TW202305339A - Vibration monitoring system for electrical machine - Google Patents

Abstract

A vibration monitoring system for electrical machine is applied on an apparatus with an electrical machine. The vibration monitoring system for electrical machine at least comprises a vibration sensing device, a signal conversion device and a server. The vibration sensing device can detect a vibration measurement signal of the electrical machine and convert the vibration measurement signal into a motor spectral feature data. The server can perform abnormality analysis, residual life analysis, health analysis, and failure analysis based on the received motor spectral feature data, and then output a notification message of the analysis result. In addition, the server can also provide corresponding maintenance guide files based on the results of the failure analysis to provide the troubleshooting order for maintenance and parts inspection.

Description

Vibration Monitoring System for Electric Motors

The present invention relates to a vibration monitoring system for electric motors, especially a vibration measurement signal that can convert the vibration measurement signal into a motor spectrum characteristic data for abnormal analysis, remaining life analysis, and health degree analysis. Analysis and failure analysis, and a system capable of issuing notifications and maintenance instructions.

Generally, machinery will gradually fail when it is running. When there are abnormal symptoms in the initial stage, if it is not dealt with in time, it may cause subsequent serious failures.

The traditional maintenance method is passive maintenance. When an abnormality occurs on the machine, the maintenance engineer will diagnose the fault based on the state of the machine. Due to the needs of maintenance technology in the industry, the research focus of maintenance has gradually shifted to the fields of condition monitoring, predictive maintenance and early fault diagnosis.

Many semiconductor manufacturing industries will develop in the direction of intelligent electronic diagnosis, so as to realize real-time monitoring and adjustment of equipment operation. The adoption of this technology will also affect the entire industry and the semiconductor manufacturing industry. It is obvious that traditional maintenance methods are no longer suitable for semiconductors. industries such as manufacturing.

In view of the above situation, this case can collect and monitor the vibration measurement signal of the motor, and can perform abnormal analysis, remaining life analysis, health analysis and failure analysis, so as to issue warnings before serious problems occur Notification, in addition, can provide maintenance guidelines for the analyzed problems, so that serious failures caused by accumulation of problems can be avoided, so the present invention should be an optimal solution.

The vibration monitoring system for motors of the present invention is applied to more than one motor facility, and the vibration monitoring system for motors includes at least one vibration sensing device connected to the motor facility to detect the The vibration measurement signal of the motor facility; at least one network device is used to receive the data and transmit it through a network transmission; at least one signal conversion device is electrically connected to the vibration sensing device and the network device , used to receive the vibration measurement signal detected by the vibration sensing device, and convert the vibration measurement signal into a motor spectrum characteristic data, and then measure the measurement time data of the vibration measurement signal and the motor spectrum The characteristic data is sent out through the network device; a server device is able to receive the measurement time data and the frequency spectrum characteristic data of the motor transmitted by the network device, and the server device has at least one processor and at least one computer The computer-readable recording medium stores at least one monitoring and analysis application program, a normal vibration data and multiple situation comparison files, wherein the computer-readable recording medium further stores computer-readable Reading instructions, when the computer-readable instructions are executed by the processors, cause the servo device to perform the following procedures: compare the received motor spectrum characteristic data with the normal vibration data through the monitoring and analysis application program , to output a judging abnormal result; to continuously store the characteristic area data of the frequency band and establish a trend model, to estimate the time trend of the total vibration value, and then according to the time trend and the measurement time The data output is the usable life data of the equipment; it is used to compare the received motor spectrum characteristic data with different situation comparison files for similar probabilities, and output the situation comparison file with the highest probability as a fault analysis and judgment result; it is used to be able to Send a notification message for the content of the judgment abnormal result, the usable life data of the equipment or/and the failure analysis judgment result.

More specifically, the vibration measurement signal is a sinusoidal vibration waveform or a shock wave waveform.

More specifically, the frequency spectrum characteristic data of the electric motor can be divided into multiple frequency band characteristic area data according to different frequency bands.

More specifically, the normal vibration data is one or more preset characteristic warning values, and the monitoring and analysis application program can compare with the motor spectrum characteristic data according to the preset characteristic warning values. If the warning value of the feature is preset, the abnormal result of the judgment will be output.

More specifically, the normal vibration data is collected under long-term normal operation, and a judgment model is trained by machine learning based on the data, and the judgment model is compared with the frequency spectrum characteristic data of the motor, If the difference is too large, an abnormal result of the judgment is output.

More specifically, the monitoring and analysis application program can continuously store the characteristic area data of the frequency band as a total vibration history data according to the measurement time data, and establish the trend model based on the total vibration history data, and use the trend The model estimates the time trend of the total vibration value, and then sets a preset total vibration upper limit value based on the equipment, and then judges based on the preset total vibration upper limit value and the time trend of the total vibration value Output the data of the upper limit time of the available equipment, and then output the usable life data of the equipment through the data of the upper limit of the available time of the equipment and the data of the measurement time.

