CN113532614A - Method, processor and weighing system for predicting sensor data - Google Patents
Method, processor and weighing system for predicting sensor data Download PDFInfo
- Publication number
- CN113532614A CN113532614A CN202110642100.4A CN202110642100A CN113532614A CN 113532614 A CN113532614 A CN 113532614A CN 202110642100 A CN202110642100 A CN 202110642100A CN 113532614 A CN113532614 A CN 113532614A
- Authority
- CN
- China
- Prior art keywords
- sensor
- predicted
- value
- prediction model
- data
- 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
- 238000000034 method Methods 0.000 title claims abstract description 69
- 238000005303 weighing Methods 0.000 title claims abstract description 42
- 238000005259 measurement Methods 0.000 claims abstract description 117
- 230000002159 abnormal effect Effects 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims description 125
- 230000006870 function Effects 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 19
- 230000008569 process Effects 0.000 description 13
- 238000003860 storage Methods 0.000 description 12
- 238000004590 computer program Methods 0.000 description 11
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01G—WEIGHING
- G01G23/00—Auxiliary devices for weighing apparatus
- G01G23/01—Testing or calibrating of weighing apparatus
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01G—WEIGHING
- G01G19/00—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups
- G01G19/22—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for apportioning materials by weighing prior to mixing them
- G01G19/32—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for apportioning materials by weighing prior to mixing them using two or more weighing apparatus
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The embodiment of the invention provides a method, a processor, a weighing system and a weighing device for predicting sensor data. The method comprises the following steps: acquiring first actual measurement values of a plurality of sensors; providing a prediction model to obtain a first prediction value aiming at a sensor to be predicted; under the condition that the difference value between the first predicted value and the first actual measured value of the sensor to be predicted is larger than a preset threshold value, predicting the state of the sensor to be predicted to be an abnormal state; second actual measurement values of the sensors are obtained again, and second predicted values for the sensors to be predicted, which are obtained after the prediction model is learned, are obtained; under the condition that the difference value between the second predicted value and the second actual measured value of the sensor to be predicted is larger than a preset threshold value, predicting the state of the sensor to be predicted to be an abnormal state; until the frequency of predicting that the state of the sensor to be predicted is an abnormal state reaches a preset frequency threshold value; and determining that the sensor to be predicted breaks down, and outputting a second predicted value as the measurement data of the sensor to be predicted.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a method, a processor, a weighing system and a weighing device for predicting sensor data.
Background
The rapid development of concrete technology has raised requirements on concrete quality, and the concrete quality has become one of the most important indexes for measuring the performance of concrete mixing plants. In the actual use process, the sensor may be out of order for various reasons, resulting in inaccurate measured data.
In the prior art, when the output of a weighing system is inaccurate and abnormal due to sensor output, normal weighing cannot be carried out.
Disclosure of Invention
It is an object of embodiments of the present invention to provide a method, processor, weighing system and weighing apparatus for predicting sensor data.
To achieve the above object, a first aspect of the present invention provides a method for predicting sensor data, comprising:
acquiring first actual measurement values of a plurality of sensors;
inputting other first actual measurement values except the first actual measurement value of the sensor to be predicted into a prediction model aiming at any sensor to be predicted in the sensor to be predicted so as to obtain a first predicted value aiming at the sensor to be predicted;
under the condition that the difference value between the first predicted value and the first actual measured value of the sensor to be predicted is larger than a preset threshold value, predicting the state of the sensor to be predicted to be an abnormal state;
acquiring second actual measurement values of the plurality of sensors again, wherein the generation time of the second actual measurement values is later than that of the first actual measurement values;
inputting other second actual measurement values except the second actual measurement value of the sensor to be predicted into the prediction model so that the prediction model learns the other second actual measurement values;
acquiring a second predicted value for the sensor to be predicted, which is obtained after the prediction model is learned;
under the condition that the difference value between the second predicted value and the second actual measured value of the sensor to be predicted is larger than a preset threshold value, predicting the state of the sensor to be predicted to be an abnormal state;
under the condition that the frequency of predicting that the state of the sensor to be predicted is in an abnormal state does not reach the preset frequency threshold value, the step of obtaining second actual measurement values of the plurality of sensors again is executed until the frequency of predicting that the state of the sensor to be predicted is in the abnormal state reaches the preset frequency threshold value;
and determining that the sensor to be predicted breaks down, and outputting a second predicted value as the measurement data of the sensor to be predicted.
In an embodiment of the present invention, the prediction model obtains the prediction value for the sensor to be predicted by calculating according to the following formula (1):
wherein, f (X)n+1) For the predicted value of the sensor to be predicted, a1，a2，...，an、β1，β2，...，βn、δ1，δ2，...，δnAre all factor parameters, X1，X2，...，XnRespectively actual measurement values of other sensors except the sensor to be predicted; y is1，Y2，...，YnAll are scale state factors, Z is a constant term, and n is the number of other sensors.
In an embodiment of the present invention, the prediction model is a radial basis function neural network, and the prediction model obtains the prediction value for the sensor to be predicted by calculating according to the following formula (2):
wherein, X'jJ ═ 1, 2.. and N, N is the number of sensors, m is the number of hidden nodes of the prediction model, W is the weight matrix of the prediction model, and W ═ W (W) is the predicted value of the sensor to be predicted0，W1，...，Wm)T，W0H is a radial basis function matrix, H is (H)0，H1，...，Hm)T，H0B is the offset of the prediction model, i is the number of hidden nodes, i is (0,1,2, …, m), ωiIs the coefficient of h, and T is the transposed sign of the matrix W.
