Production line monitoring system and production line monitoring method

Abstract

The invention discloses a production line monitoring system and a production line monitoring method. The production line monitoring system comprises a monitoring device and a screw locking device. The screw locking device is electrically connected with the monitoring device and is used for locking the target screw to generate locking data; the monitoring device performs numerical normalization operation on the locking data, and inputs the normalized locking data into a neural network model, so that the neural network model outputs feature vector data according to the normalized locking data; and the monitoring device inputs the feature vector data into an One Class Support Vector Machine (OCSVM) so as to judge whether the target screw is normally locked or abnormally locked according to an interpretation result of the one Class Support Vector Machine. According to the production line monitoring system and the production line monitoring method, whether the currently locked screw is normally locked or not can be automatically monitored.

Description

Technical Field

[0001] The present invention relates to a monitoring technology, and in particular to a production line monitoring system and a production line monitoring method. Background Art

[0002] At the screw locking station in the current production line, the production line personnel cannot immediately know the screw locking results, and need to wait until the subsequent vacuum test station to know whether the screw locking result is normal or abnormal. Therefore, there is no objective standard at the station to determine the quality of the locking. Or, after the screw is locked, the screw locking result can only be known through manual inspection.

[0003] Therefore, it is necessary to provide a production line monitoring system and a production line monitoring method to solve the above problems. Summary of the invention

[0004] The present invention provides a production line monitoring system and a production line monitoring method, which can automatically monitor whether the currently locked screws are normally locked.

[0005] The production line monitoring system of the present invention includes a monitoring device and a screw locking device. The screw locking device is electrically connected to the monitoring device and is used to lock a target screw to generate locking data. The monitoring device performs a numerical normalization operation on the locking data and inputs the normalized locking data into a neural network model so that the neural network model outputs feature vector data according to the normalized locking data. The monitoring device inputs the feature vector data into a single-class support vector machine to determine whether the target screw is normally locked or abnormally locked according to the judgment result of the single-class support vector machine.

[0006] The production line monitoring method of the present invention includes the following steps: locking a target screw by a screw locking device to generate locking data; performing numerical normalization operation on the locking data by a monitoring device; inputting the normalized locking data into a neural network model by the monitoring device, so that the neural network model outputs feature vector data according to the normalized locking data; and inputting the feature vector data into a single-class support vector machine by the monitoring device, so as to judge whether the target screw is normally locked or abnormally locked according to the judgment result of the single-class support vector machine.

[0007] Based on the above, the production line monitoring system and the production line monitoring method of the present invention can automatically analyze the locking data through the neural network model and the one-class support vector machine to determine whether the currently locked target screw is normally locked or abnormally locked.

[0008] In order to make the above features and advantages of the present invention more clearly understood, embodiments are given below with reference to the accompanying drawings for detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 is a schematic diagram of a production line monitoring system according to an embodiment of the present invention.

[0010] Figure 2 is a flow chart of a production line monitoring method according to an embodiment of the present invention.

[0011] Figure 3 is a schematic diagram of a production line monitoring system according to an embodiment of the present invention.

[0012] FIG. 4A to FIG. 4C Schematic diagram of various screw fastening examples of the present invention.

[0013] FIG. 5A to FIG. 5C It is a schematic diagram of locking data of various normal locking examples of the present invention.

[0014] FIG. 5D to FIG. 5F It is a schematic diagram of locking data of various abnormal locking examples of the present invention.

[0015] Figure 6 FIG. 4 is a schematic diagram of analyzing locking data according to an embodiment of the present invention.

[0016] Fig. 7A It is a schematic diagram of locking result information of normal locking according to an embodiment of the present invention.

[0017] Figure 7B It is a schematic diagram of locking result information of abnormal locking according to an embodiment of the present invention.

[0018] Figure 8 It is a schematic diagram of model training according to an embodiment of the present invention.

[0019] Fig. 9A as well as Fig. 9B is a flow chart of a production line monitoring method according to an embodiment of the present invention.

