Work evaluation device and work evaluation method

JPWO2024142192A5Active Publication Date: 2024-11-26MITSUBISHI ELECTRIC CORP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2024551910
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-11-26
Estimated Expiration
2042-12-27

AI Technical Summary

Technical Problem

Conventional work evaluation techniques require pre-classification of skill levels and rely on direct comparison with standard workers, making it difficult to accurately assess the quality of work without prior knowledge of the worker's skill range and reflecting the variability in work habits of standard workers.

Method used

A work evaluation device that acquires time-series target data and normative data, extracts features, calculates similarity, and assigns a work score based on the similarity with standard worker data, allowing for evaluation without pre-defined skill ranges and considering the work characteristics of standard workers.

Benefits of technology

Enables accurate and efficient evaluation of work quality by determining the work level and type, providing actionable feedback for improvement without requiring pre-classified skill ranges and reducing the need for extensive normative data preparation.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention comprises a subject data acquisition unit (111) that acquires time-series subject data, a normative data acquisition unit (112) that acquires a plurality of sets of time-series normative data, a preprocessing unit (12) that generates subject feature data relating to features extracted from the subject data, and a plurality of sets of normative feature data relating to features extracted from the plurality of sets of normative data, respectively, and a similarity level calculation unit (13) that compares the subject feature data with each of the plurality of sets of normative feature data to calculate a similarity level between the subject feature data and each of the plurality of sets of normative feature data.
Need to check novelty before this filing date? Find Prior Art

Description

Work evaluation device and work evaluation method

[0001] The present disclosure relates to an activity evaluation device and an activity evaluation method.

[0002] Conventionally, techniques for evaluating the quality of human work have been known. For example, Patent Literature 1 discloses an evaluation device that determines the skill level of a new welding worker whose skill level is to be determined based on a learning model and original data for determination, including behavior data representing the behavior of the welding worker, environmental data representing the work environment when the behavior data was measured, workpiece condition data representing the state of the workpiece when the behavior data was measured, and workpiece condition data representing the state of the workpiece when the behavior data was measured. The skill level indicates an expert, intermediate, or beginner. The learning model is generated by having welding workers with various skill levels perform welding work in advance, and performing machine learning using learning data based on data collected during the welding work as input data and the skill level as output data.

[0003] Japanese Patent Application Laid-Open No. 2021-1959

[0004] In the conventional work evaluation techniques described above, in order to evaluate the quality of a person's work, it was necessary to first classify the range of skill levels for the work performed by the person to be evaluated, such as what level of skill is required to perform the work and be considered an expert, an intermediate level, or a beginner level.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a work evaluation device that can evaluate the quality of a person's work without requiring that the range of skill required to perform the work be previously divided in accordance with the content of the work of the person to be evaluated.

[0006] The work evaluation device according to the present disclosure includes a target data acquisition unit that acquires time-series target data related to work that is generated when a person to be evaluated performs the work; a norm data acquisition unit that acquires multiple time-series norm data related to the work that is generated when multiple norm workers perform the work; a pre-processing unit that generates target feature data related to features extracted from the target data acquired by the target data acquisition unit and multiple norm feature data related to features extracted from each of the multiple norm data acquired by the norm data acquisition unit; and a similarity calculation unit that compares the target feature data with each of the multiple norm feature data and calculates the similarity of the target feature data with each of the norm feature data.

[0007] According to the present disclosure, a work evaluation device can evaluate the quality of a person's work without requiring that the range of skill required to perform the work be previously divided in accordance with the content of the work of the person to be evaluated.

[0008] Fig. 5A is a diagram showing an example of the configuration of the work evaluation device according to embodiment 1. Fig. 5B is a diagram showing an example of an evaluation result screen displayed by the work evaluation device according to embodiment 1. Fig. 5C is a diagram showing another example of the evaluation result screen displayed by the work evaluation device according to embodiment 1. Fig. 5D is a flowchart for explaining the operation of the work evaluation device according to embodiment 1. Figs. 5A and 5B are diagrams showing an example of the hardware configuration of the work evaluation device according to embodiment 1.

