Method and apparatus for production line bottleneck analysis using vibration data

Abstract

Disclosed herein are methods and apparatuses for production line bottleneck analysis using vibration data. A method according to one aspect of the disclosure includes: obtaining vibration data collected by a vibration sensor deployed at a station of a production line; performing time-domain feature extraction on the vibration data to obtain a set of time-domain feature values corresponding to the vibration data; dividing the set of time-domain feature values into a plurality of data segments based on design data and expert domain knowledge associated with a plurality of processes contained by the station, wherein each data segment of the plurality of data segments is matched to one of the plurality of processes; for each data segment of the plurality of data segments, determining whether the data segment conforms to a feature statistical distribution model of the corresponding process; and in response to determining that the data segment does not conform to the feature statistical distribution model of the corresponding process, marking the corresponding process as a bottleneck in the production line.

Description

Technical Field

[0001] This disclosure generally relates to information processing, and more specifically, to methods and apparatus for performing bottleneck analysis of a production line using vibration data. Background Technology

[0002] A production line, also known simply as a production line, is a common production method used in factories. In a factory, the efficiency of the production line is one of the key factors affecting output and quality. Therefore, for factories, identifying bottlenecks in the production line and addressing them in a targeted manner to improve production efficiency is crucial.

[0003] However, identifying bottlenecks in a production line is no easy task. A production line often comprises multiple workstations and machines from different manufacturers, and their data is typically monitored and processed by different systems. For example, the efficiency of a machine on the production line may be calculated by combining status data from the control system and production data from the Manufacturing Execution System (MES). However, the control system is usually provided by the machine's manufacturer, while the MES may come from another third-party vendor. Effectively combining data from these two different sources to identify bottlenecks affecting efficiency presents many practical difficulties.

[0004] To identify bottlenecks in the production line, one approach in existing technology is to develop a new system that combines data from multiple existing data sources and eliminates potential biases and errors. However, this approach requires significant factory investment and is difficult to apply universally across different factories due to the substantial variations in the hardware and software systems used.

[0005] Another approach is to manually identify bottlenecks. For example, experienced engineers might go to the production line to troubleshoot and find bottlenecks affecting efficiency, their causes, and then relay this information back to production staff. Clearly, this method is inefficient and does not align with the trend towards factory automation and digitalization, which aims to reduce reliance on manual labor. Summary of the Invention

[0006] This summary section is provided to introduce some selected concepts in a simplified form, which will be further described in the detailed description section below. This summary section is not intended to identify any key or essential features of the claimed subject matter, nor is it intended to help determine the scope of the claimed subject matter.

[0007] According to one aspect of this disclosure, a method for bottleneck analysis of a production line using vibration data is provided, the method comprising: acquiring vibration data collected by vibration sensors deployed at workstations in the production line; performing time-domain feature extraction on the vibration data to obtain a set of time-domain feature values ​​corresponding to the vibration data; dividing the set of time-domain feature values ​​into multiple data segments based on design data and expert domain knowledge associated with multiple processes contained in the workstation, wherein each of the multiple data segments is matched to one of the multiple processes; for each of the multiple data segments, determining whether the data segment conforms to a characteristic statistical distribution model of the corresponding process; and in response to determining that the data segment does not conform to the characteristic statistical distribution model of the corresponding process, marking the corresponding process as a bottleneck in the production line.

[0008] The above aspects of this disclosure provide an efficient and easy-to-implement mechanism for accurate bottleneck analysis of a production line. By analyzing and statistically processing vibration data from vibration sensors widely deployed throughout the production line, combined with relevant design data and expert domain knowledge, the mechanism of this disclosure can efficiently and accurately identify bottlenecks in the production line, providing a possibility for solving specific bottleneck problems and improving production line efficiency.

[0009] Furthermore, in one example of the foregoing aspects, the method further includes: identifying one or more potential causes of a bottleneck, based at least in part on design data and expert domain knowledge associated with the corresponding process; and generating and outputting a list of bottleneck causes containing the one or more causes.

[0010] Furthermore, in one example of the foregoing aspects, the sampling period of the vibration data corresponds to a complete operating cycle of the workstation.

[0011] Furthermore, in one example of the foregoing aspects, the method further includes: preprocessing the vibration data before performing time-domain feature extraction on the vibration data, wherein the preprocessing includes one of low-pass filtering and band-pass filtering, and wherein the time-domain features include one or more of the following: maximum value, peak value, root mean square, and kurtosis.

[0012] Furthermore, in one example of the foregoing aspect, the method further includes: dividing the time-domain feature set into a second plurality of pre-segments based on the data characteristics of the time-domain feature set itself, wherein the plurality of data segments are determined by updating the second plurality of pre-segments, the update including merging and / or splitting the pre-segments in the second plurality of pre-segments.

[0013] Furthermore, in one example of the foregoing aspects, the feature statistical distribution model is determined by statistical analysis of historical data, which includes a set of historical feature values ​​accumulated for the process in the current workstation and a set of historical feature values ​​accumulated for the corresponding processes in other workstations of the same type as the current workstation.

[0014] Furthermore, in one example of the foregoing aspects, determining whether the data segment conforms to the characteristic statistical distribution model of the corresponding process includes: checking whether the corresponding feature of the data segment falls within a specified deviation range based on one or more parameters of the characteristic statistical distribution model; and determining whether the data segment does not conform to the characteristic statistical distribution model of the corresponding process in response to whether the corresponding feature of the data segment falls within the specified deviation range.

