TW201619582A - Method for vibration monitoring and alarming using autoregressive models - Google Patents

Abstract

An autoregressive vibration monitoring and alarming method is to sense M fixed monitoring locations in a structure to be monitored in a monitoring period by M sensor units for detecting vibration signals and then to get M vibration signals corresponding to the M fixed monitoring locations. Then a processing unit collects the M vibration signals and constructs a vibration transformation model based on an autoregressive eXogeneous algorithm within a preparatory section of the monitoring period, where the vibration transformation model can be used to achieve another N locations to be monitored within a selected section followed by the preparatory section of the monitoring period.

Description

Vibration monitoring and alarming method with autoregressive analysis model

The invention relates to a monitoring method, in particular to a vibration monitoring and alarming method comprising an autoregressive analysis model suitable for monitoring the vibration condition of a civil structure and an industrial machine room.

With the development of precision industry and high technology, civil engineering and industrial equipment rooms are becoming more and more strict with the limitations and requirements for sensing environmental vibrations to prevent their possible harm to personnel and industry. And to summarize the sources of these environmental vibrations, the most common are natural earthquakes, or vibrations caused by human factors such as traffic and construction. When these vibrations are transmitted to the bridge or the floor of a building, they not only interfere with the personnel. Life and work, affecting physical and mental health, and serious, even bring serious threats and harm to the lives and property of personnel. In terms of industry, for the high-tech semiconductor factories that have been established in recent years, or the laboratories with high-sensitivity precision instruments, the vibration generated in the environment directly affects the yield or defect rate of the products, which in turn causes the relevant industry to Loss of profit. Therefore, based on the above, it is necessary to perform vibration monitoring on the above-described civil structure and industrial equipment room.

However, existing vibration monitoring methods are intended to be reliably monitored. The measured data must be monitored by a corresponding number of sensors at each of the locations to be monitored, and the vibration conditions in the environment are evaluated based on the sensed signals measured by the sensors at their locations. The possibility of harm. However, in actual installation, it is often limited by the floor structure of the building itself, or because of the large number of machine tools in the plant, which hinders the installation of the sensor and the layout of the circuit, which greatly affects the reliability of the monitoring data. Moreover, sensors and wires are expensive, and large-scale deployment in the above structures is not cost-effective.

Accordingly, it is an object of the present invention to provide a vibration monitoring and alarming method including an autoregressive analysis model that is not limited by the structure itself when the sensor is installed, and which can reduce the cost of deployment.

Accordingly, the present invention comprises a vibration monitoring and alarming method of an autoregressive analysis model, which monitors M fixed monitoring positions of a structure to be monitored and vibration states of N to be monitored positions different from the M fixed monitoring positions, wherein M N is a positive integer, and the vibration monitoring alarm method including the autoregressive analysis model includes a step (A), a step (B), a step (C), and a step (D).

The step (A) is to use the M sensing units for measuring the vibration to be arranged in the M fixed monitoring positions of the structure within a monitoring time, and measure the vibrations of the M corresponding fixed monitoring positions. situation.

The step (B) is a preliminary period of the monitoring time, and the sensing units of the N-measured vibrations are arranged in the N to be monitored positions of the structure, and the N corresponding to the measured position are measured. Monitoring position vibration situation.

The step (C) is to establish a vibration condition conversion model by performing a time series autocorrelation operation on the vibration conditions of the M fixed monitoring positions in the preparation period and the vibration conditions of the N positions to be monitored.

The step (D) is to measure the vibration condition of the M fixed monitoring positions by using the M sensing units in a selected period after the preliminary period, and input the vibration condition conversion model, and learn the N times of the selection period. The predicted vibration condition of the location to be monitored.

The effect of the invention is that: by the vibration condition conversion model established in the preparation period, the vibration state of the M fixed monitoring positions can be directly converted in any selected period after the preparation period, and the vibration condition can be directly converted. The predicted vibration state of the N positions to be monitored can not only monitor the position of the structure in which the structure is difficult to be mounted, but also reduce the installation cost by actually disposing the sensor at the position.

