (D)i)＝{Dai}；
And S32, discretizing and symbolizing the value of the energy consumption data attribute.
Wherein, step S4 specifically includes:
s41, obtaining a set D { D ] by adopting a parzen window estimation method1，D2，D3，D4...DnIn each node DiWherein the kernel function is of the form:
pn(x) Represents the superposition of the respective sample-centered normal probability density functions:
where n denotes the size of the sample, hnRepresents the length of the window;
s42, dividing the training data into samples with different sizes for training, wherein n represents the size of the sample, hnIndicating the length of the window, adjusting n and hnObtaining the probability density function of all nodes;
and S43, obtaining a conditional probability table of all nodes with dependency relations by using the training data.
Wherein, step S5 specifically includes:
s51, utilizing Gibbs sampling algorithm to carry out Bayesian network inference, randomly selecting an evidence variable according to the result to be inferred, defining A', and then randomly generating a starting sample S1The starting sample S1Contains an evidence variable A';
s52, copying the last initial sample data to obtain new sample data S2Setting a sampling sequence of the non-evidence variables, and sampling the non-evidence variables according to the sampling sequence; wherein non-observable variables in the bayesian network structure are non-evidence variables;
s53, updating the current sample data according to the sampling result;
s54, obtaining the distribution probability of the sampling variable according to the formula and the conditional probability table obtained in the step S4, where the formula is as follows:
if D is1And P (D)1·D2)＝P(D1)·P(D2|D1) Independently of one another, then P (D)1·D2)＝P(D1)·P(D2)
If D is1And D2If there is a dependency, P (D)1·D2)＝P(D1)·P(D2|D1)
Bayesian formula:
chain rule:
s55, circularly executing m steps S52-S54 to generate m samples of distribution probability, and calculating the posterior probability of different values of the query variable in the m samples of distribution probability, namely the posterior probability of abnormity under the currently input evidence variable.
Wherein, the logarithmic ratio in the step S6 is: the actual probability of the occurrence of an anomaly under the currently input evidence variable in the test data set is defined as B, the posterior probability in step S5 is defined as C, and the log ratio is logaB/logaC。
Example 2
Embodiment 2 is a system based on embodiment 1, as shown in fig. 2, including: the device energy flow mechanism analysis and energy consumption modeling module 1: the method is used for analyzing the energy flow mechanism of the extruder from the energy consumption angle in the using process of the extruder to obtain factors influencing energy consumption;
bayesian network construction module 2: the energy flow mechanism analysis and energy consumption modeling module is used for analyzing the factors obtained by the energy flow mechanism analysis and energy consumption modeling module of the equipment, finding the dependency relationship among the nodes of the Bayesian network, constructing a Bayesian network structure and setting an observable variable set in the Bayesian network structure as an evidence variable set;
the data processing module 3: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring energy consumption data of an extruder as a sample data set, dividing the sample data set into a training data set and a testing data set, and then processing the energy consumption data;
the parameter learning module 4: the probability density function is used for carrying out nonparametric estimation processing on the nodes of the evidence variable set to obtain reasonable probability density functions of all the nodes; obtaining a conditional probability table of all nodes by using the data;
bayesian network inference module 5: utilizing a Gibbs sampling algorithm to carry out Bayesian network reasoning to obtain the posterior probability of the occurrence of the abnormality under the currently input evidence variable;
anomaly detection analysis module 6: the system is used for calculating the actual probability of abnormality under the current input evidence in the test data set, calculating the logarithmic ratio of the actual probability to the posterior probability in the Bayesian network inference module 5, setting the confidence level according to the ratio, and judging whether the abnormality occurs according to the confidence level; and if the abnormity occurs, calculating the posterior probability of the abnormity under different evidence variables, and defining the evidence variable corresponding to the maximum value of the posterior probability as the evidence variable causing the abnormity to occur.