In an embodiment of the invention, the method further comprises: when the latest actual measurement value of the sensor is acquired again after the sensor to be predicted is determined to be out of order, the latest actual measurement value is input into the prediction model, and the prediction model does not learn the latest actual measurement value.
In an embodiment of the present invention, the method further includes a training step of the prediction model, the training step includes:
acquiring actual measurement data obtained by a plurality of sensors in a normal state within a preset historical time period as sample data;
dividing the sample data into a plurality of training data sets;
inputting each training data set into the prediction model in sequence according to the sequence of the sample data generation time, and training the prediction model;
after the prediction model outputs a sample prediction value aiming at each training data set, determining the goodness of fit of each training data set according to the training data in the training data set and the sample prediction value;
adding the training data with the later generation time to a training pool of the prediction model under the condition that the goodness of fit of the training data set with the earlier generation time is less than or equal to the goodness of fit of the training data set with the later generation time; otherwise, the training data set with the later generation time is not added into the training pool of the prediction model, so that the data in the training pool of the prediction model are ensured to be in the latest state;
and under the condition that the prediction accuracy of the prediction model reaches a preset accuracy threshold, determining that the training of the prediction model is finished.
In an embodiment of the invention, the method further comprises:
detecting whether the upper limit of the learning quantity of the prediction model reaches a preset quantity threshold value;
under the condition that a preset number threshold value is reached, according to the generation time sequence of sample data, eliminating a training data set with earlier generation time from a training pool of the prediction model;
and under the condition that the preset quantity threshold value is not reached, returning to the step of inputting each training data set into the prediction model in sequence according to the generation time sequence of the sample data and training the prediction model.
In one embodiment of the present invention, the preset number threshold is determined according to the number of sensors, the number of scale factors, and the number of constant terms.
In one embodiment of the invention, the goodness-of-fit for each training data set is determined according to equation (3) below:
wherein, f (X)n+1) For the predicted value of the sensor to be predicted, X1，X2，...，XnRespectively, the actual measured values of the other sensors than the sensor to be predicted, and n is the number of the other sensors.
In an embodiment of the invention, the method further comprises: in a case where all the sensors are determined to be in a normal state, first actual measurement values of the plurality of sensors are acquired.
In an embodiment of the invention, the method is applied to a weighing system comprising a plurality of sensors; the method further comprises the following steps: and determining the output data of the weighing system according to the second actual measurement values of the other sensors and the measurement data of the sensor to be predicted.
In an embodiment of the invention, the method further comprises: and sending a fault prompt aiming at the fault of the sensor to be predicted.
A second aspect of the invention provides a processor configured to perform the above-described method for predicting sensor data.
A third aspect of the invention provides a weighing system comprising:
the system comprises a plurality of sensors, a controller and a controller, wherein the sensors are used for measuring an article to obtain corresponding actual measurement values; and
the processor described above.
The invention provides a weighing device of a concrete mixing plant, which comprises the weighing system.
According to the technical scheme, the used prediction model can continuously update the learning data of the prediction model, so that the prediction precision of the prediction model is always maintained in a higher range, and the effectiveness of the emergency production function is guaranteed. Meanwhile, when a certain sensor is determined to have a fault through repeated verification for many times, the predicted value obtained by the prediction model can be used for replacing the measured value of the faulty sensor, so that the weight of the heavy object to be measured can be accurately determined in an emergency state.
Additional features and advantages of embodiments of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the embodiments of the invention without limiting the embodiments of the invention. In the drawings:
FIG. 1 schematically illustrates a flow diagram of a method for predicting sensor data in accordance with an embodiment of the invention;
FIG. 2 schematically illustrates a schematic diagram of the operation of a weighing system according to an embodiment of the invention;
FIG. 3 schematically shows a training flow diagram of a predictive model according to an embodiment of the invention;
FIG. 4 schematically illustrates a workflow diagram of a dynamic iterative algorithm according to an embodiment of the present invention;
FIG. 5 schematically illustrates a block diagram of a weighing system according to an embodiment of the present invention;
fig. 6 schematically shows an internal structure diagram of a computer apparatus according to an embodiment of the present invention.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating embodiments of the invention, are given by way of illustration and explanation only, not limitation.
FIG. 1 schematically shows a flow diagram of a method for predicting sensor data according to an embodiment of the invention. As shown in FIG. 1, in one embodiment of the present invention, a method for predicting sensor data is provided, comprising the steps of:
in step 101, first actual measurement values of a plurality of sensors are acquired.
And 102, inputting other first actual measurement values except the first actual measurement value of the sensor to be predicted into a prediction model aiming at any sensor to be predicted in the sensors to obtain a first predicted value aiming at the sensor to be predicted.
And 103, under the condition that the difference value between the first predicted value and the first actual measured value of the sensor to be predicted is larger than a preset threshold value, predicting that the state of the sensor to be predicted is an abnormal state.
In step 104, second actual measured values of the plurality of sensors are again acquired, wherein the second actual measured values are generated later in time than the first actual measured values.