[0020] Main component symbols:

[0021] 100, 300 production line monitoring system

[0022] 110, 310 monitoring device

[0023] 120, 320 screw locking device

[0024] 200, 400 target screws

[0025] 311 processor

[0026] 312 Storage Devices

[0027] 313 Display

[0028] 3121 Process Control Module

[0029] 3122 Front Display Module

[0030] 3123 Data Storage Module

[0031] 3124 Neural Network Model

[0032] 3125, 630 Single-class support vector machine

[0033] 330 Scanning Device

[0034] 340 fixture

[0035] 411, 412, 413 Target screws

[0036] 421, 422, 423 Lock objects

[0037] 424 Foreign matter

[0038] 610 Normalized multi-stroke torque values

[0039] 620 Autoencoder Model

[0040] 621 Embedding Layer

[0041] 631 Decision-making scope

[0042] 710, 720 User Interface

[0043] 711, 721 device information

[0044] 712, 722 Lock information

[0045] 713, 723 Lock result information

[0046] 800 Model training servers

[0047] 801_1~801_N Test lock data

[0048] 802_1~802_N Feature vector training data

[0049] 810 Data conversion module

[0050] 820 Model Training Module

[0051] Steps S210~S240, S901~S923 DETAILED DESCRIPTION

[0052] Some embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The referenced element symbols in the following description will be regarded as the same or similar elements when the same element symbols appear in different drawings. These embodiments are only part of the present invention and do not disclose all possible implementation methods of the present invention. More specifically, these embodiments are only examples of the devices and methods within the scope of the patent application of the present invention.

[0053] Figure 1 Schematic diagram of a production line monitoring system according to an embodiment of the present invention. Figure 1 , the production line monitoring system 100 includes a monitoring device 110 and a screw locking device 120. The monitoring device 110 is electrically connected to the screw locking device 120. In the present embodiment, the screw locking device 120 is used to lock the target screw 200 to generate corresponding locking data. The screw locking device 120 can provide the locking data to the monitoring device 110, so that the monitoring device 110 can automatically determine whether the target screw 200 is normally locked or abnormally locked by processing and analyzing the locking data. In the present embodiment, the monitoring device 110 can be, for example, a production line computer, and is used to monitor and control the relevant equipment and work results of the station for performing screw locking work in the production line.

[0054] Figure 2 is a flow chart of a production line monitoring method according to an embodiment of the present invention. Figure 1 as well as Figure 2 , the production line monitoring system 100 may execute steps S210 to S240. In step S210, the screw locking device 120 locks the target screw 200 to generate locking data. In step S220, the monitoring device 110 may perform a value normalization operation on the locking data. For example, the locking data may include multiple values ​​in Table 1 below.

[0055]

[0056]

[0057] Table 1

[0058] As shown in Table 1, the screw locking device 120 can record the rotation speed (rpm), torque (kgf.cm), and number of locking turns of the screwdriver of the screw locking device 120 that change with time during a locking period in the process of locking the target screw 200. The locking data may include multiple rotation speed values, multiple torque values, and multiple locking turns corresponding to the screw locking device 120 locking the target screw 200 during the locking period. In this regard, the monitoring device 110 can, for example, read the multiple torque values ​​generated by the screw locking device 120 locking the target screw 200 during the locking period in Table 1, and perform a numerical normalization operation on the multiple torque values ​​using the interval scaling method of the following formula (1), but the present invention is not limited to this.

[0059]

[0060] In formula (1), Tmin may represent the minimum torque value among the multiple torque values. Tmax may represent the maximum torque value among the multiple torque values. X represents the torque value to be converted. X' represents the normalized torque value after conversion. Therefore, the monitoring device 110 may, for example, generate the normalized torque value as shown in Table 2 below.

