TWI803396B - Anomaly diagnosis system and method for robotic arm - Google Patents

Abstract

The present invention relates to anomaly diagnosis system and method for robotic arm. Motor information of a robotic arm device traveling along a specific trail is extracted by a signal acquisition device and a diagnosis model and a health state residual are established according to the motor information of the robotic arm device traveling along a specific trail. Then, a state residual under test is obtained from new motor information according to the diagnosis model. Finally, the health state residual and the state residual under test are compared to determine whether the clamped object has an abnormal status.

Description

Mechanical arm abnormality diagnosis system and method thereof

The present invention relates to a system and method for diagnosing abnormality of a mechanical arm, in particular, a diagnostic model is established by using the motor information of the mechanical arm and the object in a healthy state without using any additional sensors, and using the established diagnostic model An abnormality diagnosis system and method thereof for diagnosing whether an abnormality occurs during the process of diagnosing a gripped object.

With the application of robotic arms in automated production and other applications, in order to improve equipment reliability or achieve industrial and business networking, abnormal diagnosis of robotic arms is becoming more and more important.

Generally, when diagnosing abnormalities or faults, using sensors for fault diagnosis is the most intuitive and simple method, so various sensors are often added to obtain abnormal signals, but because this method requires a large number of sensors, it costs a lot of money , And it is not suitable for the use of robotic arms with narrow working environment.

Therefore, the signal-type fault diagnosis is developed. Signal-based fault diagnosis uses measurement signals, such as current and voltage signals of motor drives, and uses signal processing to extract signal features, such as time domain, frequency domain, and time-frequency domain. ), wavelet transforms, and Wigner-ville distribution, but because the captured signal does not respond to any fault, the accuracy of fault diagnosis is poor.

Therefore, in order to improve the shortcomings of using sensors or signal-based fault diagnosis, this case does not need to install additional sensors, and uses model-based fault diagnosis to check the product status, and because it does not require sensors, it is applicable to any field, so , this case should be an optimal solution.

The robot arm abnormality diagnosis system of the present invention comprises: at least one robot arm device for picking and placing a gripped object; a signal acquisition device electrically connected to the robot arm device for receiving the robot arm device Motor information moving on a specific trajectory; and an analysis and judgment device connected to the signal acquisition device to establish a diagnostic model based on the motor information moving on the specific trajectory, and obtain a health status based on the diagnostic model State residual, the diagnostic model is a nonlinear model; the signal acquisition device can receive a new motor information, and obtain a state residual to be measured according to the diagnostic model, and then compare the healthy state residual with the to-be-tested The measured state residuals are compared to determine whether the gripped object has an abnormal state.

More specifically, the diagnostic model is composed of a nonlinear module and a discrete linear transfer function.

More specifically, the content of the motor information can be defined as an input signal and an output signal, and the diagnostic model is established based on the input signal and the output signal, and the input signal and the output signal are both A nonlinear signal, wherein the input signal is a speed signal, and the output signal is a torque signal.

More specifically, the analysis and judgment device can establish more than one range interval according to the different multiples of the residual of the health state, and different range intervals can be defined as different state types, and can be based on where the residual of the state to be measured is located. A range interval to determine which state type the residual of the state to be measured belongs to.

More specifically, the analysis and judgment equipment has an analysis and judgment application program, and the analysis and judgment application program includes: a receiving unit for receiving the motor information of the mechanical arm equipment moving on the specific track; a model A building unit is connected to the receiving unit to build the diagnostic model based on the motor information; a diagnostic processing unit is connected to the receiving unit and the model building unit to obtain the health model based on the diagnostic model State residual and the residual of the state to be measured, and then compare the residual of the healthy state with the residual of the state to be measured to determine whether the clamped object has an abnormal state; and a state type definition unit, which is related to The diagnostic processing unit is connected, and can establish more than one range interval according to the different multiples of the health state residual, so as to define different state types according to different range intervals, and the diagnostic processing unit can be based on the residual state of the state to be measured Which range interval is located to determine which state type the residual of the state to be measured belongs to.

A method for diagnosing abnormality of a mechanical arm, the steps of which are: (1) A robot arm device used to pick and place a gripped object receives motor information of the robot arm device moving on a specific track; (2) Establish a diagnostic model based on the motor information moving on the specific trajectory, and obtain a health state residual, the diagnostic model is a nonlinear model; and (3) Receive a new motor information, and obtain a residual of the state to be tested according to the diagnostic model, and then compare the residual of the health state with the residual of the state to be measured to determine whether the gripped object has Abnormal state.

More specifically, the diagnostic model is composed of a nonlinear module and a discrete linear transfer function.

More specifically, the content of the motor information can be defined as an input signal and an output signal, and the diagnostic model is established based on the input signal and the output signal, and the input signal and the output signal are both A nonlinear signal, wherein the input signal is a speed signal, and the output signal is a torque signal.

More specifically, the diagnostic model can generate a signal estimate, and compare the motor information for building the diagnostic model with the signal estimate to generate the state of health residual, and the new motor information It can be compared with the estimated value of the signal to generate the residual of the state under test.

More specifically, it is possible to establish more than one range interval based on the different multiples of the residual of the health state, and different range intervals can be defined as different state types, and can be based on which range interval the residual of the state to be measured is located in , to determine which state type the residual of the state to be measured belongs to.

More specifically, the diagnostic model can be established based on the motor information of the robotic arm device gripping an object without a center of mass deviation, and the health state residual is obtained, and then the new motor information is based on the diagnostic model After obtaining the residual of the state to be tested, the residual of the state to be measured is compared with thresholds established by different ratios of the health state to determine whether the gripped object has a center of mass offset state.

More specifically, the diagnostic model can be established based on the motor information of the mechanical arm equipment gripping a non-vibrating item, and the health status residual is obtained, and then the new motor information is obtained according to the diagnostic model. To measure the residual of the state, first compare the residual of the healthy state with the frequency spectrum converted from the residual of the state to be measured, so as to determine whether the clamped object has a vibration state.

More specifically, the comparison can be made based on the vibration frequency of a vibration source and the converted spectrogram of the residual error of the state to be measured, so as to determine whether the vibration state is caused by the vibration source.

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 1, 2, 3 and 4, which are schematic diagrams of the overall architecture of the robot arm abnormality diagnosis system and its method, the simplified architecture diagram of the robotic arm equipment, the architecture diagram of the analysis and judgment equipment, and the architecture diagram of the analysis and judgment application program of the present invention. , as shown in FIG. 1 , the robotic arm abnormality diagnosis system includes at least one robotic arm device 1 , a signal acquisition device 2 and an analysis and judgment device 3 .

As shown in Figure 2, the mechanical arm device 1 is used to pick and place a gripped object, and the mechanical arm device 1 at least includes an electric motor 11, a servo motor (not shown in the figure), a motor driver 12 and A gripper 13, wherein the servo motor makes the gripper 13 operate, and drives the action of the mechanical arm device 1 through the motor driver 12.

The motor driver 12 is electrically connected to the signal acquisition device 2, and the signal acquisition device 2 is electrically connected to the robot arm device 1, and when the robot arm device 1 moves on a specific trajectory, the signal acquisition The motor information (such as speed signal and torque signal) of the motor driver 12, the signal extracted by the signal acquisition device 2 is a non-linear signal.

The motor information extracted by the signal acquisition device 2 defines an input signal and an output signal. The input signal is a speed signal, and the output signal is a torque signal. The signal acquisition device 2 is a user-configurable Programmable (user-programmable) FPGA, an embedded controller (CompactRIO controller) that can perform high-speed calculations, high-precision decoding, and processing complex operations such as filtering and differentiation at the same time.

When extracting the information of the motor driver 12, the environment, electronic components, mechanisms, etc. cause the signal to generate noise, and the noise will cause signal distortion. Therefore, a filter can be used to extract the signal of a specific frequency range and filter unnecessary Noise, the filter is lowpass filter, highpass filter, bandpass filter, bandstop filter, and the signal processing application of the filter can It is installed inside the signal acquisition device 2, an additional device is connected externally, or the analysis and judgment device 3 is used for signal processing.

The analysis and judgment device 3 can compare the signals under different conditions with the residuals obtained from the model, and then diagnose different state types according to the defined threshold (the state type may be a healthy state or an unhealthy state, and can even be determined according to the unhealthy state. subdivided into different categories); therefore, after the signal is captured and processed, the analysis and judgment device 3 can establish a diagnostic model based on the processed signal to detect faults, and the diagnostic model is a nonlinear model.

The diagnostic model can generate a signal estimate, and the motor information (the actual measurement value of the health state) used to build the diagnostic model is compared with the signal estimate to generate the health state residual, and the new The motor information (actual measurement of the state under test) can be compared with the signal estimate to generate the state under test residual.

As shown in Figure 3, the analysis and judgment device 3 includes at least one processor 31 and at least one computer-readable recording medium 32, and the computer-readable recording medium 32 stores at least one analysis and judgment application program 321, The computer-readable recording medium 32 further stores computer-readable instructions, which are used to run the analysis and judgment application program 321 when the computer-readable instructions are executed by the processors 31 .

As shown in Fig. 4, the analysis and judgment application program 321 includes: (1) A receiving unit 3211, used to receive the motor information of the robot arm device 1 moving on the specific track; (2) a model building unit 3212, which is connected to the receiving unit 3211, and is used to build the diagnostic model according to the motor information; (3) A diagnostic processing unit 3213, which is connected to the receiving unit 3211 and the model building unit 3212, is used to obtain the residual of the health status and the residual of the status to be tested through the diagnostic model, and then the residual of the healthy status The difference is compared with the residual of the state to be measured to determine whether the gripped object has an abnormal state; (4) A state type definition unit 3214, which can establish more than one range interval according to the different multiples of the health state residual, so as to define different state types according to different range intervals.

The abnormal diagnosis method of the robotic arm in this case is shown in Figure 5, and the steps are as follows: (1) A robotic arm device used to pick and place a gripped object is used to receive the motor information 501 of the robotic arm device moving on a specific track; (2) Establish a diagnostic model based on the motor information moving on the specific trajectory, and obtain a health state residual, the diagnostic model is a nonlinear model 502; and (3) Receive a new motor information, and obtain a residual of the state to be tested according to the diagnostic model, and then compare the residual of the health state with the residual of the state to be measured to determine whether the gripped object has Abnormal status 503.

Since it is a common method to use linear models to estimate nonlinear systems, but the systems and capabilities that linear models can estimate are limited, and the performance is not good when used for more complex nonlinear systems, so this case uses nonlinear models to estimate nonlinear systems. The diagnostic model used in this case is the Hammerstein-Wiener (HW, Hammerstein-Wiener) model. The HW model is composed of two static nonlinear modules and a linear module. The linear module is a discrete transfer function that represents the dynamic component, as shown in Section 6 As shown in the figure, the nonlinear module forms the HW model before and after the discrete linear transfer function (linear module);

(1)

(2) 其中

(3) n表示斷點數量，指定為整數；

表示欲進行線性插值之斷點；

表示指定斷點處之對應非線性值。 The nonlinear module of this case is generated by a piecewise linear equation. The piecewise linear equation uses a linear interpolation method to perform linear interpolation at the specified breakpoint (breakpoint). After the piecewise linearization, the nonlinear signal is in each The intervals are all linear functions, also known as affine functions. The input and output nonlinear modules are as described in formulas (1)~(3), f is the nonlinear input function, and g is the nonlinear output function.

(1)

(2) of which

(3) n represents the number of breakpoints, specified as an integer;

Indicates the breakpoint for linear interpolation;

Represents the corresponding nonlinear value at the specified breakpoint.

(4) 其中

(5)

(6) 𝐵(𝑞) 表示為零點多項式； 𝑛𝑏 表示為 𝐵(𝑞) 階數+1，等同 𝐵(𝑞) 的長度； 𝐹(𝑞)表示為極點多項式； 𝑛𝑓 表示為 𝐹(𝑞) 的階數； 𝑦(𝑡) 表示估計訊號； 𝑦̂(𝑡) 表示實際量測訊號。 The discrete linear transfer function is an output error (OE) model, as described in Figure 7 and formulas (4)~(6).

(4) of which

(5)

(6) 𝐵(𝑞) is expressed as a zero-point polynomial; 𝑛𝑏 is expressed as the order of 𝐵(𝑞) + 1, which is equivalent to the length of 𝐵(𝑞); 𝐹(𝑞) is expressed as a pole polynomial; 𝑛𝑓 is expressed as the order of 𝐹(𝑞) 𝑦(𝑡) means the estimated signal; 𝑦̂(𝑡) means the actual measured signal.

(7) zero:0.882, 0.9974±0.0061i, pole:0.7148, 0.989, 0.9865± 0.0542i

(8) zero:0, 0.9977, pole:-0.8554, 0.7268, 0.9999

(9) zero:0,0 pole:0.9983, 0.9993± 0.0018i (2)     第一應用例，將建立之模型，用以判斷被夾取物品是否具有質心偏移之問題發生，說明如下： (a)      本案利用健康狀態的訊號來建立診斷模型，但診斷模型預估出來的輸出訊號預估值與健康狀態的輸出訊號實際量測值還是有差異，故透過診斷處理單元3213將輸出訊號預估值與輸出訊號實際量測值比較之後，能得到一擬合率（Fit）及比較後的殘差，然而健康狀態下的殘差較小，但若是新的馬達資訊擷取進來，且又是非健康狀態，其殘差將會非常明顯； (b)     接著診斷處理單元3213利用四分位距辨識離群值，如第8圖所示，診斷時以第三軸殘差來定義不同質心位置之閥值（如表一所示），先以1.5倍標準差做為健康閥值檢測是否產生故障（其中標準差是指健康狀態的輸出訊號模型預估值與輸出訊號實際量測值比較所得的殘差），再以此閥值之倍數判斷狀態類型（不同倍數定義多個區間以代表不同質心偏移狀態，共分為五類，左偏 20mm、左偏 5mm、中心、右偏 5mm、右偏 20mm）； 第一軸 第二軸 第三軸 偏移距離 閥值 𝑞 ₁≥ Q ₁ 𝑞 ₂≥ Q ₂ 𝑞 ₃≥ Q ₃ 中心 0 𝑞 ₁＜ Q ₁ 𝑞 ₂＜ Q ₂ 𝑞 ₃＜ Q ₃≤𝑞 ₄ 左偏5mm 3 𝑞 ₁＜ Q ₁ 𝑞 ₂＜ Q ₂ 𝑞 ₄＜ Q ₃≤𝑞 ₅ 右偏5mm 4 𝑞 ₁＜ Q ₁ 𝑞 ₂＜ Q ₂ 𝑞 ₅＜ Q ₃≤𝑞 ₆ 左偏20mm 5 𝑞 ₁＜ Q ₁ 𝑞 ₂＜ Q ₂ 𝑞 ₆＜ Q ₃ 右偏20mm 6 表一 故障診斷表 Q ₁、Q ₂、Q ₃是各軸殘差， 𝑞 ₁~𝑞 ₃（1.5σ）是馬達三軸關節判斷工件質心是否偏移的健康閥值，若殘差同時超過𝑞 ₁、𝑞 ₂、𝑞 ₃，即表示出現異常，而𝑞 ₄（10 𝑞 ₃）、𝑞 ₅（80 𝑞 ₃）、𝑞 ₆（100 𝑞 ₃）分別表示不同偏移位置，對應倍率分別為 10、80、100； (c)      本實施例中，是利用標準差判斷離群值的方法進行故障診斷，將所有故障訊號視為離群值以偵測訊號是否異常，為了偵測微小故障，故將範圍定義為 1.5 倍標準差皆屬於健康狀態，診斷狀態類型之閥值根據偏移距離、方向以 1~100倍定義，根據偏移距離診斷 5mm 倍率較低，20mm 倍率較高，根據偏移方向診斷，質心往左偏時與馬達距離比往右偏時小，力臂較短故馬達產生的扭力較小，因此質心左移倍率較低，質心右移倍率較高； (d)     假設𝑞 ₁=0.5、𝑞 ₂=0.8、𝑞 ₃=1、𝑞 ₄=10、𝑞 ₅=80、𝑞 ₆=100，當移動於相同特定軌跡的條件下，將不同的馬達資訊擷取進來後，假設第一軸殘差與第二軸殘差都超過閥值時，而第三軸殘差（以小寫的e來代表新的馬達資訊之第三軸殘差）假設為以下幾個，參考表一，就能夠產生以下幾種判斷結果： (d1) e ₁=94，根據𝑞 ₁~𝑞 ₆，判斷e ₁之質心偏移狀態為左偏20mm； (d2) e ₂=6，根據𝑞 ₁~𝑞 ₆，判斷e ₂之質心偏移狀態為左偏5mm； (d3) e ₃=0.5，根據𝑞 ₁~𝑞 ₆，判斷e ₃之質心偏移狀態為中心； (d4) e ₄=68，根據𝑞 ₁~𝑞 ₆，判斷e ₄之質心偏移狀態為右偏5mm； (d5) e ₅=120，根據𝑞 ₁~𝑞 ₆，判斷e ₅之質心偏移狀態為右偏20mm。 (3)     第二應用例，透過上述建立之模型，診斷機械手臂運行間，被夾取物品是否產生異常振動（但亦能夠依據機械手臂夾取無振動的物品的馬達資訊重新建立一新的診斷模型）。 (a)      擷取各軸馬達資訊，利用此馬達資訊之速度及扭力訊號建立診斷模型； (b)     再用快速傅立葉轉換（Fast Fourier Transform）將殘差轉為頻域以獲得頻譜圖； (c)      本案是使用模型式搭配訊號進行比對，但若是將輸出訊號本身直接進行比對，而不經過模型，如第9A圖的無振動頻譜圖與第9B圖的有振動頻譜圖可知，有些訊號經過快速傅立葉轉換後，並沒有顯示振動頻域導致誤判，如此將明顯出錯。相較於本案，當經過模型比較後之殘差再轉換為頻譜圖（如第10A圖的無振動頻譜圖與第10B圖的有振動頻譜圖）後，則不會發生沒有顯示振動頻域之問題； (d)     將健康狀態下的殘差（健康狀態的輸出訊號模型預估值與輸出訊號實際量測值比較所得的殘差）所轉換的頻譜圖，與新的訊號的殘差（健康狀態的輸出訊號模型預估值與新的輸出訊號實際量測值比較所得的殘差）所轉換的頻譜圖進行比較，以判斷是否有突波發生，如第10B圖明顯有一突波，可判斷是異常振動； (e)      再將一振動來源（例如電動馬達11本身振動、其他裝置的馬達振動或是其他外來環境振動）的振動頻率與新的訊號的殘差（健康狀態的輸出訊號模型預估值與新的輸出訊號實際量測值比較所得的殘差）所轉換的頻譜圖進行比較，以確認振動是否為該振動來源的振動頻率，以下假設可能振動來源是其他裝置的馬達，說明如下： (e1) 為了確認該突波是否為其他裝置的馬達導致的異常振動，則依據其他裝置的馬達本身資訊轉換出一頻率（例如馬達電壓1V的轉速為6000rpm，並再依據該數據轉換出轉速與頻率的換算比例或換算公式）後，則能夠確認其他裝置的馬達本身的振動頻率； (e2) 再與頻譜圖內突波之頻率比較，則能夠確定異常振動是否為其他裝置的馬達本身產生（以本應用例來看，機械手臂運行軌跡時轉速約為 5400rpm，換算成頻率約為90Hz，而第10B圖中的突波頻率振動範圍約為85Hz~90Hz，由此能夠明顯判斷異常振動是由其他裝置的馬達本身產生）。 In this case, two examples are given for implementation, and the description is as follows: (1) Establish a diagnostic model (a) Set a certain state of a gripped object without abnormality as a healthy state (for example, the center of mass at the center is not offset), and the robotic arm Pick up the item in a healthy state to move on a specific trajectory (such as circular motion), and capture the motor information of each axis, and then use the low-pass filter to perform signal post-processing on the speed and torque signals of this information (the input signal is a speed signal, the output signal is a torque signal). (b) Then use the motor information in a certain healthy state to establish a diagnostic model. The following is an example of the three-axis HW model transfer function of the established diagnostic model, as described in the following formulas (7)~(9), and the formulas are as follows:

(7) zero:0.882, 0.9974±0.0061i, pole:0.7148, 0.989, 0.9865±0.0542i

(8) zero:0, 0.9977, pole:-0.8554, 0.7268, 0.9999

(9) zero:0,0 pole:0.9983, 0.9993± 0.0018i (2) In the first application example, the model will be established to determine whether the problem of the center of mass offset of the gripped object occurs. The description is as follows: ( a) In this case, the signal of the healthy state is used to establish a diagnostic model, but there is still a difference between the estimated value of the output signal estimated by the diagnostic model and the actual measured value of the output signal of the healthy state, so the output signal is estimated by the diagnostic processing unit 3213 After the value is compared with the actual measured value of the output signal, a fitting rate (Fit) and a residual error after comparison can be obtained. However, the residual error in a healthy state is small, but if new motor information is extracted, and it is very (b) Then the diagnosis processing unit 3213 uses the interquartile range to identify outliers, as shown in Figure 8, the third axis residuals are used to define different centroid positions during diagnosis The threshold value (as shown in Table 1), first use 1.5 times the standard deviation as the health threshold to detect whether a fault occurs (the standard deviation refers to the comparison between the output signal model prediction value of the healthy state and the actual measurement value of the output signal residual), and then judge the state type by multiples of this threshold value (different multiples define multiple intervals to represent different centroid offset states, which are divided into five categories, 20mm to the left, 5mm to the left, 5mm to the center, and 5mm to the right , 20mm to the right); first axis second axis third axis offset distance Threshold 𝑞 ₁ ≥ Q ₁ 𝑞 ₂ ≥ Q ₂ 𝑞 ₃ ≥ Q ₃ center 0 𝑞 ₁ < Q ₁ 𝑞 ₂ < Q ₂ 𝑞 ₃ < Q ₃ ≤ 𝑞 ₄ 5mm to the left 3 𝑞 ₁ < Q ₁ 𝑞 ₂ < Q ₂ 𝑞 ₄ < Q ₃ ≤ 𝑞 ₅ 5mm to the right 4 𝑞 ₁ < Q ₁ 𝑞 ₂ < Q ₂ 𝑞 ₅ < Q ₃ ≤ 𝑞 ₆ 20mm to the left 5 𝑞 ₁ < Q ₁ 𝑞 ₂ < Q ₂ 𝑞 ₆ < Q ₃ 20mm to the right 6 Table 1 Fault diagnosis tables Q ₁ , Q ₂ , and Q ₃ are the residuals of each axis. 𝑞 ₁ ~ 𝑞 ₃ (1.5σ) is the health threshold for the three-axis joints of the motor to judge whether the center of mass of the workpiece is offset. If the residuals exceed 𝑞 ₁ , 𝑞 ₂ , 𝑞 ₃ , which means that there is an abnormality, and 𝑞 ₄ (10 𝑞 ₃ ), 𝑞 ₅ (80 𝑞 ₃ ), 𝑞 ₆ (100 𝑞 ₃ ) respectively represent different offset positions, and the corresponding magnification is 10 , 80, 100; (c) In this embodiment, the method of judging outliers by standard deviation is used for fault diagnosis, and all fault signals are regarded as outliers to detect whether the signals are abnormal. In order to detect small faults, the The range is defined as 1.5 times the standard deviation, which belongs to the healthy state. The threshold value of the diagnosis status type is defined by 1~100 times according to the offset distance and direction. The magnification of 5mm is low according to the offset distance, and the magnification of 20mm is high. According to the offset Direction diagnosis, when the center of mass is shifted to the left, the distance between the motor and the motor is smaller than when it is shifted to the right, and the moment arm is shorter, so the torque generated by the motor is smaller, so the center of mass shifts to the left and the center of mass moves to the right. (d ) Assuming 𝑞 ₁ =0.5, 𝑞 ₂ =0.8, 𝑞 ₃ =1, 𝑞 ₄ =10, 𝑞 ₅ =80, 𝑞 ₆ =100, when moving on the same specific trajectory, different motor information is captured Finally, assuming that both the residual error of the first axis and the residual error of the second axis exceed the threshold value, and the residual error of the third axis (the third axis residual error of the new motor information is represented by the lowercase e) is assumed to be the following, Referring to Table 1, the following judgment results can be produced: (d1) e ₁ =94, according to 𝑞 ₁ ~𝑞 ₆ , it is judged that the center of mass of e ₁ is offset by 20mm to the left; (d2) e ₂ =6, According to 𝑞 ₁ ~ 𝑞 ₆ , it is judged that the center of mass offset of e ₂ is 5mm to the left; (d3) e ₃ =0.5, according to 𝑞 ₁ ~ 𝑞 ₆ , it is judged that the center of mass offset of e ₃ is the center; (d4 ) e ₄ =68, according to 𝑞 ₁ ~𝑞 ₆ , it is judged that the centroid shift of e ₄ is 5mm to the right; (d5) e ₅ =120, according to 𝑞 ₁ ~𝑞 ₆ , it is judged that the centroid shift of e ₅ The state is 20mm to the right. (3) The second application example, through the model established above, diagnoses whether the gripped object has abnormal vibration during the operation of the robotic arm (but it is also possible to recreate a new diagnosis based on the motor information of the gripped object without vibration by the robotic arm) Model). (a) Extract the motor information of each axis, and use the speed and torque signals of the motor information to establish a diagnostic model; (b) Use Fast Fourier Transform (Fast Fourier Transform) to convert the residual into the frequency domain to obtain a spectrogram; (c ) In this case, the model is used to match the signal for comparison, but if the output signal itself is directly compared without going through the model, as shown in the no-vibration spectrum diagram in Figure 9A and the vibration spectrum diagram in Figure 9B, some signals After the fast Fourier transform, there is no display of the vibration frequency domain leading to misjudgment, so it will be obviously wrong. Compared with this case, when the residuals after model comparison are converted into spectrograms (such as the no-vibration spectrogram in Fig. 10A and the vibrating spectrogram in Fig. 10B), there will be no failure to display the vibration frequency domain. Question; (d) The residual error in the healthy state (the residual error obtained by comparing the estimated value of the output signal model in the healthy state with the actual measured value of the output signal) is transformed into a spectrogram, and the residual error of the new signal (healthy The estimated value of the output signal model of the state is compared with the residual error obtained by comparing the actual measured value of the new output signal) to compare the converted spectrogram to determine whether there is a surge. (e) Then compare the vibration frequency of a vibration source (such as the vibration of the electric motor 11 itself, the vibration of the motor of other devices, or other external environment vibration) with the residual error of the new signal (the output signal model prediction of the healthy state The residual error obtained by comparing the estimated value with the actual measured value of the new output signal) is compared to the converted spectrogram to confirm whether the vibration is the vibration frequency of the vibration source. The following assumes that the possible vibration source is the motor of other devices, as described below : (e1) In order to confirm whether the surge is abnormal vibration caused by the motor of other devices, convert a frequency based on the information of the motor itself of other devices (for example, the speed of the motor voltage 1V is 6000rpm, and then convert the speed based on this data (conversion ratio or conversion formula) to the frequency), the vibration frequency of the motor itself of other devices can be confirmed; (e2) and then compared with the frequency of the surge in the frequency spectrum, it can be determined whether the abnormal vibration is generated by the motor itself of other devices (From the perspective of this application example, the rotational speed of the robot arm is about 5400rpm, converted to a frequency of about 90Hz, and the vibration range of the surge frequency in Figure 10B is about 85Hz~90Hz, so it can be clearly judged whether the abnormal vibration is generated by the motor itself in other devices).

