Analysis device and analysis method

Abstract

To enable accurate analysis of performed work while minimizing an increase in the amount of computation.SOLUTION: An analysis device is provided, comprising a measurement data acquisition unit for acquiring measurement data indicative of chronological changes in weight of an object on a top plate, and a control unit configured to acquire the measurement data and reference data indicative of chronological changes in weight that occur when work is done on the object on the top plate through a correct work procedure, so as to evaluate the work on the basis of a difference between a feature quantity of the measurement data and a feature quantity of the reference data.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] Embodiments of the present invention relate to an analysis device and an analysis method.

Background Art

[0002] Conventionally, there is a process management system for managing work processes such as product assembly work by factory workers. The process management system aims, for example, at improving work, grasps the current work process to make improvements, and evaluates the work processes before and after the improvements. Among such process management systems, there is one that manages work processes using a weighing scale. The process management system using a weighing scale has, for example, a measuring device and a measurement control device installed on a workbench. The measuring device is equipped with a weighing scale and is a device for measuring the weight of a product being assembled on the workbench. On the other hand, the measurement control device is a device that controls the weight measurement process by the measuring device and displays the weight or the like, which is the measurement result obtained from the measuring device.

[0003] The measurement control device acquires information regarding the parts used in each work and each work content from standard work information indicating predetermined work processes, operation procedures, etc. The standard work information referred to here is, for example, a standard work procedure manual (SOP: Standard Operating Procedures), etc. Also, the information regarding parts and work content is, for example, a part ID for identifying a part, the weight of the part, and the number of operations of a power tool used for attaching a fastening part (for example, the number of times of screwing). The measurement control device collates each work included in the work process defined in the standard work information in advance with each work actually performed based on the information obtained from the standard work procedure manual and the change in the weight of the product being assembled on the workbench measured by the measuring device.

[0004] For example, the measurement and control device specifies the time required for each operation included in the work process (hereinafter referred to as "actual work time") based on the measured weight change of the product. The measurement and control device compares the work time defined in the standard work information in advance with the actual work time. When a difference is recognized between the two, for example, the measurement and control device displays information (abnormal notification) indicating that an operation different from the operation defined in the standard work information has been performed. By having such a configuration, the process management system using a weighing scale can evaluate, for example, whether the work procedure is accurate and whether the work accuracy (level) is high, and can present the evaluation results.

[0005] However, the conventional process management system as described above has a problem that cooperation with external devices such as work tools is required. For example, in order to specify the start time of screwing a part, the conventional process management system needs to receive a notification or the like from a power tool such as an electric driver at the timing when the screwing is started. Further, in the conventional process management system as described above, there is a problem that an error easily occurs in the measurement accuracy of the weight change of the product, particularly when an operation of attaching a lightweight part such as a screw is performed.

[0006] In addition to the process management system using a weighing scale as described above, there has conventionally been a process management system using a camera. The process management system using a camera photographs the state of the work and analyzes the work from the photographed video. However, in the case of the process management system using a camera, there is a problem that the amount of data (video data) processed on the edge side (measurement device side) increases, resulting in an increase in the amount of calculation. As a result, a high-performance calculation processing device is required, and the device cost and the like may increase.

Prior Art Documents

Patent Documents

[0007]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0008] The problem to be solved by the present invention is to analyze the performed work with higher accuracy while suppressing an increase in the amount of calculation.

Means for Solving the Problems

[0009] The analysis device according to the embodiment includes a measurement data acquisition unit and a control unit. The measurement data acquisition unit acquires measurement data indicating the time-series weight change of an object on the top plate. The control unit acquires the measurement data and standard data indicating the time-series weight change that occurs when the work on the object is performed in the correct work process on the top plate, and evaluates the work based on the difference between the feature amount of the measurement data and the feature amount of the standard data.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Modes for Carrying Out the Invention

[0011] Hereinafter, the analysis device and the analysis method according to the embodiment will be described with reference to the drawings.

