Method for evaluating health degree of weighing apparatus

Abstract

The invention relates to a method for evaluating the health degree of a measuring apparatus, and the method comprises the following steps: 1, respectively obtaining the weighing of a current measuring apparatus and the real-time data and result data corresponding to a sensor, and building the correlation relation between the existing data, so as to determine the quantitative relation of the mutual dependence between two or more variables, the method comprises the steps of (1) establishing a time sequence ARIMA (p, d, q) model, (2) calculating a self-adaptive threshold value of a key parameter based on historical data through a mathematical statistical method, and (3) predicting the time when a state parameter reaches an early warning threshold value through the established time sequence ARIMA (p, d, q) model, and carrying out trend early warning through the time. According to the method and the system for evaluating the health state of the measuring apparatus, data real-time automatic analysis and alarm are realized, point inspection is effectively reduced, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to a monitoring method, in particular to a method for evaluating the health of a measuring and weighing instrument, and belongs to the technical field of measuring and weighing instrument evaluation. Background Art

[0002] The weighing scale is a weighing device for continuous weighing and measuring of industrial bulk materials. Its health status will directly affect the accuracy of heavy measurement. After long-term use or its own sensitivity, the weighing scale is greatly affected by environmental factors or operation, which will cause inaccurate measurement. And usually, staff will be arranged to repair the weighing scale only after it fails. However, since it is impossible to know the type of fault or the specific location of the fault in advance, the fault repair time is long and the efficiency is low. In addition, the equipment can no longer be used after the fault occurs, and certain economic losses may also occur during the equipment maintenance period. Therefore, it is necessary to obtain the health status of the weighing scale in a timely manner.

[0003] CN109855710A-A vehicle scale weighing status detection system and detection method. Although this patent can collect and analyze analog signals from weighing equipment in real time, it is relatively complex and consists of multiple components and modules. At the same time, the signal splitter divides the signal into two paths, which may also cause data consistency problems. CN115169650A-A big data analysis equipment health prediction method. Although this patent combines historical data and applies a neural network model to evaluate the health status, it does not evaluate the equipment health status for specific application scenarios. At the same time, the neural network model is a black box model with poor interpretability.

[0004] Therefore, the present invention provides a method for monitoring the health status of a weighing scale, which evaluates the health status of a weighing scale by combining historical data and mechanism relationships and constructing an explainable model. Summary of the invention

[0005] The present invention is aimed at the problems existing in the prior art and provides a method for evaluating the health of a measuring scale. The method for evaluating the health status of a measuring scale provided by the technical solution mainly includes the following steps: respectively obtaining the real-time data, status data and result data corresponding to the current weighing of the measuring scale and the sensor, establishing the correlation relationship between the existing data, and then calculating the prediction function and the true value of the measuring scale. Based on the above steps, the present invention can obtain the core monitoring variables of the difference between the prediction function and the true value of the measuring scale, calculate the adaptive thresholds of the key parameters in the measuring scale (key parameters constructed by the total weighing weight, angle difference, linear value, internal code value, temperature, air tightness, etc.) through mathematical statistics methods and based on historical data, and then evaluate the health status of the measuring scale by comparison, that is, realize the online evaluation of the health status of the measuring scale.

[0006] In order to achieve the above object, the technical solution of the present invention is as follows: a method for evaluating the health of a measuring instrument, the method comprising the following steps:

[0007] Step 1: Obtain sampling data of the measuring instrument sensor, establish the correlation relationship between the sampling data of the sensor, and determine the quantitative relationship of mutual dependence between two or more sampling data quantities;

[0008] Step 2: Calculate the adaptive thresholds of key parameters based on historical data using mathematical statistics methods;

[0009] Step 3: Use the established time series ARIMA (p, d, q) model to predict the time when the weighing instrument status reaches the warning threshold, and use this time to issue a trend warning.