This case can be applied to the operation path of the fixed-point or non-fixed-point robotic arm for the gripped object. If it is to be applied to the non-fixed-point gripped object, it can be used with image recognition to determine the object center of the gripped object.

The advantages of the robotic arm abnormality diagnosis system and method provided by the present invention are as follows when compared with other conventional technologies: 1. There is no need to install additional sensors in this case, and the model-based fault diagnosis is used to check the product status. Since no sensors are required, it is suitable for any field. 2. This case is used to detect the internal state of the gripped object. Due to the unstable current signal, the ability to detect small abnormalities is not good. Therefore, a nonlinear model based on signals established in this case will be able to diagnose obvious abnormalities at the same time And tiny anomalies that cannot be detected by current signals. 3. The present invention uses the dynamic equation of the robotic arm as a model, matches the input and output signals, and sets corresponding thresholds for different faults, and judges whether a fault occurs based on whether the actual measurement exceeds the set threshold. As long as an effective valve can be obtained value can accurately diagnose the fault.

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: Mechanical arm equipment

11: Electric motor

12: Motor driver

13: Fixture

2: Signal acquisition equipment

3: Analysis and judgment equipment

31: Processor

32: Computer-readable recording media

321: Analysis and judgment application

3211: receiving unit

3212:Model building unit

3213: Diagnostic processing unit

3214: State type definition unit

[Fig. 1] is a schematic diagram of the overall structure of the robot arm abnormality diagnosis system and its method of the present invention. [Fig. 2] is a schematic diagram of the simple structure of the robotic arm equipment of the robotic arm abnormality diagnosis system and method of the present invention. [Fig. 3] is a schematic diagram of the structure of the analysis and judgment equipment of the robot arm abnormality diagnosis system and its method of the present invention. [Fig. 4] is a schematic diagram of the structure of the analysis and judgment application program of the robot arm abnormality diagnosis system and its method of the present invention. [Fig. 5] is a flow chart of the steps of the robot arm abnormality diagnosis system and its method of the present invention. [Fig. 6] It is a schematic diagram of the HW (Hammerstein-Wiener) model of the abnormal diagnosis system and method of the robotic arm of the present invention. [Fig. 7] is a schematic diagram of the output error model of the robot arm abnormality diagnosis system and its method of the present invention. [Fig. 8] is a schematic diagram of the third-axis trajectory residual box of the robot arm abnormality diagnosis system and its method of the present invention. [Fig. 9A] It is the signal-type non-vibration signal frequency of the abnormal diagnosis system and method of the mechanical arm of the present invention. Spectrum diagram. [Figure 9B] It is the signal type of the abnormal diagnosis system and method of the mechanical arm of the present invention with vibration signal frequency Spectrum diagram. [Fig. 10A] It is a model-type non-vibration signal frequency of the robot arm abnormality diagnosis system and its method of the present invention Spectrum diagram. [Fig. 10B] It is a model of the abnormal diagnosis system and method of the mechanical arm of the present invention with vibration signal frequency Spectrum diagram.

1: Mechanical arm equipment

2: Signal acquisition equipment

3: Analysis and judgment equipment

Claims (6)

A robotic arm abnormality diagnosis system includes: at least one robotic arm device, used to pick and place a gripped object; a signal acquisition device, electrically connected to the robotic arm device, for receiving the mechanical arm device in the Motor information moving on a specific trajectory; and an analysis and judgment device connected to the signal acquisition device to establish a diagnostic model based on the motor information moving on the specific trajectory, and obtain a health status based on the diagnostic model The residual, the diagnostic model is a nonlinear model; the signal acquisition device can receive a new motor information, and obtain a residual of the state to be measured according to the diagnostic model, and then compare the residual of the healthy state with the residual of the state to be measured The state residual is compared to determine whether the gripped item has an abnormal state. The analysis and judgment equipment can establish more than one range interval according to the different ratios of the health state residual, and different range intervals can be defined as different states. type, and according to which range interval the residual of the state to be measured is located in, it can be judged which type of state the residual of the state to be measured belongs to. The abnormal diagnosis system of the mechanical arm as described in claim 1, wherein an input signal and an output signal can be defined from the content of the motor information, and the diagnostic model is established based on the input signal and the output signal, and the input signal And the output signal is a nonlinear signal, wherein the input signal is a speed signal, and the output signal is a torque signal. The mechanical arm abnormality diagnosis system as described in claim 1, wherein the analysis and judgment equipment has an analysis and judgment application program, and the analysis and judgment application program includes: a receiving unit, used to receive the mechanical arm equipment in the specific Motor information for trajectory movement; A model building unit is connected to the receiving unit to build the diagnostic model based on the motor information; a diagnostic processing unit is connected to the receiving unit and the model building unit to obtain the diagnostic model based on the The residual of the healthy state and the residual of the state to be tested are compared with the residual of the healthy state and the residual of the state to be measured to determine whether the clamped object has an abnormal state; and a state type definition unit, The system is connected with the diagnostic processing unit, and can establish more than one range interval according to the different multiples of the health state residual, so as to define different state types according to different range intervals, and the diagnostic processing unit can be based on the state to be tested Which range interval the residual is located in is used to determine which state type the residual of the state to be measured belongs to. A method for diagnosing the abnormality of a robotic arm, the steps of which are: a robotic arm device used to pick and place a gripped object, receiving motor information of the robotic arm device moving on a specific track; according to the motor moving on the specific track information to establish a diagnostic model, and obtain a health status residual, the diagnostic model is a nonlinear model; The state residual is compared with the state residual to be tested to determine whether the gripped object has an abnormal state. The diagnostic model can generate a signal estimate, and the motor information for establishing the diagnostic model and the signal The estimated value is compared to generate the state-of-health residual, and the new motor information can be compared with the signal estimate to generate the state-to-be-measured residual. The method for diagnosing the abnormality of the mechanical arm as claimed in item 4, wherein the diagnosis model is composed of a nonlinear module and a discrete linear transfer function. The method for diagnosing the abnormality of the robotic arm as described in Claim 4, wherein the health can be based on the Different magnifications of state residuals establish more than one range interval, and different range intervals can be defined as different state types, and can be judged according to which range interval the state residual is located in to determine whether the state residual to be measured belongs to which state type.

Images