[0012] The overall configuration of the process management system 1 according to the embodiment will be described. The process management system 1 is an information processing system for managing work processes such as the assembly work of products by factory workers.

[0013] FIG. 1 is an overall configuration diagram of the process management system 1 according to the embodiment. As shown in FIG. 1, the process management system 1 includes a server 10, at least one measuring device 20, and a network 30. The server 10 is an example of the analysis device of the present invention. The server 10 and the measuring device 20 are configured to be communicably connected via the network 30. For example, the server 10 and the network 30 are communicably connected by wired communication, and the measuring device 20 and the network 30 are communicably connected by wireless communication.

[0014] Note that any communication method can be used for the communication between the server 10 and the measuring device 20, and any network configuration can be used for the network configuration of the network 30. That is, any configuration can be adopted as long as the server 10 and the measuring device 20 can communicate with each other.

[0015] In the process management system 1 of the embodiment, it is assumed that the server 10 and the measuring device 20 are installed at locations separated from each other, but it is not limited thereto. For example, an analysis device (having the same function as the server 10) may be installed near the measuring device 20, and both may be directly communicably connected without passing through the network 30. Alternatively, for example, the measuring device 20 and the analysis device may be integrated into one device.

[0016] The server 10 is an information processing device such as a general-purpose computer. Further, as shown in FIG. 1, the measuring device 20 is a device attached to the upper part of the top plate 501 of the workbench 50 or the upper part of the base of the cart 60. The measuring device 20 includes a sensor (strain gauge) for measuring the weight of an object placed on the device itself. The server 10 evaluates the work actually performed by the operator by analyzing the weight change of the object measured by the measuring device 20.

[0017] Hereinafter, the configuration of the measuring device 20 will be described in more detail with reference to FIG. 2. FIG. 2 is an overall configuration diagram of the measuring device 20 of the embodiment. As an example, FIG. 2 shows the configuration when the measuring device 20 is attached to the upper part of the top plate 501 of the working machine 50 shown in FIG. 1. As shown in FIG. 2, the measuring device 20 includes a top plate 201, a support column 202, a strain gauge 203, and a control device 204. In order to distinguish from the top plate 501 originally provided in the working machine 50 itself, hereinafter, the top plate 201 provided in the measuring device 20 is referred to as the "measurement top plate 201".

[0018] The measurement top plate 201 is a general top plate made of, for example, resin, wood, or metal. The support column 202 has a shape with a mechanism for attaching the self-device (measuring device 20) to the upper part of the top plate 501 of the working machine 50 and a mechanism for supporting the measurement top plate 201 from below. The strain gauge 203 is attached to the support column 202 and is a weight sensor that measures the amount of strain generated in the support column 202. For example, the heavier the weight of the object placed on the measurement top plate 201, the greater the strain generated in the support column 202. The strain gauge 203 has its resistance value fluctuated by the strain of the support column 202. The control device 204 has an arithmetic function and a communication function. It observes the amount of change in the resistance value of the strain gauge 203 by the arithmetic function, and calculates the amount of change in the weight on the top plate from this amount of change. Based on the time-series weight change from this calculation result, waveform data is generated, and the waveform data is transmitted to the server 10 by the communication function.

[0019] In this way, since the measuring device 20 has a configuration in which the measurement top plate 201, the strain gauge 203, and the control device 204 are integrated, it can be easily attached to the upper part of the top plate of an existing general working machine or the upper part of a general cart by simply attaching it retrofitting, and a general working machine or cart can be easily changed into a working machine or cart capable of measuring weight. Further, by having such a configuration, the measuring device 20 can enable weight measurement without significantly changing the height of the top plate of the existing working machine or the loading height of the cart.

[0020] Figure 3 is a diagram showing an example of the measurement result of the measurement by the measurement device 20 of the embodiment. In the graph shown in Figure 3, the horizontal axis represents time, and the vertical axis represents the weight of the product placed on the measurement top plate 201. This weight can be calculated based on the amount of strain measured by the strain gauge 203. Note that the measurement device 20 may be configured to output the value of the amount of strain as it is instead of the weight.