The terms describing positional relationships in the drawings are for illustrative purposes only and are not to be construed as limiting the patent;
it should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (8)
1. An extruder anomaly detection method based on energy consumption data and a Bayesian network is characterized by comprising the following steps:
s1, analyzing an energy flow mechanism of an extruder from the energy consumption angle in the using process of the extruder to obtain factors influencing energy consumption; the factors comprise piston diameter D, piston rod diameter D and actual flow Q of the hydraulic pumpmFlow Q of input extrusion cylinder, moving speed V of extrusion rod, and energy loss Delta P of hydraulic pumpmhEnergy loss delta P of hydraulic valve groupfhMechanical efficiency eta of hydraulic pumpmVolumetric efficiency eta of hydraulic pumpvVolumetric efficiency of the extrusion cylinder, total power P for the extrusion cycle, master cylinder pressure PaPump port pressure P of hydraulic pump, extrusion cylinder thrust F and extrusion cylinder input power PciRotational speed of hydraulic pumpnvInput torque T of hydraulic pumpn；
S2, finding out the dependency relationship among the nodes of the Bayes network by combining the Bayes algorithm and the factors in the step S1, constructing a Bayes network structure and setting an observable variable set therein as an evidence variable set;
s3, collecting energy consumption data of the extruder as a sample data set, dividing the sample data set into a training data set and a testing data set, and then processing the energy consumption data;
s4, carrying out nonparametric estimation processing on the nodes of the evidence variable set to obtain probability density functions of all the nodes; obtaining a conditional probability table of all nodes by using the data of the training data set;
s5, carrying out Bayesian network reasoning by utilizing a Gibbs sampling algorithm to obtain a posterior probability about the occurrence of an anomaly under the currently input evidence variable;
s6, calculating the actual probability of the occurrence of the abnormality under the evidence variable currently input in the test data set, calculating the logarithmic ratio of the actual probability to the posterior probability in the step S5, setting a confidence coefficient according to the ratio, and judging whether the abnormality occurs according to the confidence coefficient; and if the abnormity occurs, calculating the posterior probability of the abnormity under different evidence variables, and defining the evidence variable corresponding to the maximum value of the posterior probability as the evidence variable causing the abnormity to occur.
2. The abnormality detection method according to claim 1, characterized in that step S1 specifically includes:
the total power used for the extrusion cycle is expressed as:
P＝ΔPmh+ΔPfh+Pci＝UI
wherein, Δ PmhRepresenting the energy loss, Δ P, of the hydraulic pumpfhRepresenting the energy loss, P, of the hydraulic valve groupciThe method comprises the steps of representing the input power of an extrusion oil cylinder, representing a current voltage value by U, and representing a current value by I;
input power P of hydraulic pumpmiIs represented by Pmi＝2·π·nv·Tn
Wherein,nvrepresenting the speed of rotation of the hydraulic pump, TnRepresenting the input torque of the hydraulic pump;
p for output power of hydraulic pumpmoIs represented by Pmo＝p·Qm
Wherein p represents the pump port pressure of the hydraulic pump, QmRepresenting the actual flow rate of the hydraulic pump, can be calculated by:
wherein, VmIs the traveling speed of the hydraulic pump, DmIs the diameter of the hydraulic pump;
the energy loss Δ P of the hydraulic pumpmhExpressed as:
ΔPmh＝Pmi-Pmo＝2·π·nv·Tn-p·Qm
the energy consumption of the hydraulic valve group is expressed as Δ Pfh＝n·Δpf·ΔQf
Wherein n represents the number of valves, Δ pfRepresenting the valve port pressure difference, Δ QfRepresenting the valve port flow;
p for power input of extrusion oil cylinderciIs represented by Pci＝F·V
Wherein F represents the thrust of the extrusion oil cylinder, and V represents the moving speed of the extrusion rod;
Q＝Qm
wherein, in the formula: a represents the effective area of the extrusion oil cylinder; eta 'of'mThe mechanical efficiency of the extrusion oil cylinder is shown; eta 'of'vThe volumetric efficiency of the extrusion cylinder is shown; d representsPiston diameter; d represents the piston rod diameter; q represents the flow rate of the input extrusion oil cylinder; p is a radical ofaIndicates master cylinder pressure; indicating the volumetric efficiency of the squeeze cylinders.
3. The abnormality detection method according to claim 2, characterized in that in the bayesian network structure described in step S2, the piston diameter D and the piston rod diameter D are fixed parameters, and the actual flow rate Q of the hydraulic pump is set to be equal to or lower than the actual flow rate Q of the hydraulic pumpm(ii) a Flow Q of input extrusion cylinder, moving speed V of extrusion rod, and energy loss Delta P of hydraulic pumpmhEnergy loss delta P of hydraulic valve groupfhMechanical efficiency eta of hydraulic pumpmVolumetric efficiency eta of hydraulic pumpvVolumetric efficiency of the extrusion cylinder, total power P for the extrusion cycle, master cylinder pressure PaPump port pressure P of hydraulic pump, extrusion cylinder thrust F and extrusion cylinder input power PciIs a variable with cognitive uncertainty, in which the master cylinder pressure paPump port pressure p of hydraulic pump, and rotation speed n of hydraulic pumpvAnd input torque T of hydraulic pumpnIs an observable variable.