And 105, inputting other second actual measurement values except the second actual measurement values of the sensor to be predicted into the prediction model so that the prediction model learns the other second actual measurement values.
And 106, acquiring a second predicted value for the sensor to be predicted, which is obtained after the prediction model is learned.
And 107, under the condition that the difference value between the second predicted value and the second actual measured value of the sensor to be predicted is larger than a preset threshold value, predicting the state of the sensor to be predicted to be an abnormal state.
And step 109, determining that the sensor to be predicted has a fault, and outputting a second predicted value as the measurement data of the sensor to be predicted.
In this embodiment, the method for predicting sensor data described above is applied to a weighing system, in particular, a concrete batching plant weighing system. The weighing system comprises a scale body, and the scale body comprises a plurality of sensors, so that the weight of materials in the scale body can be identified in real time. Specifically, each sensor measures the material to be weighed in the weighing scale body, and an actual measurement value corresponding to each sensor can be obtained. For convenience of description, the measurement value obtained by the first measurement may be referred to as a first actual measurement value. For all the sensors, any one sensor can be selected as the sensor to be predicted, that is, the measured value of the sensor can be predicted through the measured data of other sensors. For example, assuming that there are 10 sensors, Q1, Q2, Q3, and Q10, respectively, any one of the 10 sensors may be used as the sensor to be predicted. Assuming that Q1 is the sensor to be predicted, the measurement of Q1 can be predicted from the actual measurements of the other 9 sensors (Q2, Q3, a.., Q10).
Specifically, for any one sensor to be predicted in all the sensors, the other first actual measurement values except the first actual measurement value of the sensor to be predicted may be input into the prediction model, so as to obtain the first predicted value for the sensor to be predicted through the prediction model. That is, the actual measurement values of the other sensors may be input into the trained prediction model, and the prediction model may predict the measurement value of the sensor to be predicted according to the input actual measurement values of the other sensors, so as to obtain the predicted value of the sensor to be predicted. The predicted value predicted from the first actual measured value may be referred to as a first predicted value.
Further, the first predicted value may be compared with the first actual measurement value of the sensor to be predicted. Specifically, a difference between the first predicted value and the first actual measured value may be calculated. If the difference value is larger than the preset threshold value, the difference between the actual measurement value of the sensor to be predicted and the corresponding predicted value is larger, and the state of the sensor to be predicted can be predicted to be an abnormal state. The preset threshold is preset by a technician and can be adjusted according to actual conditions. For example, assuming that the preset threshold is set to 1.0, the measurement requirement of the sensor is increased in the subsequent process, i.e. the measurement error of the sensor is expected to be as small as possible, the preset threshold may be adjusted to 0.5. The actual measurement of the sensor continues to be acquired, which may be referred to as a second actual measurement. Wherein the generation time of the second actual measurement value is later than the generation time of the first actual measurement value. For example, the first actual measurement value is the measurement value measured by the sensor 10 minutes ago, and the second actual measurement value is the measurement value measured by the sensor 5 minutes ago. Likewise, other second actual measurement values except for the second actual measurement value of the sensor to be predicted can be input into the prediction model, and the second predicted value of the sensor to be predicted is obtained through the prediction model again. Meanwhile, the prediction model can learn other second actual measurement values to update the learning data of the prediction model, so that the learning precision of the prediction model is always maintained in a higher range.
Furthermore, the second predicted value can also be compared with a second actual measured value of the sensor to be predicted. Specifically, a difference between the second predicted value and the second actual measured value may be calculated. If the difference value is larger than the preset threshold value, the difference between the actual measurement value of the sensor to be predicted and the corresponding predicted value is larger, and the state of the sensor to be predicted can be predicted to be an abnormal state. And judging whether the frequency of predicting the abnormal state of the sensor to be predicted reaches a preset frequency threshold value, if so, determining that the sensor to be predicted breaks down, and taking the second predicted value as the measurement data of the sensor to be predicted. Namely, when the sensor to be predicted is determined to be out of order, the actual measurement value of the sensor to be predicted can be replaced by the predicted value, and therefore the emergency production function of the weighing system is achieved. If the frequency of predicting that the state of the sensor to be predicted is an abnormal state does not reach the preset frequency threshold value, whether the sensor to be predicted is abnormal or not needs to be continuously judged. The actual measurement value of the sensor can be obtained again, the steps are repeated again, and the predicted value is continuously compared with the actual measurement value of the sensor to be predicted so as to determine whether the sensor is abnormal or not.
It can be seen that, in the actual prediction process, the sensor is determined to have a fault immediately without predicting that the state of the sensor is an abnormal state. But rather continuously verifies that the most recent actual measurement of the sensor is being predicted for the sensor to be predicted. It is understood that in the first step loop, i.e. the first execution of step 108, if the number of times that the state of the sensor to be predicted is an abnormal state is predicted does not reach the preset number threshold, step 104 is executed. At this time, the actual measurement value of the acquired sensor is also generated later than the second actual measurement value. For the sake of descriptive simplicity only, it is not named third actual measurement value, fourth actual measurement value, etc. here. If the sensor to be predicted is determined to be abnormal for multiple times, namely the number of times of predicting that the state of the sensor to be predicted is an abnormal state reaches a preset number threshold, at the moment, the sensor to be predicted can be determined to have a fault. Then, the latest obtained predicted value can be determined as the measured value of the sensor to be predicted with a fault, and then the weight of the heavy object to be predicted can be determined together according to the actual measured values of other sensors and the measured value of the sensor to be predicted. That is, the output data of the weighing system may be determined according to the second actual measurement values of the other sensors and the measurement data of the sensor to be predicted.