[0015] Furthermore, in one example of the foregoing aspects, the characteristic statistical distribution model includes a normal distribution, and wherein the parameters of the normal distribution include one or more of the following: mean, variance.

[0016] Furthermore, in one example of the foregoing aspects, identifying one or more possible causes of a bottleneck includes: determining the possible causes of a bottleneck by comparing the data segment with other data segments within the current sampling period; and / or determining the possible causes of a bottleneck by comparing the data segment with corresponding data segments within other sampling periods.

[0017] According to another aspect of this disclosure, a computing device is provided, the computing device comprising: at least one processor; and a memory coupled to the at least one processor and used to store instructions, wherein, when executed by the at least one processor, the at least one processor causes the at least one processor to: acquire vibration data collected by vibration sensors deployed at a workstation in a production line; perform time-domain feature extraction on the vibration data to obtain a set of time-domain feature values ​​corresponding to the vibration data; divide the set of time-domain feature values ​​into multiple data segments based on design data and expert domain knowledge associated with multiple processes contained in the workstation, wherein each of the multiple data segments is matched to one of the multiple processes; for each of the multiple data segments, determine whether the data segment conforms to a feature statistical distribution model of the corresponding process; and in response to determining that the data segment does not conform to the feature statistical distribution model of the corresponding process, mark the corresponding process as a bottleneck in the production line.

[0018] According to another aspect of this disclosure, an apparatus for performing bottleneck analysis on a production line using vibration data is provided. The apparatus includes: a module for acquiring vibration data collected by vibration sensors deployed at workstations on the production line; a module for performing time-domain feature extraction on the vibration data to obtain a set of time-domain feature values ​​corresponding to the vibration data; a module for dividing the set of time-domain feature values ​​into multiple data segments based on design data and expert domain knowledge associated with multiple processes contained in the workstation, wherein each of the multiple data segments is matched to one of the multiple processes; a module for determining, for each of the multiple data segments, whether the data segment conforms to a characteristic statistical distribution model of the corresponding process; and a module for marking the corresponding process as a bottleneck in the production line in response to determining that the data segment does not conform to the characteristic statistical distribution model of the corresponding process.

[0019] According to another aspect of this disclosure, a computer-readable storage medium is provided having instructions stored thereon that, when executed by at least one processor, cause the at least one processor to perform any of the methods described herein.

[0020] According to another aspect of this disclosure, a computer program product is provided, comprising instructions that, when executed by at least one processor, cause the at least one processor to perform any of the methods described herein. Attached Figure Description

[0021] Implementations of this disclosure are illustrated in the accompanying drawings by way of example rather than limitation, and similar reference numerals in the drawings denote the same or similar parts, wherein:

[0022] Figure 1 Exemplary environments in which some implementations of this disclosure may be implemented are shown;

[0023] Figure 2 Exemplary fragments of vibration data acquired by vibration sensors according to some implementations of this disclosure are shown;

[0024] Figure 3 Exemplary structures of production lines according to some implementations of this disclosure are shown;

[0025] Figure 4 A flowchart of an exemplary method according to some implementations of this disclosure is shown;

[0026] Figure 5 Exemplary architectures for some implementations according to this disclosure are shown;

[0027] Figure 6 Block diagrams of exemplary devices according to some implementations of this disclosure are shown; and

[0028] Figure 7 A block diagram of an exemplary computing device according to some implementations of this disclosure is shown.

[0029] List of reference numerals

[0030] 110: Vibration sensor; 120: Processing equipment; 130: Network

[0031] 310-330: Workstation 1-Workstation 3; 340-360: Machine 1-Machine 3

[0032] 370-390: Process 1-Process 3

[0033] 410-470: Steps

[0034] 510: Vibration Analysis Stage 520: Data Matching Stage 530: Statistical Analysis Stage

[0035] 540: Vibration data; 550: Design data; 560: Expert domain knowledge

[0036] 570: List of Bottleneck Causes

[0037] 610-650: Modules

[0038] 710: Processor; 720: Memory Detailed Implementation

[0039] In the following specification, numerous specific details are set forth for illustrative purposes. However, it should be understood that implementations of this disclosure can be carried out without these specific details. In other instances, well-known circuits, structures, and techniques have not been shown in detail so as not to affect the understanding of the specification.

[0040] Throughout this specification, references to "an implementation," "implementation," "exemplary implementation," "some implementations," "various implementations," etc., indicate that the implementations of this disclosure described may include specific features, structures, or characteristics. However, it is not implied that every implementation must include these specific features, structures, or characteristics. Furthermore, some implementations may have some, all, or none of the features described for other implementations.

[0041] In a manner most conducive to understanding the claimed subject matter, various operations may be described as a series of discrete actions or operations in sequence. However, the order in which they are described should not be construed as implying that these operations are necessarily order-dependent. In particular, these operations may be performed in a manner other than that presented. In other implementations, various additional operations may be performed, and / or various operations already described may be omitted.

[0042] In the specification and claims, the phrase "A and / or B" may appear to mean one of the following: (A), (B), (A and B). Similarly, the phrase "A, B and / or C" may appear to mean one of the following: (A), (B), (C), (A and B), (A and C), (B and C), (A and B and C).

[0043] In the specification and claims, the terms “coupled” and “connected” and their derivatives may be used. It is important to understand that these terms are not intended to be synonyms. Rather, in a particular implementation, “connected” is used to indicate that two or more components are in direct physical or electrical contact with each other, while “coupled” is used to indicate that two or more components cooperate or interact with each other, but they may or may not be in direct physical or electrical contact.