More specifically, the monitoring and analysis application program can be based on the ratio of the total vibration history data to the preset total vibration upper limit as a first judgment value, and then adapt a time trend according to the total vibration value. Simple linear regression to obtain a degree of stability, and use the degree of stability as a second judgment value, and then use the ratio of the equipment's usable life data to the equipment's usable upper limit time data as a third judgment value, and finally The first judgment value, the second judgment value and the third judgment value are weighted to obtain a piece of health data.

More specifically, the monitoring and analysis application program can compare the received frequency band characteristic area data with different situation comparison files, and output the fault analysis and judgment result according to the situation comparison file with the highest probability, and if it is judged The similarity probability between the received frequency band characteristic area data and each situation comparison file is lower than a set standard, then the received frequency band characteristic area data can be established as a new situation comparison file.

More specifically, the monitoring and analysis application program can provide a report interface for reporting a successful judgment result or a judgment failure result through the reporting interface after the monitoring and analysis application program provides the fault analysis and judgment result. , and the monitoring and analyzing application program can report according to the judging success result or the judging failure result, so as to improve the accuracy of the fault analyzer.

More specifically, the computer-readable recording medium stores maintenance guide files based on different situational comparison files. If the fault analysis and judgment result is analyzed, the monitoring and analysis application can be stored in the maintenance suggestion. The device finds the corresponding maintenance guide file to provide the order of maintenance and parts inspection.

More specifically, the network transmission method is a wireless network transmission method or a wired network transmission method.

More specifically, the notification message can be sent through mail, communication software or short message technology.

Other technical contents, features and effects of the present invention will be clearly presented in the following detailed description of preferred embodiments with reference to the drawings.

Please refer to Figures 1A to 1D, which are schematic diagrams of the equipment configuration of the vibration monitoring system for electric motors, the connection diagram between the network device and the servo equipment, the internal structure diagram of the servo equipment, and the architecture diagram of the monitoring and analysis application program of the present invention. It can be seen from the above that the vibration monitoring system for motors is applied to the motor facility 1 on the equipment machine in a factory building. The motor facility 1 is connected to the vibration sensing device 2, and the connection method is not limited to screwing on the On the motor or bonded to the surface of the motor, there are different connection methods mainly depending on the type of motor facility 1 (the connection is mainly close to the vibration source of the motor), wherein the motor facility 1 can be a vibration motor or an air motor, and The vibration measurement signal of the vibration motor is a sinusoidal vibration waveform, and the vibration measurement signal of the air motor is a shock wave waveform.

The motor device 1 can be a constant pressure pump, fan motor, compressor, induction motor, and the motor device 1 is not limited to steady-state & transient motors.

The other end of the vibration sensing device 2 is connected to the signal conversion device 3, the signal conversion device 3 is used to receive the vibration measurement signal detected by the vibration sensing device 2, and convert the vibration measurement signal It is a motor spectrum characteristic data, and then the measurement time data of the vibration measurement signal and the motor spectrum characteristic data are connected with the network device 4 (the network device 4 refers to any device with Internet transmission function, such as Ethernet gateway or/and network sharer) are connected through the connection line 31, and transmitted through the network transmission mode (wireless network transmission mode or wired network transmission mode), as shown in Figure 2A , after the vibration sensing device 2 obtains the vibration measurement signal, it converts a time domain diagram (Time Domain) into the motor frequency spectrum characteristic data through the signal conversion device 3, as shown in Figure 2B, the motor frequency spectrum characteristic data is a frequency domain Frequency Domain (Frequency Domain), and the conversion of a time-domain image into a frequency-domain image can use an algorithm such as Fourier transform to convert a time-domain signal into the corresponding amplitude and phase at different frequencies, and its spectrum is the time-domain signal. performance in the domain.

In addition, the frequency spectrum feature data of the electric motor can be extracted according to different frequency bands to extract multiple frequency band feature area data, and the spectrum feature extraction can be performed by the signal conversion device 3 or the servo device 5, as shown in Figures 2B and 2C As shown, the frequency spectrum characteristic data of the motor is divided into multiple sections by frequency range. Figure 2C clearly marks the amplitude value of each frequency band area, and the frequency range of Band1 refers to the total amount of vibration stipulated by ISO Frequency range, and the frequency range of Band2~8 is defined as follows (the following definitions of different frequency ranges are just examples of one of the implementation modes, and the actual implementation will have different Hz definition ranges depending on different devices): (1) Band2: 55Hz~59Hz (2) Band3: 115Hz~117Hz (3) Band4: 170Hz~174Hz (4) Band5: 227Hz~231Hz (5) Band6: 285Hz~290Hz (6) Band7: 300Hz~1000Hz (7) Band8: 100Hz~2000Hz