[0009] Embodiments of the present disclosure will be described in detail below with reference to the drawings. Embodiment 1. An activity evaluation device according to Embodiment 1 evaluates an activity performed by a worker (hereinafter referred to as a "activity evaluation subject") whose activity is to be evaluated. Specifically, the activity evaluation device according to Embodiment 1 evaluates the activity performed by the activity evaluation subject by determining to what extent the level of the activity performed by the activity evaluation subject (hereinafter referred to as the "activity level") reaches the activity levels of multiple model workers (hereinafter referred to as "model workers"). Note that in Embodiment 1, the model workers are assumed to be workers skilled in the activity, so-called veterans. Furthermore, in Embodiment 1, the activity level is indicated, for example, by an activity score. In Embodiment 1, the activity evaluation device evaluating the activity performed by the activity evaluation subject means calculating an activity score indicating the activity level of the activity performed by the activity evaluation subject. In the following Embodiment 1, as an example, the activity score is represented by a numerical value between "0" and "1." For example, the closer the task level of the task performed by the task evaluation subject is estimated to be to the task levels of tasks performed by multiple standard workers, the larger the task score calculated by the task evaluation device. In other words, the closer the task score calculated by the task evaluation device is to "1," the closer the task performed by the task evaluation subject to that task score is estimated to be to the task performed by standard workers, or in other words, the more skilled the task is estimated to be.

[0010] In the first embodiment, as an example, the work to be evaluated by the work evaluation device is assumed to be brazing work. The work evaluation device evaluates the brazing work performed by the work evaluation subject by determining to what extent the work level of the brazing work performed by the work evaluation subject on an object to be brazed (hereinafter referred to as a "work") reaches the work level of the brazing work performed by a standard worker who is skilled in the brazing work.

[0011] The work processes until the brazing of the workpiece is completed include a heating process for heating the workpiece and a brazing process for supplying brazing material to the workpiece after the heating process. The worker performing the brazing completes the brazing work on the workpiece by performing the work required in the heating process and the work required in the brazing process. In this way, the series of work performed by the worker until the brazing of the workpiece is completed includes multiple different types of work corresponding to each work process. In the first embodiment, the type of work corresponding to each work process is referred to as a work type. In the first embodiment, the brazing work includes two work types: work of the work type "heating" (work required in the heating process) and work of the work type "brazing" (work required in the brazing process).

[0012] 1 is a diagram showing an example configuration of an work evaluation device 1 according to embodiment 1. The work evaluation device 1 is connected to a sensor 2, a normative data storage device 3, and an output device 4. The work evaluation device 1, the sensor 2, the normative data storage device 3, and the output device 4 constitute a work evaluation system.

[0013] The sensor 2 is placed around the work evaluation subject or the workpiece, such as a brazing device (not shown), and acquires time-series data (hereinafter referred to as "target data") related to the work (here, brazing work) generated when the work evaluation subject performs the work. The target data includes, for example, data indicating changes in the state of the workpiece due to the brazing work performed by the work evaluation subject. More specifically, the target data includes data indicating the location of a fire on the workpiece, data indicating the position of the brazing filler metal, or data indicating the surface temperature of the workpiece. In the first embodiment, the sensor 2 includes a camera. The camera, for example, three-dimensionally measures the movement of a gas burner (not shown) or flame and the movement of the brazing filler metal when the evaluation subject brazes the workpiece, and acquires the data as digital data. The sensor 2 outputs the acquired target data to the work evaluation device 1.

[0014] The normative data storage device 3 stores time-series data (hereinafter referred to as "normative data") related to a task (here, brazing) generated by a standard worker performing the task. The content of the data included in the normative data is the same as the content of the data included in the target data. That is, the normative data is data indicating changes in the state of the workpiece associated with brazing work performed by the standard worker, including data indicating the location of a fire on the workpiece, data indicating the location of the brazing insert, and data indicating the surface temperature of the workpiece. For example, a manager or the like who evaluates the work evaluation subjects may have multiple standard workers perform the same task (here, brazing work) as the work to be evaluated by the work evaluation subjects in advance to collect multiple normative data, and store the collected multiple normative data in the normative data storage device 3. Note that, although the normative data is assumed to be stored in the normative data storage device 3 here, this is merely an example. For example, the normative data may be stored in a database (not shown). For example, a manager or the like may collect multiple normative data in advance and generate a database in which the normative data is stored.

[0015] The output device 4 is, for example, a display device. The output device 4 may be, for example, a storage device such as a server. In the first embodiment, as an example, the output device 4 is a display device.