1‧‧‧First steps

2‧‧‧Second step

3‧‧‧ third step

4‧‧‧ fourth step

5‧‧‧ fifth step

6‧‧‧ Structure

61‧‧‧Fixed monitoring position

62‧‧‧Location to be monitored

7‧‧‧Sensor unit

8‧‧‧Processing unit

81‧‧‧Channel Integrator

82‧‧‧ data extractor

83‧‧‧ computer

Other features and effects of the present invention will be apparent from the following description of the drawings. FIG. 1 is a flow chart illustrating an embodiment of the vibration monitoring and alarming method of the present invention including an autoregressive analysis model; 2 is a schematic view, and FIG. 3 is a schematic view, and FIG. 3 is a schematic view. FIG. 2 illustrates a vibration monitoring system used in the embodiment.

Referring to FIG. 1 , FIG. 2 and FIG. 3 , an embodiment of the vibration monitoring and alarming method of the present invention including an autoregressive analysis model includes a first step 1, a second step 2, a third step 3, and a fourth step. 4 and a fifth step 5, and monitoring the M fixed monitoring positions 61 of the structure 6 with a vibration monitoring system as shown in FIG. 3 (only four fixed monitoring positions 61 are shown in the figure) and different from the M vibration conditions of N fixed monitoring positions 61 (only three of which are to be monitored 62 are shown in the figure), wherein M and N are positive integers, and the vibration monitoring system includes most of the supply a vibration sensing unit 7 and a processing unit 8 for processing the vibration state of the structure into monitoring data, the sensing unit 7 being selected from at least one of an accelerometer and a speedometer, the processing The unit 8 has a channel integrator 81, a data extractor 82, and a computer 83. For convenience of explanation, the function will be described later.

The first step 1 is to use the M sensing units 7 to be disposed in the M fixed monitoring positions 61 of the structure 6 during a monitoring time, and measure the vibration status of the M fixed monitoring positions 61 corresponding thereto. And according to the accented sensing unit 7 as an accelerometer or a speedometer, corresponding to an acceleration signal or a speed signal reflecting the vibration condition of the M fixed monitoring positions 61, but for the following description, the following only senses The unit 7 is an accelerometer, and the acceleration signal is taken as an example.

The second step 2 is a preliminary period of the monitoring time, and the N sensing units 7 are disposed on the N to-be-monitored locations 62 of the structure 6 and the N positions to be monitored are measured. The acceleration signal of the vibration condition of 62.

The third step 3 is to use the channel integrator 81 to filter and integrate the acceleration signals corresponding to the vibration conditions of the M fixed monitoring positions 61 and the corresponding N positions to be monitored 62 during the preliminary period to improve the signal miscellaneous. The signal ratio is used to perform discrete time sampling on the filtered and integrated processed acceleration signals, and to obtain a plurality of vibration states reflecting the M fixed monitoring positions 61 according to the set number of sampling points. The external variable parameter data and a plurality of autoregressive parameter data reflecting the vibration conditions of the N to-be-monitored positions 62 are input to the computer 83. Next, the computer 83 linearly superimposes the two in a chronological order to establish an autoregressive external variable model and further represents as follows: Where Y is an autoregressive external variable model, q is an order, S is the number of sampling points, { y _t } is the autoregressive parameter data of the t-th discrete time index, and { x _t } is the t-th discrete time Variable parameter data, For the i-th autoregressive operation coefficient, θ _j is the j-th external variable operation coefficient, W is a combination of the vibration state of the M fixed monitoring positions 61 and the vibration state of the plurality of to-be-monitored positions 62, P is The vibration conditions of the M fixed monitoring positions 61 are converted into the transfer coefficients of the vibration conditions of the N to-be-monitored positions 62. In this way, the linear regression characteristic of the integrated autoregressive operation is easy to predict the characteristics of the vibration condition parameter data corresponding to each discrete time index by the past discrete time indicators, and the external variable parameter operation is not affected by the discrete time indicators and corresponding The vibration condition parameter data can randomly provide the advantage of stable prediction, and after statistical sorting, the transfer coefficient of the vibration state of the M fixed monitoring positions 61 into the vibration condition of the N to-be-monitored positions 62 is obtained.

It is worth mentioning that the computer 83 also confirms that the plurality of extrinsic parameter data and the plurality of autoregressive parameter data are matched by the autoregressive external variable model to match the appropriate fit (Best fit), so that the Predicting vibration conditions is more reliable. After confirming that the autoregressive external variable model can provide a reliable prediction result, the N sensing units 7 disposed at the position to be monitored 62 can be removed after the preliminary period, leaving only the M The sensing unit 7 is disposed at the fixed monitoring position 61.