[0061] Normalized torque value 1 1 0.235 0.294 0.529 0.588 0.529 0.529 0.558 0.235 0.118 0 0.235 0.941 0.882 0.118

[0062] Table 2

[0063] In step S230, the monitoring device 110 may input the normalized locking data into the neural network model so that the neural network model (NN) outputs feature vector data according to the normalized locking data. In step S240, the monitoring device 110 may input the feature vector data into the support vector machine model (SVM) to determine whether the target screw 200 is normally locked or abnormally locked according to the judgment result of the support vector machine model. In one embodiment, the support vector machine model may be a one-class support vector machine (OCSVM) or other models that can distinguish a single decision boundary for multiple data. In this embodiment, the neural network model and the one-class support vector machine are respectively trained models. Therefore, the production line monitoring system 100 of this embodiment can automatically and instantly determine whether the target screw 200 is normally locked or abnormally locked.

[0064] In one embodiment, the training data of the single-class support vector machine has only one category. The single-class support vector machine learns a decision boundary through the characteristics of normal samples (for example, the decision boundary can be applied as Figure 6The decision range 631) is then used to determine whether the new data point is similar to the training data. If it exceeds the boundary, it is considered abnormal. Since the single-class support vector machine is a known algorithm, it will not be described here.

[0065] Figure 3 Schematic diagram of a production line monitoring system according to an embodiment of the present invention. Figure 3 The production line monitoring system 300 includes a monitoring device 310, a screw locking device 320, a code scanning device 330, and a fixture 340. The monitoring device 310 includes a processor 311, a storage device 312, and a display 313. The processor 311 is electrically connected to the storage device 312, the display 313, the screw locking device 320, the code scanning device 330, and the fixture 340. The storage device 312 can store a process control module 3121, a front-end display module 3122, a data storage module 3123, a neural network model 3124, and a single-class support vector machine 3125. In one embodiment, the process control module 3121, the front-end display module 3122, the data storage module 3123, the neural network model 3124, and the single-class support vector machine 3125 can be implemented by circuits, firmware, or software.

[0066] In this embodiment, the processor 311 may be, for example, a central processing unit (CPU), a graphics processing unit (GPU), or other programmable general-purpose or special-purpose microprocessor, a digital signal processor (DSP), a programmable controller, an application specific integrated circuit (ASIC), a programmable logic device (PLD), other similar processing devices, or a combination of these devices.

[0067] The storage device 312 may include, for example, a dynamic random access memory (DRAM), a flash memory, or a non-volatile random access memory (NVRAM). The storage device 312 may be suitable for storing the relevant programs and algorithms of the process control module 3121, the front-end display module 3122, the data storage module 3123, the neural network model 3124, and the single-class support vector machine 3125, and may also store relevant locking data and other relevant programs and algorithms for implementing the production line monitoring method of the present invention.

[0068] In this embodiment, before the screw locking operation is performed, the product will be placed on the fixture 340, and the monitoring device 310 can first scan the barcode of the product through the barcode scanning device 330 to obtain the product serial number. The monitoring device 310 can read the corresponding driving information (such as the driving voltage and driving instructions of the screw locking device 320), and obtain the locking time and locking data. In this embodiment, the processor 311 can execute the data storage module 3123 to store the product serial number, locking time and locking data of the target screw 400. In one embodiment, the locking object can be fixedly set on the fixture 340, and the screw locking device 320 will lock the target screw 400 towards the locking object to generate corresponding locking data.

[0069] First, it should be noted that the target screw 400 may have a variety of locking conditions. FIG. 4A to FIG. 4C For example, FIG. 4A to FIG. 4C Schematic diagram of various screw fastening examples of the present invention. Figure 4A This is the normal screw fastening operation completed state. FIG. 4B to FIG. 4C It is an abnormal operation state of screw locking. More specifically, if Figure 4A As shown, the target screw 411 may be normally fastened to the screw hole of the fastening object 421. Figure 4B As shown, the target screw 412 may be broken and cannot be normally fastened to the screw hole of the fastening object 422. Figure 4C As shown, there may be a foreign object 424 in the screw hole of the locking object 423, which causes the target screw 413 to be unable to be normally locked into the screw hole of the locking object 423. However, the present invention is not limited to the specific circumstances of normal locking and abnormal locking.