[0016] The work evaluation device 1 is mounted on, for example, a brazing device. The work evaluation device 1 evaluates the brazing work performed by the work evaluation subject based on target data acquired from a sensor 2 and normative data stored in a normative data storage device 3. The work evaluation device 1 then outputs the results of the evaluation of the brazing work performed by the work evaluation subject to an output device 4. In detail, the work evaluation device 1 displays the results of the evaluation of the brazing work performed by the work evaluation subject on a display device.

[0017] The task evaluation device 1 includes a data acquisition unit 11, a pre-processing unit 12, a similarity calculation unit 13, a score calculation unit 14, a task type estimation unit 15, a representative normative data selection unit 16, and an output control unit 17. The data acquisition unit 11 includes a target data acquisition unit 111 and a normative data acquisition unit 112.

[0018] The data acquisition unit 11 acquires target data and normative data.

[0019] The target data acquisition unit 111 of the data acquisition unit 11 acquires, from the sensor 2, time-series target data related to the work, which is generated when the work evaluation subject performs the work. Specifically, the target data acquisition unit 111 acquires, from the sensor 2, time-series target data related to the brazing work, which is generated when the work evaluation subject performs the brazing work. The data length of the target data may be any length. The target data acquired by the target data acquisition unit 111 from the sensor 2 may be multidimensional data including multiple different types of data, such as data indicating a fire location, data indicating a brazing insert location, and data indicating a surface temperature, or may be one-dimensional data including only one type of data, such as data indicating a surface temperature. For example, the target data acquisition unit 111 may acquire multidimensional target data from the sensor 2 and extract one type of data from the multidimensional target data to create one-dimensional target data. In the following first embodiment, as an example, the target data acquired by the target data acquisition unit 111 from the sensor 2 is one-dimensional data indicating a surface temperature. The target data acquisition unit 111 outputs the acquired target data to the pre-processing unit 12.

[0020] The normative data acquisition unit 112 of the data acquisition unit 11 acquires, from the normative data storage device 3, multiple time-series normative data related to the brazing work, generated by multiple reference workers performing the work. Specifically, the normative data acquisition unit 112 acquires, from the normative data storage device 3, multiple time-series normative data related to the brazing work, generated by multiple reference workers performing the brazing work. Note that, like the target data, the normative data acquired by the normative data acquisition unit 112 from the normative data storage device 3 may be multidimensional data including multiple different types of data, or one-dimensional data including only one type of data. For example, the normative data acquisition unit 112 may acquire multidimensional target data from the normative data storage device 3 and extract one type of data from the multidimensional target data to create one-dimensional normative data. It is sufficient that the content of the target data and the content of the normative data are the same. In the following first embodiment, as an example, the normative data acquired by the normative data acquisition unit 112 from the normative data storage device 3 is one-dimensional data indicating surface temperature. The normative data acquisition unit 112 outputs the acquired normative data to the pre-processing unit 12.

[0021] The pre-processing unit 12 performs pre-processing to generate data related to features extracted from the target data acquired by the target data acquisition unit 111 (hereinafter referred to as "target feature data") and multiple data related to features extracted from each of the multiple sets of normative data acquired by the normative data acquisition unit 112 (hereinafter referred to as "normative feature data"). Specifically, the pre-processing unit 12 first chronologically segments the target data of any length into multiple sets of fixed-length data (hereinafter referred to as "segmented target data"). The pre-processing unit 12 then extracts features from each of the multiple sets of segmented target data and generates target feature data for each of the extracted features. The target feature data is, for example, data in the form of a vector. The pre-processing unit 12 also segments each of the multiple sets of normative data of any length into multiple sets of fixed-length data (hereinafter referred to as "segmented normative data"). The pre-processing unit 12 then extracts features from each of the multiple sets of segmented normative data and generates normative feature data for each of the extracted features. The normative feature data is, for example, data in the form of a vector. Note that, if the plurality of normative data are data that have already been segmented into fixed lengths, the pre-processing unit 12 may simply treat the plurality of normative data as the plurality of segmented normative data. The pre-processing unit 12 outputs the generated plurality of target feature data and the generated plurality of normative feature data to the similarity calculation unit 13. The pre-processing unit 12 associates each of the generated plurality of target feature data with the target data from which the features were extracted, and outputs the corresponding data to the similarity calculation unit 13. The pre-processing unit 12 also associates each of the generated plurality of normative feature data with the normative data from which the features were extracted, and outputs the corresponding data to the similarity calculation unit 13.