4. The abnormality detection method according to claim 1, characterized in that said data processing of step S3 includes:
s31, setting D as the set of all nodes, pi (D)i) Is DiOf parent node of (2), pi (D)i)＝{Dai}；
And S32, discretizing and symbolizing the value of the energy consumption data attribute.
5. The abnormality detection method according to claim 4, characterized in that step S4 specifically includes:
s41, obtaining a set D { D ] by adopting a parzen window estimation method1，D2，D3，D4...DnIn each node DiWherein the kernel function is of the form:
pn(x) Represents the superposition of the respective sample-centered normal probability density functions:
where n denotes the size of the sample, hnRepresents the length of the window;
s42, dividing the training data into samples with different sizes for training, wherein n represents the size of the sample, hnIndicating the length of the window, adjusting n and hnObtaining the probability density function of all nodes;
and S43, obtaining a conditional probability table of all nodes with dependency relations by using the data of the training data set.
6. The abnormality detection method according to claim 1, characterized in that step S5 specifically includes:
s51, utilizing Gibbs sampling algorithm to carry out Bayesian network inference, randomly selecting an evidence variable according to the result to be inferred, defining A', and then randomly generating a starting sample S1The starting sample S1Contains an evidence variable A';
s52, copying the last initial sample data to obtain new sample data S2Setting a sampling sequence of the non-evidence variables, and sampling the non-evidence variables according to the sampling sequence; wherein non-observable variables in the bayesian network structure are non-evidence variables;
s53, updating the current sample data according to the sampling result;
s54, obtaining the distribution probability of the sampling variable according to the formula and the conditional probability table obtained in the step S4, where the formula is as follows:
if D is1And D2Independently of one another, then P (D)1·D2)＝P(D1)·P(D2)
If D is1And D2If there is a dependency, P (D)1·D2)＝P(D1)·P(D2|D1)
Bayesian formula:
chain rule:
s55, circularly executing m steps S52-S54 to generate m samples of distribution probability, and calculating the posterior probability of different values of the query variable in the m samples of distribution probability, namely the posterior probability of abnormity under the currently input evidence variable.
7. The abnormality detection method according to claim 1, characterized in that the log ratio in step S6 is: the actual probability of the occurrence of an anomaly under the currently input evidence variable in the test data set is defined as B, the posterior probability in step S5 is defined as C, and the log ratio is logaB/logaC。
8. The system of any one of claims 1-7, comprising:
the device energy flow mechanism analysis and energy consumption modeling module (1): the method is used for analyzing the energy flow mechanism of the extruder from the energy consumption angle in the using process of the extruder to obtain factors influencing energy consumption;
bayesian network construction module (2): the energy flow mechanism analysis and energy consumption modeling module is used for analyzing the factors obtained by the energy flow mechanism analysis and energy consumption modeling module of the equipment, finding the dependency relationship among the nodes of the Bayesian network, constructing a Bayesian network structure and setting an observable variable set in the Bayesian network structure as an evidence variable set;
data processing module (3): the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring energy consumption data of an extruder as a sample data set, dividing the sample data set into a training data set and a testing data set, and then processing the energy consumption data;
a parameter learning module (4): the probability density function is used for carrying out nonparametric estimation processing on the nodes of the evidence variable set to obtain reasonable probability density functions of all the nodes; obtaining a conditional probability table of all nodes by using the data of the training data set;
bayesian network inference module (5): utilizing a Gibbs sampling algorithm to carry out Bayesian network reasoning to obtain the posterior probability of the occurrence of the abnormality under the currently input evidence variable;
anomaly detection analysis module (6): the system is used for calculating the actual probability of abnormality under the current input evidence in the test data set, calculating the logarithmic ratio of the actual probability to the posterior probability in the Bayesian network inference module (5), setting confidence level according to the ratio, and judging whether abnormality occurs according to the confidence level; and if the abnormity occurs, calculating the posterior probability of the abnormity under different evidence variables, and defining the evidence variable corresponding to the maximum value of the posterior probability as the evidence variable causing the abnormity to occur.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910028040.XA CN109894495B (en) | 2019-01-11 | 2019-01-11 | Extruder anomaly detection method and system based on energy consumption data and Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910028040.XA CN109894495B (en) | 2019-01-11 | 2019-01-11 | Extruder anomaly detection method and system based on energy consumption data and Bayesian network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109894495A CN109894495A (en) | 2019-06-18 |