[0070] by FIG. 5A to FIG. 5C For example, FIG. 5A to FIG. 5C is a schematic diagram of locking data of various normal locking examples of the present invention. When the target screw is normally locked into the screw hole of the locking object, the screw locking device 320 can obtain, for example, FIG. 5A to FIG. 5C It is worth noting that in the case of normal locking, since the target screw and / or locking object for each locking may have slight structural differences or different locking conditions, the screw locking device 320 may obtain the following torque data: FIG. 5A to FIG. 5C The different torque changes corresponding to the normal locking conditions are shown in FIG. 1 , wherein the locking time length refers to the time for screw locking. FIG. 5A to FIG. 5C As shown, when the screw locking device 320 detects that the torque value is 2.1 (kgf.cm), the screw locking device 320 may determine that the locking is completed and stop locking.

[0071] by FIG. 5D to FIG. 5F For example, FIG. 5D to FIG. 5F is a schematic diagram of locking data of various abnormal locking examples of the present invention. When the target screw is abnormally locked into the screw hole of the locking object, the screw locking device 320 can obtain, for example, FIG. 5D to FIG. 5F The torque data is shown in FIG. 1 , but the present invention is not limited thereto. Figure 5D As shown, when the target screw Figure 4B or Figure 4C When the abnormal locking condition occurs, the screw locking device 320 can obtain Figure 5D The torque value changes shown in the figure may be due to premature locking or foreign matter causing the maximum torque value to occur earlier. Figure 5E As shown, when the target screw is, for example, skewed in the locking angle during the locking process or the locking operation is not performed continuously, the torque value is higher in the early stage (for example, 150-200ms) compared to the normal locking state, or the torque value approaches 2.1 (kgf.cm) earlier than the normal locking state. Fig. 5F As shown, when the target screw is not actually screwed in due to, for example, the head or thread slipping occurs during the screwing process, the torque value has not been able to reach the maximum torque value. FIG. 5D to FIG. 5F The abnormal locking waveform and FIG. 5A to FIG. 5C The normal locking waveform has obvious differences.

[0072] Refer to Figure 3 In this embodiment, the processor 311 can normalize the continuous multiple torque values ​​in the locking data to generate the normalized multiple torque values. Then, the processor 311 can execute the process control module 3121 to read the neural network model 3124 and the one-class support vector machine 3125. The processor 311 can process and analyze the locking data through the neural network model 3124 and the one-class support vector machine 3125. Figure 6 For example, Figure 6 FIG. 3 is a schematic diagram of analyzing the lock data according to an embodiment of the present invention. The neural network model 3124 may be as follows: Figure 6 An auto-encoder model (auto-encoder) 620 is shown.

[0073] The autoencoder model 620 can be used to perform data compression on the normalized multiple torque values ​​610, and generate corresponding feature vector data through the embedding layer 621 of the autoencoder model 620. For example, the autoencoder model 620 can compress 400 torque values ​​into 64-dimensional feature parameters and extract the corresponding feature vector data from the feature layer. In this regard, the autoencoder model 620 can effectively reduce the amount of data input to the single-class support vector machine 3125. The single-class support vector machine 3125 can be as follows Figure 6 The single-class support vector machine 630 shown in FIG. The single-class support vector machine 630 can determine whether the target screw is normally locked or abnormally locked according to whether the coordinate point of the feature vector data in the feature space is located within the decision range 631. Figure 6 As shown, if the feature vector data is represented by, for example, coordinate point 632 in the feature space, since coordinate point 632 is within decision range 631, the current locking result can be determined to be normal locking. On the contrary, if the feature vector data is represented by, for example, coordinate point 633 in the feature space, since coordinate point 633 is outside decision range 631, the current locking result can be determined to be abnormal locking.

[0074] Refer to Figure 3 In this embodiment, in response to the single-class support vector machine 3125 determining that the target screw 400 is normally locked, the process control module 3121 can continue to perform the locking operation and locking judgment of the next target screw. On the contrary, in response to the single-class support vector machine 3125 determining that the target screw 400 is abnormally locked, the process control module 3121 can suspend the operation of the screw locking device 320. In addition, the processor 311 can also execute the front-end display module 3122 to display the locking result information on the user interface through the display 313 according to whether the target screw 400 is normally locked or abnormally locked.

[0075] by Fig. 7A For example, Fig. 7A FIG. 3 is a schematic diagram of the locking result information of a normal locking according to an embodiment of the present invention. The display 313 may display the following information: Fig. 7AThe user interface 710 includes display device information 711, locking information 712, and locking result information 713. The device information 711 may include information such as the model, line, site, and mode of the screw locking device 320, but the present invention is not limited thereto. The locking information 712 may include, for example, the number of screws produced, the number of good products, the screw locking yield, the whole machine locking yield, the current product serial number, the standard number of screws, the number of screws locked, the preset threshold (torque), the real-time status, the locking result, the locking angle, and the locking torque, but the present invention is not limited thereto. The locking result information 713 may be presented as a smiley face icon to effectively inform the user that the current locking result is normal locking, but the present invention is not limited thereto. In one embodiment, the locking result information 713 may also be displayed in an enlarged manner in the user interface 710, so that the person who is lowering his head to lock the screws can see that the locking is normal (for example, an enlarged smiley face) as soon as he looks up, thereby effectively prompting the user.

[0076] by Figure 7B For example, Figure 7B FIG. 3 is a schematic diagram of the locking result information of abnormal locking according to an embodiment of the present invention. The display 313 may display the following information: Figure 7B The user interface 720 includes display device information 721, locking information 722, and locking result information 723. The locking result information 723 can be presented as a crying face icon and / or sound effect to effectively notify the user that the current locking result is abnormal locking, but the present invention is not limited to this. In one embodiment, the locking result information 723 is displayed in an enlarged manner in the user interface 720, so that the person who lowers his head to lock the screws can see the locking abnormality (for example, an enlarged crying face) as soon as he looks up. In one embodiment, the crying face icon can be further accompanied by a sound effect to effectively remind the user of the locking abnormality.

[0077] In one embodiment, the monitoring device 310 can also operate in an offline mode, without executing the neural network model 3124 and the single-class support vector machine 3125. The offline mode can be applied to new product modeling and / or when there is an abnormality in the production line. In one embodiment, in the offline mode, the monitoring device 310 can still continue the locking operation, and the monitoring device 310 can display the display 313 such as Fig. 7A as well as Figure 7B The positions of the user interface 710, the locking result information 713, and the locking result information 723 of the user interface 720 are changed to display the torque change curve of the current locking object in real time (for example, FIG. 5A to FIG. 5F As shown), production line personnel can watch the real-time monitoring screen. In addition, Fig. 7A as well as Figure 7BThe user interface 710 and the user interface 720 may also display an additional storage icon. When the production line personnel determine that the locking result of the current locking object is abnormal through the torque change curve displayed on the user interface 710 and the user interface 720, the production line personnel can click the storage icon to save the relevant locking data and the corresponding torque change curve to the folder. In this way, the monitoring device 310 can still perform the locking action.

[0078] Figure 8 FIG. 1 is a schematic diagram of model training according to an embodiment of the present invention. Figure 8 , the single-class support vector machine 3125 can be pre-trained by the model training server 800. In this embodiment, the model training server 800 may include a data conversion module 810 and a model training module 820. The model training server 800 may have a processor and a storage device as described above, and may store and execute the data conversion module 810 and the model training module 820. The model training server 800 may pre-train the single-class support vector machine 3125, and load the trained single-class support vector machine 3125 into the above-mentioned Figure 1 and Figure 3 The monitoring device 110 and the monitoring device 310 described in the embodiments.

[0079] Specifically, the model training server 800 can obtain multiple test locking data 801_1~801_N generated by normal locking of multiple test screws, where N is a positive integer. The test locking data 801_1~801_N can be, for example, multiple data sets each having multiple torque values. The data conversion module 810 can perform a numerical normalization operation on the test locking data 801_1~801_N to generate a normalized multiple test locking data. The data conversion module 810 can convert the multiple torque values ​​into a numerical value between 0 and 1, and input them into the neural network model as a feature parameter to effectively eliminate the data differences between the values ​​due to units. The data conversion module 810 can, for example, execute the autoencoder model as described above to generate multiple feature vector training data 802_1~802_N through the embedding layer of the autoencoder model, where N is a positive integer. Next, the model training module 820 may train the single-class support vector machine 3125 according to the feature vector training data 802_1 ˜ 802_N.