[0044] Bottleneck analysis of production lines to improve efficiency and reduce costs remains a hot topic in industrial production and other fields. This disclosure provides an efficient and easy-to-implement mechanism for bottleneck analysis of production lines. By analyzing and statistically processing data collected by vibration sensors deployed in the production line, combined with relevant design data and expert domain knowledge, bottlenecks in the production line can be accurately identified. Furthermore, one or more possible causes of the bottleneck can be identified, thereby making it possible to solve specific bottleneck problems and further improve production line efficiency.

[0045] First refer to Figure 1 This illustrates an exemplary operating environment 100 in which some implementations of this disclosure may be implemented. For example... Figure 1 As shown, in some implementations, the operating environment 100 may include at least one vibration sensor 110 and a processing device 120, which can be communicatively coupled to each other via a network 130.

[0046] Processing device 120 is used to implement the various schemes described in this disclosure based at least in part on vibration data acquired by at least one vibration sensor 110. Each of the vibration sensors 110 may be an Internet of Things (IoT) sensor, which may collectively form part of the Internet of Things. The vibration sensor 110 is deployed at appropriate locations in a production line, for example, it may be deployed at a workstation in the production line, and more specifically, it may be deployed on a specific machine (such as a transmission system, machine tool, etc.) at that workstation, to detect mechanical vibration at the workstation / machine, and to convert the detected mechanical vibration into current or voltage or other signal forms for output. Such vibration signals may typically be a superposition of many sine waves. An exemplary fragment of vibration data acquired by the vibration sensor is shown below. Figure 2As shown, the horizontal axis represents time (seconds), and the vertical axis represents the unit of vibration, such as meters per second squared (m / s²). 2 ).

[0047] Currently, vibration sensors are installed on new intelligent equipment or existing equipment in factory production lines for digital transformation, enabling equipment management and predictive maintenance. For example, vibration sensors installed on motors can monitor motor operation to detect motor malfunctions promptly. Unlike these and other common uses of vibration sensors, some implementations according to this disclosure use vibration data collected by the sensors for production line bottleneck analysis.

[0048] In some implementations, vibration sensor 110 continuously collects vibration data and provides the collected vibration data for a specific time period (sampling period) to processing device 120 in response to a triggering event or an indication signal from an external source (e.g., processing device 120). In one example, the sampling period of the vibration data corresponds to a complete operating cycle of the workstation on which the vibration sensor is deployed. In other words, whenever the workstation completes a complete operating cycle, the corresponding vibration sensor provides the vibration data collected for that time period to processing device 120.

[0049] Furthermore, in some implementations, the vibration sensor 110 may not directly provide the collected vibration data to the processing device 120, but instead transmit the vibration data to other devices (not shown), such as databases, storage devices, etc. Accordingly, the processing device 120 can obtain the stored vibration data from the vibration sensor 110 from such other devices during operation.

[0050] Examples of processing device 120 may include, but are not limited to: mobile devices, personal digital assistants (PDAs), wearable devices, smartphones, cellular phones, handheld devices, messaging devices, computers, personal computers (PCs), desktop computers, laptop computers, notebook computers, handheld computers, tablet computers, workstations, minicomputers, mainframe computers, supercomputers, network devices, web devices, processor-based systems, multiprocessor systems, consumer electronics devices, programmable consumer electronics devices, televisions, digital televisions, set-top boxes, or any combination thereof. In some implementations, processing device 120 may be deployed at a location remote from vibration sensor 110. Furthermore, in some implementations, the functionality of processing device 120 may be implemented by an application running on it; however, this disclosure is not limited thereto.

[0051] Furthermore, despite Figure 1The processing device 120 is shown as a single device, but those skilled in the art will understand that it can also be implemented as a group of devices. In some implementations, the processing device 120 can be implemented as a server, a server array, or a server cluster. Furthermore, in some implementations, the processing device 120, or at least a portion thereof, can be deployed in a distributed computing environment. In some implementations, the processing device 120, or at least a portion thereof, can be deployed in the cloud, employing cloud computing technology.

[0052] Network 130 may include any type of wired or wireless communication network, or a combination of wired and wireless communication networks. Examples of communication networks may include, but are not limited to: Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), Public Switched Telephone Network (PSTN), Internet, Intranet, Internet of Things (IoT), Infrared (IR) network, Bluetooth network, Near Field Communication (NFC) network, ZigBee network, etc. Furthermore, although... Figure 1 A single network 130 is shown, but network 130 can be configured to include multiple networks. In some implementations, the processing device 120 may also be directly coupled to at least one vibration sensor 110 without going through network 130.

[0053] Next reference Figure 3 This illustrates an exemplary structure 300 of a production line according to some implementations of this disclosure. Typically, the structure of a production line, from top to bottom, involves stations, machines, and specific processes. Figure 3 In the example, the production line is shown as having multiple workstations 310-330, namely workstation 1, workstation 2, and workstation 3. Each workstation may have one or more machines. For example, workstation 1 (310) may have one machine 1 (340), and workstation 2 (320) may have two machines, namely machine 2 (350) and machine 3 (360). Workstation 3 (330) also has its own machine, which is not shown here for simplicity. Each machine involves several different processes. Here, we only take machine 1 (340) as an example, which involves process 1 (370), process 2 (380), and process 3 (390). Each process may correspond to a different working state of the machine, such as loading, assembly, and unloading. For the purpose of simplicity, the processes involved in machine 2 and machine 3 are also not shown here.

[0054] In some implementations of this disclosure Figure 1The vibration sensor 110 shown is deployed at a workstation, for example, a vibration sensor 110 may be deployed at workstation 1 (310), and more specifically, may be deployed on machine 1 (340) to detect mechanical vibration at the corresponding location. As previously described, in one example, the vibration data acquired by processing unit 120 each time it is to be processed may correspond to a complete operating cycle of workstation 1 (310), which is the total time required for all three processes (370-390) under workstation 1 (310) to be executed sequentially once. However, the vibration data to be processed may also correspond to a longer time period (e.g., multiple complete operating cycles of a workstation) or a shorter time period (e.g., the time corresponding to executing one or more processes in a complete operating cycle of a workstation), and this disclosure is not limited thereto.

[0055] Figure 4 A flowchart of an exemplary method 400 according to some implementations of this disclosure is shown. Method 400 can use vibration data for production line bottleneck analysis. Method 400 can, for example, perform... Figure 1 Implemented in the processing device 120 shown or any similar or related entity.

[0056] like Figure 4 As shown, method 400 begins at step 410, in which vibration data is acquired by vibration sensors deployed at workstations on the production line. To illustrate this in conjunction with the preceding examples, processing device 120 may receive vibration data acquired by vibration sensor 110 deployed at workstation 1 (310) on the production line. In some implementations, the vibration data may be continuously acquired by vibration sensor 110 and actively provided to processing device 120 for processing in response to certain triggering conditions; alternatively, the vibration data may be provided by vibration sensor 110 to processing device 120 in response to an indication signal or acquisition request from an external source (e.g., from processing device 120). Furthermore, in some implementations, the vibration data acquired by vibration sensor 110 may be provided to a device such as a database for storage, from which processing device 120 may acquire the vibration data as needed.

[0057] After acquiring the vibration data, method 400 proceeds to step 420, in which time-domain feature extraction is performed on the vibration data to obtain a set of time-domain feature values ​​corresponding to the vibration data. As an example, the vibration data acquired in step 410 could be a 60-second segment of vibration data, which might correspond to a complete operating cycle of workstation 1 (310). The time-domain vibration analysis in step 420 can involve one or more of various time-domain features, including but not limited to maximum value, peak value, root mean square, kurtosis, etc.

[0058] Taking the maximum value as the extracted time-domain feature as an example, in step 420, for vibration data with a total length of 60 seconds, the maximum vibration value per unit length (e.g., every 1 second) can be calculated, and the 60 maximum values ​​obtained are used as the set of time-domain feature values ​​for the vibration data. In an alternative implementation, a sliding window can be used to calculate the maximum vibration value in each window. For example, the window size can be set to 5 seconds, and the step size to 1 second. The window slides along the time sequence according to the set step size, thus obtaining 56 maximum values ​​for the 60-second vibration data as the corresponding set of time-domain feature values. Other implementation methods are also feasible and can be selected depending on the needs of the system implementation.

[0059] Furthermore, in some implementations, method 400 may include preprocessing the vibration data acquired in step 410 prior to step 420. For example, the preprocessing may include low-pass filtering or band-pass filtering. This preprocessing effectively reduces the interference that noise in the vibration data might cause to the temporal feature extraction described above.

[0060] In some implementations, various specific time-domain processing algorithms that may be used for time-domain feature extraction in step 420, as well as various processing algorithms that may be used in the preprocessing process described above, can be included in a single time-domain analysis container. Accordingly, when specifically implementing step 420 and the preprocessing preceding it, one or more appropriate algorithms can be selected from this container as needed.

[0061] After processing in step 420, the raw vibration data is transformed into a specific set of eigenvalues. However, these eigenvalues ​​can only reflect the characteristics of the vibration data itself, and cannot provide more detailed information related to workstations / machines / processes on the production line.

[0062] Next, method 400 proceeds to step 430, in which the time-domain feature set is divided into multiple data segments based on design data and expert domain knowledge associated with the multiple processes contained in the workstation, wherein each of the multiple data segments is matched to one of the multiple processes.

[0063] Design data represents the specifications followed during the initial design of the entire system. For example, for each workstation-machine-process, there are corresponding process design parameters. These may include the process sequence, the corresponding working state of the process (such as loading, assembly (automatic), unloading, etc.), the preset time required to execute the process (or the allowable time range), and the upper / lower limits of acceptable vibration data characteristic values ​​(such as maximum values) during the execution of the process. These design parameters correspond one-to-one with each process and are stored in a database for use when needed.

[0064] However, design data merely represents fixed values / ranges that the production line's control system should follow. This typically corresponds to ideal conditions. In actual operation at a workstation or by the machines within it, especially over time, these design parameters often differ from their actual values. Therefore, relying solely on design data can introduce significant errors into subsequent calculations and processing.

[0065] Therefore, some implementations of this disclosure further incorporate expert domain knowledge, which may be gradually accumulated during production line operation, thus more accurately reflecting the actual situation of the workstation-machine-process. For example, expert domain knowledge may include supplementary information on the working status corresponding to a process reflected in the design data. For instance, the design data may indicate... Figure 3 The process 2 (380) shown corresponds to the automated assembly state; while the expert domain knowledge indicates that, in process 2, after automated assembly, a manual adjustment state is required for the assembly result, only in this way can the assembly result of process 2 be complete in actual production. Such knowledge may not have been considered in the design data, but it is necessary for actual production. In addition, the expert domain knowledge may also include one or more permissible time ranges (e.g., corresponding to the manual adjustment state supplemented to process 2 (380) above), some characteristics of abnormal jitter when the corresponding machine malfunctions, etc. This expert domain knowledge also corresponds one-to-one with each process and can be stored in a database for use when needed.