[0021] Note that the measurement device 20 may be provided with weight measurement means of a method other than the strain gauge. That is, the measurement device 20 can have an arbitrary configuration as long as it can grasp the weight change of the object on the measurement top plate 201.

[0022] Figure 3 shows an example of the weight change measured when the product assembly work is performed on the measurement top plate 201. Since a plurality of parts are sequentially attached to the product placed on the measurement top plate 201, for example, as shown in Figure 3, the measured weight increases with time. In addition, the graph shown in Figure 3 shows that an event in which the weight instantaneously increases greatly has occurred 10 times ((a) to (j) in Figure 3). This is caused, for example, when an operator attaches a fastening part such as a screw to the product and a temporary load is applied to the product and the measurement top plate 201.

[0023] By analyzing the waveform of the graph when such an event of a large change in weight occurs, it becomes possible to analyze the work performed by the operator. For example, based on the shape and size of the waveform, it becomes possible to estimate what kind of work was performed. Also, based on the interval between waveforms, it becomes possible to estimate the time spent on each work. In addition, based on the order in which waveforms of each shape occur, it becomes possible to estimate whether the work was performed in the correct order.

[0024] For example, the four waveforms in FIGS. 3(b) to 3(e) showing four consecutive weight changes are waveforms with similar shapes and similar magnitudes. This kind of waveform data occurs, for example, when screws of the same weight are respectively attached to four locations of a product. Thus, it can be inferred from the waveform that the same operation has been performed. Similarly, since the three waveforms in FIGS. 3(f) to 3(h) and the two waveforms in FIGS. 3(i) to 3(j) are also waveforms with similar shapes and similar magnitudes, it can be inferred that the same operation has been performed.

[0025] Also, for example, the four waveforms in FIGS. 3(b) to 3(e) are larger in magnitude than the three waveforms in FIGS. 3(f) to 3(h). From this, it can be inferred that in the operations corresponding to the four waveforms in FIGS. 3(b) to 3(e), operations that impose a greater load on the measurement top plate 201 have been performed.

[0026] Also, if a waveform does not occur at the time when a waveform should occur in the correct operation process, it can be inferred that the operator has not performed the necessary operation.

[0027] Hereinafter, the functional configuration of the server 10 will be described in more detail with reference to FIG. 4. FIG. 4 is a block diagram showing the functional configuration of the server 10 according to the embodiment. As shown in FIG. 4, the server 10 includes a learning unit 110 and an inference unit 120.

[0028] The learning unit 110 acquires waveform data indicating the weight change of an object on the measurement top plate 201 when an operation is performed in a typical operation process. The learning unit 110 performs machine learning using the data based on the acquired waveform data as teacher data. The inference unit 120 evaluates the operation to be evaluated by inputting the waveform data of the operation to be evaluated into the learned learning model generated by the machine learning by the learning unit 110. Thereby, the inference unit 120 can determine whether the operation to be evaluated is an operation performed in the correct operation process. Hereinafter, the configurations of the learning unit 110 and the inference unit 120 will be described in more detail.

[0029] As shown in FIG. 4, the learning unit 110 includes a learning data acquisition unit 111, a learning data editing unit 112, a learning execution unit 113, and a learning model storage unit 114.

[0030] The learning data acquisition unit 111 acquires, for example, from the control device 204, waveform data indicating the time-series change of the strain amount of the column 202 when work is performed in a typical work process. The learning data acquisition unit 111 outputs the acquired waveform data to the learning data editing unit 112.

[0031] The learning data editing unit 112 acquires the waveform data output from the learning data acquisition unit 111. The learning data editing unit 112 performs data editing so that the acquired waveform data can be used as teacher data for machine learning performed by the learning execution unit 113.