And the network device 4 can be connected with a servo device 5, so that the signal conversion device 3 can transmit the measurement time data and the frequency spectrum characteristic data of the motor to the servo device 5 through the network device 4, the The servo device 5 is provided with a processor 51, an information receiver/transmitter 53, and a computer-readable recording medium 52, wherein the information receiver/transmitter 53 receives the measurement time data and the motor frequency spectrum in a network transmission mode Characteristic data, and at least one monitoring analysis application program 521 and a data storage unit 522 are stored in the computer-readable recording medium 52, and the internal system of the data storage unit 522 stores various normal vibration data, multiple situation comparison files, Total vibration history data, various maintenance guide files (established according to different situational comparison files);

(c12)        非線性: Exponential Model 衰退成指數分配

(c2) 以建模模型估算達到上限值之時間，此時間即為 End Time（TEnd）。 (c3) RUL（剩餘壽命） =  TFail – Tnow (4)     一健康度判斷器5214，係與該資料處理器5211及該剩餘壽命判斷器5213相連接，能夠依據該總振動歷史數據與該預設總振動上限值的比率做為一第一判斷值，並再依據該總振動值的時間趨勢配適一簡單線性回歸，以取得一穩定度，並依該穩定度做為一第二判斷值，之後再以該設備可用壽命數據與該設備可用上限時間數據的比率做為一第三判斷值，最後再將該第一判斷值、該第二判斷值及該第三判斷值以權重分配取得一健康度數據，說明如下： (a)      本案以設備總振度 / 部件振動值，模型預估與 ISO 規範計算健康度，如第6圖所示，先記錄電動機設備歷史總振動量與開始運作日期601後，再進行資料更新602（若有新的資料進來，則將舊資料與新資料合併），之後進行資料清洗603（用來排除停機狀態資料與排除人為造成異常資料），最後進行模型配適，估算健康度、剩餘壽命604； (b)     第一判斷值（H1）: 以總振動值為參考依據 (b1) ISO 規範總振動值上限為 ( ex: calss 1, ＜15kw, 4.8mm/s) (b2) 健康度：

(c)      第二判斷值（H2）: 以 Model R^2 解釋穩定度 (c1) 定義振動總量上限值 (c2) 採振幅 vs. 時間配適一簡單線性回歸, 並以 R^2 值轉換表示相依時間之穩定度，以第7A圖為例，量測曲線所配適出的配適線之R ²為0.0316，而H2=96.4%，再以第7B圖為例，剛開始為穩定，而一定時間後，曲線往上走則表示為不穩定，量測曲線所配適出的配適線之R ²為0.8484，而H2=15.1%， (c3) 簡單線性回歸（

）說明如下： 樣本資料

誤差：

最小平方法求誤差最小值，計算

而輸出運算範例如下： lm(formula = Value ~ time2value_1, data = VMS) Residuals: Min      1Q          Median      3Q     Max -0.47368 -0.14947     -0.00553   0.13945  1.51415 Coefficients: Estimate             Std. Error    t value   Pr(＞|t|) (Intercept)      1.768e+00     5.545e-03   318.89     ＜2e-16 *** time2value_1  4.953e-07      5.922e-09   83.63      ＜2e-16 *** --- Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 0.1837 on 10358 degrees of freedom Multiple R-squared:  0.403,      Adjusted R-squared:  0.403 F-statistic: 6993 on 1 and 10358 DF,  p-value: ＜2.2e-16 (c4) R^2: 值介於 0 ~ 1之間: 0: 表示穩定，與時間無關，故效能穩定 1: 表示不穩定，與時間相依，故效能不穩定 (d)     第三判斷值（H3）: 以剩餘壽命估計，如第7C圖所示，其量測資料為不規則的量測曲線，而另一趨勢曲線為弧線向上，並於碰觸到上限值之時間設為T _End，而T ₀為開始使用時間，T ₁為當下量測時間，而估計公式如下：