According to the method for predicting the sensor data, the used prediction model can continuously update the learning data of the sensor data, so that the prediction accuracy of the sensor data is always maintained in a higher range, and the effectiveness of an emergency production function is guaranteed. Meanwhile, when a certain sensor is determined to have a fault through repeated verification for many times, the predicted value obtained by the prediction model can be used for replacing the measured value of the faulty sensor, so that the weight of the heavy object to be measured can be accurately determined in an emergency state.
In one embodiment, the method further comprises: and after determining that the sensor to be predicted fails, when the latest actual measurement value of the sensor is acquired again, and the latest actual measurement value is input into the prediction model, the prediction model does not learn the latest actual measurement value.
After determining that the sensor to be predicted fails, or after determining that any one sensor among all the sensors fails, acquiring an actual measurement value newly measured by the sensor again, inputting the actual measurement value into the prediction model, and when predicting any one sensor again through the prediction model, the prediction model does not learn the data. Therefore, the predicted value obtained by the prediction model is close to the correct measured value of the sensor under normal conditions in order to avoid the prediction model from learning wrong data.
In one embodiment, the method further comprises: in a case where all the sensors are determined to be in a normal state, first actual measurement values of the plurality of sensors are acquired.
At the beginning of the method, it should be ensured that all sensors are normal. That is, it can be ensured that the first actual measurement values obtained from the initial measurements of all the sensors acquired should be accurate. In the subsequent process, the actual measurement values of the sensors are obtained all the time, and each sensor is also predicted to detect whether the sensor is abnormal or not. That is to say, in the method, in order to find the abnormal sensor in time during the operation of the weighing system, and use the predicted value to replace the measured value of the sensor, so that the weighing system can still be used normally in an emergency state, and still can provide more accurate data to the user. Instead of finding out which sensor is abnormal, if any. If it has been initially determined that there is a fault with the sensor, the fault should be repaired before the weighing system is restarted.
Also, in one embodiment, the method further comprises: and sending a fault prompt aiming at the fault of the sensor to be predicted.
That is, if it is determined that a certain sensor has a fault by the above method, a corresponding fault reminder may be sent for the fault condition of the sensor, so that the fault may be handled in time. While still using the predicted values instead of the measured values of the sensors while waiting for the fault to be repaired, so that the weighing system can be used normally in an emergency situation.
In one embodiment, as shown in FIG. 2, a schematic diagram of the operation of a weighing system is provided. A plurality of sensors are arranged in the weighing system, and after the sensors acquire actual measurement values of objects in the scale body, the data acquisition unit can acquire the measurement data of the sensors and then transmit the measurement data to mechanical parts such as an industrial personal computer. The weighing system includes sensors that each measure an item being weighed to obtain its corresponding actual measurement. Meanwhile, for any one of the sensors to be predicted, the predicted value of the sensor can be obtained through the prediction model, so that when the sensor fails, the predicted value can be used for replacing the measured value of the sensor with the failure, and normal use under an emergency state can be met. The prediction model adopts a dynamic iteration algorithm, so that the training data of the prediction model can be updated and iterated in time, namely the original prediction model can be compensated, and the prediction accuracy of the prediction model can be maintained at a higher level for a long time.
Specifically, in one embodiment, as shown in fig. 3, the method further includes a training step of the prediction model, where the training step includes:
And 303, sequentially inputting each training data set into the prediction model according to the sequence of the sample data generation time, and training the prediction model.
And step 304, after the prediction model outputs the sample prediction value aiming at each training data set, determining the goodness of fit of each training data set according to the training data in the training data set and the sample prediction value.
In step 305, if the goodness-of-fit of the training data set generated earlier in time is less than or equal to the goodness-of-fit of the training data set generated later in time, the training data generated later in time is added to the training pool of the predictive model.
In step 306, if the goodness of fit of the training data set with the earlier generation time is greater than the goodness of fit of the training data set with the later generation time, the training data set with the later generation time is not added to the training pool of the prediction model, so as to ensure that the data in the training pool of the prediction model keeps the latest state.
And 307, under the condition that the prediction accuracy of the prediction model reaches a preset accuracy threshold, determining that the training of the prediction model is finished.
Before the prediction model is put into practical use, the prediction model can be trained, and after the training is finished, the trained prediction model is put into use. Specifically, actual measurement data obtained by a plurality of sensors in a normal state in a preset historical time period may be acquired from a database, and the actual measurement data may be used as sample data. In the training data, accurate data measured by the sensor in a normal state are adopted, so that the prediction accuracy of the prediction model is reduced because the prediction model learns inaccurate data.