[0022] The similarity calculation unit 13 compares each of the target feature data with the plurality of normative feature data based on the plurality of target feature data and the plurality of normative feature data output from the pre-processing unit 12. The similarity calculation unit 13 performs the above comparison for each of the target feature data generated by the pre-processing unit 12 and calculates the similarity of the target feature data with each normative feature data.

[0023] The similarity calculation unit 13 may use a known method to compare the target feature data with each piece of normative feature data and calculate the similarity of the target feature data with each piece of normative feature data. For example, the similarity calculation unit 13 may input the target feature data and the normative feature data output from the pre-processing unit 12 into a trained model (hereinafter referred to as a "machine learning model") that receives the target feature data and the normative feature data as input and outputs a similarity, thereby obtaining the similarity and calculating the similarity. For example, the similarity calculation unit 13 may use an attention method to calculate the attention_weight between the waveform of the target feature data and the waveform of the normative feature data as the similarity. For example, the similarity calculation unit 13 may calculate the similarity according to rules that are generated in advance by an administrator or the like and stored in a location that the similarity calculation unit 13 can refer to.

[0024] The similarity calculation unit 13 outputs data on multiple similarities calculated for each piece of object feature data (hereinafter referred to as "similarity data") to the score calculation unit 14, the task type estimation unit 15, and the representative normative data selection unit 16. The similarity data is data in which, for each piece of object feature data, the object feature data, the target data, and multiple similarities are associated with each other. Note that the similarity calculation unit 13 simply associates the object data output from the pre-processing unit 12 in association with the object feature data with the object feature data in the similarity data. The similarity data may also be data in which, for each piece of object feature data, the object feature data, the target data, multiple similarities, the normative feature data used to calculate the similarities, and the normative data are associated with each other. Note that the similarity calculation unit 13 simply associates the normative data output from the pre-processing unit 12 in association with the normative feature data with the normative feature data in the similarity data.

[0025] The score calculation unit 14 calculates a work score that indicates how closely the work level of the work performed by the work evaluation subject has reached the work level of work performed by multiple standard workers, based on the similarity data output from the similarity calculation unit 13.

[0026] Specific examples of the method for calculating the task score by the score calculation unit 14 will be described below. In the following specific examples (<Specific Example 1> and <Specific Example 2>), it is assumed that the similarity calculation unit 13 calculates the similarity of certain object feature data (first object feature data) with five pieces of normative feature data (first normative feature data, second normative feature data, third normative feature data, fourth normative feature data, and fifth normative feature data), and the results are that the similarity with the first normative feature data is "0.8," the similarity with the second normative feature data is "0.8," the similarity with the third normative feature data is "0.9," the similarity with the fourth normative feature data is "0.7," and the similarity with the fifth normative feature data is "0.3." That is, it is assumed that similarity data in which first object feature data, object data corresponding to the first object feature data, and values ​​such as "0.8," "0.8," "0.9," "0.7," and "0.3" are associated with each other is output from the similarity calculation unit 13 to the score calculation unit 14. The score calculation unit 14 calculates an activity score corresponding to the first object feature data based on the similarity data described above. In detail, the score calculation unit 14 calculates an activity score for an activity indicated by the first object feature data based on the similarity data described above. The activity indicated by the first object feature data is an activity performed by the person being evaluated from which the object data from which the first object feature data was extracted was obtained.

[0027] <Specific Example 1> For example, for first object feature amount data, the score calculation unit 14 calculates the task score from the largest value among the similarities associated with the first object feature amount data. In this case, the score calculation unit 14 first selects "0.9," the largest value among the similarities associated with the first object feature amount data. Then, the score calculation unit 14 calculates the task score for the first object feature amount data from the selected "0.9." More specifically, the score calculation unit 14 calculates the task score for the task indicated by the first object feature amount data from the selected "0.9." The score calculation unit 14 calculates the task score by converting the selected similarity into the task score. For example, the score calculation unit 14 multiplies the selected similarity by 100. Specifically, the score calculation unit 14 calculates 0.9 x 100 = 90 as the task score. Note that this is merely an example, and the score calculation unit 14 may, for example, use the selected similarity as the task score as is. That is, the score calculation unit 14 may set “0.9” as the work score. A conversion formula for converting the selected similarity into the work score is set in advance and stored in a location that can be referenced by the score calculation unit 14.