(e)      結合 H1 & H2 & H3，並由業主設定權重比例

,

(f)       實例舉例如第8圖所示，說明如下： (f1)  以某半導體廠機台實例，其中總振動量上限值為6，而圖中y軸是表示實際振動資料點，x軸是時間，而橫線則是本例配適之線性趨勢線（y=2.228 +3.505x10 ^-7x），數據顯示此設備振動趨於穩定，預估209天後達警戒上限； (f2)  而健康度計算如下，H1 = 52.3%，H2 = 99.2%，H3 = 93.7%，採平均計算後，整體健康度估計為 81.8%； (f3)  而本例的模型配適運算如下： lm(formula = Value ~ time2value_1, data = VMS) Residuals: Min      1Q      Median      3Q       Max -0.5818 -0.3820 -0.2258    0.2087  3.3649 Coefficients: Estimate      Std. Error     t value     Pr(＞|t|) (Intercept)    2.228e+00    3.909e-02    57.003     ＜2e-16 *** time2value_ 1 3.505e-07  1.420e-07    2.469       0.0138 * --- Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 0.5703 on 763 degrees of freedom Multiple R-squared:  0.007927,   Adjusted R-squared:  0.006627 F-statistic: 6.097 on 1 and 763 DF,  p-value: 0.01376 (5)     一故障分析器5215，係與該資料處理器5211相連接，能夠將接收之頻段特徵區域資料與不同的情境比對檔進行比對，並依據最高機率的情境比對檔輸出為該故障分析判斷結果，且若是判斷該接收之頻段特徵區域資料與每一個情境比對檔的相近機率低於一設定標準之下，則能夠將該接收之頻段特徵區域資料建立為一新的情境比對檔，說明如下： (a)      而本案故障分析程序，如第9圖所示，進行電動機設備振動資料模擬901，建立多個情境檔902，再進行資料清洗、標籤化、標準化等處理903之後，進行神經網路訓練、建模、更新904，最後再將模型輸出905，並將模型套用分析906； (b)     而當電動機設備新資料輸入907，之後則透過該模型進行故障分析預測908； (c)      而當預測失誤後，系統能夠提供失效預測進行更新模組909，之後則能夠於該資料儲存單元522內的情境庫910內進行更新情境911，用以提高故障分析器的準確度。 (6)     一警報通知器5216，係與異常偵測器5212、剩餘壽命判斷器5213、健康度判斷器5214、故障分析器5215相連接，用以當判斷有異常情況時，則能夠透過該資訊接收/傳輸器53以mail、通訊軟體或是簡訊訊息等技術發出通知訊息或是直接顯示於回報介面上。 (7)     一使用介面器5217，係與異常偵測器5212、剩餘壽命判斷器5213、健康度判斷器5214、故障分析器5215及警報通知器5216相連接，該使用介面器5217能夠提供一回報介面，用以提供該故障分析判斷結果後，能夠透過該回報介面進行回報一判斷成功結果或是一判斷失效結果，而該監測分析應用程式能夠依據該判斷成功結果或是判斷失效結果進行回報，用以提高故障分析的準確度。 Wherein the monitoring analysis application program 521 includes: (1) a data processor 5211, used to receive the measurement time data and the frequency spectrum characteristic data of the motor, and can divide the frequency spectrum characteristic data of the motor into multiple (2) An anomaly detector 5212, which is connected with the data processor 5211, is used for quantitative analysis or qualitative analysis of the frequency band characteristic area data, as shown in the first paragraph 3, descriptions and examples are as follows: (a) Quantitative analysis: (a1) Carry out quantitative analysis 302 on the frequency spectrum characteristic data 301 of the motor, and then determine the normal vibration data, and the normal vibration data is one or A plurality of preset characteristic warning values 303, and finally compare the preset characteristic warning values with multiple frequency band characteristic area data of the motor spectrum characteristic data, and output the judgment abnormal result if the preset characteristic warning values are reached And issue a warning signal 304; (a2) As shown in Figure 4, it is the result of quantitative analysis based on the total vibration of a certain ice water pump motor. The process is as follows: A. Record the real-time data of the total vibration (not in the figure Regular oscillation curve); B. Based on the ISO-10816 vibration control limit (unit: mm/s), where the vibration value <= 0.7 means Good, if 0.7 < vibration value <= 1.8 means Acceptable, if 1.8 < The vibration value <= 4.5 means Unsatisfactory, and if 4.5 < the vibration value means Unacceptable; C. The figure shows that the vibration value exceeds 1.8mm/s for a total of 52 times (the area above the horizontal line), which means that part of the operation presents a large vibration , although it does not happen often, but it should be noted; D. Real-time system records one by one and warns relevant personnel to intervene for confirmation; E. In addition to using ISO as a reference, it also provides several logics for formulating norms: average value *n, n=1 ,2...; mean + n*standard deviation, n=3,4... ; median + n*IQR, n=1.5,3,...; or custom. (b) Qualitative analysis: (b1) Perform qualitative analysis 305 on the motor frequency spectrum characteristic data 301, wherein the normal vibration data is collected under long-term normal operating conditions 306, and train a machine learning method based on the data. After judging the model 307, and providing the user with an interface to select the sensitivity (high, standard, low) 308, the judgment model can be compared with the new motor frequency spectrum characteristic data 309, if the difference is too large (more than the model decision Boundary), then it is judged as abnormal 310 and output and record the abnormal result of the judgment and a warning signal 311, on the contrary, if the comparison result is within the decision boundary of the model, it is judged as no abnormality 312; (b2) and qualitative analysis used The method is Isolation Forest, which is briefly described as follows: A. Outliers are those that are easy to be isolated; data that are sparsely distributed and far away from high density are outliers; B. Continuously and randomly cut the data set , until there is 1 point left in each subspace; C. Repeat the above data cutting behavior for many times; D. After multiple random cutting, calculate the abnormal score. If it is closer to 1, it is more likely to be an abnormal point. If all the scores are 0.5 It can be explained that there may be no abnormal points in the data; (b3) As shown in Figures 5A~5D, it is the result of qualitative analysis based on the total vibration of a certain ice water pump motor. The process is as follows: A. Take the vibration data of the impeller of the ice water pump motor as an example, and collect two data sets: Normal Set (operating data in normal state, as shown in Figure 5A) and Testing Set (data in abnormal state of impeller, referring to impeller Unbalanced and load data, as shown in Figure 5C); B. The model is obtained with the Training Set and the Isolation Forest machine learning method, and its Anomaly Score is calculated (for the impeller in normal state, as shown in Figure 5B ), select the critical value from Anomaly Score, this case is 0.7236, and record the model; C. Apply the model to the Testing Set and calculate the Anomaly Score (Table 1 below) value Update time score 0.3001953 2021-02-19 15:46:36 0.724 0.2445313 2021-02-19 15:46:37 0.620 0.3119141 2021-02-19 15:46:38 0.724 0.3158203 2021-02-19 15:46:39 0.724 0.3382812 2021-02-19 15:46:40 0.724 0.2855469 2021-02-19 15:46:41 0.724 0.2835937 2021-02-19 15:46:42 0.724 0.2010742 2021-02-19 15:46:43 0.587 0.1849609 2021-02-19 15:46:44 0.596 0.1844727 2021-02-19 15:46:45 0.594 Table 1 Vibration value vs Anomaly Score D. With 0.7236 as the threshold value, if the Anomaly Score of the Testing Set is greater than 0.7236, it is an abnormal data point (such as the data point in the upper area in Figure 5D), otherwise, if it is less than 0.7236, it is a normal point ; E. Record the abnormal data points in the database and notify the relevant engineering units to inquire; F. This example shows that if the abnormal data points continue to appear, you should pay attention to whether there is any abnormal condition of the impeller that causes a large difference from the original normal data. (3) A remaining life judging device 5213 is connected with the data processor 5211, which can continuously store the characteristic area data of the frequency band as a total vibration history data according to the measurement time data, and establish a total vibration history data based on the total vibration history data. The trend model, and use the trend model to estimate the time trend of the total vibration value, and then set a preset total vibration upper limit according to the equipment, and then use the preset total vibration upper limit and The time trend of the total vibration value is used to determine the upper limit time data of the equipment, and then output the usable life data of the equipment based on the upper limit time data of the equipment and the measurement time data. The description is as follows: (a) In this case, the amplitude value is Based on, fit the best trend model (linear or nonlinear model) from historical data. Establish the relationship between the vibration OA value and the time series, and convert the measured data into predicted time series information as the basis for predicting the performance degradation of the machine; (b) The operating conditions are as follows: (b1) Need to define the warning value: the total vibration amount is The upper limit of ISO is a reference, and other vibration quantities can be customized by the field, and the field is defined as follows: (b11) Maximum amplitude (Max)*n, n=2,3,…; (b12) Average value of amplitude*n , n=2,3,…; (b13) or self-defined (b2) long-term data modeling is more robust; (c) and the operation method is as follows (the following operation example is just an example of one of the implementation modes, and the actual Execution, there will be different nonlinear algorithms depending on different devices): (c1) A trend model is equipped in a linear and nonlinear manner (c11) Linear: glm