Then, the sample data may be divided intoA plurality of training data sets. Each training data set may be considered as a matrix, which includes a plurality of data. For example, the training dataset 1 is a matrix a, where a ═ X0，1，X0，2，...，X0，n]Wherein, X represents the actual measurement value of the sensor, the first corner mark of X represents the value of the sensor for the second time, and the second corner mark represents the number of the sensor. For example, if 10 sensors in the scale body were all measuring for the same weight measuring item P1, 10 actual measurements for item P1 would be obtained. If the No. 10 sensor is taken as a sensor to be predicted, actual measurement data of the No. 1-10 sensors are all taken as learning data, and a predicted value for the No. 10 sensor is obtained. That is, the actual measurement value of each sensor is taken a plurality of times to predict the data of other sensors. Thus, X0，1The actual measurement for sensor number 1 is shown and this is the first value for sensor number 1. X is2，5The actual measurement of sensor No. 5 is shown, which is the 3 rd value for sensor No. 1.
Each training data set may be input into the predictive model in turn in order of sample data generation time to train the predictive model. The prediction model may output a corresponding sample prediction value for each training data set. Specifically, the predicted value for the sensor to be predicted can be calculated according to the following formula (1):
wherein, f (X)n+1) For the predicted value of the sensor to be predicted, a1，a2，...，an、β1，β2，...，βn、δ1，δ2，...，δnAre all factor parameters, X1，X2，...，XnRespectively actual measurement values of other sensors except the sensor to be predicted; y is1，Y2，...，YnAll are scale state factors, Z is a constant term, and n is the number of other sensors.
In the training process of the prediction model, the fact is to determine a1，a2，...，an、β1，β2，...，βn、δ1，δ2，...，δnThe process of (1). That is, the values of α, β, and δ are all set as initial default values initially during the training process, i.e., the training process is the process of determining these parameters. When the solved value is more accurate, f (X) is finally calculatedn+1) The closer the value of (c) is to the actual measurement value of the sensor. The number of actual measured values is n, X is the actual measured value of each sensor and can be called a factor, Y is the scale state factor which is already determined, and alpha, beta and delta belong to the parameters of X and Y, namely called factor parameters. The unknown numbers of α, β, and δ are all n, and 3n +1 unknown numbers are obtained by adding the constant term Z. Therefore, at least 3n +1 equations are required to solve α, β, δ. Therefore, it should be ensured that at least the number of training data in the training pool of the predictive model is 3n + 1. Under the condition that the data quantity in the training pool is larger than 3n +1, the more training data, the higher the prediction accuracy of the trained prediction model is.
In another embodiment, the predictive model may be a radial basis function neural network. The prediction model can calculate the predicted value of the sensor to be predicted by the following formula (2):
wherein, X'jJ ═ 1, 2.. and N, N is the number of sensors, m is the number of hidden nodes of the prediction model, W is the weight matrix of the prediction model, and W ═ W (W) is the predicted value of the sensor to be predicted0，W1，...，Wm)T，W0H is a radial basis function matrix, H is (H)0，H1，...，Hm)T，H0B is the bias value of the prediction model, i is hiddenThe serial number of the layer node, i ═ 0,1,2, …, m, ωiIs the coefficient of h, and T is the transposed sign of the matrix W.
Further, after the prediction model outputs the sample prediction value for each training data set, the goodness of fit of each training data set can be determined according to the training data in the training data set and the sample prediction value. Specifically, the goodness-of-fit for each training data set may be determined according to equation (3) below:
wherein, f (X)n+1) For the predicted value of the sensor to be predicted, X1，X2，...，XnRespectively, the actual measured values of the other sensors than the sensor to be predicted, and n is the number of the other sensors.
Goodness of fit R refers to the degree of fit of the regression line to the observed value. The statistic for measuring goodness of fit is a coefficient of solution (or determining coefficient) R2. Wherein R is2The maximum value is 1. R2The closer the value of (1) is, the better the fitting degree of the regression straight line to the observed value is; otherwise, R2The smaller the value of (a) is, the worse the fitting degree of the regression line to the observed value is. Thus, the goodness-of-fit for each set of training data sets may be determined continuously.
And comparing the magnitude of the goodness of fit of the training data set with the earlier generation time and the training data set with the later generation time. If the goodness of fit of the training data set with the earlier generation time is less than or equal to the goodness of fit of the training data set with the later generation time, the goodness of fit of the training data set with the later generation time is higher, the data is more accurate, and the prediction model is more suitable for training. In this case, later in time training data may be generated and added to the training pool of the predictive model. Conversely, if the goodness of fit of the training data set generated earlier in time is greater than the goodness of fit of the training data set generated later in time, it indicates that the goodness of fit of the training data set generated earlier in time is higher, while the data of the training data set generated later in time is less accurate. In this case, it is not necessary to add training data that is generated later in time to the training pool of the predictive model. In this way, it is ensured that the data in the training pool of the predictive model remains up to date.
Further, in one embodiment, the method further comprises: detecting whether the upper limit of the learning quantity of the prediction model reaches a preset quantity threshold value; under the condition that a preset number threshold value is reached, according to the generation time sequence of sample data, eliminating a training data set with earlier generation time from a training pool of the prediction model; and under the condition that the preset quantity threshold value is not reached, returning to the step of inputting each training data set into the prediction model in sequence according to the generation time sequence of the sample data and training the prediction model.