[0010] As an improvement of the present invention, step 1 is specifically as follows: using regression analysis, according to the weighing value of the measuring scale and the state data corresponding to each sensor, the correlation relationship between the parameters is calculated:

[0011]

[0012]

[0013] Among them, Q k is the measurement value of the kth group of sensors of the measuring instrument, is the measurement prediction value of the kth group of sensors, (C k ,T k ,G k ,A k , L k ) is the internal code value C of the corresponding group of sensors k , Temperature T k , air tightness G k , Angle difference A k and linear value L k , k represents the sensor serial number of the measuring instrument (k = 1, 2, ..., n), n is the number of sensors of the measuring instrument, that is, there are n groups of sensors in total. W k1 , W k2 ,...,W kl is the coefficient of each corresponding characteristic parameter of the kth group of sensors, l is the number of characteristics involved in the model calculation (in the present invention, the number of l is 5, representing the characteristic internal code value, temperature, air tightness, angle difference value, and linear value respectively). i is the historical data sample number, and m is the total number of historical data samples.

[0014] Then, based on the historical sample data, the least squares method is used to minimize the objective function Solve to calculate the correlation coefficients in the above formula. After calculating the correlation coefficients, the measurement value Q of the kth group of sensors can be calculated according to the following formula: k Make predictions:

[0015]

[0016] According to the predicted value and the actual value Q k , and calculate the core monitoring variables of the measuring instrument according to the following formula:

[0017]

[0018] Among them, ε k It represents the difference between the measurement prediction value and the actual measurement value of the k-th group of sensors, and serves as the core monitoring variable of the k-th group of sensors.

[0019] The health status of the measuring instrument is evaluated based on the core monitoring variables.

[0020] As an improvement of the present invention, step 2 is specifically as follows: the sensor state corresponding to the kth group of sensors can be obtained by The evaluation is mainly carried out through the core monitoring variable ε k The size of the core monitoring variable ε k If it is within a reasonable range, it means that the status of the kth group of sensors is normal. Otherwise, it means that there may be a problem and feedback is needed for investigation.

[0021] Reasonable range setting for all core monitoring variables that can be calculated from historical data According to the distribution, statistics are collected and warning thresholds are set to warn the status of the corresponding sensors. The setting of warning thresholds is completed through the 3-sigma criterion:

[0022]

[0023]

[0024] Among them, ε ki is the difference between the predicted value and the actual value of the kth sensor in the i-th historical data sample, μ k is the mean of the core monitoring variables of all k-th group of sensors in the historical data, σ k is the mean square error of the core monitoring variables of the kth group of sensors in the historical data, i is the sample number of the historical data, and m is the total number of samples of the historical data.

[0025] According to the 3-sigma criterion, the warning threshold of the kth group of sensors can be set as:

[0026] εk alarm =μ k +3σ k

[0027] When the core monitoring variable ε is calculated from the real-time collected sample data k Beyond ε k alarm When , the status of the kth group of sensors sends out a warning signal.

[0028] As an improvement of the present invention, in step 3, the time series ARIMA (p, d, q) model is established to predict the time when the state parameter reaches the warning threshold, where the historical data step length is l h and prediction step length l p It can be specified according to the actual situation. By predicting the results of historical data in the future, the possible development level of key indicators can be judged, and the prediction results based on the time series can be evaluated to achieve the purpose of trend warning.

[0029] Each accumulation step is l h The data is predicted by the trained time series model with a step length of l p The future core monitoring variable data is compared with the size of the predicted data and the set threshold to determine whether the alarm condition is met and the time to reach the warning threshold, so as to provide trend warning.

[0030] Compared with the prior art, the present invention has the following advantages:

[0031] 1) The method and system for evaluating the health status of measuring and weighing instruments provided by the present invention realize real-time automatic data analysis and alarm, effectively reduce point inspections, and improve work efficiency;

[0032] 2) Construct key parameters of measuring instruments through dispersed internal code values ​​and other measuring point information, and reflect the overall health status of measuring instruments through the numerical changes of key parameters;

[0033] 3) Conduct data distribution analysis on the constructed key parameters based on historical data and construct warning thresholds according to statistical methods, which is more scientific and effective. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Model training process

[0035] Figure 2 Real-time data model processing flow DETAILED DESCRIPTION

[0036] In order to deepen the understanding of the present invention, the present embodiment is described in detail below with reference to the accompanying drawings.