[0066] Using the design data and expert domain knowledge corresponding to these processes, step 430 can divide the time-domain feature value set output in step 420 into multiple data segments. For example, the design data of process 1 (370) in the database indicates that its duration is 10 seconds, corresponding to the working state of loading; the design data of process 2 (380) indicates that its duration is 20 seconds, corresponding to the working state of automatic assembly; and the design data of process 3 (390) indicates that its duration is 20 seconds, corresponding to the working state of unloading. In addition, the database also contains expert domain knowledge for process 2 (380), indicating that a duration of 10 seconds is also required here, corresponding to the manual adjustment state after automatic assembly. Based on at least the above information, the time-domain feature value set consisting of the 60 maximum values ​​mentioned above can be divided into three data segments in chronological order, with lengths of 10 seconds, 30 seconds, and 20 seconds, respectively, and matched to process 1 (370), process 2 (380), and process 3 (390).

[0067] Although the above example only describes segmenting the set of time-domain feature values ​​by considering the duration associated with each process for the purpose of simplification, it should be noted that such segmentation / matching operations can also additionally or alternatively consider other information contained in the design data and expert domain knowledge, such as the acceptable range of values ​​for the corresponding time-domain feature values ​​indicated in the design data and expert domain knowledge.

[0068] Furthermore, in some implementations, before performing step 430, method 400 may also include a pre-division operation, wherein the set of time-domain feature values ​​is divided into a second or more pre-segments based on the data characteristics of the set itself. For example, unlike the three data segments ultimately divided by step 430 in the example above, the pre-division operation may first divide the 60-second time-domain feature value set into six pre-segments of unequal length in chronological order. Such pre-division may only consider some characteristics or patterns that can be seen from the data itself in the set, such as the feature values ​​within 60 seconds exhibiting a certain periodic change, or the feature value size in one time period being significantly different from the feature value size in another time period, etc., without establishing a correlation with the corresponding design data and expert domain knowledge as in step 430.

[0069] Accordingly, in the case of pre-division, the multiple data segments finally obtained in step 430 may be determined by updating the second multiple pre-segments obtained by pre-division, wherein the update may include merging and / or splitting the pre-segments in the second multiple pre-segments.

[0070] Then, method 400 proceeds to step 440, in which for each of the plurality of data segments, it is determined whether the data segment conforms to the characteristic statistical distribution model of the corresponding process.

[0071] In some implementations, the characteristic statistical distribution model used in step 440 can be determined by statistical analysis of historical data of the corresponding process. This historical data can include data specific to the current workstation (e.g., ...). Figure 3 The process in station 1 (310) shown (e.g.) Figure 3 The set of historical feature values ​​accumulated for process 1 (370) shown can also include the set of historical feature values ​​accumulated for the corresponding process in other workstations of the same type as the current workstation. The latter is particularly applicable to scenarios where multiple identical production lines are running in parallel in a factory. Other methods for determining the feature statistical distribution model used in step 440 are also possible, and this disclosure is not limited to the specific implementation described above.

[0072] More specifically, in some implementations, determining whether the data segment conforms to the characteristic statistical distribution model of the corresponding process in step 440 may include: checking whether the corresponding feature of the data segment falls within a specified deviation range based on one or more parameters of the characteristic statistical distribution model; and determining that the data segment does not conform to the characteristic statistical distribution model of the corresponding process in response to the corresponding feature of the data segment not falling within the specified deviation range.

[0073] For example, a pre-determined characteristic statistical distribution model, based on statistical analysis of historical data, can characterize a feature of the corresponding process (such as the duration / cycle of the process) as conforming to a normal distribution, which has one or more parameters, such as mean, variance, etc. Therefore, in step 440, it can be checked whether the duration of the currently processed data segment (e.g., 10 seconds for the first data segment matched to process 1 (370)) conforms to the normal distribution defined by the mean and variance. For example, it can be required that the 10-second duration of the first data segment deviates from the mean by no more than three standard deviations (calculated from the variance); otherwise, it is considered that the duration of the first data segment does not fall within the specified deviation range, and thus the data segment is determined not to conform to the characteristic statistical distribution model of the corresponding process 1 (370).

[0074] Additionally or alternatively, the feature statistical distribution model used in step 440 can also be associated with other features. For example, for the previously extracted time-domain features (such as the maximum value), there may also be a pre-determined statistical distribution model based on historical data for a specific process (such as process 1 (370)). For example, such a statistical distribution model also follows a normal distribution and therefore has corresponding parameters. Therefore, it is also possible to determine whether the corresponding time-domain features of the currently processed first data segment also conform to the specific feature statistical distribution model in a similar manner as described above.

[0075] After step 440, method 400 proceeds to step 450, where, in response to the determination in step 440 that the data segment does not conform to the characteristic statistical distribution model of the corresponding process, the corresponding process is marked as a bottleneck in the production line.

[0076] Understandably, for each of the multiple data segments divided in step 430, the operations in steps 440-450 must be performed to identify whether the corresponding process is a bottleneck in the production line.

[0077] Through the aforementioned operations, bottlenecks in the production line can be efficiently and accurately identified based on the processing of vibration data collected by vibration sensors. Building upon this, in some implementations, method 400 may further include an additional step 460, wherein one or more potential causes of the bottleneck are identified, at least in part, based on design parameters associated with the corresponding process and expert domain knowledge; and an additional step 470, wherein a bottleneck cause list containing the one or more causes is generated and output. The output information helps technicians locate and resolve specific bottleneck problems, thereby enabling further improvements in production line efficiency.