| CN109894495B true CN109894495B (en) | 2020-12-22 |
Family
ID=66943660
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910028040.XA Active CN109894495B (en) | 2019-01-11 | 2019-01-11 | Extruder anomaly detection method and system based on energy consumption data and Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109894495B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110738242B (en) * | 2019-09-25 | 2021-08-10 | 清华大学 | Bayes structure learning method and device of deep neural network |
| CN111122199A (en) * | 2019-12-31 | 2020-05-08 | 新奥数能科技有限公司 | Boiler fault diagnosis method and device |
| CN111800389A (en) * | 2020-06-09 | 2020-10-20 | 同济大学 | Port network intrusion detection method based on Bayesian network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014188555A (en) * | 2013-03-27 | 2014-10-06 | Ube Machinery Corporation Ltd | Depressurization method in extrusion press |
| CN107295622A (en) * | 2017-06-06 | 2017-10-24 | 重庆邮电大学 | A kind of wlan system WAP energy-conservation anti-interference method based on Bayesian game |
| CN107729381A (en) * | 2017-09-15 | 2018-02-23 | 广州嘉影软件有限公司 | Interactive multimedia resource polymerization method and system based on multidimensional characteristic identification |
| CN108691944A (en) * | 2018-06-18 | 2018-10-23 | 重庆大学 | Based on the asymmetric MR damper for squeezing enhancement effect |
-
2019
- 2019-01-11 CN CN201910028040.XA patent/CN109894495B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014188555A (en) * | 2013-03-27 | 2014-10-06 | Ube Machinery Corporation Ltd | Depressurization method in extrusion press |
| CN107295622A (en) * | 2017-06-06 | 2017-10-24 | 重庆邮电大学 | A kind of wlan system WAP energy-conservation anti-interference method based on Bayesian game |
| CN107729381A (en) * | 2017-09-15 | 2018-02-23 | 广州嘉影软件有限公司 | Interactive multimedia resource polymerization method and system based on multidimensional characteristic identification |
| CN108691944A (en) * | 2018-06-18 | 2018-10-23 | 重庆大学 | Based on the asymmetric MR damper for squeezing enhancement effect |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109894495A (en) | 2019-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ren et al. | A novel model with the ability of few-shot learning and quick updating for intelligent fault diagnosis | |
| WO2023197461A1 (en) | Gearbox fault early warning method and system based on working condition similarity evaluation | |
| CN111639430B (en) | Natural gas pipeline leakage identification system driven by digital twinning | |
| CN107941537B (en) | A kind of mechanical equipment health state evaluation method | |
| CN109894495B (en) | Extruder anomaly detection method and system based on energy consumption data and Bayesian network | |
| Li et al. | Data-driven bearing fault identification using improved hidden Markov model and self-organizing map | |
| CN111311059A (en) | Knowledge graph-based water mill room fault diagnosis method | |
| Baraldi et al. | Reconstruction of missing data in multidimensional time series by fuzzy similarity | |
| Yu | A nonlinear probabilistic method and contribution analysis for machine condition monitoring | |
| CN111190349A (en) | Method, system and medium for monitoring state and diagnosing fault of ship engine room equipment | |
| CN109829561B (en) | Accident prediction method based on smoothing processing and network model machine learning | |
| CN114528547B (en) | ICPS (information and communication system) unsupervised online attack detection method and equipment based on community feature selection | |
| CN111899905A (en) | Fault diagnosis method and system based on nuclear power device | |
| CN114444582A (en) | Mechanical equipment fault diagnosis method based on convolutional neural network and Bayesian network | |
| Tang et al. | Review and perspectives of machine learning methods for wind turbine fault diagnosis | |
| Li et al. | Fault prognosis of hydraulic pump based on bispectrum entropy and deep belief network | |
| CN111126477A (en) | Learning and reasoning method of hybrid Bayesian network | |
| Guo et al. | Operation stage division and RUL prediction of bearings based on 1DCNN-ON-LSTM | |
| CN113092083B (en) | Machine pump fault diagnosis method and device based on fractal dimension and neural network | |
| Yafei et al. | Fault Diagnosis of Axial Piston Pump Based on Extreme‐Point Symmetric Mode Decomposition and Random Forests | |
| CN107527093B (en) | Wind turbine generator running state diagnosis method and device | |
| Wang et al. | Particle filtering-based system degradation prediction applied to jet engines | |
| CN115660198B (en) | Method for predicting residual service life of rolling bearing | |
| CN109210268B (en) | Big data processing method based on ultralow-power electromagnetic valve | |
| Martí et al. | YASA: yet another time series segmentation algorithm for anomaly detection in big data problems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |