CN117864892A - Elevator fault prediction system and method - Google Patents
Elevator fault prediction system and method Download PDFInfo
- Publication number
- CN117864892A CN117864892A CN202410118871.7A CN202410118871A CN117864892A CN 117864892 A CN117864892 A CN 117864892A CN 202410118871 A CN202410118871 A CN 202410118871A CN 117864892 A CN117864892 A CN 117864892A
- Authority
- CN
- China
- Prior art keywords
- elevator
- data
- operation process
- process data
- maintenance
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B66—HOISTING; LIFTING; HAULING
- B66B—ELEVATORS; ESCALATORS OR MOVING WALKWAYS
- B66B5/00—Applications of checking, fault-correcting, or safety devices in elevators
- B66B5/0006—Monitoring devices or performance analysers
- B66B5/0037—Performance analysers
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B66—HOISTING; LIFTING; HAULING
- B66B—ELEVATORS; ESCALATORS OR MOVING WALKWAYS
- B66B5/00—Applications of checking, fault-correcting, or safety devices in elevators
- B66B5/0006—Monitoring devices or performance analysers
- B66B5/0018—Devices monitoring the operating condition of the elevator system
- B66B5/0025—Devices monitoring the operating condition of the elevator system for maintenance or repair
Landscapes
- Elevator Control (AREA)
Abstract
The invention discloses an elevator fault prediction system and method, wherein the system comprises: the elevator data acquisition module is used for acquiring elevator operation process data and elevator maintenance data; the data preprocessing module is in communication connection with the elevator data acquisition module, and the elevator operation process data and the elevator maintenance data establish a neural network topological structure to obtain associated elevator operation process data which causes the elevator maintenance data to be related to the elevator operation process data; the fault prediction module is in communication connection with the data preprocessing module, the elevator operation process data is input into the neural network model, the weight of elevator maintenance data corresponding to the elevator operation process data is calculated, the elevator maintenance data is combined to extract a maintenance time sequence and input into the machine learning model, and the residual service life of the elevator is obtained; and the emergency processing module is used for making corresponding standby emergency processing means according to the residual service life of the elevator internal equipment corresponding to the elevator maintenance data. And the prediction of elevator fault data and the prediction of service life are realized through the RBF neural network model.
Description
Technical Field
The invention belongs to the technical field of elevator fault detection, and particularly relates to an elevator fault prediction system and method.
Background
With the development demands of "smart cities" and "smart elevators", elevator operation system failure prediction problems have been widely focused and discussed. Current elevator maintenance is typically based on periodic inspection and maintenance planning or repair after an elevator fails. The problem with this approach is that potential faults cannot be predicted in advance, resulting in elevator faults and increased downtime, affecting user experience and operating costs.
In the Chinese patent of the invention with the patent number of CN201910297745.1, a fault prediction method based on elevator operation parameters is disclosed, elevator parameters sensitive to elevator faults in elevator data are obtained, and an elevator parameter set is formed; carrying out data processing on the elevator parameter set, removing abnormal values of parameters in the elevator parameter set, and filling missing data of the elevator parameter by adopting an interpolation method; and taking the elevator parameter set Q of M time points as a training data set, and taking the same mechanism of the elevator at the moment M as an elevator parameter vector with N dimensions to process the elevator parameters. Constructing a model of a multi-layer convolutional neural network; and carrying out fault prediction on the elevator parameters. Based on the fault prediction of the elevator operation parameters, the current acquired elevator parameters can be utilized, prediction can be performed based on the previous elevator parameters, a perfect elevator fault prediction model is established, and fault prediction is performed.
The defect of the existing patent is that the elevator acquisition parameters are trained by adopting a knowledge base rule reasoning mode so as to achieve the fault prediction conforming to the elevator parameters; but a single knowledge base rule needs a large amount of historical data or expert knowledge base data to support so as to realize the effect of predicting the training of the multi-layer convolutional neural network model; meanwhile, the prediction neural network model only predicts the elevator operation parameters which are linearly increased or have certain regular changes, and can not predict the elevator operation parameters and time sequences with complex change rules.
Disclosure of Invention
Aiming at the problem that the existing elevator fault prediction method adopts a reasoning mode of a knowledge base rule to train elevator acquisition parameters so as to achieve the fault prediction conforming to elevator parameters, but omits a large amount of historical data or expert knowledge base data required by the knowledge base rule to support so as to achieve the effect of predicting the training of the multi-layer convolutional neural network model, the invention provides an elevator fault prediction system and an elevator fault prediction method.
In order to achieve the technical purpose, the invention adopts the following technical scheme:
an elevator fault prediction system comprises an elevator data acquisition module, a data preprocessing module, a fault prediction module and an emergency processing module;
The elevator data acquisition module is used for acquiring elevator operation process data and elevator maintenance data;
the data preprocessing module is in communication connection with the elevator data acquisition module, a neural network topological structure is established through elevator operation process data and elevator maintenance data, and the elevator operation process data which causes the association of the elevator maintenance data with the elevator operation process data is obtained through analysis;
the fault prediction module is in communication connection with the data preprocessing module, the elevator operation process data is input into the neural network model, the weight of elevator maintenance data corresponding to the elevator operation process data is calculated, and then the elevator maintenance data is combined to extract a maintenance time sequence and input into the machine learning model, so that the residual service life of the elevator is obtained;
and the emergency processing module is used for making corresponding standby emergency processing means according to the residual service life of the internal equipment of the elevator, which is specifically corresponding to the maintenance data of each elevator.
Further, the elevator operation process data comprises uplink and downlink signal data, door zone signal data, uplink and downlink direction limit signal data, car top emergency stop switch signal data, safety gear switch signal data, car speed limiter switch signal data, thermal protection relay signal data, each floor door lock switch signal data, car door lock switch signal data, elevator start-stop speed data, elevator uniform speed operation speed data, elevator start-stop acceleration data, brake voltage data, brake current data and each temperature data.
Further, the elevator maintenance data comprises a limit switch contact failure, a band-type brake clearance too small or a brake screw failure, a motor contactor main contact pressure insufficient failure, a potentiometer adjustment improper of an SWD electronic board, a car top magnetic switch box failure, a power supply system sudden power failure, a control power fuse blowing or control switch contact failure, a total fuse breaking failure, a speed changing relay failure or a speed changing circuit open line to enable a stop relay failure and a floor relay contact failure.
Further, the detailed steps of establishing the neural network topology from the elevator run process data and the elevator repair data include:
elevator operation process data as input layer x of topological structure i (i=1, 2,3., n), hidden layer R i The activation function of the (i=1, 2,3., p) node is constituted by a gaussian function, the elevator repair data being the output layer y k (k＝1，2，3...，m)；
The input layer transmits an input signal to the hidden layer, the hidden layer determines the center of a radial basis function, vector data input by the input layer is directly mapped to a hidden layer space, the mapping relation from the hidden layer space to an output layer space is linear, and the output layer is a linear weighted sum of hidden layer outputs.
Further, the radial basis function center uses the formula of a gaussian function:
wherein c i Is the center vector, sigma, of the ith basis function i The variance (width parameter) of the radial basis function, p, is the number of hidden functions. The hidden layer adopts a nonlinear optimization methodThe output layer then needs to implement the function R from the Gaussian function i (x)→y k I.e. the output is the output that forms the neural network after linear weighted combination of the output layers.
Output layer y k The calculation formula of (2) is as follows:
wherein m is the number of nodes of the output layer, w ik Weights from the ith hidden layer node to the kth output node.
Further, the training method of the whole neural network model adopts a nearest neighbor cluster learning algorithm, and the nearest neighbor cluster learning algorithm flow comprises the following steps:
s1001, selecting a Gaussian function as a radial basis function, setting a vector a (l) for storing the sum of various output vectors, and counting the number of samples by a design counter b (l), wherein l represents the number of categories;
S1002、(x 1 ，y 1 ) Initializing c as an initialization data pair 1 ＝x 1 ，a(1)＝y 1 B (1) =l, x 1 Establishing RBF network for cluster center, and initially implying layer as c 1 The output layer weight vector is initially w 1 ＝a(1)/b(1)；
S1003, adding a sample data pair (x 2 ，y 2 ) Thereby finding x 2 To c 1 Center distance |x 2 -c 1 I (I); when |x 2 -c 1 |≤γ，c 1 That is x 2 Nearest neighbor cluster of (1) =y at this time 1 +y 2 ，b(1)＝2，w 1 =a (1)/b (2); when |x 2 -c 1 | > γ, x 2 As a new cluster center, get c 2 ＝x 2 ，a(2)＝y 2 B (2) =1; will hide the unit c 2 Adding to the network established in step S1002, c 2 The weight vector of the output layer is w 2 ＝a(2)/b(2)。
S1004, when the mth data pair (x m ，y m ) m=3, 4,5., N is falseLet M cluster centers exist in the network, M center points are represented as c 1 ，c 2 ，c 3 ，...c M At this time, M hidden units exist in the already established RBF network, and data pairs (x m ，y m ) Distance |x to cluster center m -c i I, i=1, 2, 3..m, find M distances minimum, assume |x m -x j Minimum then c j Is x m Is the nearest neighbor cluster of (a);
then judge when |x 2 -c 1 When I > gamma, X is m As a new cluster center, let c M+1 ＝x m ，M＝M+1，a(M)＝y m B (M) =1. The values of each a (i), b (i) (i=1, 2,3,..m-1) were kept unchanged. And then the hidden unit c M Added to the established RBF network.
When |x 2 -c 1 And (3) the I is less than or equal to gamma, and the following treatment is carried out: a (j) =a (j) +y m B (j) =b (j) +1. When i+.j, i=1, 2,3,..m, each a (i) is kept when in use, the value of b (i) (i=1, 2,3,) M-1 is unchanged. Let the weight vector of the output layer be w i ＝a(i)/b(i)i＝1，2，3，...，M。
S1005, according to the steps, the established RBF network output function is as follows:
In the algorithm, the smaller the gamma is, the more clusters are obtained, the larger the calculated amount is, and the magnitude of the gamma can be combined with the actual situation to adjust the initialization value.
The prediction algorithm of the whole neural network model can also adopt the following steps:
s101, initializing, and determining N vectors as initial clustering center vectors c i I.e. c 1 ，c 2 ，...，c N ；
S102, carrying out sample normalization processing,
s103, calculating the Euclidean distance, and solving the minimum distance:
d i (k)＝||X m -c i (k)||，m＝1，2，3...M
d min ＝min||X m -c i (k)||
wherein d is min Is the minimum Euclidean distance (nearest neighbor rule);
s104, updating center C i Calculating a sample mean value by adopting a mean value method;
s105, judging whether the distribution change of the clustering center is smaller than a preset value, if so, calculating d from the new calculation min The k value is added with 1, the step S1003 is returned, if the k value is smaller than the preset value, the step S1006 is entered;
s106, after confirming the center of the whole neural network, determining the mean square error:
s107, calculating linear weight from the hidden layer to the output layer, and completing the calculation by using an error correction learning algorithm, wherein the actual output is set as y k The calculated output isThen->
And adjusting the weight of the output layer, calculating by adopting a minimum mean square error rule, wherein the square difference between the actual output and the expected output is minimum, namely, the target signal is set as:
r＝d i -W i T X
the weight vector is adjusted as:
ΔW i ＝η(d j -W i T X)X。
Further, analyzing to obtain detailed steps of related elevator operation process data which cause the elevator maintenance data to be related to the elevator operation process data, adopting the elevator operation process data which judges that the elevator operation process data and a preset initial value generate abnormal changes as the reason data for generating the elevator maintenance data faults, observing whether the reasons for generating the elevator maintenance data faults all have the reasons for generating the elevator operation process data abnormality or not for many times, judging that the elevator operation process data is an input layer of a neural network topological structure if the occurrence frequency is higher than the preset frequency of a system, and judging that the elevator operation process data is not the reason for causing the elevator maintenance data if the occurrence frequency is lower than the preset frequency of the system.
Further, the weight of the elevator operation process data corresponding to the elevator maintenance data is the weight vector value of the output layer in step S104 or the linear weight from the hidden layer to the output layer in step S1007.
Further, the elevator operation process data further comprises an elevator number and an elevator installation position, the elevator maintenance data further comprises an elevator number, an elevator installation position and an elevator maintenance time, and the elevator remaining service life is analyzed in the machine learning model through the same type of elevator maintenance data and the sequence of the type of elevator maintenance time.
Further, the detailed steps of analyzing the remaining service life of the elevator in the machine learning model by inputting the elevator maintenance data of the same type and the sequence of the elevator maintenance time of the same type are as follows:
s2001, calculating a time sequence difference, and calculating and analyzing whether an elevator time maintenance sequence is regular or not by adopting a first-order difference, a second-order difference and a K-step difference mode;
s2002, selecting different time sequence models for learning according to different rules; the time sequence model comprises a white noise sequence model, an autoregressive model, a moving average model, an autoregressive moving average model and an ARIMA model;
s2003, predicting a next time fault node, namely the residual service life of the elevator, by adopting different time sequence models according to different types of elevator maintenance data;
and S2004, in combination with the next time fault node generated by prediction, pre-judging elevator maintenance data generated by the elevator is sent to the emergency processing module.
The elevator fault prediction system also comprises an elevator data identification module, wherein different elevator numbers are marked into the same type of elevator by identifying whether elevator numbers or collected elevator running process data are matched with elevator maintenance data field items; when the same elevator number lacks historical elevator operation process data and elevator maintenance data to train the neural network model, the historical elevator operation process data and elevator maintenance data are built for the same type of elevator to train the neural network model;
The method for identifying the elevator numbers to be marked into the same type of elevator comprises the following steps:
acquiring an elevator number, and converting the elevator number into a character string;
according to the total length value of the character string, intercepting the character of 50% length value of the character string to be matched with all written elevator numbers;
if the same character string is matched, judging that the elevator operation process data and the elevator maintenance data are used for training the neural network model by using the historical data written with the elevator number when the historical data are lack of training the neural network model, predicting the elevator maintenance data value by taking the weight value of the hidden layer as the weight value during fault prediction, and judging whether the elevator maintenance data value of the elevator number exceeds the fault early-warning value of the elevator synchronously matched with the same character string.
The method for identifying the same type of elevator by matching the collected elevator operation process data with the elevator maintenance data field items comprises the following steps:
acquiring elevator operation process data and elevator maintenance data fields;
judging whether fields of each data item of the elevator operation process data and the elevator maintenance data fields are matched;
if the fields of each data item of the elevator operation process data and the elevator maintenance data fields are matched, the same type of elevator is judged.
If the field part of the elevator operation process data and the elevator maintenance data input data item of the new elevator are matched, the same item of the historical elevator operation process data and the elevator maintenance data which are matched with the field part of the field part is used as a neural network model input layer and an output layer for training the new elevator, the node weight value of the hidden layer after training is used as the elevator operation process data of the new elevator to predict the weight value of the hidden layer, the elevator operation process data of the new elevator is multiplied by the weight value of the hidden layer to obtain predicted elevator maintenance data, and whether the maintenance data should be early-warned or not is judged.
The system also comprises a data storage management module, a historical elevator operation process data table and a historical elevator maintenance data table are built according to the same elevator numbers, the collected elevator operation process data and elevator maintenance data are automatically written into the corresponding historical elevator operation process data table and historical elevator maintenance data table after being predicted by the fault prediction module, and meanwhile, the historical elevator operation process data table and the historical elevator maintenance data table build corresponding index association relations according to elevator number fields.
An elevator failure prediction method, comprising the steps of:
s1, collecting relevant historical data of an elevator operation process and elevator maintenance;
S2, analyzing and obtaining historical data of the elevator maintenance data and the related correlation elevator operation process;
s3, building an RBF neural network topological structure or an RBF neural network model according to the association relation between elevator operation process data and elevator maintenance data;
s4, inputting the collected relevant historical data of the elevator operation process and elevator maintenance into an RBF neural network model, training the RBF neural network model by adopting a nearest neighbor clustering learning algorithm or a prediction algorithm, acquiring weights from hidden layer nodes to output nodes, and setting the weights as weight vectors of all hidden layers when the elevator operation process data are taken as an input layer and the elevator maintenance data are taken as an output layer;
s5, acquiring the latest elevator operation process data as an input layer in the RBF neural network model, and calculating to obtain predicted latest elevator maintenance data through implicit layer weight vector conversion obtained by calculation in the step S4;
s6, comparing the predicted latest elevator maintenance data with the elevator maintenance data of the same type in the same period history, if the elevator maintenance data of the same type in the same period history falls in the range or exceeds the range value, sending out corresponding elevator maintenance fault early warning, and if the elevator maintenance data of the same type in the same period history does not fall in the range or is lower than the minimum value of the range value, entering a step S7;
S7, inputting the predicted elevator maintenance data extraction time sequence into a machine learning model to obtain the residual service life of the elevator;
and S8, judging whether the residual service life of the elevator is close to the preset maintenance data of the whole elevator, if so, informing corresponding maintenance personnel to go to the door for maintenance through the elevator number and the elevator installation position, and if not, continuing to monitor the data of the operation process of the elevator.
Compared with the prior art, the invention has the following beneficial effects:
inputting relevant historical elevator maintenance data of the same type corresponding to elevator operation process data into an RBF neural network model, establishing an RBF neural network topological structure or the RBF neural network model, calculating and searching weights from hidden layer nodes to output nodes in the RBF neural network model, so that new elevator operation process data can be quickly converted into predicted elevator maintenance data; and comparing with synchronous real elevator maintenance data, judging whether the early warning is needed or not, and predicting the residual service life of the elevator. The method achieves the prejudgment of the maintenance data of the related historical elevators of the same type, only requires the training of synchronous partial data, not only can achieve the weight training effect from hidden layer nodes to output nodes in the RBF neural network model, but also can achieve the aim of predicting the training of the multi-layer convolutional neural network model without supporting a large amount of historical data or expert knowledge base data.
Drawings
Fig. 1 is a block diagram of an overall structure of an elevator failure prediction system according to an embodiment of the present invention;
fig. 2 is an overall flowchart of an elevator failure prediction method according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a neural network topology according to an embodiment of the present invention.
Detailed Description
The invention will be further described with reference to examples and drawings, to which reference is made, but which are not intended to limit the scope of the invention.
As shown in fig. 1, the embodiment provides an elevator fault prediction system, which comprises an elevator data acquisition module, a data preprocessing module, a fault prediction module and an emergency processing module;
the elevator data acquisition module is used for acquiring elevator operation process data and elevator maintenance data;
the data preprocessing module is in communication connection with the elevator data acquisition module, a neural network topological structure is established through elevator operation process data and elevator maintenance data, and the elevator operation process data which causes the association of the elevator maintenance data with the elevator operation process data is obtained through analysis;
the fault prediction module is in communication connection with the data preprocessing module, the elevator operation process data is input into the neural network model, the weight of elevator maintenance data corresponding to the elevator operation process data is calculated, and then the elevator maintenance data is combined to extract a maintenance time sequence and input into the machine learning model, so that the residual service life of the elevator is obtained;
And the emergency processing module is used for making corresponding standby emergency processing means according to the residual service life of the internal equipment of the elevator, which is specifically corresponding to the maintenance data of each elevator.
The elevator operation process data comprise uplink and downlink signal data, door zone signal data, uplink and downlink direction limit signal data, car top emergency stop switch signal data, safety clamp switch signal data, car speed limiter switch signal data, thermal protection relay signal data, all floor door lock switch signal data, car door lock switch signal data, elevator start-stop speed data, elevator uniform speed operation speed data, elevator start-stop acceleration data, brake voltage data, brake current data and all temperature data. A total of 15 items.
The elevator maintenance data comprise limit switch contact failure, brake clearance undersize or brake screw failure, main contact pressure failure of a motor contactor, improper potentiometer adjustment of an SWD electronic board, car top magnetic switch box failure, sudden power failure of a power supply system, control power fuse blowing or control switch contact failure, total fuse breaking failure, speed changing relay failure or speed changing circuit line disconnection to enable stop relay failure and floor relay contact failure. 11 items in total.
The detailed steps of building the neural network topology structure by the elevator operation process data and the elevator maintenance data comprise:
elevator operation process data as input layer x of topological structure i (i=1, 2,3., n), hidden layer R i The activation function of the (i=1, 2,3., p) node is constituted by a gaussian function, the elevator repair data being the output layer y k (k＝1，2，3...，m)；
As shown in fig. 3, the input layer delivers the input signal to the hidden layer, the hidden layer determines the center of the radial basis function, the vector data input by the input layer is directly mapped to the hidden layer space, the mapping relationship from the hidden layer space to the output layer space is linear, and the output layer is the linear weighted sum of the hidden layer outputs.
Further, the radial basis function center uses the formula of a gaussian function:
wherein c i Is the center vector, sigma, of the ith basis function i The variance (width parameter) of the radial basis function, p, is the number of hidden functions. The hidden layer adopts a nonlinear optimization method, and the output layer needs to realize the function R from the Gaussian function i (x)→y k I.e. the output is the output that forms the neural network after linear weighted combination of the output layers.
Output layer y k The calculation formula of (2) is as follows:
wherein m is the number of nodes of the output layer, w ik Weights from the ith hidden layer node to the kth output node.
Further, the training method of the whole neural network model adopts a nearest neighbor cluster learning algorithm, and the nearest neighbor cluster learning algorithm flow comprises the following steps:
s1001, selecting a Gaussian function as a radial basis function, setting a vector a (l) for storing the sum of various output vectors, and counting the number of samples by a design counter b (l), wherein l represents the number of categories;
S1002、(x 1 ，y 1 ) Initializing c as an initialization data pair 1 ＝x 1 ，a(1)＝y 1 B (1) =l, x 1 Establishing RBF network for cluster center, and initially implying layer as c 1 The output layer weight vector is initially w 1 ＝a(1)/b(1)；
S1003, adding a sample data pair (x 2 ，y 2 ) Thereby finding x 2 To c 1 Center distance |x 2 -c 1 I (I); when |x 2 -c 1 |≤γ，c 1 That is x 2 Nearest neighbor cluster of (1) =y at this time 1 +y 2 ，b(1)＝2，w 1 =a (1)/b (2); when |x 2 -c 1 | > γ, x 2 As a new cluster center, get c 2 ＝x 2 ，a(2)＝y 2 B (2) =1; will hide the unit c 2 Adding to the network established in step S1002, c 2 The weight vector of the output layer is w 2 ＝a(2)/b(2)。
S1004, when the mth data pair (x m ，y m ) m=3, 4,5, N, it is assumed that there are M cluster centers in the network at this time, M center points denoted as c 1 ，c 2 ，c 3 ，...c M At this time, M hidden units exist in the already established RBF network, and data pairs (x m ，y m ) Distance |x to cluster center m -c i I, i=1, 2, 3..m, find M distances minimum, assume |x m -x j Minimum then c j Is x m Is the nearest neighbor cluster of (a);
then judge when |x 2 -c 1 When I > gamma, X is m As a new cluster center, let c M+1 ＝x m ，M＝M+1，a(M)＝y m B (M) =1. The values of each a (i), b (i) (i=1, 2,3,..m-1) were kept unchanged. And then the hidden unit c M Added to the established RBF network.
When |x 2 -c 1 And (3) the I is less than or equal to gamma, and the following treatment is carried out: a (j) =a (j) +y m B (j) =b (j) +1. When i+.j, i=1, 2,3,..m, each a (i) is kept when in use, the value of b (i) (i=1, 2,3,) M-1 is unchanged. Let the weight vector of the output layer be w i ＝a(i)/b(i)i＝1，2，3，...，M。
S1005, according to the steps, the established RBF network output function is as follows:
in the algorithm, the smaller the gamma is, the more clusters are obtained, the larger the calculated amount is, and the magnitude of the gamma can be combined with the actual situation to adjust the initialization value.
The prediction algorithm of the whole neural network model can also adopt the following steps:
s101, initializing, and determining N vectors as initial clustering center vectors c i I.e. c 1 ，c 2 ，...，c N ；
S102, carrying out sample normalization processing,
s103, calculating the Euclidean distance, and solving the minimum distance:
d i (k)＝||X m -c i (k)||，m＝1，2，3...M
d min ＝min||X m -c i (k)||
wherein d is min Is the minimum Euclidean distance (nearest neighbor rule);
s104, in updatingHeart C i Calculating a sample mean value by adopting a mean value method;
s105, judging whether the distribution change of the clustering center is smaller than a preset value, if so, calculating d from the new calculation min The k value is added with 1, the step S1003 is returned, if the k value is smaller than the preset value, the step S1006 is entered;
s106, after confirming the center of the whole neural network, determining the mean square error:
s107, calculating linear weight from the hidden layer to the output layer, and completing the calculation by using an error correction learning algorithm, wherein the actual output is set as y k The calculated output isThen->
And adjusting the weight of the output layer, calculating by adopting a minimum mean square error rule, wherein the square difference between the actual output and the expected output is minimum, namely, the target signal is set as:
r＝d i -W i T X
the weight vector is adjusted as:
ΔW i ＝η(d j -W i T X)X。
analyzing to obtain detailed step of related elevator operation process data which causes elevator maintenance data to be related to the elevator operation process data, wherein the elevator operation process data which is used for judging abnormal changes of the elevator operation process data and a preset initial value is adopted as cause data for generating elevator maintenance data faults;
and whether the cause of the failure of the elevator maintenance data is generated by multiple matching search or not is judged, the frequency of the elevator maintenance data when generated is judged, if the frequency of the elevator maintenance data is higher than the preset frequency of the system, the elevator maintenance data is judged to be an input layer of a neural network topological structure, and if the frequency of the elevator maintenance data is lower than the preset frequency of the system, the elevator maintenance data is judged not to be the cause of the failure. The elevator operation process data of the input layer has no mapping relation with the output layer of the neural network topological structure, and has no corresponding hidden layer. The multiple matching search is generally limited to be within 10 times in the system, the index search for reducing the data volume is reduced, and meanwhile, the related elevator operation process data for generating elevator maintenance data faults can be obtained through a small amount of data matching analysis, so that an input layer and an output layer of the neural network topological structure are established.
The weight of the elevator maintenance data corresponding to the elevator operation process data is the weight vector value of the output layer in the step S104 or the linear weight from the hidden layer to the output layer in the step S1007.
The elevator operation process data also comprises an elevator number and an elevator installation position, the elevator maintenance data also comprises an elevator number, an elevator installation position and an elevator maintenance time, and the elevator residual service life is analyzed in the machine learning model by inputting the same type of elevator maintenance data and the sequence of the type of elevator maintenance time. And the elevator installation position is collected, so that the maintenance address can be rapidly positioned when faults occur.
The sequence input machine learning model of the elevator maintenance data of the same type and the elevator maintenance time of the same type is used for analyzing the residual service life of the elevator, and the method comprises the following detailed steps:
s2001, calculating a time sequence difference, and calculating and analyzing whether an elevator time maintenance sequence is regular or not by adopting a first-order difference, a second-order difference and a K-step difference mode;
s2002, selecting different time sequence models for learning according to different rules; the time sequence model comprises a white noise sequence model, an autoregressive model, a moving average model, an autoregressive moving average model and an ARIMA model;
S2003, predicting a next time fault node, namely the residual service life of the elevator, by adopting different time sequence models according to different types of elevator maintenance data;
and S2004, in combination with the next time fault node generated by prediction, pre-judging elevator maintenance data generated by the elevator is sent to the emergency processing module. The emergency processing module prepares emergency processing means in advance according to fault elevator maintenance data possibly generated by different elevator operation process data. The same type of elevator maintenance data refers to the reason that a single elevator maintenance data analysis leads to the generation of the elevator maintenance data, namely, a plurality of elevator operation process data related to the elevator maintenance data, a neural network topological structure or a neural network model is established to calculate the weight value of a corresponding hidden layer of a single elevator maintenance data output layer, so that the influence of a plurality of irrelevant elevator operation process data on the predicted elevator maintenance data value is conveniently removed.
The elevator fault prediction system also comprises an elevator data identification module, wherein different elevator numbers are marked into the same type of elevator by identifying whether elevator numbers or collected elevator running process data are matched with elevator maintenance data field items; when the same elevator number lacks historical elevator operation process data and elevator maintenance data to train the neural network model, the historical elevator operation process data and elevator maintenance data are built for the same type of elevator to train the neural network model;
The method for identifying the elevator numbers to be marked into the same type of elevator comprises the following steps:
acquiring an elevator number, and converting the elevator number into a character string;
according to the total length value of the character string, intercepting the character of 50% length value of the character string to be matched with all written elevator numbers; the method for intercepting the character string adopts a shift intercepting mode, firstly intercepts from the first position of the character string, if the character string is matched with all written elevator numbers, the intercepting is stopped when the same character string is searched, and if the same character string is not searched, the character string with the same length value is continuously intercepted from the second position of the character string and is matched with all written elevator numbers until the total length value of the whole character string is intercepted.
If the same character string is matched, judging that the elevator operation process data and the elevator maintenance data are used for training the neural network model by using the historical data written with the elevator number when the historical data are lack of training the neural network model, predicting the elevator maintenance data value by taking the weight value of the hidden layer as the weight value during fault prediction, and judging whether the elevator maintenance data value of the elevator number exceeds the fault early-warning value of the elevator synchronously matched with the same character string. And if the detected value exceeds the preset value, early warning is carried out. The elevator numbers to be identified are not compared with contemporaneous historical data, so that the elevator maintenance data of the same type cannot be extracted and the maintenance time sequence is input into a machine learning model to obtain the remaining service life of the elevator, the elevator numbers of a new type cannot be input to predict the remaining service life of the elevator, and the possibility of overrun of the elevator maintenance data value and the possibility of fault early warning in the elevator operation process data of the same type can be found out. The elevator fault prediction system can achieve the effect of predicting elevator faults when different elevator numbers are achieved or new elevator data are written into the system.
The method for identifying the same type of elevator by matching the collected elevator operation process data with the elevator maintenance data field items comprises the following steps:
acquiring elevator operation process data and elevator maintenance data fields;
judging whether fields of each data item of the elevator operation process data and the elevator maintenance data fields are matched; if the elevator running process data and the elevator maintenance data field input by the type A elevator are completely matched with the type B elevator, the failure prediction neural network model of the type B elevator is judged to be suitable for the type A elevator. A large amount of historical data or expert knowledge base data required by the input of a new elevator type is reduced to support, so that the aim of predicting the training of the multi-layer convolutional neural network model can be achieved; meanwhile, if the existing elevator corresponding data with different elevator numbers can be divided into the same type of elevators, the memory storage space capacity occupied by the model calculation and the model is reduced.
If the fields of each data item of the elevator operation process data and the elevator maintenance data fields are matched, the same type of elevator is judged.
If the field part of the elevator operation process data and the elevator maintenance data input data item of the new elevator are matched, the same item of the historical elevator operation process data and the elevator maintenance data which are matched with the field part of the field part is used as a neural network model input layer and an output layer for training the new elevator, the node weight value of the hidden layer after training is used as the elevator operation process data of the new elevator to predict the weight value of the hidden layer, the elevator operation process data of the new elevator is multiplied by the weight value of the hidden layer to obtain predicted elevator maintenance data, and whether the maintenance data should be early-warned or not is judged.
The system also comprises a data storage management module, a historical elevator operation process data table and a historical elevator maintenance data table are built according to the same elevator numbers, the collected elevator operation process data and elevator maintenance data are automatically written into the corresponding historical elevator operation process data table and historical elevator maintenance data table after being predicted by the fault prediction module, and meanwhile, the historical elevator operation process data table and the historical elevator maintenance data table build corresponding index association relations according to elevator number fields. And the accurate acquisition and calculation of historical data are facilitated when the neural network model is built by the follow-up fault prediction module.
As shown in fig. 2, an elevator failure prediction method includes the steps of:
s1, collecting relevant historical data of an elevator operation process and elevator maintenance;
s2, analyzing and obtaining historical data of the elevator maintenance data and the related correlation elevator operation process;
s3, building an RBF neural network topological structure or an RBF neural network model according to the association relation between elevator operation process data and elevator maintenance data;
s4, inputting the collected relevant historical data of the elevator operation process and elevator maintenance into an RBF neural network model, training the RBF neural network model by adopting a nearest neighbor clustering learning algorithm or a prediction algorithm, acquiring weights from hidden layer nodes to output nodes, and setting the weights as weight vectors of all hidden layers when the elevator operation process data are taken as an input layer and the elevator maintenance data are taken as an output layer;
s5, acquiring the latest elevator operation process data as an input layer in the RBF neural network model, and calculating to obtain predicted latest elevator maintenance data through implicit layer weight vector conversion obtained by calculation in the step S4;
s6, comparing the predicted latest elevator maintenance data with the elevator maintenance data of the same type in the same period history, if the elevator maintenance data of the same type in the same period history falls in the range or exceeds the range value, sending out corresponding elevator maintenance fault early warning, and if the elevator maintenance data of the same type in the same period history does not fall in the range or is lower than the minimum value of the range value, entering a step S7;
S7, inputting the predicted elevator maintenance data extraction time sequence into a machine learning model to obtain the residual service life of the elevator;
and S8, judging whether the residual service life of the elevator is close to the preset maintenance data of the whole elevator, if so, informing corresponding maintenance personnel to go to the door for maintenance through the elevator number and the elevator installation position, and if not, continuing to monitor the data of the operation process of the elevator. The approach judgment method can adopt the elevator numbers to inquire the elevator maintenance data maintenance time average value or the equal proportion change value of the elevators with the same elevator model, if the residual service life (residual time) is more than one third of the whole normal elevator maintenance data maintenance time average value or the equal proportion change value, no early warning is needed, the approach judgment method judges that the elevators are not close, and if the residual service life (residual time) is less than one third of the whole normal elevator maintenance data maintenance time average value or the equal proportion change value, the approach judgment method judges that the elevators are close. Sending out an early warning report; meanwhile, according to different types of elevator maintenance data (the bad contact faults of the limit switch are caused by sundry dust), the sundry dust accumulation part of the elevator can be dedusted through a preset dedusting device, and the limit switch is ensured to be contacted more fully. The dust removing device is used as emergency treatment equipment connected with the emergency treatment module.
Compared with the prior art, the invention has the following beneficial effects:
inputting relevant historical elevator maintenance data of the same type corresponding to elevator operation process data into an RBF neural network model, establishing an RBF neural network topological structure or the RBF neural network model, calculating and searching weights from hidden layer nodes to output nodes in the RBF neural network model, so that new elevator operation process data can be quickly converted into predicted elevator maintenance data; and comparing with synchronous real elevator maintenance data, judging whether the early warning is needed or not, and predicting the residual service life of the elevator. The method achieves the prejudgment of the maintenance data of the related historical elevators of the same type, only requires the training of synchronous partial data, not only can achieve the weight training effect from hidden layer nodes to output nodes in the RBF neural network model, but also can achieve the aim of predicting the training of the multi-layer convolutional neural network model without supporting a large amount of historical data or expert knowledge base data.
The elevator fault prediction system and the elevator fault prediction method provided by the application are described in detail. The description of the specific embodiments is only intended to facilitate an understanding of the method of the present application and its core ideas. It should be noted that it would be obvious to those skilled in the art that various improvements and modifications can be made to the present application without departing from the principles of the present application, and such improvements and modifications fall within the scope of the claims of the present application.
Claims (10)
1. The elevator fault prediction system is characterized by comprising an elevator data acquisition module, a data preprocessing module, a fault prediction module and an emergency processing module;
the elevator data acquisition module is used for acquiring elevator operation process data and elevator maintenance data;
the data preprocessing module is in communication connection with the elevator data acquisition module, a neural network topological structure is established through elevator operation process data and elevator maintenance data, and the elevator operation process data which causes the association of the elevator maintenance data with the elevator operation process data is obtained through analysis;
the fault prediction module is in communication connection with the data preprocessing module, the elevator operation process data is input into the neural network model, the weight of elevator maintenance data corresponding to the elevator operation process data is calculated, and then the elevator maintenance data is combined to extract a maintenance time sequence and input into the machine learning model, so that the residual service life of the elevator is obtained;
and the emergency processing module is used for making corresponding standby emergency processing means according to the residual service life of the internal equipment of the elevator, which is specifically corresponding to the maintenance data of each elevator.
2. An elevator failure prediction system according to claim 1, characterized in that the detailed step of establishing a neural network topology from elevator operation process data and elevator maintenance data comprises:
Elevator operation process data as input layer x of topological structure i (i=1, 2,3., n), hidden layer R i The activation function of the (i=1, 2,3., p) node is constituted by a gaussian function, the elevator repair data being the output layer y k (k＝1，2，3...，m)；
The input layer transmits an input signal to the hidden layer, the hidden layer determines the center of a radial basis function, vector data input by the input layer is directly mapped to a hidden layer space, the mapping relation from the hidden layer space to an output layer space is linear, and the output layer is a linear weighted sum of hidden layer outputs.
3. An elevator failure prediction system according to claim 2, characterized in that the radial basis function center uses a gaussian function formula:
wherein c i Is the center vector, sigma, of the ith basis function i For variance of radial basis function, p is the number of hidden functions, the hidden layer adopts nonlinear optimization method, and the output layer needs to realize the function R from the Gaussian function i (x)→y k I.e. the output is the output that forms the neural network after linear weighted combination of the output layers.
4. The elevator failure prediction system according to claim 3, wherein the training method of the entire neural network model adopts a nearest neighbor cluster learning algorithm, and the nearest neighbor cluster learning algorithm flow comprises:
S1001, selecting a Gaussian function as a radial basis function, setting a vector a (l) for storing the sum of various output vectors, and counting the number of samples by a design counter b (l), wherein l represents the number of categories;
S1002、(x 1 ，y 1 ) Initializing c as an initialization data pair 1 ＝x 1 ，a(1)＝y 1 B (1) =l, x 1 Establishing RBF network for cluster center, and initially implying layer as c 1 The output layer weight vector is initially w 1 ＝a(1)/b(1)；
S1003, adding a sample data pair (x 2 ，y 2 ) Thereby finding x 2 To c 1 Center distance |x 2 -c 1 I (I); when |x 2 -c 1 |≤γ，c 1 That is x 2 Nearest neighbor cluster of (1) =y at this time 1 +y 2 ，b(1)＝2，w 1 =a (1)/b (2); when |x 2 -c 1 | > γ, x 2 As a new cluster center, get c 2 ＝x 2 ，a(2)＝y 2 B (2) =1; will hide the unit c 2 Adding to the network established in step S1002, c 2 The weight vector of the output layer is w 2 ＝a(2)/b(2)；
S1004, when the mth data pair (x m ，y m ) m=3, 4,5, N, it is assumed that there are M cluster centers in the network at this time, M center points denoted as c 1 ，c 2 ，c 3 ，...c M At this time, M hidden units exist in the already established RBF network, and data pairs (x m ，y m ) Distance |x to cluster center m -c i I, i=1, 2, 3..m, find M distances minimum, assume |x m -x j Minimum then c j Is x m Is the nearest neighbor cluster of (a);
then judge when |x 2 -c 1 When I > gamma, X is m As a new cluster center, let c M+1 ＝x m ，M＝M+1，a(M)＝y m B (M) =1, each a (i), b (i) (i=1, 2,3,) M-1) the value is unchanged and the hidden unit c is again replaced M Adding the RBF network to the established RBF network;
when |x 2 -c 1 And (3) the I is less than or equal to gamma, and the following treatment is carried out: a (j) =a (j) +y m B (j) =b (j) +1, when i+.j, i=1, 2,3,..m., each a (i) is kept when in use, b (i) (i=1, 2,3,., M-1) is unchanged, and the weight vector of the output layer is set to w i ＝a(i)/b(i)i＝1，2，3，...，M；
S1005, according to the steps, the established RBF network output function is as follows:
5. an elevator failure prediction system according to claim 3, characterized in that the prediction algorithm of the whole neural network model adopts the following steps:
s101, initializing, and determining N vectors as initial clustering center vectors c i I.e. c 1 ，c 2 ，...，c N ；
S102, carrying out sample normalization processing,
s103, calculating the Euclidean distance, and solving the minimum distance:
d i (k)＝||X m -c i (k)||，m＝1，2，3...M
d min ＝min||X m -c i (k)||
wherein d is min Is the minimum Euclidean distance;
s104, updating center C i Calculating a sample mean value by adopting a mean value method;
s105, judging whether the distribution change of the clustering center is smaller than a preset value, if so, calculating d from the new calculation min The k value is added with 1, the step returns to the step S1003, if the k value is smaller than the preset valueSetting, and entering step S1006;
s106, after confirming the center of the whole neural network, determining the mean square error:
S107, calculating linear weight from the hidden layer to the output layer, and completing the calculation by using an error correction learning algorithm, wherein the actual output is set as y k The calculated output isThen->
And adjusting the weight of the output layer, calculating by adopting a minimum mean square error rule, wherein the square difference between the actual output and the expected output is minimum, namely, the target signal is set as:
r＝d i -W i T X
the weight vector is adjusted as:
ΔW i ＝η(d j -W i T X)X。
6. the elevator failure prediction system according to claim 4 or 5, wherein the detailed step of analyzing the associated elevator operation process data that causes the elevator maintenance data to be related to the elevator operation process data is performed by using the elevator operation process data that determines that the elevator operation process data and the preset initial value are abnormally changed as the cause data that causes the elevator maintenance data to fail, observing whether the cause that causes the elevator maintenance data to fail is all the cause of the elevator operation process data abnormality multiple times, if the occurrence frequency is higher than the preset frequency of the system, determining that the elevator operation process data is an input layer of a neural network topology structure, and if the occurrence frequency is lower than the preset frequency of the system, determining that the elevator operation process data is not the cause of the failure that causes the elevator maintenance data.
7. An elevator failure prediction system according to claim 6, characterized in that the elevator run process data further comprises an elevator number and an elevator installation position, the elevator service data further comprises an elevator number, an elevator installation position and an elevator service time, and the remaining service life of the elevator is analyzed by inputting the same type of elevator service data and the sequence of the type of elevator service time into the machine learning model.
8. The elevator failure prediction system according to claim 7, wherein the step of analyzing remaining service life of the elevator in the machine learning model is performed by inputting the same type of elevator maintenance data and the sequence of the type of elevator maintenance time in the machine learning model, wherein the step of:
s2001, calculating a time sequence difference, and calculating and analyzing whether an elevator time maintenance sequence is regular or not by adopting a first-order difference, a second-order difference and a K-step difference mode;
s2002, selecting different time sequence models for learning according to different rules; the time sequence model comprises a white noise sequence model, an autoregressive model, a moving average model, an autoregressive moving average model and an ARIMA model;
s2003, predicting a next time fault node, namely the residual service life of the elevator, by adopting different time sequence models according to different types of elevator maintenance data;
and S2004, in combination with the next time fault node generated by prediction, pre-judging elevator maintenance data generated by the elevator is sent to the emergency processing module.
9. An elevator failure prediction system according to claim 8, characterized in that,
the elevator data identification module is used for identifying whether the elevator numbers or the collected elevator running process data are matched with elevator maintenance data field items or not, and dividing different elevator numbers into the same type of elevators; when the same elevator number lacks historical elevator operation process data and elevator maintenance data to train the neural network model, the historical elevator operation process data and elevator maintenance data are built for the same type of elevator to train the neural network model;
The method for identifying the elevator numbers to be marked into the same type of elevator comprises the following steps:
acquiring an elevator number, and converting the elevator number into a character string;
according to the total length value of the character string, intercepting the character of 50% length value of the character string to be matched with all written elevator numbers;
if the same character string is matched, judging that the elevator operation process data and the elevator maintenance data are used for training a neural network model by using the historical data written with the elevator number when the historical data are lack of training the neural network model, taking the weight value of the hidden layer as the weight value during fault prediction, predicting the elevator maintenance data value, and judging whether the elevator maintenance data value of the elevator number exceeds the fault early-warning value of the elevator synchronously matched with the same character string;
the method for identifying the same type of elevator by matching the collected elevator operation process data with the elevator maintenance data field items comprises the following steps:
acquiring elevator operation process data and elevator maintenance data fields;
judging whether fields of each data item of the elevator operation process data and the elevator maintenance data fields are matched;
if the fields of each data item of the elevator operation process data and the elevator maintenance data fields are matched, the same type of elevator is judged.
10. An elevator failure prediction method, characterized by comprising the steps of:
s1, collecting relevant historical data of an elevator operation process and elevator maintenance;
s2, analyzing and obtaining historical data of the elevator maintenance data and the related correlation elevator operation process;
s3, building an RBF neural network topological structure or an RBF neural network model according to the association relation between elevator operation process data and elevator maintenance data;
s4, inputting the collected relevant historical data of the elevator operation process and elevator maintenance into an RBF neural network model, training the RBF neural network model by adopting a nearest neighbor clustering learning algorithm or a prediction algorithm, acquiring weights from hidden layer nodes to output nodes, setting the weights as elevator operation process data, and taking the elevator operation process data and the elevator maintenance data as weight vectors of all hidden layers when taking the input layer and the elevator maintenance data as the output layer;
s5, acquiring the latest elevator operation process data as an input layer in the RBF neural network model, and calculating to obtain predicted latest elevator maintenance data through implicit layer weight vector conversion obtained by calculation in the step S4;
s6, comparing the predicted latest elevator maintenance data with the elevator maintenance data of the same type in the same period history, if the elevator maintenance data of the same type in the same period history falls in the range or exceeds the range value, sending out corresponding elevator maintenance fault early warning, and if the elevator maintenance data of the same type in the same period history does not fall in the range or is lower than the minimum value of the range value, entering a step S7;
S7, inputting the predicted elevator maintenance data extraction time sequence into a machine learning model to obtain the residual service life of the elevator;
and S8, judging whether the residual service life of the elevator is close to the preset maintenance data of the whole elevator, if so, informing corresponding maintenance personnel to go to the door for maintenance through the elevator number and the elevator installation position, and if not, continuing to monitor the data of the operation process of the elevator.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410118871.7A CN117864892A (en) | 2024-01-29 | 2024-01-29 | Elevator fault prediction system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410118871.7A CN117864892A (en) | 2024-01-29 | 2024-01-29 | Elevator fault prediction system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117864892A true CN117864892A (en) | 2024-04-12 |
Family
ID=90580987
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410118871.7A Pending CN117864892A (en) | 2024-01-29 | 2024-01-29 | Elevator fault prediction system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117864892A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118396820A (en) * | 2024-06-26 | 2024-07-26 | 天津科技大学 | Laboratory safety management method and system based on Internet of things |
-
2024
- 2024-01-29 CN CN202410118871.7A patent/CN117864892A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118396820A (en) * | 2024-06-26 | 2024-07-26 | 天津科技大学 | Laboratory safety management method and system based on Internet of things |
| CN118396820B (en) * | 2024-06-26 | 2024-09-24 | 天津科技大学 | Laboratory safety management method and system based on Internet of things |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102092185B1 (en) | Platform for analyzing electric motor health and analysis method using the same | |
| RU2313815C2 (en) | Device and method for controlling technical plant, which contains a set of systems, in particular, electric power plant | |
| CN117390536B (en) | Operation and maintenance management method and system based on artificial intelligence | |
| CN107358366B (en) | A kind of distribution transformer fault risk monitoring method and system | |
| CN115833400A (en) | Monitoring and early warning method and system for power equipment of transformer substation | |
| CN109061341A (en) | Kalman filtering transformer fault prediction technique and system neural network based | |
| CN117408162B (en) | Power grid fault control method based on digital twin | |
| CN109583520B (en) | State evaluation method of cloud model and genetic algorithm optimization support vector machine | |
| JP6856443B2 (en) | Equipment abnormality diagnosis system | |
| CN117308275B (en) | Temperature difference-based pipeline connection abnormality detection method and system | |
| Borchersen et al. | Model‐based fault detection for generator cooling system in wind turbines using SCADA data | |
| EP3968479A1 (en) | Systems and methods for automatic power topology discovery | |
| KR101941854B1 (en) | System and method of estimating load with null data correction | |
| RU2707423C2 (en) | Method and system for diagnostics of industrial facility | |
| CN107291991A (en) | A kind of Wind turbines early defect method for early warning based on dynamic network mark | |
| CN117864892A (en) | Elevator fault prediction system and method | |
| JPH11212637A (en) | Preventive maintenance method and device | |
| CN118395358B (en) | Intelligent anti-misoperation topology analysis method for transformer substation | |
| CN119010109A (en) | Energy storage system self-adaptive power execution device and distribution method | |
| CN117639228A (en) | Distribution network operating status prediction method and system based on digital twins | |
| Lutska et al. | Forecasting the efficiency of the control system of the technological object on the basis of neural networks | |
| CN118411161B (en) | Electromechanical equipment operation and maintenance method and system based on data analysis | |
| CN117272844B (en) | Method and system for predicting service life of distribution board | |
| Farahat et al. | Similarity-based feature extraction from vibration data for prognostics | |
| CN118297224A (en) | Multi-equipment linkage fault prediction method, medium and system for refrigeration machine room |
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 |