The fourth step 4 is to measure the vibration condition of the M fixed monitoring positions 61 by using the M sensing units 7 in a selected period after the preparation period, and correspondingly obtain an acceleration signal, and then the computer 83 performs discrete time sampling. The plurality of extrinsic parameter data matrices are processed to input the autoregressive outer variable model to obtain the predicted vibration condition Y' of the N to-be-monitored positions 62 of the selected time period.

The fifth step 5 is to use the computer 83 to issue an alarm when the acceleration signal amplitude value of the M fixed monitoring positions 61 and the acceleration signal amplitude value of the N to-be-monitored positions 62 exceed a preset value. Moreover, the computer 83 also performs fast Fourier transform on the vibration conditions of the M fixed monitoring positions 61 and the predicted vibration conditions of the N to-be-monitored positions 62 to obtain a one-eighth audio spectrum for the personnel to observe the spectrum. Amplitude Quickly know which monitoring location has an abnormal vibration condition. In addition, the data extractor 82 can also upload the plurality of external variable parameter data obtained in the third step 3 and the plurality of autoregressive parameter data to a cloud server for immediate knowledge by a remote person. Vibration information.

In the vibration monitoring and alarming method of the present invention including the autoregressive analysis model, although it is not known in advance how the vibration wave enters the structure 6, the characteristics of the general vibration wave are generated via the circumference of the structure 6 and the column. Entering the transmission path of the structure 6, and according to the assumption of the rigid flexible plate, the vibration mode adjacent to the center of the structure 6 is similar to the vibration mode around the structure 6 or the column, and thus generally includes autoregression by the present invention. When the vibration monitoring alarm method of the analysis model monitors the vibration condition of the structure 6, the M fixed monitoring positions 61 are set at the corner of the structure 6 or near the column, and the N positions to be monitored 62 It is disposed in the center of the structure 6. However, the positional arrangement is not limited thereto, and may be set around the transmission path where the vibration is generated depending on the monitoring requirements, for example, may be set on the desktop of the semiconductor machine or on the column head. Good results can achieve good monitoring results.

When the N positions to be monitored 62 are determined, the processing unit 8 first arranges the M sensing units 7 at the M fixed monitoring positions 61 and the N positions to be monitored 62 in the preliminary period. The N sensing units 7 are measured. It is worth mentioning that the M sensing units 7 and the N sensing units 7 are capable of visually monitoring, and the accelerometers or measuring the vertical vibrations are measured. An accelerometer that measures horizontal vibration, the computer 83 and the obtained acceleration signal is determined by the order q and then a large number of statistics are established. The autoregressive external variable model, and after confirming the appropriate degree of the autoregressive external variable model, the processing unit 8 obtains the M fixed monitoring positions again in the selected time period after any of the selected time periods after the preliminary period When the acceleration signal of 61 is input, the acceleration signals of the N to-be-monitored positions 62 in the selected period are obtained directly by inputting the autoregressive external variable model. At this time, the N to be monitored are originally scheduled to be in the preliminary period. The N sensing units 7 of the position 62 can be removed, so that the existing vibration monitoring alarm method with the autoregressive analysis model needs to be arranged according to the number of the positions to be monitored to save the sense of the missing number of sensing units. In addition to the cost of measuring the cost of the unit, in addition to overcoming the problem that the location to be monitored to be monitored in the past is difficult to install due to the floor structure of the building, or is limited by the need for the installation of the machine, and then Saving the cost of deployment is not limited by the environment of the structure to be monitored, and the convenience of monitoring is increased.

It is to be further explained that, in the second step 2, when the number of sensing units 7 cannot be disposed in the N to-be-monitored positions 62 corresponding to the sensing time and is not limited by the sensing cost, In the second step 2, it is decided to pay attention to the N' positions of the N to-be-monitored positions 62, N' is a positive integer smaller than N, and the third step 3 is performed to establish a vibration state conversion model corresponding to the N' positions. Then, returning to the second step 2, it is decided to pay attention to N" positions, N" is a positive integer smaller than N, and then performing the third step 3 to establish a vibration condition conversion model corresponding to the N" positions, and then repeating Several times until the establishment of the vibration condition conversion model for the N positions to be monitored 62 is completed, the processing unit 8 can also obtain the vibration status of the M fixed monitoring positions 61 in any of the selected periods after the preliminary period. Then, the vibration condition conversion model is directly input to obtain the predicted vibration state of the N to-be-monitored positions 62, thereby achieving the effect of saving the layout cost.