[0032] For example, the learning data editing unit 112 converts the acquired waveform data indicating the time-series change of the strain amount of the column 202 into waveform data indicating the time-series change of the weight. Further, the learning data editing unit 112 divides the waveform data indicating the weight change of the entire series of work processes into waveform data for each work included in the work process (for example, divides the series of waveform data shown in FIG. 3 into waveform data for each of the events (a) to (j)). The learning data editing unit 112 outputs the waveform data of the entire series of work processes and the waveform data for each work to the learning execution unit 113, respectively.

[0033] The learning execution unit 113 acquires the waveform data of the entire series of work processes and the waveform data for each work output from the learning data editing unit 112. The learning execution unit 113 performs machine learning using the acquired waveform data of the entire series of work processes as teacher data. Thereby, the learning execution unit 113 generates a learning model that outputs an evaluation result of the entire work process for the input of the waveform data of the entire series of work processes. Also, the learning execution unit 113 performs machine learning using the waveform data for each work as teacher data. Thereby, the learning execution unit 113 generates a learning model that outputs an evaluation result for each work for the input of the waveform data of each work. The learning execution unit 113 stores the generated learned learning model in the learning model storage unit 114.

[0034] Here, the learning execution unit 113 performs machine learning using the waveform data of the entire series of work processes and the waveform data for each work, respectively, but a configuration in which machine learning is performed using only one of the waveform data may be used.

[0035] The learning model storage unit 114 stores the learned learning model generated by the learning execution unit 113. The learning model storage unit 114 is configured to include a storage medium such as a magnetic disk or a semiconductor memory.

[0036] As shown in FIG. 4, the inference unit 120 includes an inference data acquisition unit 121, an inference data editing unit 122, an inference execution unit 123, and an inference result output unit 124.

[0037] The inference data acquisition unit 121 acquires from the control device 204 waveform data showing the time-series change of the strain amount of the support column 202, which is the waveform data of the work process to be evaluated. The inference data acquisition unit 121 outputs the acquired waveform data to the inference data editing unit 122.

[0038] The inference data editing unit 122 acquires the waveform data output from the inference data acquisition unit 121. The inference data editing unit 122 performs data editing on the acquired waveform data so that it can be used for inference using machine learning performed by the inference execution unit 123.

[0039] For example, the inference data editing unit 122 converts the waveform data indicating the time-series change of the strain amount of the column 202 that has been acquired into waveform data indicating the time-series change of the weight. Further, the inference data editing unit 122 divides the waveform data indicating the weight change of the entire series of work processes into waveform data for each work included in the work process (for example, divides the series of waveform data shown in FIG. 3 into waveform data for each event (a) to (j)). The inference data editing unit 122 outputs the waveform data of the entire series of work processes and the waveform data for each work to the inference execution unit 123, respectively.

[0040] The inference execution unit 123 acquires the waveform data of the entire series of work processes and the waveform data for each work output from the inference data editing unit 122. Further, the inference execution unit 123 acquires a learned learning model that has been machine-learned using the waveform data of the entire series of work processes as teacher data from the learning model storage unit 114. The inference execution unit 123 performs inference by inputting the waveform data of the entire series of work processes acquired from the inference data editing unit 122 into the acquired learned learning model. Thereby, an evaluation of the entire work process to be evaluated is performed. The inference execution unit 123 obtains, for example, an evaluation result indicating whether a series of operations of the work process to be evaluated are performed in the correct order and at the correct intervals.

[0041] In addition, the inference execution unit 123 acquires a learned learning model, in which machine learning has been performed using the waveform data for each operation as teacher data, from the learning model storage unit 114. The inference execution unit 123 executes inference by inputting the waveform data for each operation acquired from the inference data editing unit 122 into the acquired learned learning model. Thereby, an evaluation for each operation is performed. The inference execution unit 123 obtains, for example, an evaluation result indicating whether each operation in the operation process to be evaluated has been performed in the correct method and with the correct content. The inference execution unit 123 outputs information indicating the above evaluation result output from the learned learning model and the waveform data used for the inference (that is, the waveform data input into the learned learning model) to the inference result output unit 124.