In the training process of the prediction model, the training data set is continuously input into the prediction model, so that the prediction model can continuously learn the input data. But it is necessary to ensure that the training data in the training pool of the predictive model is up-to-date. Thus, a preset number threshold may be set, wherein the preset number threshold is determined based on the number of sensors, the number of scale factors, and the number of constant terms. That is, as mentioned in the above embodiment, the preset number threshold may be determined according to the number of unknown terms of α, β, δ and the number of constant terms. Whether the upper limit of the learning quantity of the prediction model reaches a preset quantity threshold value or not can be detected, and if the upper limit of the learning quantity of the prediction model is determined to reach the preset quantity threshold value, a training data set with earlier generation time can be removed from a training pool of the prediction model according to the generation time sequence of sample data; and under the condition that the preset number threshold is not reached, continuously returning to execute the steps of inputting each training data set into the prediction model in sequence according to the generation time sequence of the sample data and training the prediction model. The method for dynamically updating the learning data has the function of keeping the data learned by the prediction model to be the latest data all the time, namely the data closest to the current state of the weighing system, so that the predicted value obtained by the prediction model can be ensured to be closer to the actual measured value of the sensor.
As shown in fig. 4, a workflow diagram of a dynamic iterative algorithm employed by the predictive model is provided. Firstly, the sensors are all in a normal state, and actual measurement values measured by the sensors in the normal state are obtained. Suppose the training dataset is matrix a ═ X0，1，X0，2，...，X0，n]. Setting prediction model to learn m each timek(k. epsilon. N) group data, wherein M > Mk> 3n + 1. K denotes the number of sets of prediction mode learning data, and M denotes the amount of data in the training pool of the prediction model. First, the number of sets of calculations needs to be regressed. According to a mathematical habit calculation method, the initial value is defaulted to 0. That is, when the initialization k is 0, the prediction model should learn mk group data. Then, aiming at each group of data, the prediction model obtains a corresponding predicted value, so that the goodness of fit R of each group of data can be calculatedk。RkThe calculation formula of (2) is the above formula (3), and is not described herein again. Then, a set of data after the interval time t may be taken for input into the predictive model. Wherein the time interval t may be 10-15 mins. It can be seen that the generation time of the set of training data at this time is later than the generation time of the set of training data that has been learned before. For each learning set of data, the value of k is increased by 1, i.e., k equals k + 1. Similarly, goodness-of-fit R 'corresponding to the group of data is calculated again'k. And compare RkAnd R'kThe value of (c). If R iskIs greater than R'kThen, it means that the training data set data generated later is not accurate, and the step of inputting the execution data to the prediction model for learning and prediction can be returned. If R iskIs less than or equal to R'kThe data of this learning can be added to the training pool of the prediction model. It can be determined that the number of training pools, M, of the predictive model should be at least greater than 3n + 1. At this time, whether the data in the training pool of the prediction model exceeds M or not can be detected, and if yes, the training data with the earliest generation time in the training pool can be removed according to a first-in first-out principle. If the data in the training pool of the predictive model does not exceed M, the data input to the predictive model can be returned to for learning and predictionThe step (2). The goodness-of-fit of this set of training data continues to be compared to the goodness-of-fit of the next set of training data. In this way, the predictive model is continuously trained.
After a period of training, the prediction accuracy of the prediction model may be determined. Specifically, the predicted value of the prediction model can be compared with the predicted actual measurement value of the sensor, whether the difference value between the predicted value and the actual measurement value is within a preset difference value range is judged, and if yes, the predicted value of the prediction model is accurate; otherwise, the prediction value of the prediction model is inaccurate. In this way, the prediction accuracy of the prediction model is verified through multiple groups of data until the prediction accuracy of the prediction model reaches a preset accuracy threshold, and the prediction model can be determined to be trained completely. For example, 100 sets of data may be used to verify the prediction accuracy of the predictive model. Assuming that the error between the predicted value of 88 groups of the 100 groups of data and the actual measured value of the sensor is within the preset difference range (0, 0.5), the prediction model indicates that the prediction model predicts the 88 groups of data accurately. The prediction accuracy is 88%.
Therefore, the original system is compensated through dynamic iteration typical data, so that the prediction accuracy of the prediction model is maintained at a high level for a long time, the emergency production function is effectively realized, the running stability of the weighing system is ensured, the prediction accuracy of the prediction model can be maintained at a high level for a long time, and the effectiveness of emergency production is further ensured.
An embodiment of the present invention provides a processor, which is configured to execute a program, where the program executes the above method for predicting sensor data when running.
In one embodiment, as shown in fig. 5, there is also provided a weighing system comprising:
the system comprises a plurality of sensors 501-1, 501-2, 501-n, a plurality of sensors and a plurality of sensors, wherein the sensors are used for measuring the articles to obtain corresponding actual measurement values; and
the processor 502 described above configured to perform any of the methods described above for predicting sensor data.
In one embodiment, a weighing device of a concrete mixing station is further provided, and the weighing device comprises the weighing system, a scale body and the like.
The memory may include volatile memory in a computer readable medium, Random Access Memory (RAM) and/or nonvolatile memory such as Read Only Memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.
Embodiments of the present invention provide a storage medium having stored thereon a program that, when executed by a processor, implements the above-described method for predicting sensor data.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 6. The computer device includes a processor a01, a network interface a02, a memory (not shown), and a database (not shown) connected by a system bus. Wherein processor a01 of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises an internal memory a03 and a non-volatile storage medium a 04. The non-volatile storage medium a04 stores an operating system B01, a computer program B02, and a database (not shown in the figure). The internal memory a03 provides an environment for the operation of the operating system B01 and the computer program B02 in the nonvolatile storage medium a 04. The database of the computer device is used for storing actual measurement data of the sensors and the like. The network interface a02 of the computer device is used for communication with an external terminal through a network connection. The computer program B02 is executed by the processor a01 to implement a method for predicting sensor data.