(c12) Nonlinear: Exponential Model decays into an exponential distribution

(c2) The time to reach the upper limit is estimated by the modeling model, and this time is End Time (TEnd). (c3) RUL (remaining life) = TFail - Tnow (4) A health degree determiner 5214 is connected with the data processor 5211 and the remaining life determiner 5213, and can be based on the total vibration history data and the preset The ratio of the total vibration upper limit value is used as a first judgment value, and then a simple linear regression is fitted according to the time trend of the total vibration value to obtain a stability, and the stability is used as a second judgment value , and then use the ratio of the equipment usable life data to the equipment usable upper limit time data as a third judgment value, and finally obtain the first judgment value, the second judgment value and the third judgment value by weight distribution 1. Health degree data, explained as follows: (a) In this case, the health degree is calculated based on the total vibration degree of the equipment/component vibration value, model estimation and ISO specifications. As shown in Figure 6, first record the historical total vibration of the motor equipment and start operation After the date 601, perform data update 602 (if new data comes in, merge the old data with new data), then perform data cleaning 603 (used to exclude shutdown status data and human-caused abnormal data), and finally model Fitting, estimating health degree and remaining life 604; (b) First judgment value (H1): based on the total vibration value (b1) The upper limit of the total vibration value of the ISO specification is (ex: calss 1, <15kw, 4.8mm /s) (b2) Health:

(c) The second judgment value (H2): Use Model R^2 to explain the stability (c1) and define the upper limit of the total vibration value (c2). The amplitude vs. time is matched with a simple linear regression, and the R^2 value is used The conversion indicates the stability of time dependence. Taking Figure 7A as an example, the ^R2 of the fitting line fitted by the measurement curve is 0.0316, and H2=96.4%. Taking Figure 7B as an example, it is stable at the beginning , and after a certain period of time, if the curve goes up, it means instability. The R ² of the fitting line fitted by the measurement curve is 0.8484, and H2=15.1%. (c3) Simple linear regression (

) are described as follows: Sample data

error:

The least square method finds the minimum value of the error, and calculates

The output calculation example is as follows: lm(formula = Value ~ time2value_1, data = VMS) Residuals: Min 1Q Median 3Q Max -0.47368 -0.14947 -0.00553 0.13945 1.51415 Coefficients: Estimate Std. Error t value Pr(＞|t|) (Intercept ) 1.768e+00 5.545e-03 318.89 ＜2e-16 *** time2value_1 4.953e-07 5.922e-09 83.63 ＜2e-16 *** --- Signif. codes: 0 '***' 0.001 '* *' 0.01 '*' 0.05 '.' 0.1 '' 1 Residual standard error: 0.1837 on 10358 degrees of freedom Multiple R-squared: 0.403, Adjusted R-squared: 0.403 F-statistic: 6993 on 1 and 10358 DF, p- value: <2.2e-16 (c4) R^2: The value is between 0 and 1: 0: means stable, independent of time, so the performance is stable 1: means unstable, dependent on time, so the performance is unstable ( d) The third judgment value (H3): Estimated by the remaining life, as shown in Figure 7C, the measurement data is an irregular measurement curve, and the other trend curve is an upward arc, and when it touches the upper limit The value time is set to T _End , and T ₀ is the start time, T ₁ is the current measurement time, and the estimation formula is as follows:

(e) Combine H1 & H2 & H3, and set the weight ratio by the owner

,

(f) An example is shown in Figure 8, which is described as follows: (f1) Take a semiconductor factory machine as an example, where the upper limit of the total vibration is 6, and the y-axis in the figure represents the actual vibration data points, and the x-axis is the time, and the horizontal line is the linear trend line (y=2.228 +3.505x10 ^-7 x) of this example. The data shows that the vibration of this equipment tends to be stable, and it is estimated that it will reach the upper limit of warning in 209 days; (f2) and The health degree is calculated as follows, H1 = 52.3%, H2 = 99.2%, H3 = 93.7%, after the average calculation, the overall health degree is estimated to be 81.8%; (f3) The model fitting calculation in this example is as follows: lm(formula = Value ~ time2value_1, data = VMS) Residuals: Min 1Q Median 3Q Max -0.5818 -0.3820 -0.2258 0.2087 3.3649 Coefficients: Estimate Std. Error t value Pr(＞|t|) (Intercept) 2.228e+00 3.909e.03 57 ＜2e-16 *** time2value_ 1 3.505e-07 1.420e-07 2.469 0.0138 * --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 '' 1 Residual standard error: 0.5703 on 763 degrees of freedom Multiple R-squared: 0.007927, Adjusted R-squared: 0.006627 F-statistic: 6.097 on 1 and 763 DF, p-value: 0.01376 (5) A fault analyzer 5215, with The data processor 5211 is connected, and can compare the received frequency band characteristic area data with different situation comparison files, and output the fault analysis and judgment result according to the situation comparison file with the highest probability, and if it is judged that the received If the similarity probability between the frequency band characteristic area data and each situational comparison file is lower than a set standard, then the received frequency band characteristic area data can be established as a new situational comparison file, as explained below: (a) In this case The fault analysis program, as shown in Fig. 9, performs vibration data simulation 901 of the motor equipment, establishes multiple situation files 902, and then proceeds After data cleaning, labeling, and standardization 903, neural network training, modeling, and updating 904 are performed, and finally the model is output 905, and the model is applied for analysis 906; (b) when new data of the motor equipment is input 907, Then use the model to perform failure analysis and prediction 908; (c) when the prediction is wrong, the system can provide failure prediction to update the module 909, and then update the situation 911 in the situation library 910 in the data storage unit 522 , to improve the accuracy of the fault analyzer. (6) An alarm notifier 5216 is connected with the abnormality detector 5212, the remaining life judging device 5213, the health degree judging device 5214, and the failure analyzer 5215, so that when it is judged that there is an abnormal situation, it can pass through the information The receiver/transmitter 53 sends a notification message by means of mail, communication software, or short message message or directly displays it on the return interface. (7) A user interface device 5217 is connected with the abnormal detector 5212, the remaining life judgment device 5213, the health degree judgment device 5214, the failure analyzer 5215 and the alarm notification device 5216, and the user interface device 5217 can provide a return interface, after providing the failure analysis and judgment result, a successful judgment result or a failure judgment result can be reported through the reporting interface, and the monitoring and analysis application program can report according to the judgment success result or the failure judgment result, To improve the accuracy of fault analysis.