[0042] The inference result output unit 124 acquires information indicating the evaluation result and the waveform data used for the inference output from the inference execution unit 123. The inference result output unit 124 outputs the information indicating the evaluation result. The inference result output unit 124 includes a display device such as a liquid crystal display (LCD), an organic EL (Electroluminescence) display, or a cathode-ray tube (CRT) monitor, and displays the information indicating the evaluation result. Note that, for example, the inference result output unit 124 may be configured to include a communication interface and output the information indicating the evaluation result to an external device (not shown).

[0043] In addition, when configuring the process management system 1 to perform real-time evaluation of the operation while the operation by the operator is being performed, the inference result output unit 124 may be a notification device (warning device) such as a buzzer. That is, it is also possible to configure to perform real-time notification (warning) to the operator at the timing when it is detected that the operation by the operator is incorrect.

[0044] In addition, when the evaluation result based on the acquired information is an evaluation result indicating that the work has been performed in the correct work process (when no abnormality is detected), the inference result output unit 124 outputs the waveform data used in the inference to the learning execution unit 113. The learning execution unit 113 further performs machine learning using this waveform data as teacher data. In this way, by using the waveform data when it is determined that the work has been performed in the correct work process (that is, when no abnormality is detected in the work to be evaluated) as teacher data, the number of samples of the teacher data can be increased, and it becomes possible to generate a learning model capable of performing inference with higher accuracy.

[0045] Hereinafter, the operation during learning and the operation during inference of the server 10 will be described.

[0046] Hereinafter, an example of the operation during learning of the server 10 will be described. FIG. 5 is a flowchart showing the operation during learning by the server 10 of the embodiment.

[0047] The learning data acquisition unit 111 acquires waveform data indicating the time-series change of the strain amount of the column 202 when the work is performed in a typical work process from the control device 204 (ACT001). Next, the learning data editing unit 112 performs data editing so that the waveform data acquired from the learning data acquisition unit 111 can be used as teacher data for machine learning performed by the learning execution unit 113 (ACT002). The learning data editing unit 112 converts the acquired waveform data indicating the time-series change of the strain amount of the column 202 into waveform data indicating the time-series weight change. In addition, the learning data editing unit 112 divides the waveform data indicating the weight change of the entire series of work processes into waveform data for each work included in the work process.

[0048] Next, the learning execution unit 113 performs machine learning using the waveform data of the entire series of work processes obtained from the learning data editing unit 112 as teacher data. As a result, the learning execution unit 113 generates a learning model that outputs an evaluation result of the entire work process for an input of the waveform data of the entire series of work processes. Also, the learning execution unit 113 performs machine learning using the waveform data for each work obtained from the learning data editing unit 112 as teacher data. As a result, the learning execution unit 113 generates a learning model that outputs an evaluation result for each work for an input of the waveform data of each work (ACT003).

[0049] The learned model storage unit 114 stores the learned learning model generated by the learning execution unit 113 (ACT004). Thus, the operation of the server 10 during learning shown in the flowchart of FIG. 5 ends.

[0050] Hereinafter, an example of the operation of the server 10 during inference will be described. FIG. 6 is a flowchart showing the operation during inference by the server 10 of the embodiment.

[0051] The inference data acquisition unit 121 acquires, from the control device 204, waveform data showing the time-series change in the amount of distortion of the support column 202, which is the waveform data of the work process to be evaluated (ACT101). Next, the inference data editing unit 122 performs data editing on the waveform data acquired from the inference data acquisition unit 121 so that it can be used for inference using machine learning performed by the inference execution unit 123 (ACT102). The inference data editing unit 122 converts the acquired waveform data showing the time-series change in the amount of distortion of the support column 202 into waveform data showing the time-series weight change. Also, the inference data editing unit 122 divides the waveform data showing the weight change of the entire series of work processes into waveform data for each work included in the work process.