[0078] More specifically, identifying one or more possible causes of the bottleneck in step 460 may include determining the possible causes of the bottleneck by comparing the data segment with other data segments within the current sampling period.

[0079] For example, suppose the current processing is the second data segment matched to process 2 (380). After it is identified as a bottleneck in the production line in step 450, the specific cause of the bottleneck cannot be determined at this time. This is because, as described in the previous example, the process is specifically associated with two states: an automated assembly state (as indicated by the design data) and a manual adjustment state (as indicated by expert domain knowledge). The former is controlled by the control system, while the latter is affected by manual operation.

[0080] In some implementations of this disclosure, for example in step 460, the second data segment can be compared with relevant data from the first data segment preceding it (matched to process 1 (370)) and / or the third data segment following it (matched to process 3 (390)) within the current sampling period. If only the second data segment exhibits the aforementioned deviation and is therefore identified as a bottleneck, while the first and third data segments do not exhibit similar deviations, then there is a high probability that the problem lies in the manual operation portion involved in process 2 (380), since the control logic of the control systems of process 2 (380) and the processes preceding and following it should generally be consistent.

[0081] Accordingly, in step 470, in one example of the bottleneck cause list generated and output, for process 2 (380), only the bottleneck cause with the highest probability, namely, the bottleneck caused by manual operation, can be included; while in another example of such a bottleneck cause list, both the bottleneck caused by manual operation and the bottleneck caused by control system logic can be listed as possible causes, but the former is placed first because it has a higher probability.

[0082] Additionally or alternatively, in some implementations of step 460, it may be considered to determine potential bottleneck causes by comparing the data segment with corresponding data segments in other sampling periods. For example, the currently processed second data segment may be compared with related data of a second data segment at a corresponding position in one or more preceding and / or subsequent sampling periods, and one or more potential bottleneck causes identified thereby may be included in the final generated bottleneck cause list in a similar manner to the foregoing.

[0083] Figure 5 Examples of architectures according to some implementations of this disclosure are shown, which generally may include a vibration analysis phase 510, a data matching phase 520, and a statistical analysis phase 530. Specifically, the vibration analysis phase 510 may correspond to... Figure 4 The steps 420 and related operations shown in the diagram involve taking vibration data 540 from sources such as vibration sensor 110 as input, performing vibration analysis on the vibration data by selecting a suitable time-domain feature extraction algorithm, and obtaining the corresponding set of time-domain feature values. The data matching stage 520 can correspond to... Figure 4 The steps 430 and related operations shown here involve design data 550 and expert domain knowledge 560 being provided as input to a data matching stage 520. The latter uses this data to match the output of the vibration analysis stage 510, i.e., the set of time-domain eigenvalues, to divide it into multiple data segments corresponding to different processes. The statistical analysis stage 530 can then correspond to... Figure 4The steps 440-470 shown in the diagram involve statistical analysis of each of the multiple data segments divided in the data matching stage 520 to determine whether they conform to the characteristic statistical distribution model of the corresponding process and thereby mark bottlenecks in the production line. Furthermore, the statistical analysis stage 540 can also use design data 550 and expert domain knowledge 560 as input to identify one or more possible causes of the marked bottlenecks and generate and output a corresponding bottleneck cause list 570.

[0084] Those skilled in the art will understand that Figure 5 The exemplary architecture shown can be implemented using software, hardware, firmware, or any combination thereof. In one example, it can be implemented in... Figure 1 In the processing device 120 shown or any similar or related entity.

[0085] The following is for reference. Figure 6 The diagram illustrates a block diagram of an exemplary device 600 according to some implementations of this disclosure. Device 600 can use vibration data for production line bottleneck analysis. Device 600 can, for example, [perform...]. Figure 1 Implemented in the processing device 120 shown or any similar or related entity.

[0086] like Figure 6 As shown, device 600 may include module 610 for acquiring vibration data collected by vibration sensors deployed at workstations on the production line. Device 600 may also include module 620 for performing time-domain feature extraction on the vibration data to obtain a set of time-domain feature values ​​corresponding to the vibration data. Device 600 may further include module 630 for dividing the set of time-domain feature values ​​into multiple data segments based on design data and expert domain knowledge associated with multiple processes contained in the workstation, wherein each of the multiple data segments is matched to one of the multiple processes. Device 600 may further include module 640 for determining, for each of the multiple data segments, whether the data segment conforms to the characteristic statistical distribution model of the corresponding process. Furthermore, device 600 may further include module 650 for marking the corresponding process as a bottleneck in the production line in response to determining that the data segment does not conform to the characteristic statistical distribution model of the corresponding process.

[0087] Furthermore, in some implementations, one or more of the above-described modules of device 600 may further include further sub-modules, and / or device 600 may also include additional modules for performing other operations already described in the specification, such as in combination with... Figure 4The exemplary method 400 is described in detail with reference to the flowchart. For example, in one implementation, the apparatus 600 may further include one or more modules for identifying one or more potential causes of bottlenecks, based at least in part on design data and expert domain knowledge associated with the corresponding process, and generating and outputting a bottleneck cause list containing the one or more causes. Furthermore, in some implementations, the various modules of the apparatus 600 may be combined or separated depending on actual needs. All of the above and other variations fall within the scope of this disclosure.

[0088] Those skilled in the art will understand that the exemplary device 600 can be implemented using software, hardware, firmware, or any combination thereof.

