CN111639842B - Equipment health assessment method, assessment system and equipment health prediction method - Google Patents
Equipment health assessment method, assessment system and equipment health prediction method Download PDFInfo
- Publication number
- CN111639842B CN111639842B CN202010431329.9A CN202010431329A CN111639842B CN 111639842 B CN111639842 B CN 111639842B CN 202010431329 A CN202010431329 A CN 202010431329A CN 111639842 B CN111639842 B CN 111639842B
- Authority
- CN
- China
- Prior art keywords
- evaluation
- data
- vibration
- temperature
- equipment
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention belongs to the field of equipment health assessment and prediction, and particularly relates to an equipment health assessment method, an assessment system and an equipment health prediction method, which comprise the following steps: synchronously acquiring vibration data of different vibration acquisition positions on the equipment at each vibration acquisition time, and synchronously acquiring temperature data of different temperature acquisition positions on the equipment at each temperature acquisition time; obtaining an evaluation compensation factor based on all the temperature data; based on all vibration data, evaluating by using a vibration evaluation state model to obtain an evaluation state vector, performing weighted average on probability densities corresponding to all evaluation states in the evaluation state vector, and compensating a weighted average result by using an evaluation compensation factor to complete equipment health evaluation. The invention collects data from two view angles of time and space, adopts the compensation factors to compensate the output of the vibration evaluation state model, avoids the problem that the traditional evaluation method has stronger hysteresis on the working temperature change of the key execution mechanism of the equipment, and improves the accuracy of the health evaluation and prediction of the equipment.
Description
Technical Field
The invention belongs to the field of equipment health assessment and prediction, and particularly relates to an equipment health assessment method, an assessment system and an equipment health prediction method.
Background
With the rapid development of industrial big data and artificial intelligence technology, industrial production and manufacture are upgraded towards intellectualization. Once any one of the devices or links fails, the system is disabled, normal production is affected, great economic loss of enterprises is caused, personnel safety accidents are caused when serious, and losses are brought to the country and people. Therefore, from the viewpoints of safe production and economic benefits of enterprises, by collecting operation data of industrial production key equipment, the health assessment and prediction of the equipment are necessary by adopting big data and artificial intelligence technology. The health evaluation and prediction is to monitor the health state of the equipment in real time, quantitatively evaluate the health state of the equipment and predict the health of the equipment.
At present, equipment health management technology has been paid attention to, and among equipment key monitoring data, data closely related to the health state of equipment comprises temperature data and vibration data. Based on the equipment operation data, the health evaluation is carried out on the equipment by adopting a statistical analysis method, and the method is easy to realize, has strong limitation, is excessively dependent on manual experience, has great relevance between the accuracy of an evaluation result and the experience of an evaluator, and has poor portability. In the prior art, the equipment state is mainly evaluated and predicted by a machine learning model driven by bearing vibration data, and the analysis result is only related to the vibration data and cannot reflect the influence of the temperature change of the measuring point on the equipment health state in time. In addition, the existing method often collects single-point data (such as bearing data of a driving end of the equipment) of the equipment, and does not combine vibration data and temperature data measured by multiple points such as a motor, a transmission mechanism and an executing mechanism of the equipment to carry out equipment health assessment and prediction. Therefore, the existing equipment health evaluation method has the problem of low evaluation prediction accuracy.
Disclosure of Invention
The invention provides a device health assessment method, an assessment system and a device health prediction method, which are used for solving the technical problem that the accuracy of assessment and prediction is not high due to the fact that the influence of temperature is not considered in the existing device health assessment and prediction method.
The technical scheme for solving the technical problems is as follows: a method of device health assessment, comprising:
s1, synchronously acquiring vibration data of different vibration acquisition positions on equipment to be evaluated at each vibration acquisition time, and synchronously acquiring temperature data of different temperature acquisition positions on the equipment at each temperature acquisition time;
s2, carrying out average value calculation on all the temperature data, and taking a calculation result as an evaluation compensation factor;
and S3, based on all vibration data, adopting a trained vibration evaluation state model to evaluate to obtain an evaluation state vector formed by the number of data samples in each evaluation state, carrying out weighted average on probability densities corresponding to each evaluation state in the evaluation state vector, and adopting an evaluation compensation factor to compensate the weighted average result to obtain an evaluation value, thereby completing equipment health evaluation.
The beneficial effects of the invention are as follows: the method can realize the equipment health assessment driven by data and models, because the equipment health is assessed by adopting vibration data alone, the influence of temperature on the equipment health is ignored, so that the assessment result has deviation and needs to be compensated by adopting temperature data. In addition, for the data of different categories of temperature data and vibration data, data acquisition at a plurality of different moments is respectively carried out on a plurality of monitoring points of the equipment from two view angles of time and space, the acquired data more comprehensively reflect the running condition of the equipment, and the accuracy of health assessment is further improved. By adopting the data acquisition and processing mode and compensating the model output, the accuracy of the equipment health assessment is effectively improved.
Based on the technical scheme, the invention can be improved as follows.
Further, the time interval between two adjacent temperature acquisition moments is in the second level, and the time interval between two adjacent vibration acquisition moments is in the millisecond level;
the S1 further includes: cleaning and denoising all the temperature data; all of the vibration data are cleaned to remove outliers.
The invention has the further beneficial effects that: the method comprises the steps of collecting key monitoring point data of equipment, adopting different time granularity according to different types of data, and setting the collection time granularity to be millisecond level for data with high change frequency such as vibration data; for data with low change frequency such as temperature data, the acquisition time granularity is set to the second level. The temperature data with the acquisition period of second level is acquired through data acquisition with different time granularity, the vibration data with the sampling period of millisecond level is acquired, the acquisition point of each type of data comprises a plurality of different measurement points, and the equipment health assessment and prediction are carried out from two view angles of space and time, so that the comprehensiveness and reliability of the equipment health assessment and prediction are improved, and the problem that the assessment accuracy of the equipment health is influenced only by the second level (lower acquisition frequency) data or by the single measurement point data is also solved. In addition, the temperature data are cleaned to remove noise, so that the quality of the equipment temperature data is improved, and the accuracy of an evaluation result obtained through compensation is further improved.
Further, the vibration evaluation state model is obtained by training a Catboost model based on a device vibration data training set.
Further, in the training process of the vibration evaluation state model, a firefly algorithm is adopted, a vector is formed by a model learning rate parameter and a maximum depth parameter of a tree, the vector represents the position of fireflies, an objective function is defined, the input of the objective function is the position of the fireflies, the output of the objective function is an evaluation standard of the vibration evaluation state model, and the optimal solution of the learning rate parameter and the maximum depth parameter of the tree is synchronously obtained through iterative calculation.
The invention has the further beneficial effects that: the key parameters affecting the performance of the vibration evaluation state model comprise a learning rate learning_rate and a maximum depth of tree, the specific effects of the two parameters are different along with the difference of data, and in order to enable the vibration evaluation state model to automatically adjust parameters for different data when performing health evaluation and prediction, a method for optimizing the vibration evaluation state model parameters based on a firefly algorithm is adopted, so that the purpose that the vibration evaluation state model can automatically match with the optimal training parameters under different data conditions is achieved, and the adaptability of the vibration evaluation state model evaluation and prediction algorithm is improved. In addition, 2 key parameters of the state model were evaluated for vibration using FA (Firefly Algorithm ): the learning rate leanning_rate and the maximum depth of the tree are optimized, and meanwhile, the optimal 2 key parameters are obtained instead of optimizing one by one, so that the optimization efficiency is improved.
Further, in the step S3, before weighted averaging the probability density corresponding to each evaluation state in the evaluation state vector, according to the probability density and the model diagnosis error rate of each evaluation state in the evaluation state vector, part of the evaluation states and the number of data samples corresponding to the evaluation states are filtered from the evaluation state vector, so as to obtain a new evaluation state vector.
The invention has the further beneficial effects that: before weighted averaging, filtering is performed, so that the accuracy of evaluation is further improved.
Further, in S3, the result of compensating the weighted average by using the evaluation compensation factor is specifically: adding the evaluation compensation factor to the result of the weighted average to make a correction;
in the weighted average, the weight corresponding to each evaluation state is: the ratio of the number of data samples of the evaluation state to the total number of data samples in the evaluation state vector; alternatively, a weighted average method based on fuzzy membership is used for determination.
The invention has the further beneficial effects that: the compensation of the model output is carried out in an addition mode, so that the problem that the traditional equipment health assessment method has strong hysteresis on the working temperature change of the equipment key actuating mechanism can be effectively solved, and the accuracy of equipment health assessment is improved.
Further, the S2 includes:
carrying out standardization processing on a group of temperature data which is formed by collecting all temperature collecting moments at each temperature collecting position, wherein the temperature data in the group of temperature data is converted into 0 and the temperature data which is larger than an alarm value is converted into 1, and the temperature data between the standard value and the alarm value is converted into a ratio of a first difference value to a second difference value, wherein the first difference value is the difference value between the temperature data and the standard value, and the second difference value is the difference value between the standard value and the alarm value;
averaging the normalized temperature data of each temperature acquisition position at each temperature acquisition time to obtain a temperature data compensation value at the temperature acquisition time;
multiplying the temperature data compensation value at each temperature acquisition time with the recommended compensation value at the temperature acquisition time, and taking the average value of the products as the compensation factor.
The invention has the further beneficial effects that: by adopting the mean value calculation mode, the influence of the temperature of the equipment corresponding to the health state of the equipment can be effectively reflected, and when the mean value calculation mode is used for evaluating and outputting a compensation model, the compensation effect is good, and the evaluation accuracy is improved.
The invention also provides a device health assessment system, comprising:
the data acquisition and processing module is used for executing S1 in the equipment health assessment method; the compensation factor calculation module is used for executing S2 in the equipment health assessment method based on all the temperature data obtained by the data acquisition and processing module; and the equipment health evaluation model is used for executing the S3 in the equipment health evaluation method based on all vibration data obtained by the data acquisition and processing module and the evaluation compensation factors obtained by the compensation factor calculation module.
The beneficial effects of the invention are as follows: the system comprises a device health evaluation model which realizes the device health evaluation driven by the data and the model under the assistance of a data acquisition and processing module and a compensation factor calculation module, and because the vibration data is adopted to evaluate the device health, the influence of the temperature on the device health is ignored, so that the evaluation result has deviation and the temperature data is required to be compensated. In addition, the data acquisition and processing module respectively performs data acquisition at a plurality of different moments on a plurality of monitoring points of the equipment from two view angles of time and space, the acquired data more comprehensively reflects the running condition of the equipment, and the accuracy of health assessment is further improved. By adopting the system, the accuracy of equipment health assessment is effectively improved.
The invention also provides a device health prediction method, which adopts the device health evaluation method to perform multiple health evaluations on the device to be predicted, and performs health prediction on the device to be predicted based on the results of the multiple health evaluations.
The beneficial effects of the invention are as follows: by adopting the equipment health assessment method, the accuracy of the result for health prediction is ensured, and the accuracy of equipment health prediction is further improved.
The present invention also provides a machine-readable storage medium storing machine-executable instructions that, when invoked and executed by a processor, cause the processor to implement a device health assessment method as described above.
Drawings
FIG. 1 is a block flow diagram of a method for evaluating device health according to an embodiment of the present invention;
FIG. 2 is a flowchart of another method for evaluating device health according to an embodiment of the present invention;
FIG. 3 is a flowchart of a sliding window algorithm based on median filtering according to an embodiment of the present invention;
FIG. 4 is an effect diagram of local mean decomposition of a high temperature fan in a healthy state according to an embodiment of the present invention;
FIG. 5 is a schematic flow chart of a firefly algorithm optimization model parameter provided by an embodiment of the invention;
FIG. 6 is a schematic diagram of classifying health states of a device according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of a vibration evaluation state model with data and model combination according to an embodiment of the present invention;
FIG. 8 is a panoramic view of a device health assessment method according to an embodiment of the present invention;
FIG. 9 is a schematic diagram of a plurality of measurement points of an apparatus according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention. In addition, the technical features of the embodiments of the present invention described below may be combined with each other as long as they do not collide with each other.
Example 1
A device health assessment method 100, as shown in fig. 1, includes:
step 110, synchronously acquiring vibration data of different vibration acquisition positions on equipment to be evaluated at each vibration acquisition time, and synchronously acquiring temperature data of different temperature acquisition positions on the equipment at each temperature acquisition time;
step 120, performing average value calculation on all temperature data, wherein the calculation result is used as an evaluation compensation factor;
and 130, based on all vibration data, evaluating by adopting a trained vibration evaluation state model to obtain an evaluation state vector formed by the number of data samples in each evaluation state, carrying out weighted average on probability densities corresponding to each evaluation state in the evaluation state vector, and adopting an evaluation compensation factor to compensate the weighted average result to obtain an evaluation value, thereby completing the equipment health evaluation.
The method for acquiring and processing temperature and vibration data of the training vibration evaluation state model is the same as the method for acquiring and processing data in the evaluation method of the embodiment, wherein a training data set is formed by vibration signals and corresponding health evaluation labels, the vibration evaluation state model is trained, and evaluation state vectors output by the model during training are subjected to the weighted average of the method to obtain evaluation values, so that the evaluation values are used for determining health evaluation in the evaluation method of the embodiment.
The method can realize the equipment health assessment and prediction driven by data and models, the assessment compensation factor c is adopted to compensate the output of the vibration assessment state model, preferably, the vibration assessment state model is obtained by training the Catboost model based on an equipment vibration data training set, and further, the assessment compensation factor c is adopted to compensate the output of the vibration assessment state model, then, the c-Catboost model is constructed, the data is combined with the equipment operation model, the problem that the traditional equipment health assessment method has strong hysteresis on the working temperature change of an equipment key executing mechanism can be solved, and the accuracy of equipment health assessment is improved. In addition, for the data of different categories of temperature data and vibration data, data acquisition at a plurality of different moments is respectively carried out on a plurality of monitoring points of the equipment from two view angles of time and space, the acquired data more comprehensively reflect the running condition of the equipment, and the accuracy of health assessment is further improved.
Preferably, as shown in fig. 2, the time interval between two adjacent temperature acquisition moments is in the order of seconds, and the time interval between two adjacent vibration acquisition moments is in the order of milliseconds.
The method comprises the steps of collecting key monitoring point data of equipment, adopting different time granularity according to different types of data, and setting the collection time granularity to be millisecond level for data with high change frequency such as vibration data; for data with low change frequency such as temperature data, the acquisition time granularity is set to the second level. And acquiring data with different time granularity, obtaining temperature data with a second acquisition period, obtaining vibration data with a millisecond acquisition period, wherein the acquisition point of each type of data comprises a plurality of different measurement points, and comprehensively acquiring monitoring data of a plurality of key measurement points of the equipment from two view angles of space and time. For example, the temperature and vibration data of the motor bearing of the high-temperature fan in the cement production industry are collected, the vibration data can be obtained once every 15 minutes, the time window of the vibration data is 1 minute, and the sampling rate is 2kHz.
The temperature acquisition time and the vibration acquisition time may be the same or different, and the temperature acquisition position and the vibration acquisition position may be the same or different. For example, at the same time, vibration data and temperature data measured at multiple points of the device are collected, which combines the vibration data and the temperature data measured at multiple points of the device motor, the transmission mechanism, the execution mechanism and the like; meanwhile, the data with different acquisition cycle time granularity are combined, so that the second-level temperature data in the SCADA system and the vibration data with the acquisition frequency of 1 KHz-10 KHz are available. Under the two perspectives of space and time, the device health assessment and prediction are carried out, the comprehensiveness and reliability of the device health assessment and prediction are improved, and the problem that the assessment accuracy of the device health is influenced only by single second-level (lower acquisition frequency) data or single measuring point data is also solved.
Preferably, before the average value calculation is performed on all the temperature data, cleaning and denoising are performed on all the temperature data.
For example, a sliding window algorithm based on median filtering can be used for cleaning the equipment temperature data, so that the quality of the equipment temperature data is improved. The sliding window algorithm based on the median filtering is characterized in that: constructing a slider with a fixed size, gradually sliding the slider downwards along the temperature data chain, reordering the temperature data in the slider once sliding, taking the median value of the temperature data after sequencing as the true value of the slider, and then moving the slider downwards until the tail of the data chain. The fixed slider size is preset to be 3-7 data units, for example, the fixed slider size is set to be 5 data units, and fig. 3 shows a specific flow of a sliding window algorithm based on median filtering.
Preferably, the method for calculating the evaluation compensation factor comprises the following steps:
carrying out standardization processing on a group of temperature data which is formed by collecting all temperature collecting moments at each temperature collecting position, wherein the temperature data in the group of temperature data is converted into 0, the temperature data which is larger than the standard value is converted into 1, and the temperature data which is between the standard value and the alarm value is converted into the ratio of a first difference value to a second difference value, wherein the first difference value is the difference value between the temperature data and the standard value, and the second difference value is the difference value between the standard value and the alarm value;
averaging the normalized temperature data of each temperature acquisition position at each temperature acquisition time to obtain a temperature data compensation value at the temperature acquisition time;
multiplying the temperature data compensation value at each temperature acquisition time with the recommended compensation value at the temperature acquisition time, and taking the average value of the products as an evaluation compensation factor.
For example, the high temperature fan temperature data is characterized in that the data point distribution is concentrated between the standard value Ta and the alarm value Tb, a small part of the data points are distributed between the alarm value and the high alarm value, and a small part of the data points are distributed below and close to the standard value. The specific flow of the temperature data mechanism model is as follows by combining expert knowledge:
first, the temperature data is normalized according to the following formula: i represents the i-th temperature acquisition time.
Then, taking the average value of the normalized values of the temperature measuring points as the compensation value c of the sample data, and calculating the compensation value c by the following formula, wherein i is used for marking different data acquisition points. Specifically, in the present embodiment, a total of 7 temperature data acquisition points are included:
finally, the compensation factor Fa is calculated as follows.
Wherein s is a For the proposed compensation value set in this state, a sample setThe compensation value of (2) is the product of the compensation factor obtained by the set of samples and the suggested compensation value of the corresponding state, and n is the total acquisition times.
In addition, based on all vibration data, the state probability vector is obtained by adopting a trained vibration evaluation state model, and the method can be specifically as follows: cleaning all vibration data by adopting a local anomaly factor algorithm, and performing feature extraction on the cleaned vibration data by adopting a local mean decomposition algorithm to obtain a vibration feature evaluation set (the vibration data can be processed in step 110 or step 130 without limitation); based on the vibration characteristic evaluation set, a trained vibration evaluation state model is adopted to obtain a state probability vector.
Judging the abnormality degree of a sample point by calculating a local abnormality factor of the sample point, wherein the greater the value is, the greater the possibility that the point is an abnormal point is, the closer the value is to 1 or less than 1, the greater the possibility that the point is a normal point is, and the quality of vibration data of equipment is improved by adopting a local abnormality factor algorithm; extracting the characteristics of equipment vibration data by utilizing a local mean decomposition algorithm, and continuously decomposing original vibration data into products of a plurality of envelope signals and pure frequency modulation signals until an error function of the signals after decomposition and the original signals is a monotonic function; and selecting the decomposed previous k-dimensional product function component as a characteristic quantity of the vibration signal, and taking the extracted previous k-dimensional product function component as a characteristic quantity of the vibration data of the equipment.
Specifically, firstly, data cleaning is performed on vibration data, and the data cleaning step comprises the following steps:
the k-neighborhood of the sample point p is defined as the distance between the data set point p and the point closest to the k-th point (excluding p) of the point p. The distance may be calculated according to the characteristics of different data sets, and euclidean distance is generally adopted by default. The sample points are respectively marked as x i I=1, 2,3, …, point x when k=t i The k distance of (2) is:
wherein x is k＝t Representative distance point x i The kth near point.
Calculate the point x i And point x j The reachable distance between the two is as follows:
RD k (x i ，x j )＝max(D k (x i ),||x i -x j ||)
let us assume at point x i N points in total in k neighborhood of (a) by x N(m) M=1, 2,3 … N, point x i The calculation formula of the local reachable density of (2) is as follows:
the point x can be further obtained by the following formula i Is a local anomaly factor of (a):
the denominator of this is the point x i Is the local reachable density of (2), the molecule is the point x i The average value of the local reachable densities of all sample points in the k neighborhood of (a), and judging the abnormality degree of the point by the ratio. If point x i Local reachable density LRD of (2) k (x i ) The smaller the point x i Local abnormality factor LOF of (2) k (x i ) The larger the point x i The more likely an outlier is. If point x i Local reachable density LRD of (2) k (x i ) The larger the point x i Local abnormality factor LOF of (2) k (x i ) The closer to 1 or less than 1, at point x i The smaller the probability of being an outlier.
After data cleaning, the cleaned vibration data is subjected to local mean decomposition to be decomposed into products of a plurality of envelope signals and pure frequency modulation signals until the error function of the signals after decomposition and the original signals is a monotonic function. The specific decomposition steps comprise:
finding all extreme points n (i) of the sample points, and calculating an average value m (i) between adjacent extreme points:
after smoothing the curves formed by all M (i), a local mean function M can be obtained 11 (i) A. The invention relates to a method for producing a fibre-reinforced plastic composite And calculates an envelope estimate l (i) from n (i):
similarly, an envelope estimate function L can be obtained 11 (i)。
From X (i) M 11 (i) Separating to obtain H 11 (i)：
H 11 (i)＝X(i)-M 11 (i)
By L 11 (i) For H 11 (i) Further modulating to obtain pure FM signal S 11 (i)：
S 11 (i)＝H 11 (i)/L 11 (i)
Ideal S 11 (i) Is a pure frequency modulation signal, i.e. requires L 11 (i) =1, otherwise will S 11 (i) Repeating the above steps as a new X (i) until S 11 (i) Is a pure frequency modulation signal. The iteration number is recorded as n, and the iteration is continuously carried out according to the steps to obtain H 1n (i) And S is 1n (i)。
Multiplying all envelope estimation value functions generated in the process to obtain an envelope signal L 1 (i)：
Thus, from L 1 (i) And S is 1n (i) Multiplication gives a first PF component, denoted PF1:
PF1(i)＝L 1 (i)S 1n (i)
the component comprising the signal of highest frequency of the original signal, which is instantaneousThe amplitude is L 1 (i) Its instantaneous frequency can be defined by S 1n (i) The calculation results are that:
PF1 is separated from X (i) to obtain a new signal u (i), and the above steps are repeated until u m (i) The number of repeated iterations is noted as m, which is a monotonic function.
Decomposing the original signal into m-dimensional PF components and a monotone signal u by m-step decomposition m (i) And:
fig. 4 shows an effect diagram of local mean decomposition of the high temperature fan in a healthy state, and the product function component of the first k dimensions is selected as a feature quantity of the vibration signal, where k=5 in this embodiment.
When the equipment health assessment and prediction are carried out by a machine learning-based method, the diagnosis effect and accuracy depend on the parameter optimization of a machine learning algorithm, models trained by different parameters have differences, and if the parameters are not optimized or the optimization effect is not good, the diagnosis is inaccurate. Meanwhile, the parameter optimization method of the existing method also has the problem of low efficiency. Therefore, preferably, in the training process of the vibration evaluation state model, a firefly algorithm is adopted, a vector is formed by a model learning rate parameter and a maximum depth parameter of a tree, the vector represents the position of fireflies, an objective function is defined, the input of the objective function is the position of the fireflies, the output of the objective function is an evaluation standard of the vibration evaluation state model, and the optimal solution of the learning rate parameter and the maximum depth parameter of the tree is obtained synchronously through iterative calculation.
Specifically, as shown in fig. 5, the method for parameter tuning by adopting the firefly algorithm comprises the following steps:
(S1) generating n fireflies in a feasible region, and randomly initializing an initial position x of each firefly i Endowed withThe fluorescence value of each firefly is l 0 The dynamic decision domain is r 0 . Initializing a fluorescence extinction rate rho, a fluorescence value update rate gamma, and defining the maximum iteration number. Defining an objective function f (x), wherein the input of the function is the position of fireflies, the output of the function is the evaluation standard of a Catboost algorithm, namely the output of the Catboost class is the model classification accuracy and the single-group sample evaluation time consumption.
(S2) calculating the fluorescence value l of the firefly at the time t according to the following formula i (t),i＝1,2,3,…,n。
l i (t)＝(1-ρ)l i (t-1)+γf(x i (t))
(S3) each firefly is in its dynamic decision domain r i In (t+1), firefly monomers with fluorescence values higher than that of firefly monomers are selected to form a neighborhood set N i (t) calculating the probability p of moving firefly i to each firefly j in its neighborhood by the following formula ij (t)。
(S4) each firefly moves according to the probability, the fireflies at the optimal positions randomly move, and the position x of each firefly is updated i (t)。
(S5) recalculating the fluorescence value of each firefly according to the formula in S1, when the maximum iteration number is met, the brightest firefly position is the optimal solution for the search, otherwise, returning to the step (S3).
And (S6) outputting all the individual optimal solutions, and comparing all the individual optimal solutions to obtain a global optimal solution.
Fig. 6 is a diagram showing the classification of the health status of the device according to the present embodiment, including health, sub-health, danger early warning, and high-risk alarm.
The key parameters affecting the performance of the c-Catboost/Catboost algorithm comprise a learning rate leanning_rate and a maximum depth of tree, the specific effects of the two parameters are different along with the difference of data, and in order to enable the c-Catboost/Catboost to automatically adjust parameters for different data during health evaluation and prediction, a method for optimizing the parameters of the c-Catboost/Catboost model based on a firefly algorithm is adopted, so that the purpose that the c-Catboost/Catboost model can automatically match the optimal training parameters under different data conditions is achieved, and the adaptability of the c-Catboost/Catboost evaluation and prediction algorithm is improved. In addition, 2 key parameters for the c-Catoost/Catoost model were determined using FA (Firefly Algorithm ): the learning rate leanning_rate and the maximum depth of the tree are optimized, and meanwhile, the optimal 2 key parameters are obtained instead of optimizing one by one, so that the optimization efficiency is improved.
Preferably, as shown in fig. 2, the method adds a filter based on probability density, filters the output (estimated state vector) of the Catboost model, then performs weighted average, and corrects the filtered and weighted average output of the Catboost model by using a compensation factor to form a c-Catboost model.
Specifically, as shown in fig. 7, before the probability density corresponding to each evaluation state in the evaluation state vector is weighted and averaged, according to the probability density and the model diagnosis error rate of each evaluation state in the evaluation state vector, part of the evaluation states and the number of data samples corresponding to the evaluation states are filtered from the evaluation state vector, so as to obtain a new evaluation state vector.
Specifically, the filtering rule is that when the number of data samples in a certain evaluation state is not more than one ten thousandth of the total number of data samples, the result in the evaluation state is regarded as a diagnosis error, and the diagnosis error is removed.
Preferably, the result of compensating the weighted average by using the evaluation compensation factor is specifically: the result of the evaluation compensation factor and the weighted average is added for correction.
Preferably, in the weighted average, the weight corresponding to each evaluation state is: the ratio of the number of data samples to the total number of data samples for the evaluation state in the evaluation state vector; alternatively, a weighted average method based on fuzzy membership is used for determination.
In summary, the method of the embodiment performs data acquisition on a plurality of execution mechanisms of the equipment under different time granularity to obtain temperature data with a second acquisition period and vibration data with a millisecond sampling period, and then a sliding window algorithm and a local anomaly factor algorithm based on median filtering can be respectively adopted to perform data cleaning on the temperature data and the vibration data of the equipment; carrying out local mean decomposition on vibration signal data of equipment to be evaluated to obtain characteristic data, and constructing an original input sample set; establishing a temperature data mechanism model of the equipment based on expert knowledge, calculating a compensation factor through the mechanism model, and constructing a c-Catboost model by utilizing the compensation factor and adding a filter based on probability density; then, performing equipment health assessment by adopting a trained and optimized c-Catboost model to obtain a health assessment value; and finally, predicting the future health of the equipment according to the obtained historical health evaluation value to obtain an evaluation and prediction result. Furthermore, a firefly algorithm can be adopted, two key parameters of the learning rate of the c-Catboost model and the maximum depth of the tree are optimized at the same time, the equipment health assessment precision and the health prediction accuracy are respectively improved, and the quantitative assessment of the equipment health state can be provided for the predictive maintenance of the equipment by the method of the embodiment.
For a better illustration of the evaluation method according to the invention, the following is given by way of example:
as shown in fig. 8, the above-mentioned evaluation method is adopted to perform equipment health evaluation, wherein, as shown in fig. 9, vibration data and temperature data of multipoint measurement of equipment are collected at the same moment, vibration data and temperature data of multipoint measurement of equipment motors, transmission mechanisms, execution mechanisms and the like are combined, and by combining data with different collection cycle time granularity, second-level temperature data and vibration data with collection frequency of 1 KHz-10 KHz in an SCADA system are formed, and further based on the method of the embodiment, equipment health evaluation and prediction are performed.
The example apparatus of fig. 9 is a high temperature fan, which is a key apparatus in cement production, and has a very important meaning for the judgment of the operation condition thereof. The measuring point 1 is a non-driving end of the motor, the measuring point 2 is a driving end of the motor, the measuring point 3 is a driving end of the fan, and the measuring point 4 is a non-driving end of the fan. The high-temperature fan motor is arranged between the measuring point 1 and the measuring point 2, the high-temperature fan transmission mechanism is arranged between the measuring point 2 and the measuring point 3, and the high-temperature fan executing mechanism (fan blade) is arranged between the measuring point 3 and the measuring point 4.
The vibration sensor is deployed for 4 measuring points of the high-temperature fan, so that the vibration condition of the corresponding part of the fan is comprehensively reflected, and vibration data in 3 directions are acquired at each measuring point.
The temperature data are shown in table 1:
table 1 parameters relating to high temperature fan performance in a centralized report
Defining the health of the device includes: the A level, the B level, the C level and the D level respectively correspond to health, sub-health, danger early warning and high-danger warning.
Wherein: class a = new delivery status;
class B = long-term operating state (long-term continuous operation);
class C = unfavorable continuous operation (short term continuous operation);
class D = severe vibration, equipment may be damaged at any time.
According to the 4 measuring point data, the method of the embodiment is adopted to carry out equipment health assessment.
For example, if the health evaluation result is level C, if the evaluation is specifically located at the measuring point 2 and the measuring point 3, it is necessary to check or update whether the coupling and the motor mounting screw are reliable or not, and whether a loosening condition exists or not.
Example two
A device health assessment system, comprising:
a data acquisition and processing module for performing step 110 of a device health assessment method as described in the first embodiment; a compensation factor calculating module, configured to perform step 120 of an apparatus health assessment method according to the first embodiment, based on all the temperature data obtained by the data collecting and processing module; the device health evaluation model is configured to perform step 130 in a device health evaluation method according to the first embodiment, based on all vibration data obtained by the data acquisition and processing module and the evaluation compensation factor obtained by the compensation factor calculation module. The related technical solution is the same as the first embodiment, and will not be described herein.
It should be noted that the data acquisition and processing model may clean and dry the temperature data, to transmit to the compensation factor calculation module to calculate the compensation factor, and may clean and dry the vibration data and extract the characteristics, to transmit to the equipment health evaluation model to calculate the evaluation state vector.
The system comprises a device health evaluation model, wherein the device health evaluation model is driven by the data acquisition and processing module and the compensation factor calculation module, and is used for evaluating the device health by adopting vibration data alone, so that the influence of temperature on the device health is ignored, the evaluation result has deviation and is required to be compensated by adopting temperature data. In addition, the data acquisition and processing module respectively performs data acquisition at a plurality of different moments on a plurality of monitoring points of the equipment from two view angles of time and space, the acquired data more comprehensively reflects the running condition of the equipment, and the accuracy of health assessment is further improved. By adopting the system, the accuracy of equipment health assessment is effectively improved.
Example III
A device health prediction method, which is implemented by adopting the device health evaluation method according to the first embodiment, performs multiple health evaluations on a device to be predicted, and performs health prediction on the device to be predicted based on the results of the multiple health evaluations. The related technical solution is the same as the first embodiment, and will not be described herein.
By adopting the equipment health assessment method, the accuracy of the result for health prediction is ensured, and the accuracy of equipment health prediction is further improved.