[0052] Next, the inference execution unit 123 acquires the waveform data of the entire series of work processes and the waveform data for each work acquired from the inference data editing unit 122. Also, the inference execution unit 123 acquires from the learned model storage unit 114 a learned learning model that has been machine-learned using the waveform data of the entire series of work processes as teacher data. The inference execution unit 123 executes inference by inputting the waveform data of the entire series of work processes acquired from the inference data editing unit 122 into the acquired learned learning model, and obtains an evaluation result. Further, the inference execution unit 123 acquires from the learned model storage unit 114 a learned learning model that has been machine-learned using the waveform data for each work as teacher data. The inference execution unit 123 executes inference by inputting the waveform data for each work acquired from the inference data editing unit 122 into the acquired learned learning model, and obtains an evaluation result (ACT103).

[0053] Next, the inference result output unit 124 outputs information indicating the evaluation result acquired from the inference execution unit 123 (ACT104). Also, when the evaluation result based on the acquired information is an evaluation result indicating that the work was performed in the correct work process (when no abnormality was detected in the work) (ACT105·YES), the inference result output unit 124 outputs the waveform data used for the inference to the learning execution unit 113.

[0054] Next, the learning execution unit 113 uses this waveform data as teacher data and performs machine learning (ACT106). Thus, the operation of the server 10 during inference shown in the flowchart of FIG. 6 ends.

[0055] Next, a modification example of the embodiment will be described.

[0056] As a first modification example, for example, a configuration may be adopted in which a camera (not shown) for photographing the state of work is further added to the configuration of the above-described process management system 1. In this case, the process management system is configured such that, for example, the time of the waveform data measured by the measuring device 20 is synchronized with the time of the video data photographed by the camera. When the server detects an abnormality (for example, that an incorrect operation has been performed) from the waveform data, the server specifies the time at which the abnormality occurred. The server cuts out the video data for a predetermined period (for example, 10 seconds before and after the specified time) including the specified time. Then, the server outputs (displays) the cut-out video data together with the information indicating the evaluation result.

[0057] By providing such a configuration, the process management system of the first modification example can easily realize the function of more detailedly checking the status of the work performed by the operator while checking the video later.

[0058] As a second modification example, for example, instead of performing the evaluation of the work using machine learning, the process management system may be configured to analyze the waveform data by a predetermined method and evaluate the work. The predetermined method is, for example, calculating and evaluating the degree of coincidence between the waveform of the exemplary waveform data (hereinafter referred to as "exemplary data") and the actually measured waveform by an arbitrary method. Note that the predetermined method may be, for example, evaluating the degree of coincidence between the intervals of the waveform of the exemplary data and the intervals of the actually measured waveform.

[0059] In this case, the exemplary data may be generated using machine learning. For example, a large number of waveform data obtained when work is performed in the correct work process are prepared as input data, and the server generates waveform data having typical features using machine learning. The server may use the generated waveform data as the exemplary data.

[0060] As described above, the server 10 of the process management system 1 of the embodiment is configured to analyze the work performed by the operator using the waveform data indicating the weight change of the product on the measurement top plate 201. By having such a configuration, the process management system 1 of the embodiment does not require cooperation with external devices such as work tools that are necessary in conventional process management systems. Also, by having such a configuration, the process management system 1 of the embodiment does not have the problem that occurs in conventional process management systems using cameras, such as an increase in the amount of calculation due to an increase in the amount of data processed on the edge side (measurement device side). As a result, the process management system 1 of the embodiment does not require a high-performance arithmetic processing device and can suppress an increase in equipment costs and the like.

[0061] In this way, the process management system 1 of the embodiment can analyze the work performed with higher accuracy while suppressing an increase in the amount of calculation.

[0062] Also, since the process management system 1 of the embodiment can suppress an increase in the amount of calculation, it can also perform the analysis of the work process in real time. Further, since the process management system 1 of the embodiment analyzes the work performed by the operator using the waveform data indicating the weight change, it can also analyze how the force is applied to the product and the top plate, which cannot be detected from the video by the camera. In addition, since the process management system 1 of the embodiment does not require the state of the work to be photographed by a camera, it can reduce the psychological burden on the operator caused by being monitored.