[0080] The model training module 820 can train the single-class support vector machine 3125 according to the feature vector training data 802_1-802_N and the objective function of the following formula (2) and formula (3). In the following formula (2) and formula (3), w is the normal vector of the hyperplane. m is the number of observations. φ(x i ) is the data point x of the training data i The function mapped to the hyperplane. v is the upper limit of the proportion of outliers in all training samples and the lower limit of the proportion of support vectors in all samples. ρ is the threshold and is used to represent the distance between the hyperplane and the origin. ξ i is the slack variable. To this end, the user can set v as the ratio of the sample boundary judgment for normal locking to draw a Figure 6 Decision range 631 is shown.

[0081]

[0082] subject to: <w,φ(x i )>≥ρ-ξ i ,ξ i ≥0......Formula (3)

[0083] The single-class support vector machine 3125 can be described by the prediction function f(x) of the following formula (4). In the following formula (4), α i are Lagrange multipliers. k(x i ,x) is the kernel function and is used to calculate the data point x of the training data i and the similarity between the new data point x. In this regard, f(x) can be used to describe Figure 6 The decision range 631 is shown. If the new data point x falls into Figure 6 If the new data point x falls within the decision range 631 shown in FIG. 6 , then f(x) = 1, and this represents that the training data is a positive sample (ie, normal locking). On the contrary, if the new data point x falls within the decision range 631 shown in FIG. Figure 6 If the value is outside the decision range 631 shown, f(x)=-1, which means that the training data is a negative sample (ie, an abnormal lock).

[0084]

[0085] In other words, the single-class support vector machine 3125 can be pre-trained by the feature vector training data 802_1-802_N to define the decision range in the feature space. And, it is worth noting that the feature vector training data 802_1-802_N is generated according to the test lock data 801_1-801_N corresponding to normal lock. And, in this embodiment, the model training server 800 can further export the trained single-class support vector machine 3125 to the monitoring device (i.e., the production line computer).

[0086] Fig. 9A as well as Fig. 9B is a flow chart of a production line monitoring method according to an embodiment of the present invention. Figure 3 , Fig. 9A as well as Fig. 9B , the production line monitoring system 300 can execute the monitoring process of the following steps S901 to S923. In step S901, the user can start the screw locking device 320. In step S902, the user can input the work number so that the monitoring device 310 starts and executes the process control module 3121 and the front-end display module 3122. In step S903, the process control module 3121 can start the neural network model 3124 and the single-class support vector machine 3125. In step S904, the user can put the product (i.e., the locking object) into the fixture 340. In step S905, the user can press the start button. In step S906, the barcode scanning device 330 can be started to scan the barcode of the target screw 400, the screw locking device 320 and / or the fixture 340, and obtain the product serial number, device number, device model and / or site information of the screw locking device 320. In step S907, the process control module 3121 can determine whether the barcode scanning is successful.

[0087] If the code scanning fails, in step S908, the front-end display module 3122 may display a failure warning (e.g. Figure 7B In step S909, the user can perform a device check. If the barcode is scanned successfully, in step S910, the process control module 3121 can further determine whether the barcode is successfully determined.

[0088] If the determination fails, in step S908, the front-end display module 3122 may display a failure warning through the display 313. If the determination succeeds, in step S911, the process control module 3121 may read the parameters (i.e., the corresponding driving information, such as the driving voltage and driving instructions, etc.) through the data storage module 3123. In step S912, the process control module 3121 may determine whether the parameters are successfully obtained.

[0089] If the parameters are not successfully acquired, in step S908, the front-end display module 3122 may display a failure warning through the display 313. If the parameters are successfully acquired, in step S913, the monitoring device 310 may power on the screw locking device 320 for forward rotation. In step S914, the screw locking device 320 may perform forward rotation locking and generate locking data. In step S915, the neural network model 3124 and the single-class support vector machine 3125 may process and analyze the locking data. In step S916, the process control module 3121 may determine whether the determination is passed.