Example IV
A machine-readable storage medium storing machine-executable instructions that, when invoked and executed by a processor, cause the processor to implement a device health assessment method as described in embodiment one above.
The related technical solutions are the same as the embodiments, and are not described herein again.
It will be readily appreciated by those skilled in the art that the foregoing description is merely a preferred embodiment of the invention and is not intended to limit the invention, but any modifications, equivalents, improvements or alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (10)
1. A method of device health assessment, comprising:
s1, synchronously acquiring vibration data of different vibration acquisition positions on equipment to be evaluated at each vibration acquisition time, and synchronously acquiring temperature data of different temperature acquisition positions on the equipment at each temperature acquisition time;
s2, carrying out average value calculation on all the temperature data, and taking a calculation result as an evaluation compensation factor;
s3, based on all vibration data, adopting a trained vibration evaluation state model to evaluate to obtain an evaluation state vector composed of the number of data samples in each evaluation state, carrying out weighted average on probability densities corresponding to each evaluation state in the evaluation state vector, and adopting an evaluation compensation factor to compensate the weighted average result to obtain an evaluation value, thereby completing equipment health evaluation;
the step S2 comprises the following steps:
carrying out standardization processing on a group of temperature data which is formed by collecting all temperature collecting moments at each temperature collecting position, wherein the temperature data in the group of temperature data is converted into 0 and the temperature data which is larger than an alarm value is converted into 1, and the temperature data between the standard value and the alarm value is converted into a ratio of a first difference value to a second difference value, wherein the first difference value is the difference value between the temperature data and the standard value, and the second difference value is the difference value between the standard value and the alarm value;
averaging the normalized temperature data of each temperature acquisition position at each temperature acquisition time to obtain a temperature data compensation value at the temperature acquisition time;
multiplying the temperature data compensation value at each temperature acquisition time with the recommended compensation value at the temperature acquisition time, and taking the average value of the products as the compensation factor.
2. The apparatus health assessment method according to claim 1, wherein the time interval between adjacent two of the temperature acquisition times is of the order of seconds, and the time interval between adjacent two of the vibration acquisition times is of the order of milliseconds;
the S1 further includes: cleaning and denoising all the temperature data; all of the vibration data are cleaned to remove outliers.
3. The device health assessment method of claim 1, wherein the vibration assessment state model is obtained by training a Catboost model based on a device vibration data training set.
4. The method for evaluating equipment health according to claim 1, wherein in the training process of the vibration evaluation state model, a firefly algorithm is adopted to form a vector of a model learning rate parameter and a maximum depth parameter of a tree, the vector represents a position of fireflies, an objective function is defined, an input of the objective function is the position of the fireflies, an output of the objective function is an evaluation standard of the vibration evaluation state model, and an optimal solution of the learning rate parameter and the maximum depth parameter of the tree is obtained synchronously through iterative calculation.
5. The method according to claim 1, wherein in the step S3, before weighted averaging the probability densities corresponding to the respective evaluation states in the evaluation state vector, a part of the evaluation states and the number of data samples corresponding to the evaluation states are filtered from the evaluation state vector according to the probability density and the model diagnosis error rate of each evaluation state in the evaluation state vector, so as to obtain a new evaluation state vector.
6. The method according to claim 1, wherein in S3, the result of the weighted average is compensated by using the evaluation compensation factor, specifically: adding the evaluation compensation factor to the result of the weighted average to make a correction;
in the weighted average, the weight corresponding to each evaluation state is: the ratio of the number of data samples of the evaluation state to the total number of data samples in the evaluation state vector; alternatively, a weighted average method based on fuzzy membership is used for determination.
7. The method according to any one of claims 1 to 6, wherein the estimating state vector is obtained by using a trained vibration estimating state model based on all vibration data, specifically:
cleaning all vibration data by adopting a local anomaly factor algorithm, and extracting characteristics of the cleaned vibration data by adopting a local mean decomposition algorithm to obtain a vibration characteristic evaluation set; based on the vibration characteristic evaluation set, a trained vibration evaluation state model is adopted to obtain an evaluation state vector.
8. A device health assessment system, comprising:
a data acquisition and processing module for performing S1 in a device health assessment method according to any one of claims 1 to 7; a compensation factor calculation module for performing S2 in a device health assessment method according to any one of claims 1 to 7 based on all the temperature data obtained by the data acquisition and processing module; an equipment health assessment model for executing S3 in an equipment health assessment method according to any one of claims 1 to 7 based on all vibration data obtained by the data acquisition and processing module and the assessment compensation factor obtained by the compensation factor calculation module.
9. A device health prediction method, characterized in that a device health evaluation method according to any one of claims 1 to 7 is adopted, a plurality of health evaluations are performed on a device to be predicted, and health prediction is performed on the device to be predicted based on the result of the plurality of health evaluations.