[0063] Note that the measuring device 20 may transmit to the server 10, among the measurement data obtained by measurement, the measurement data having a value deviated from the average value excluded. Thereby, the amount of data of the measurement data to be transmitted can be reduced. Also, the measuring device 20 may temporarily store the measurement data obtained by measurement in a temporary register, and transmit it to the server 10 all at once when a predetermined amount of measurement data is accumulated. Thereby, the number of communications can be reduced.

[0064] Further, the measurement device 20 may calculate the difference between the measurement data obtained by measurement and the measurement data obtained by measurement at the previous timing, and transmit the measurement data to the server 10 when the difference amount is more than a predetermined threshold value. Thereby, the data amount of the measurement data to be transmitted can be reduced. Further, the measurement device 20 may consider that there is a variation in the measurement data only when the state where the difference amount does not exceed the predetermined threshold value continues for a predetermined number of times (for example, 10 times). Thereby, the data amount of the measurement data to be transmitted can be reduced.

[0065] Note that the process management system 1 of the embodiment can specify the time when each operation is started and the time required for each operation based on the shape of the waveform included in the waveform data obtained by measurement and the interval between the waveforms. Thereby, the process management system 1 can easily consider the points for improving the work process.

[0066] For example, as points for improving the work process, the following points can be considered. First, since it becomes possible to grasp the time taken for the element work, it becomes possible to specify the work that takes a long time. Furthermore, as the number of assembled products increases and the measurement results of the time of the element work are accumulated, it becomes possible to grasp the variation in the time taken for each work. The greater the variation in the work, the higher the possibility that the work difficulty is high or the difference due to the skill level of the worker is large, and it also becomes possible to specify such work.

[0067] According to the above-described embodiment, the analysis device includes a measurement data acquisition unit and a control unit. For example, the analysis device is the server 10 in the embodiment, and the measurement data acquisition unit is the inference data acquisition unit 121 in the embodiment. The measurement data acquisition unit acquires measurement data indicating the time-series weight change of an object on the top plate. For example, the top plate is the measurement top plate 201 in the embodiment, the object is the product in the embodiment, and the measurement data indicating the weight change is the measurement data indicating the amount of distortion of the support column 202 that instructs the measurement top plate 201 in the embodiment. The control unit acquires the measurement data and standard data indicating the time-series weight change that occurs when the work on the object is performed in the correct work process on the top plate. The control unit evaluates the work based on the difference between the feature amount of the measurement data and the feature amount of the standard data.

[0068] In the above analysis device, the control unit may perform an evaluation to identify whether the work on the object is performed in the correct work process.

[0069] The above analysis device may further include a learning unit. For example, the learning unit is the learning execution unit 113 in the embodiment. The learning unit generates a learning model that outputs an evaluation result for the input of the measurement data by performing machine learning using the standard data. The control unit performs the evaluation by inputting the acquired measurement data into the learning model.

[0070] In the above analysis device, when it is determined by the evaluation that there is no abnormality in the work, the control unit may output the measurement data input to the learning model to the learning unit. The learning unit further performs machine learning using the measurement data acquired from the control unit.

[0071] In the above analysis device, the measurement data and the standard data may be waveform data. The control unit performs the evaluation based on the similarity between the feature amount of the waveform included in the standard data and the feature amount of the waveform included in the standard data.

[0072] In the above analysis device, the control unit may perform an evaluation based on at least one of the shape, size, and the interval between waveforms included in the waveform data.

[0073] In the above analysis device, the measurement data may be data indicating a change in the amount of strain measured by a measuring device that measures the amount of strain of the top plate support portion that supports the top plate. For example, the top plate support portion is the support column 202 in the embodiment, and the measuring device is the measuring device 20 in the embodiment.