In summary, when the vibration state of the M fixed monitoring positions 61 is measured in any of the selected time periods after the preliminary period by the vibration condition conversion model established in the preliminary period, the converter is more directly converted. As the predicted vibration conditions of the N positions to be monitored 62, it is not necessary to actually arrange the sensing unit 7 at the N positions to be monitored 62 to predict the vibration condition of the positions, thereby not only reducing the installation cost, but also simultaneously Overcoming the problem that the location of the building to be monitored is difficult to install due to the floor structure of the building or is limited by the need for the machine to be installed, so it is not limited by the environment of the structure to be monitored, and the convenience of monitoring is increased. It is indeed possible to achieve the object of the invention.

The above is only the preferred embodiment of the present invention, and the scope of the present invention is not limited thereto, that is, the simple equivalent changes and modifications made by the patent application scope and patent specification content of the present invention, All remain within the scope of the invention patent.

1‧‧‧First steps

2‧‧‧Second step

3‧‧‧ third step

4‧‧‧ fourth step

5‧‧‧ fifth step

Claims (8)

A vibration monitoring and alarming method comprising an autoregressive analysis model, which monitors M fixed monitoring positions of a structure to be monitored and vibration states of N to be monitored positions different from the M fixed monitoring positions, wherein M and N are both As a positive integer, the vibration monitoring alarm method including the autoregressive analysis model includes: (A) arranging, by using M sensing vibrations, a sensing unit disposed at the M fixed monitoring positions of the structure within a monitoring time, And measuring M vibration conditions corresponding to the fixed monitoring position where the M is located; (B) using the N sensing electrodes for measuring the vibration to be disposed in the N of the structure during a preliminary period of the monitoring time The position to be monitored is measured, and N vibration conditions corresponding to the position to be monitored are measured; (C) the vibration state of the M fixed monitoring positions in the preparation period by a processing unit, and the N to be monitored The vibration state of the position is subjected to a time series autocorrelation operation to establish a vibration condition conversion model; and (D) measuring, by the M sensing units, the vibration state of the M fixed monitoring positions in a selected time period after the preliminary time period and inputting The Moving status transition model, and the prediction that the N oscillation condition of the selected positions to be monitored period. The vibration monitoring and alarming method including the autoregressive analysis model according to claim 1, wherein the step (C) is to process the vibration state of the M fixed monitoring positions into a plurality of external variable parameter data, and the N waiting The vibration condition of the monitoring position is processed into a plurality of autoregressive parameter data, and both An autoregressive external variable model is established by linear superposition in chronological order, and then the transfer coefficient of the vibration state of the M fixed monitoring positions into the vibration state of the N to be monitored positions is obtained. The vibration monitoring and alarming method according to claim 2, wherein the step (D) is that the processing unit processes the vibration state of the M fixed monitoring positions into a plurality of external variable parameters during the selecting period. The data is input into the autoregressive external variable model to know the predicted vibration conditions of the N to-be-monitored locations of the selected time period. The vibration monitoring and alarming method including the autoregressive analysis model according to any one of claims 1 to 3, wherein the processing unit in the step (C) comprises performing filtering and integration processing to generate an acceleration signal or a speed signal. A channel integrator that outputs analog data, a data extractor that performs discrete time sampling of analog data, and a computer that processes digital data into a matrix. The vibration monitoring and alarming method including the autoregressive analysis model according to claim 1, further comprising a step (E), when the vibration state of the M fixed monitoring positions and the vibration state of the N to be monitored positions exceed a preset An alert is issued when the value is reached. The vibration monitoring and alarming method including the autoregressive analysis model according to claim 1, further comprising a step (F), the processing unit vibrating the vibration status of the M fixed monitoring positions and the predicted vibration state of the N to be monitored positions Perform a fast Fourier transform to get a one-eighth audio spectrum. The vibration monitoring and alarming method including the autoregressive analysis model according to claim 1, further comprising a step (G), when the processing unit can further fix the M The vibration condition of the monitoring position and the vibration condition of the N to be monitored positions are uploaded to a cloud server. The vibration monitoring and alarming method of claim 1, wherein the sensing unit is at least one of an accelerometer and a speedometer.