[0037] Embodiment 1 A method for evaluating the health of a measuring instrument, the method comprising the following steps: The instrument type of the molten iron scale is Toledo, and there are 8 sensors, each of which can collect temperature value, internal code value, air tightness, angle difference value, weighing weight, and linear value type data. The historical data of a corresponding period of time can be obtained, and the correlation relationship between the existing data can be established to determine the quantitative relationship between the measurement value and other variables. Take the first sensor of the molten iron scale as an example, and explain it in detail in combination with this embodiment.

[0038] 1. Using regression analysis, according to the weighing value of the measuring scale and the status data corresponding to each sensor, calculate the correlation between the parameters:

[0039]

[0040]

[0041] Among them, Q is the measurement value of the first sensor of the molten iron scale, is the measurement prediction value of the first group of sensors, (C, T, G, A, L) is the internal code value C, temperature T, air tightness G, angle difference value A and linear value L of the corresponding sensor. W1, W2, ..., W l is the coefficient of each corresponding characteristic parameter of the first group of sensors of the molten iron scale, l is the number of features involved in the model calculation (in this example, the number of l is 5, representing the characteristic internal code value, temperature, air tightness, angle difference value, and linear value respectively). i is the historical data sample number, and m is the total number of historical data samples.

[0042] Then, according to the historical sample data, the objective function is Solve to calculate the correlation coefficients in the above formula. After calculating the correlation coefficients, the measurement value Q of the k-th group of sensors can be predicted according to the following formula:

[0043]

[0044] According to the predicted value and the actual value Q, and calculate the core monitoring variable of the measuring instrument according to the following formula:

[0045]

[0046] Among them, ε represents the difference between the measurement prediction value and the actual measurement value of the first sensor of the molten iron scale, which serves as the core monitoring variable.

[0047] The health status of the measuring instrument is evaluated based on the core monitoring variables.

[0048] 2. Evaluate the health status of the measuring instrument according to the core monitoring variables, that is, calculate the adaptive thresholds of key parameters based on historical data through mathematical statistics methods.

[0049] The sensor status corresponding to the first sensor of the hot metal scale can be The evaluation is mainly based on the size of the core monitoring variable ε. If the core monitoring variable ε is within a reasonable range, it means that the status of the kth group of sensors is normal. Otherwise, there may be a problem and feedback is required for investigation.

[0050] Reasonable range setting for all core monitoring variables that can be calculated from historical data According to the distribution, statistics are collected and warning thresholds are set to warn the status of the corresponding sensors. The setting of warning thresholds is completed through the 3-sigma criterion:

[0051]

[0052]

[0053] Among them, ε i is the difference between the predicted value and the actual value of the first group of sensors in the i-th historical data sample of the molten iron scale, μ is the mean of all core monitoring variables in the historical data, σ is the mean square error of the core monitoring variables, i is the historical data sample number, and m is the total number of historical data samples.

[0054] According to the 3-sigma criterion, the warning threshold can be set as:

[0055] ε alarm =μ+3σ

[0056] When the core monitoring variable ε calculated from the real-time collected sample data exceeds ε alarm When an emergency occurs, an early warning signal is issued.

[0057] According to the core monitoring variables calculated from historical data, μ is calculated to be 53218.03, σ is calculated to be 40419.31, so the calculated ε alarm It is 174475.97.

[0058] 3. Based on the trend change of the historical core monitoring variables of the molten iron weighing scale, a time series ARIMA (p, d, q) model is established, where the historical data step length is l h and prediction step length l p Can be specified according to actual situation (in this example, l h =50,l p=10, that is, according to the trained time series model, 50 core monitoring variables in history are used each time to predict the 10 core monitoring variables in the future). By predicting the results of historical data in the future, the possible development level of key indicators is judged, and the prediction results based on the time series are evaluated to achieve the purpose of trend warning.