[0074] In the above analysis device, the measurement data may be data indicating a change in weight measured by a measuring device that measures the weight of an object on the top plate.

[0075] In the above analysis device, the object may be an industrial product, and the working process may be a process of assembling the industrial product.

[0076] The above analysis device may further include a camera. The camera captures the state of the work. When the control unit determines that there is an abnormality in the work based on the evaluation, it identifies the time when the abnormality occurred based on the feature amount of the measurement data, and outputs video data including the video at the identified time.

[0077] Part of the functions of the process management system 1 in each of the above-described embodiments may be implemented by a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed to be realized. Here, the “computer system” shall include hardware such as an OS and peripheral devices. Further, the “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, etc., and a storage device such as a hard disk built in a computer system. Furthermore, the “computer-readable recording medium” also includes, like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line, something that dynamically holds a program for a short time, and something that holds a program for a certain period of time, like a volatile memory inside a computer system that becomes a server or a client in that case. Also, the above program may be for realizing part of the functions described above, and may further be something that can be realized in combination with a program already recorded in a computer system for realizing the functions described above.

[0078] Although some embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalent scope thereof.

Explanation of Reference Numerals

[0079] 1…Engineering management system, 10…Server, 20…Measuring device, 30…Network, 50…Work machine, 60…Cart, 110…Learning unit, 111…Learning data acquisition unit, 112…Learning data editing unit, 113…Learning execution unit, 114…Learning model storage unit, 120…Inference unit, 121…Inference data acquisition unit, 122…Inference data editing unit, 123…Inference execution unit, 124…Inference result output unit, 201…Measuring top plate, 202…Support column, 203…Strain gauge, 204…Control device, 501…Top plate

Claims

1. A measurement data acquisition unit that acquires measurement data indicating the time-series weight change of an object on a top plate, A control unit that acquires the measurement data and standard data indicating the time-series weight change that occurs when the work on the object is performed in the correct work process on the top plate, and evaluates the work based on the difference between the feature amount of the measurement data and the feature amount of the standard data, An analysis device comprising:

2. The control unit performs the evaluation of specifying whether the work on the object is performed in the correct work process The analysis device according to claim 1.

3. A learning unit that generates a learning model that outputs the result of the evaluation for the input of the measurement data by performing machine learning using the standard data further comprising, The control unit performs the evaluation by inputting the acquired measurement data into the learning model The analysis device according to claim 1.

4. When the control unit determines that there is no abnormality in the work based on the evaluation, the control unit outputs the measurement data input to the learning model to the learning unit, The learning unit further performs the machine learning using the measurement data acquired from the control unit The analysis device according to claim 3.

5. The measurement data and the standard data are waveform data, The control unit performs the evaluation based on the similarity between the feature amount of the waveform included in the standard data and the feature amount of the waveform included in the standard data The analysis device according to claim 1.

6. The control unit performs the evaluation based on at least one of the shape, size, and interval between the waveforms included in the waveform data The analysis device according to claim 5.

7. The measurement data is data indicating a change in the amount of distortion measured by a measuring device that measures the amount of distortion of a top plate support portion that supports the top plate The analysis device according to claim 1.

8. The measurement data is data indicating the weight change measured by a measuring device that measures the weight of the object on the top plate The analysis device according to claim 1.

9. The object is an industrial product, The work process is a process of assembling the industrial product The analysis device according to claim 1.

10. A camera that photographs the state of the work further comprising, When it is determined that there is an abnormality in the operation based on the evaluation, the control unit identifies the time at which the abnormality occurred based on the feature amount of the measurement data, and outputs video data including the video at the identified time. The analysis device according to claim 1.

11. An analysis method by a computer, comprising: measuring the weight of an object on a top plate; generating measurement data indicating changes in the weight over time; acquiring standard data indicating changes in the weight over time that occur when an operation on the object is performed in a correct operation process on the top plate; evaluating the operation based on a difference between a feature amount of the measurement data and a feature amount of the standard data; An analysis method having the above steps.

Images