[0089] Figure 7 A block diagram of an exemplary computing device 700 according to some implementations of this disclosure is shown. The computing device 700 can optimize machine learning classification tasks for anomaly handling. The computing device 700 can, for example, […]. Figure 1 Implemented in the processing device 120 shown or any similar or related entity.

[0090] like Figure 7 As shown, the computing device 700 may include at least one processor 710. The processor 710 may include any type of general-purpose processing unit (e.g., CPU, GPU, etc.), dedicated processing unit, core, circuitry, controller, etc. Furthermore, the computing device 700 may also include a memory 720 coupled to the processor 710. The memory 720 may include any type of medium that can be used to store data. In some implementations, the memory 720 is configured to store instructions that, when executed, cause at least one processor 710 to perform the operations described in this disclosure, for example, in conjunction with... Figure 4 The exemplary method 400 is described in detail in the flowchart.

[0091] Furthermore, in some implementations, the computing device 700 may also be coupled to or equipped with one or more peripheral components, which may include, but are not limited to, a display, speakers, a mouse, a keyboard, etc. Additionally, in some implementations, the computing device 700 may also be equipped with a communication interface that supports various types of wired / wireless communication protocols for communicating with a communication network. Examples of communication networks may include, but are not limited to: Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), Public Telephone Network, Internet, Intranet, Internet of Things (IoT), Infrared Network, Bluetooth Network, Near Field Communication (NFC) Network, ZigBee Network, etc.

[0092] Furthermore, in some implementations, the above and other components can communicate with each other via one or more buses / interconnects that can support any suitable bus / interconnect protocol, including Peripheral Component Interconnect (PCI), High-Speed ​​PCI, Universal Serial Bus (USB), Serial Attached SCSI (SAS), Serial ATA (SATA), Fibre Channel (FC), System Management Bus (SMBus), or other suitable protocols.

[0093] Those skilled in the art will understand that the above description of the structure of computing device 700 is merely exemplary and not restrictive, and other structures are also feasible, as long as they can be used to implement the functions discussed in this disclosure.

[0094] Various implementations of this disclosure can be implemented using hardware units, software units, or combinations thereof. Examples of hardware units may include devices, components, processors, microprocessors, circuits, circuit elements (e.g., transistors, resistors, capacitors, inductors, etc.), integrated circuits, application-specific integrated circuits (ASICs), programmable logic devices (PLDs), digital signal processors (DSPs), field-programmable gate arrays (FPGAs), memory cells, logic gates, registers, semiconductor devices, chips, microchips, chipsets, etc. Examples of software units may include software components, programs, applications, computer programs, application programs, system programs, machine programs, operating system software, middleware, firmware, software modules, routines, subroutines, functions, methods, procedures, software interfaces, application programming interfaces (APIs), instruction sets, computational code, computer code, code segments, computer code segments, words, values, symbols, or any combination thereof. Determining whether an implementation is carried out using hardware units and / or software units can vary depending on a variety of factors, such as desired computing speed, power level, thermal tolerance, processing cycle budget, input data rate, output data rate, memory resources, data bus speed, and other design or performance constraints, as is expected of a given implementation.

[0095] Some implementations of this disclosure may include an article of writing. The article of writing may include a storage medium for storing logic. Examples of storage media may include one or more types of computer-readable storage media capable of storing electronic data, including volatile or non-volatile memory, removable or non-removable memory, erasable or non-erasable memory, writable or rewritable memory, and so on. Examples of logic may include various software units, such as software components, programs, applications, computer programs, application programs, system programs, machine programs, operating system software, middleware, firmware, software modules, routines, subroutines, functions, methods, procedures, software interfaces, application programming interfaces (APIs), instruction sets, computational code, computer code, code segments, computer code segments, words, values, symbols, or any combination thereof. In some implementations, for example, the article of writing may store executable computer program instructions that, when executed by a processor, cause the processor to perform the methods and / or operations described herein. Executable computer program instructions may include any suitable type of code, such as source code, compiled code, interpreted code, executable code, static code, dynamic code, and so on. Executable computer program instructions can be implemented according to a predefined computer language, method, or syntax used to command the computer to perform specific functions. These instructions can be implemented using any suitable high-level, low-level, object-oriented, visual, compiled, and / or interpreted programming language.

[0096] The examples described above include those of the disclosed architecture. It is certainly impossible to describe every conceivable combination of components and / or methods, but those skilled in the art will understand that many other combinations and arrangements are also possible. Therefore, this novel architecture is intended to cover all such alternatives, modifications, and variations that fall within the spirit and scope of the appended claims.

Claims

1. A method for bottleneck analysis of a production line using vibration data, comprising: Acquire vibration data collected by vibration sensors deployed at workstations on the production line; The vibration data is subjected to time-domain feature extraction to obtain a set of time-domain feature values ​​corresponding to the vibration data; Based on design data and expert domain knowledge associated with multiple processes contained in the workstation, the time-domain feature value set is divided into multiple data segments, wherein each of the multiple data segments is matched to one of the multiple processes; the expert domain knowledge includes supplements to the working state corresponding to the process reflected by the design data, and / or, one or more allowable time ranges, and / or, features of abnormal jitter when the corresponding machine fails; For each of the multiple data segments, determine whether the data segment conforms to the characteristic statistical distribution model of the corresponding process; and In response to the determination that the data segment does not conform to the characteristic statistical distribution model of the corresponding process, the corresponding process is marked as a bottleneck in the production line.

2. The method according to claim 1, further comprising: Based at least in part on design data and expert domain knowledge associated with the corresponding process, identify one or more possible causes of the bottleneck; as well as Generate and output a list of bottleneck causes that includes one or more of the stated reasons.