[0090] If the determination is passed (i.e., it is determined to be normally locked), then in step S922, the process control module 3121 can determine whether the number of locks has been reached. If so, in step S921, the process control module 3121 can power off the screw locking device 320. In step S923, the user can remove the product. If the determination is not passed (i.e., it is determined to be abnormally locked), then in step S908, the front display module 3122 can display a failure warning (e.g., Figure 7B Furthermore, in step S917, the process control module 3121 may determine whether the number of locking times exceeds a preset number of times.

[0091] If the number of locking times exceeds the preset number, in step S918, the process control module 3121 stops the locking operation. Also, in step S921, the process control module 3121 may power off the screw locking device 320. In step S923, the user may remove the product. If the number of locking times does not exceed the preset number, in step S919, the process control module 3121 may operate the screw locking device 320 to reverse. In step S920, the process control module 3121 may determine whether the reversal is successful.

[0092] If the reverse rotation is unsuccessful, in step S921, the process control module 3121 may power off the screw locking device 320. In step S923, the user may remove the product. If the reverse rotation is successful, in step S913, the monitoring device 310 may power on the screw locking device 320 again to rotate forward, so that the screw locking device 320 can rotate forward and lock again.

[0093] In other words, in response to the single-class support vector machine 3125 determining that the target screw 400 is abnormally locked, the process control module 3121 can control the screw locking device 320 to perform a reverse operation and re-lock the target screw. In addition, when the process control module 3121 determines that the number of re-locking times of the target screw 400 exceeds a preset number, the process control module 3121 stops the locking operation, and the equipment process engineer removes the product from the fixture 340.

[0094] Therefore, the production line monitoring system 300 of this embodiment can automatically determine whether the target screw 400 is normally locked (i.e., locked successfully) or abnormally locked (i.e., locked failed), and can immediately display a failure warning. In addition, the production line monitoring system 300 and the production line monitoring method of this embodiment can also automatically re-lock the target screw that is abnormally locked within a preset number of times to effectively avoid misjudgment or misoperation and stop the production line operation.

[0095] In summary, the production line monitoring system and production line monitoring method of the present invention can accurately analyze the locking data generated by the target screw during the locking process through the trained neural network model and the single-class support vector machine, so as to instantly determine whether the currently locked target screw is normally locked or abnormally locked. The production line monitoring system and production line monitoring method of the present invention can also collect locking data, such as torque trend and maximum torque data, and perform abnormal detection based on the locking data to confirm the locking quality. The production line monitoring system and production line monitoring method of the present invention can also remind the operator by displaying a reminder icon (such as a crying face or a smiling face) and / or a sound prompt, so that the production line personnel can quickly confirm whether the current locking result is normal locking or abnormal locking. The production line monitoring system and production line monitoring method of the present invention can also collect historical locking data, such as historical locking torque and historical locking results, etc., and record and upload them to the cloud system or the central console, and provide historical data for the front-end production line personnel to consult. In addition, the production line monitoring system and the production line monitoring method of the present invention can also adjust and modify the locking parameters through the front-end computer to integrate the consistency of the front-end computer and the locking device.

[0096] Although the present invention has been disclosed as above by way of embodiments, it is not intended to limit the present invention. Any person skilled in the art should be able to make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the attached claims.

Claims

1. A production line monitoring system, the production line monitoring system comprising: a monitoring device; as well as a screw locking device, the screw locking device is electrically connected to the monitoring device and is used to lock a target screw to generate locking data, The monitoring device performs a numerical normalization operation on the locking data, and inputs the normalized locking data into a neural network model so that the neural network model outputs a feature vector data according to the normalized locking data, and the monitoring device inputs the feature vector data into a single-class support vector machine to determine whether the target screw is normally locked or abnormally locked according to the judgment result of the single-class support vector machine. 2 . The production line monitoring system as claimed in claim 1 , wherein the locking data comprises a plurality of torque values ​​corresponding to the screw locking device locking the target screw during a locking period.