In addition, the data storage unit 522 of the computer-readable recording medium 52 stores maintenance guide files based on different situation comparison files. If the failure analysis and judgment result is analyzed, the monitoring and analysis application program 521 can be used for the maintenance. It is recommended that the memory find the corresponding maintenance guide file to provide the order of maintenance and parts inspection.

When the vibration monitoring system for electric motors provided by the present invention is compared with other conventional technologies, its advantages are as follows: (1) The present invention can collect and monitor the motor through the vibration measurement signal of the motor, and can perform abnormal analysis, remaining life analysis, health analysis and fault analysis, so as to issue warnings before serious problems occur notify. (2) The present invention can provide maintenance guidelines for the analyzed problems, so as to avoid the occurrence of serious failures caused by the accumulation of problems.

The present invention has been disclosed above through the above-mentioned embodiments, but it is not intended to limit the present invention. Anyone who is familiar with this technical field and has common knowledge can understand the foregoing technical characteristics and embodiments of the present invention without departing from the present invention. Within the spirit and scope, some changes and modifications can be made, so the patent protection scope of the present invention must be defined by the claims attached to this specification.

1: Motor facility 2: Vibration sensing device 3: Signal conversion device 31: Connecting line 4: Network device 5:Servo equipment 51: Processor 52: Computer-readable recording media 521:Monitoring analysis application 5211: data processor 5212: anomaly detector 5213: remaining life judger 5214: health degree judge 5215: Fault Analyzer 5216: Alert Notifier 5217: use the interface device 522: data storage unit 53: Information receiver/transmitter

[Fig. 1A] is a schematic diagram of the equipment configuration of the vibration monitoring system for electric motors according to the present invention. [Figure 1B] is a schematic diagram of the connection between the network device and the servo equipment used in the vibration monitoring system of the motor according to the present invention. [Figure 1C] is a schematic diagram of the internal structure of the servo equipment used in the motor vibration monitoring system of the present invention. [Figure 1D] is a schematic diagram of the structure of the monitoring and analysis application program used in the vibration monitoring system of the motor according to the present invention. [Fig. 2A] is a schematic diagram of the data processing of the vibration monitoring system for the motor according to the present invention. [Fig. 2B] is a schematic diagram of the data processing of the vibration monitoring system used in the motor according to the present invention. [Fig. 2C] is a schematic diagram of data processing for the vibration monitoring system of the motor according to the present invention. [Fig. 3] is the flow chart of abnormal detection and analysis of the vibration monitoring system for electric motors according to the present invention. [Fig. 4] is a schematic diagram of an example of quantitative analysis for abnormal detection and analysis of the motor vibration monitoring system according to the present invention. [Fig. 5A] is a schematic diagram of an example of qualitative analysis of the abnormality detection analysis of the vibration monitoring system of the motor used in the present invention. [Fig. 5B] is a schematic diagram of an example of qualitative analysis for the abnormal detection analysis of the vibration monitoring system of the motor according to the present invention. [Fig. 5C] is a schematic diagram of an example of qualitative analysis for the abnormal detection analysis of the motor vibration monitoring system according to the present invention. [Fig. 5D] is a schematic diagram of an example of qualitative analysis for the abnormal detection analysis of the vibration monitoring system of the motor according to the present invention. [Fig. 6] is a flow chart of the remaining life and health analysis of the vibration monitoring system for electric motors according to the present invention. [Fig. 7A] is a schematic diagram illustrating the health analysis of the vibration monitoring system for electric motors according to the present invention. [Fig. 7B] is a schematic diagram illustrating the health analysis of the vibration monitoring system for electric motors according to the present invention. [Fig. 7C] is a schematic diagram illustrating the health analysis of the vibration monitoring system for electric motors according to the present invention. [Fig. 8] is a schematic diagram of an example of the health analysis of the vibration monitoring system for electric motors according to the present invention. [Fig. 9] is a flow chart of fault analysis for the vibration monitoring system of the motor according to the present invention.