Those skilled in the art will appreciate that the architecture shown in fig. 6 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
Embodiments of the present invention provide an apparatus, which includes a processor, a memory, and a program stored in the memory and executable on the processor, and when the program is executed by the processor, the steps of the method for predicting sensor data described above are implemented.
The present application further provides a computer program product adapted to perform a program of initializing method steps for predicting sensor data when executed on a data processing device.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. The use of the phrase "including an" as used herein does not exclude the presence of other, identical elements, components, methods, articles, or apparatus that may include the same, unless expressly stated otherwise.
The above are merely examples of the present application and are not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims (14)
1. A method for predicting sensor data, the method comprising:
acquiring first actual measurement values of a plurality of sensors;
inputting other first actual measurement values except the first actual measurement value of the sensor to be predicted into a prediction model aiming at any sensor to be predicted in the sensors to obtain a first predicted value aiming at the sensor to be predicted;
under the condition that the difference value between the first predicted value and the first actual measured value of the sensor to be predicted is larger than a preset threshold value, predicting that the state of the sensor to be predicted is an abnormal state;
acquiring second actual measurement values of the plurality of sensors again, wherein the second actual measurement values are generated later than the first actual measurement values;
inputting other second actual measurement values except for the second actual measurement value of the sensor to be predicted into the prediction model so that the prediction model learns the other second actual measurement values;
acquiring a second predicted value for the sensor to be predicted, which is obtained after the prediction model is learned;
under the condition that the difference value between the second predicted value and the second actual measured value of the sensor to be predicted is larger than the preset threshold value, predicting that the state of the sensor to be predicted is an abnormal state;
under the condition that the frequency of predicting that the state of the sensor to be predicted is in an abnormal state does not reach a preset frequency threshold value, the step of obtaining second actual measurement values of the plurality of sensors again is executed until the frequency of predicting that the state of the sensor to be predicted is in the abnormal state reaches the preset frequency threshold value;
and determining that the sensor to be predicted has a fault, and outputting the second predicted value as the measurement data of the sensor to be predicted.
2. The method for predicting sensor data as set forth in claim 1, wherein the predictive model calculates a predicted value for the sensor to be predicted by the following equation (1):
wherein, f (X)n+1) For the predicted value of the sensor to be predicted, a1，a2，…，an、β1，β2，…，βn、δ1，δ2，…，δnAre all factor parameters, X1，X2，…，XnRespectively actual measurement values of other sensors except the sensor to be predicted; y is1，Y2，…，YnAll are scale state factors, Z is a constant term, and n is the number of other sensors.
3. The method for predicting sensor data as set forth in claim 1, wherein the prediction model is a radial basis neural network, and the prediction model obtains a prediction value for the sensor to be predicted by calculating the following formula (2):
wherein, X'jJ is (1,2, …, N), N is the number of the sensors, m is the number of hidden nodes of the prediction model, W is the weight matrix of the prediction model, and W is (W)0,W1,…,Wm)T，W0H is a radial basis function matrix, H is (H)0,H1,…,Hm)T，H0B is a bias value of the prediction model, i is a sequence number of the hidden node, i is (0,1,2, …, m), ω isiIs the coefficient of h, and T is the transposed sign of the matrix W.
4. The method for predicting sensor data as recited in claim 1, further comprising:
and after determining that the sensor to be predicted fails, when the latest actual measurement value of the sensor is acquired again, and the latest actual measurement value is input into the prediction model, the prediction model does not learn the latest actual measurement value.
5. The method for predicting sensor data as recited in claim 1, further comprising a training step of the predictive model, the training step comprising:
acquiring actual measurement data obtained by a plurality of sensors in a normal state within a preset historical time period as sample data;
dividing the sample data into a plurality of training data sets;
inputting each training data set into the prediction model in sequence according to the sequence of the sample data generation time, and training the prediction model;
after the prediction model outputs a sample prediction value aiming at each training data set, determining the goodness of fit of each training data set according to the training data in the training data set and the sample prediction value;
adding the training data with the later generation time to a training pool of the prediction model under the condition that the goodness of fit of the training data set with the earlier generation time is less than or equal to the goodness of fit of the training data set with the later generation time; otherwise, the training data set with the later generation time is not added into the training pool of the prediction model so as to ensure that the data in the training pool of the prediction model keeps the latest state;
and under the condition that the prediction accuracy of the prediction model reaches a preset accuracy threshold, determining that the prediction model is trained completely.
6. The method for predicting sensor data as recited in claim 5, further comprising:
detecting whether the upper limit of the learning quantity of the prediction model reaches a preset quantity threshold value;
under the condition that the preset quantity threshold value is reached, according to the generation time sequence of the sample data, eliminating a training data set with an early generation time from a training pool of the prediction model;
and under the condition that the preset quantity threshold value is not reached, returning to the generation time sequence according to the sample data, sequentially inputting each training data set into the prediction model, and training the prediction model.
7. The method for predicting sensor data as set forth in claim 6, wherein said preset number threshold is determined as a function of a number of said sensors, a number of scale factors, and a number of constant terms.