3. The production line monitoring system as described in claim 1, wherein the neural network model is an autoencoder model, the autoencoder model is used to compress the normalized locked data, and generate the feature vector data through an embedding layer of the autoencoder model.

4. The production line monitoring system as claimed in claim 1, wherein the single-class support vector machine determines whether the target screw is normally locked or abnormally locked according to whether a coordinate point of the feature vector data in a feature space is located within a decision range.

5. The production line monitoring system as claimed in claim 4, wherein the single-class support vector machine is pre-trained via a plurality of feature vector training data to define the decision range in the feature space, wherein the feature vector training data are generated based on a plurality of test locking data corresponding to normal locking.

6. The production line monitoring system as claimed in claim 1, wherein the monitoring device is further used to execute a process control module, and the process control module reads and executes the single-class support vector machine, In response to the single-class support vector machine determining that the target screw is abnormally locked, the process control module suspends the operation of the screw locking device. 7 . The production line monitoring system of claim 6 , wherein in response to the single-class support vector machine determining that the target screw is abnormally locked, the process control module controls the screw locking device to perform a reverse operation and re-lock the target screw. 8 . The production line monitoring system as claimed in claim 7 , wherein when the process control module determines that a re-tightening number of the target screw exceeds a preset number, the process control module stops the tightening operation.

9. The production line monitoring system as claimed in claim 1, wherein the monitoring device is further used to execute a front-end display module to display a locking result information on a user interface according to whether the target screw is normally locked or abnormally locked. 10 . The production line monitoring system as claimed in claim 1 , wherein the monitoring device is further used to execute a data storage module to store a product serial number, a locking time and the locking data of the target screw.

11. A production line monitoring method, the production line monitoring method comprising: A target screw is locked by a screw locking device to generate locking data; Performing a value normalization operation on the locked data through a monitoring device; Inputting the normalized locking data into a neural network model through the monitoring device, so that the neural network model outputs a feature vector data according to the normalized locking data; as well as The monitoring device inputs the feature vector data into a single-class support vector machine to determine whether the target screw is normally locked or abnormally locked according to the judgment result of the single-class support vector machine. 12 . The production line monitoring method according to claim 11 , wherein the locking data comprises a plurality of torque values ​​corresponding to the screw locking device locking the target screw during a locking period.

13. The production line monitoring method of claim 11 , wherein the neural network model is an autoencoder model, and the step of generating the feature vector data comprises: Performing data compression on the normalized locked data by using the autoencoder model; as well as The feature vector data is generated via an embedding layer of the autoencoder model.

14. The production line monitoring method according to claim 11, wherein the step of the single-class support vector machine determining whether the target screw is normally locked or abnormally locked comprises: The single-class support vector machine determines whether the target screw is normally locked or abnormally locked according to whether a coordinate point of the feature vector data in a feature space is located within a decision range.

15. The production line monitoring method as claimed in claim 14, wherein the single-class support vector machine is pre-trained via a plurality of feature vector training data to define the decision range in the feature space, wherein the feature vector training data are generated based on a plurality of test locking data corresponding to normal locking.

16. The production line monitoring method according to claim 11, further comprising: Executing a process control module through the monitoring device to read and execute the single-class support vector machine; as well as In response to the single-class support vector machine determining that the target screw is abnormally locked, the process control module suspends the operation of the screw locking device.

17. The production line monitoring method according to claim 16, further comprising: In response to the single-class support vector machine determining that the target screw is abnormally locked, the process control module controls the screw locking device to perform a reverse operation and re-lock the target screw.

18. The production line monitoring method according to claim 17, further comprising: When the process control module determines that a re-locking number of the target screw exceeds a preset number, the locking operation is stopped by the process control module.

19. The production line monitoring method according to claim 11, further comprising: A front-end display module is executed by the monitoring device to display a locking result information on a user interface according to whether the target screw is normally locked or abnormally locked.

20. The production line monitoring method according to claim 11, further comprising: The monitoring device also executes a data storage module to store a product serial number, a locking time and the locking data of the target screw.

Images