10. A machine-readable storage medium storing machine-executable instructions which, when invoked and executed by a processor, cause the processor to implement a device health assessment method according to any one of claims 1 to 7 and/or a device health prediction method according to claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010431329.9A CN111639842B (en) | 2020-05-20 | 2020-05-20 | Equipment health assessment method, assessment system and equipment health prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010431329.9A CN111639842B (en) | 2020-05-20 | 2020-05-20 | Equipment health assessment method, assessment system and equipment health prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111639842A CN111639842A (en) | 2020-09-08 |
| CN111639842B true CN111639842B (en) | 2023-06-20 |
Family
ID=72329515
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010431329.9A Active CN111639842B (en) | 2020-05-20 | 2020-05-20 | Equipment health assessment method, assessment system and equipment health prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111639842B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112561230A (en) * | 2020-11-06 | 2021-03-26 | 国网浙江慈溪市供电有限公司 | Environment-friendly equipment state monitoring method based on electrical characteristics |
| CN113656238B (en) * | 2021-08-17 | 2023-07-04 | 北京京航计算通讯研究所 | AI industrial application capability test method and system of intelligent terminal |
| CN115809178B (en) * | 2022-11-29 | 2023-10-17 | 国网山东省电力公司应急管理中心 | Emergency equipment health state monitoring and early warning method and device based on data analysis |
| CN117670260B (en) * | 2024-01-31 | 2024-04-30 | 成都三一能源环保技术有限公司 | Downhole drilling instrument use period management system based on data analysis |
Family Cites Families (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102032943B (en) * | 2010-12-15 | 2012-07-04 | 常州赛杰电子信息有限公司 | Self-calibrated three-shaft vibration monitoring and environment monitoring acquisition system |
| CN103645052B (en) * | 2013-12-11 | 2016-03-30 | 北京航空航天大学 | A kind of gearbox of wind turbine remote online status monitoring and lifetime estimation method |
| US10042341B1 (en) * | 2015-02-19 | 2018-08-07 | State Farm Mutual Automobile Insurance Company | Systems and methods for monitoring building health |
| CN105160489B (en) * | 2015-09-28 | 2018-07-13 | 国家电网公司 | A kind of Hydropower Unit variable weight deterioration assessment system and appraisal procedure |
| CN107146004B (en) * | 2017-04-20 | 2018-02-16 | 浙江大学 | A kind of slag milling system health status identifying system and method based on data mining |
| CN107451402A (en) * | 2017-07-13 | 2017-12-08 | 北京交通大学 | A kind of equipment health degree appraisal procedure and device based on alarm data analysis |
| CN109425483B (en) * | 2017-09-04 | 2021-05-25 | 锐电科技有限公司 | Wind turbine generator running state evaluation and prediction method based on SCADA and CMS |
| CN109141881B (en) * | 2018-07-06 | 2020-03-31 | 东南大学 | Rotary machine health assessment method of deep self-coding network |
| CN108984893B (en) * | 2018-07-09 | 2021-05-07 | 北京航空航天大学 | Gradient lifting method-based trend prediction method |
| JP7143148B2 (en) * | 2018-08-23 | 2022-09-28 | 三菱重工業株式会社 | Prediction device, prediction method, and program |
| CN109711636A (en) * | 2019-01-09 | 2019-05-03 | 南京工业大学 | A kind of river level prediction technique promoting tree-model based on chaos firefly and gradient |
| CN110363339B (en) * | 2019-07-05 | 2022-03-08 | 南京简睿捷软件开发有限公司 | Method and system for performing predictive maintenance based on motor parameters |
| CN110794683A (en) * | 2019-11-27 | 2020-02-14 | 中国大唐集团科学技术研究院有限公司华中电力试验研究院 | Wind power gear box state evaluation method based on deep neural network and kurtosis characteristics |
-
2020
- 2020-05-20 CN CN202010431329.9A patent/CN111639842B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111639842A (en) | 2020-09-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111639842B (en) | Equipment health assessment method, assessment system and equipment health prediction method | |
| CN111275288B (en) | XGBoost-based multidimensional data anomaly detection method and device | |
| CN109146246B (en) | Fault detection method based on automatic encoder and Bayesian network | |
| CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
| WO2019080367A1 (en) | Method for evaluating health status of mechanical device | |
| CN111178553A (en) | Industrial equipment health trend analysis method and system based on ARIMA and LSTM algorithms | |
| CN110685868A (en) | Wind turbine generator fault detection method and device based on improved gradient elevator | |
| CN112414694B (en) | Equipment multistage abnormal state identification method and device based on multivariate state estimation technology | |
| CN108801950B (en) | Ultraviolet spectrum water quality abnormity detection method based on sliding window multi-scale principal component analysis | |
| CN114623050A (en) | Method, system and medium for detecting abnormal operation data of wind turbine generator | |
| CN113033011B (en) | Ship mechanical health state assessment method and system | |
| CN110570013B (en) | Single-station online wave period data prediction diagnosis method | |
| CN117290800B (en) | Timing sequence anomaly detection method and system based on hypergraph attention network | |
| CN114358435A (en) | Pollution source-water quality prediction model weight influence calculation method of two-stage space-time attention mechanism | |
| CN111861206A (en) | Industrial industry prosperity index obtaining method based on enterprise electric power big data | |
| CN113158448A (en) | Self-adaptive calculation method for equipment threshold of ship system | |
| CN117350377A (en) | Knowledge graph driving-based equipment fault diagnosis method and device | |
| CN116070669B (en) | Workshop energy consumption prediction method and management system based on improved deep belief network | |
| CN116933003A (en) | Remaining service life prediction method of unmanned aerial vehicle engine based on DaNet | |
| CN116070368A (en) | Method for predicting residual life of high-speed bearing of offshore wind turbine | |
| CN113212244B (en) | New energy vehicle power battery life prediction method and system | |
| CN115936473A (en) | Unsupervised KPI (Key performance indicator) abnormity detection method combining prediction and reconstruction | |
| Li et al. | Fault Feature Extraction Method of Reciprocating Compressor Valve Based on SPA-MF | |
| CN114819261A (en) | Fault prediction method and system based on real-time running state of equipment | |
| CN114201825A (en) | Method and system for evaluating equipment performance degradation state based on combination characteristics |
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 |