8. The method for predicting sensor data as set forth in claim 5, wherein the goodness-of-fit for each training data set is determined according to the following equation (3):
wherein, f (X)n+1) For the predicted value of the sensor to be predicted, X1，X2，…，XnRespectively, the actual measured values of other sensors except the sensor to be predicted, and n is the number of other sensors.
9. The method for predicting sensor data as recited in claim 1, further comprising:
in a case where all the sensors are determined to be in a normal state, first actual measurement values of the plurality of sensors are acquired.
10. The method for predicting sensor data as set forth in claim 1, wherein the method is applied to a weighing system comprising a plurality of sensors; the method further comprises the following steps:
and determining the output data of the weighing system according to the second actual measurement values of other sensors and the measurement data of the sensor to be predicted.
11. The method for predicting sensor data as recited in claim 1, further comprising:
and sending a fault prompt aiming at the fault of the sensor to be predicted.
12. A processor configured to perform the method for predicting sensor data according to any one of claims 1 to 11.
13. A weighing system, characterized in that the weighing system comprises:
the system comprises a plurality of sensors, a controller and a controller, wherein the sensors are used for measuring an article to obtain corresponding actual measurement values; and
the processor of claim 12.
14. A concrete batching plant weighing apparatus, characterized in that it comprises a weighing system according to claim 13.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110642100.4A CN113532614A (en) | 2021-06-09 | 2021-06-09 | Method, processor and weighing system for predicting sensor data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110642100.4A CN113532614A (en) | 2021-06-09 | 2021-06-09 | Method, processor and weighing system for predicting sensor data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113532614A true CN113532614A (en) | 2021-10-22 |
Family
ID=78095722
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110642100.4A Pending CN113532614A (en) | 2021-06-09 | 2021-06-09 | Method, processor and weighing system for predicting sensor data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113532614A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114002574A (en) * | 2021-10-29 | 2022-02-01 | 上海华力微电子有限公司 | Method for testing semiconductor structure |
| CN118228173A (en) * | 2024-03-20 | 2024-06-21 | 北京暖流科技有限公司 | Terminal acquisition system and data acquisition method for judging abnormal sensor data acquisition |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109582793A (en) * | 2018-11-23 | 2019-04-05 | 深圳前海微众银行股份有限公司 | Model training method, customer service system and data labeling system, readable storage medium storing program for executing |
-
2021
- 2021-06-09 CN CN202110642100.4A patent/CN113532614A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109582793A (en) * | 2018-11-23 | 2019-04-05 | 深圳前海微众银行股份有限公司 | Model training method, customer service system and data labeling system, readable storage medium storing program for executing |
Non-Patent Citations (2)
| Title |
|---|
| 林海军 等: ""基于信息融合的汽车衡称重传感器故障诊断"", 《控制理论与应用》 * |
| 蒋敦 等: "混凝土搅拌站计量传感器故障实时诊断研究", 《建设机械技术与管理》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114002574A (en) * | 2021-10-29 | 2022-02-01 | 上海华力微电子有限公司 | Method for testing semiconductor structure |
| CN118228173A (en) * | 2024-03-20 | 2024-06-21 | 北京暖流科技有限公司 | Terminal acquisition system and data acquisition method for judging abnormal sensor data acquisition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20190019096A1 (en) | Estimator, estimation method, program and storage medium where program stored for model parameter estimation and model parameter estimation system | |
| JP6404832B2 (en) | Method of determining control technical observer for SoC | |
| CN113532614A (en) | Method, processor and weighing system for predicting sensor data | |
| CN107480028B (en) | Method and device for acquiring usable residual time of disk | |
| Sushchenko et al. | Processing of redundant information in airborne electronic systems by means of neural networks | |
| CN113205187A (en) | Learning device, learning method, computer-readable medium, determination device, determination method, and computer-readable medium | |
| JP2020047847A (en) | Data processing method, data processor and data processing program | |
| CN109814101B (en) | Method and device for predicting position of aircraft | |
| CN104991581A (en) | Method and device for controlling pressure of process chamber | |
| CN113254873A (en) | Satellite clock error forecasting method and system thereof | |
| CN109213057B (en) | Intelligent diagnosis device and method | |
| CN112230146A (en) | Method, system and equipment for predicting battery charging remaining time | |
| CN110245147B (en) | Block chain data processing method and device, readable storage medium and computer equipment | |
| CN117952377A (en) | Digital twin discrete manufacturing workshop synchronous evolution method, device and medium | |
| CN111832693A (en) | Neural network layer operation and model training method, device and equipment | |
| US20230252357A1 (en) | Learning with moment estimation using different time constants | |
| CN110865530A (en) | Atomic time calculation method | |
| CN117092975A (en) | Production control method and system based on digital twin | |
| US20200363208A1 (en) | Factory-specific inertial measurement unit error model | |
| US20190235128A1 (en) | Determination of virtual process parameters | |
| US20230266158A1 (en) | Method and system for eccentric load error correction | |
| CN114117656A (en) | Method, program and apparatus for predicting wall thickness reduction amount | |
| CN115952853B (en) | Method and device for constructing ore pulp density detection model and ore pulp density detection system | |
| CN117606590B (en) | Calibration method and device of steam instrument, electronic equipment and storage medium | |
| Cherepanska et al. | Artificial Neural Network as a Basic Element of the Automated Goniometric System |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211022 |