1: Motor facility

2: Vibration sensing device

3: Signal conversion device

31: Connecting line

4: Network device

Claims (10)

A vibration monitoring system for electric motors is applied to more than one electric motor installation, and the vibration monitoring system for electric motors comprises: At least one vibration sensing device is connected with the electric motor installation to detect the vibration measurement signal of the electric motor installation; At least one network device for receiving data and sending it out in a network transmission mode; At least one signal conversion device is electrically connected with the vibration sensing device and the network device, used to receive the vibration measurement signal detected by the vibration sensing device, and convert the vibration measurement signal into a motor Spectrum characteristic data, and then transmit the measurement time data of the vibration measurement signal and the frequency spectrum characteristic data of the motor through the network device; and A servo device capable of receiving the measured time data and the frequency spectrum characteristic data of the motor transmitted by the network device, and the servo device has at least one processor and at least one computer-readable recording medium, and the computers can The readable recording medium stores at least one monitoring and analysis application program, a normal vibration data and multiple situation comparison files, wherein the computer-readable recording medium further stores computer-readable instructions, when the processors When the computer-readable instructions are executed, the servo equipment will perform the following procedures: compare the received motor frequency spectrum characteristic data with the normal vibration data through the monitoring and analysis application program, so as to output a judgment result for judging abnormality; To continuously store the characteristic area data of the frequency band and establish a trend model to estimate the time trend of the total vibration value, and then output a usable life data of the equipment according to the time trend and the measurement time data; for Compare the received motor spectrum characteristic data with different situational comparison files for similar probability, and output the situational comparison file with the highest probability as a fault analysis and judgment result; it is used to judge the abnormal result of the judgment and the availability of the equipment A notification message is issued for the life data or/and the content of the failure analysis and judgment result. The vibration monitoring system for motors as described in Claim 1, wherein the vibration measurement signal is a sinusoidal vibration waveform or a shock wave waveform. The vibration monitoring system for motors as described in Claim 1, wherein the frequency spectrum characteristic data of the motor can be divided into multiple frequency band characteristic area data according to different frequency bands. The vibration monitoring system for motors as described in claim 1, wherein the normal vibration data is one or more preset characteristic warning values, and the monitoring analysis application program can be based on the preset characteristic warning values, and the The frequency spectrum characteristic data of the electric motor is compared, and if the preset characteristic warning value is reached, the judgment abnormal result is output. The vibration monitoring system for motors as described in claim 1, wherein the normal vibration data is collected under long-term normal operation, and a judgment model is trained by machine learning based on the data, and the judgment model is used Compared with the frequency spectrum characteristic data of the electric motor, if the difference is too large, the abnormal judgment result is output. The vibration monitoring system for motors as described in claim 1, wherein the monitoring and analysis application program can continuously store the characteristic area data of the frequency band as a total vibration history data according to the measurement time data, and establish based on the total vibration history data Draw out the trend model, and use the trend model to estimate the time trend of the total vibration value, and then set a preset total vibration upper limit value according to the equipment, and then use the preset total vibration upper limit value and the time trend of the total vibration value to determine the data of the upper limit time available for the equipment, and then output the usable life data of the equipment based on the data of the upper limit time available for the equipment and the data of the measurement time. The vibration monitoring system for motors as described in claim 6, wherein the monitoring analysis application program can be based on the ratio of the total vibration history data and the preset total vibration upper limit value as a first judgment value, and then based on The time trend of the total vibration value is fitted with a simple linear regression to obtain a degree of stability, and the degree of stability is used as a second judgment value, and then the ratio of the usable life data of the equipment to the usable upper limit time data of the equipment is used As a third judging value, at last the first judging value, the second judging value and the third judging value are weighted to obtain a piece of health data. The vibration monitoring system for motors as described in Claim 1, wherein the monitoring analysis application program can compare the received frequency band characteristic area data with different situation comparison files, and output according to the situation comparison files with the highest probability For the failure analysis and judgment result, and if it is judged that the similar probability of the received frequency band characteristic area data and each situation comparison file is lower than a set standard, then the received frequency band characteristic area data can be established as a new Situational comparison files. The vibration monitoring system for electric motors as described in claim 1, wherein the monitoring and analysis application program can provide a report interface for reporting through the report interface after the monitoring and analysis application program provides the fault analysis and judgment results A result of judging success or a result of judging failure, and the monitoring and analysis application program can report according to the result of judging success or judging failure, so as to improve the accuracy of failure analysis. The vibration monitoring system for electric motors as described in claim 1, wherein the computer can read the maintenance guide files stored in the recording medium based on different situational comparison files, if the failure analysis and judgment results are analyzed, the The monitoring analysis application program can find the corresponding maintenance guide file in the maintenance suggestion storage, so as to provide the order of troubleshooting for maintenance and parts inspection.

Images