3. The method of claim 1, wherein, The sampling period of the vibration data corresponds to a complete operating cycle of the workstation.

4. The method according to claim 1, further comprising: Before performing time-domain feature extraction on the vibration data, the vibration data is preprocessed. The preprocessing includes either low-pass filtering or band-pass filtering. The time-domain features include one or more of the following: maximum value, peak value, root mean square, and kurtosis.

5. The method according to claim 1, further comprising: Based on the data characteristics of the time-domain feature value set itself, the time-domain feature value set is divided into a second set of pre-segments. The plurality of data segments are determined by updating the second plurality of pre-segments, wherein the update includes merging and / or splitting the pre-segments in the second plurality of pre-segments.

6. The method of claim 1, wherein, The feature statistical distribution model is determined by statistical analysis of historical data, which includes a set of historical feature values ​​accumulated for the process in the current workstation, and a set of historical feature values ​​accumulated for the corresponding processes in other workstations of the same type as the current workstation.

7. The method of claim 1, wherein, Determining whether the data segmentation conforms to the characteristic statistical distribution model of the corresponding process includes: Based on one or more parameters of the feature statistical distribution model, check whether the corresponding feature of the data segment falls within a specified deviation range; and If the corresponding feature of the data segment does not fall within the specified deviation range, the data segment is determined to be inconsistent with the feature statistical distribution model of the corresponding process.

8. The method of claim 7, wherein, The characteristic statistical distribution model includes a normal distribution, wherein the parameters of the normal distribution include one or more of the following: mean and variance.

9. The method of claim 2, wherein, Identifying one or more potential causes of the bottleneck includes: The potential causes of bottlenecks are determined by comparing the data segments with other data segments within the current sampling period; and / or The causes of potential bottlenecks are determined by comparing the data segments with corresponding data segments in other sampling periods.

10. A computing device, comprising: At least one processor; as well as A memory coupled to the at least one processor and used to store instructions, wherein, when executed by the at least one processor, the at least one processor causes the at least one processor to: Acquire vibration data collected by vibration sensors deployed at workstations on the production line; The vibration data is subjected to time-domain feature extraction to obtain a set of time-domain feature values ​​corresponding to the vibration data; Based on design data and expert domain knowledge associated with multiple processes contained in the workstation, the time-domain feature value set is divided into multiple data segments, wherein each of the multiple data segments is matched to one of the multiple processes; the expert domain knowledge includes supplements to the working state corresponding to the process reflected by the design data, and / or, one or more allowable time ranges, and / or, features of abnormal jitter when the corresponding machine fails; For each of the multiple data segments, determine whether the data segment conforms to the characteristic statistical distribution model of the corresponding process; and In response to the determination that the data segment does not conform to the characteristic statistical distribution model of the corresponding process, the corresponding process is marked as a bottleneck in the production line.

11. The computing device of claim 10, wherein, The memory is also used to store instructions that, when executed by the at least one processor, cause the at least one processor to: Based at least in part on design data and expert domain knowledge associated with the corresponding process, identify one or more potential causes of the bottleneck; and Generate and output a list of bottleneck causes that includes one or more of the stated reasons.

12. The computing device of claim 10, wherein, Determining whether the data segmentation conforms to the characteristic statistical distribution model of the corresponding process includes: Based on one or more parameters of the feature statistical distribution model, check whether the corresponding feature of the data segment falls within a specified deviation range; and If the corresponding feature of the data segment does not fall within the specified deviation range, the data segment is determined to be inconsistent with the feature statistical distribution model of the corresponding process.

13. The computing device of claim 12, wherein, The characteristic statistical distribution model includes a normal distribution, wherein the parameters of the normal distribution include one or more of the following: mean and variance.

14. The computing device of claim 11, wherein, Identifying one or more potential causes of the bottleneck includes: The potential causes of bottlenecks are determined by comparing the data segments with other data segments within the current sampling period; and / or The causes of potential bottlenecks are determined by comparing the data segments with corresponding data segments in other sampling periods.

15. An apparatus for performing bottleneck analysis on a production line using vibration data, comprising: Module used to acquire vibration data collected by vibration sensors deployed at workstations on the production line; A module for extracting time-domain features from the vibration data to obtain a set of time-domain feature values ​​corresponding to the vibration data; A module for dividing the time-domain feature set into multiple data segments based on design data and expert domain knowledge associated with multiple processes contained in the workstation, wherein each of the multiple data segments is matched to one of the multiple processes; the expert domain knowledge includes supplements to the working state corresponding to the process reflected by the design data, and / or, one or more permissible time ranges, and / or, features of abnormal jitter when the corresponding machine fails; A module for determining whether each of the plurality of data segments conforms to the characteristic statistical distribution model of the corresponding process; and A module for marking the corresponding process as a bottleneck in the production line in response to determining that the data segment does not conform to the characteristic statistical distribution model of the corresponding process.

16. The apparatus of claim 15, further comprising: A module for identifying one or more potential causes of bottlenecks, based at least in part on design data and expert domain knowledge associated with the corresponding process; as well as A module for generating and outputting a list of bottleneck causes containing one or more of the aforementioned reasons.

17. A computer-readable storage medium having instructions stored thereon, which, when executed by at least one processor, cause the at least one processor to perform the method according to any one of claims 1-9.

18. A computer program product comprising instructions that, when executed by at least one processor, cause the at least one processor to perform the method according to any one of claims 1-9.

Images