[0059] Each accumulation step is l h =50 data, and use the trained time series model to predict the step size of l p = 10 future core monitoring variable data, and by comparing the predicted data with the set threshold, determine whether the alarm condition is reached and the time to reach the warning threshold, and perform trend warning. In this example, when the scheduled task predicts that the size of the next 10 core monitoring variables exceeds ε continuously, alarm When the set number of times is five, a warning signal is issued.

[0060] It should be noted that the above embodiments are not intended to limit the protection scope of the present invention, and equivalent changes or substitutions made on the basis of the above technical solutions all fall within the protection scope of the claims of the present invention.

Claims

1. A method for evaluating the health of a weighing instrument, characterized in that: The method comprises the following steps: Step 1: Obtain the real-time data and result data corresponding to the current weighing instrument and the sensor, establish the correlation between the existing data, and determine the quantitative relationship between two or more variables. Step 2: Calculate the adaptive thresholds of key parameters based on historical data using mathematical statistics methods. Step 3: Use the established time series ARIMA (p, d, q) model to predict the time when the state parameter reaches the warning threshold, and use this time to issue a trend warning.

2. The method for evaluating the health of a weighing instrument according to claim 1, characterized in that: Step 1 is as follows: Using regression analysis, the correlation between the parameters is calculated based on the weighing value of the measuring scale and the status data corresponding to each sensor: Among them, Q k is the measurement value of the kth group of sensors of the measuring instrument, is the measurement prediction value of the kth group of sensors, (C k ,T k ,G k ,A k ,L k ) is the internal code value C of the corresponding group of sensors k, , Temperature T k , Air tightness G k , Angle difference A k and linear value L k , k represents the sensor serial number of the measuring instrument (k = 1, 2, ..., n), n is the number of sensors of the measuring instrument, that is, there are n groups of sensors in total, W k1 , W k2 ,...,W kl is the coefficient of each corresponding characteristic parameter of the kth group of sensors, l is the number of characteristics involved in the model calculation (in the present invention, the number of l is 5, representing the characteristic internal code value, temperature, air tightness, angle difference value, and linear value respectively), i is the historical data sample number, m is the total number of historical data samples, Then, according to the historical sample data, the objective function is Solve to calculate the correlation coefficients in the above formula. After calculating the correlation coefficients, the measurement value Q of the kth group of sensors can be calculated according to the following formula: k Make predictions: According to the predicted value and the actual value Q k , and calculate the core monitoring variables of the measuring instrument according to the following formula: Among them, ε k represents the difference between the measurement prediction value and the actual measurement value of the k-th group of sensors, which serves as the core monitoring variable of the k-th group of sensors. The health status of the measuring instrument is evaluated based on the core monitoring variables.

3. The method for monitoring the health status of a weighing scale according to claim 2, characterized in that: Step 2 is as follows: The sensor status corresponding to the kth group of sensors can be obtained by The evaluation is mainly carried out through the core monitoring variable ε k The size of the core monitoring variable ε k If the value is within a reasonable range, it means that the status of the kth group of sensors is normal. Otherwise, it means that there may be a problem and feedback is needed for investigation. Reasonable range setting for all core monitoring variables that can be calculated from historical data According to the distribution, statistics are collected and warning thresholds are set to warn the status of the corresponding sensors. The setting of warning thresholds is completed through the 3-sigma criterion: Among them, ε ki is the difference between the predicted value and the actual value of the kth sensor in the i-th historical data sample, μ k is the mean of the core monitoring variables of all k-th group of sensors in the historical data, σ k is the mean square error of the core monitoring variables of the kth group of sensors in the historical data, i is the sample number of the historical data, m is the total number of samples of the historical data, According to the 3-sigma criterion, the warning threshold of the kth group of sensors can be set as: e k alarm =μ k +3s k When the core monitoring variable ε is calculated from the real-time collected sample data k Beyond ε k alarm When , the status of the kth group of sensors sends out an early warning signal. The established time series ARIMA (p, d, q) model is used to predict the time when the state parameters reach the warning threshold, and this time is used to make trend warnings. The possible development level of various indicators is judged by predicting the results of historical data in the future, and an evaluation is performed based on the prediction results of the time series to achieve the purpose of trend warning.

Images