[0028] <Specific Example 2> For example, the score calculation unit 14 may calculate the task score for a first object feature amount data from the average value of the similarities associated with the first object feature amount data. In this case, the score calculation unit 14 first calculates the average value of the similarities "0.8," "0.8," "0.9," "0.7," and "0.3" associated with the first object feature amount data. In this example, the score calculation unit 14 calculates the average value "0.7." The score calculation unit 14 calculates the task score by converting the calculated average value of the similarities into the task score. For example, the score calculation unit 14 multiplies the calculated average value of the similarities by 100. Specifically, the score calculation unit 14 calculates 0.7 x 100 = 70 as the task score. Note that this is merely an example, and the score calculation unit 14 may use "0.7" as the task score. A conversion formula for converting the calculated average value of the similarities into the task score is set in advance and stored in a location that can be referenced by the score calculation unit 14.

[0029] The above-mentioned <Specific Example 1> and <Specific Example 2> are merely examples. The score calculation unit 14 may calculate the task score by a method other than the above-mentioned <Specific Example 1> and <Specific Example 2>. For example, the score calculation unit 14 may calculate the task score by a method such as that shown in the following <Specific Example 3>.

[0030] <Specific Example 3> For example, the score calculation unit 14 may calculate the task score by weighting the similarity associated with certain object feature data (second object feature data). Here, weighting refers to, for example, padding. In detail, for example, if all similarities associated with the second object feature data are smaller than a predetermined target similarity (hereinafter referred to as "target similarity") but are close to the target similarity, the score calculation unit 14 sets the target similarity as the similarity associated with the second object feature data. If the difference between the similarity and the target similarity is within a predetermined threshold (hereinafter referred to as "difference determination threshold"), the score calculation unit 14 determines that the similarity is close to the target similarity. For example, suppose that the similarity calculation unit 13 calculates the similarities of the second object feature data with the first normative feature data, the second normative feature data, the third normative feature data, the fourth normative feature data, and the fifth normative feature data, respectively, and finds that the similarity with the first normative feature data is "0.8," the similarity with the second normative feature data is "0.8," the similarity with the third normative feature data is "0.8," the similarity with the fourth normative feature data is "0.8," and the similarity with the fifth normative feature data is "0.8." That is, suppose that the similarity calculation unit 13 outputs to the score calculation unit 14 similarity data in which the second object feature data and the object data corresponding to the second object feature data are associated with each other as "0.8," "0.8," "0.8," "0.8," and "0.8." Furthermore, suppose that the target similarity is "0.9," and the difference determination threshold is "0.1." In this case, although all similarities ("0.8") are smaller than "0.9", the difference from "0.9" is within "0.1". Therefore, the score calculation unit 14 sets the similarity corresponding to the second object feature amount data to "0.9". In other words, the score calculation unit 14 weights the similarity associated with the second object feature amount data to set the similarity corresponding to the second object feature amount data to "0.9". Then, the score calculation unit 14 converts the weighted similarity into an action score to calculate the action score. For example, the score calculation unit 14 multiplies the weighted similarity by 100.Specifically, the score calculation unit 14 calculates 0.9 x 100 = 90 as the work score. Note that this is merely an example, and the score calculation unit 14 may also use "0.9" as the work score. A conversion formula for converting the weighted similarity into the work score is set in advance and stored in a location that can be referenced by the score calculation unit 14.

[0031] The rules for how the score calculation unit 14 calculates the task score based on the similarity data output from the similarity calculation unit 13 can be set appropriately by an administrator or the like.

[0032] The score calculation unit 14 outputs data related to the calculated task score (hereinafter referred to as "task score data") to the output control unit 17. The task score data is data in which, for each piece of object feature data, object data, and task score are associated with each other. The task score data may also be data in which, for each piece of object feature data, object feature data, object data, task score, normative feature data and normative data corresponding to the similarity used to calculate the task score are associated with each other. Note that when the feature data is data in which object feature data, multiple similarities, object data, normative feature data and normative data are associated with each other, the score calculation unit 14 can identify, from the feature data, the normative feature data and normative data corresponding to the similarity used to calculate the task score.

[0033] The task type estimation unit 15 estimates the task type of the task performed by the task evaluation subject. For each piece of target feature amount data, the task type estimation unit 15 estimates the task type of the task performed by the task evaluation subject corresponding to the target feature amount data. In the first embodiment, the task type estimation unit 15 estimates whether the task performed by the task evaluation subject corresponding to the target feature amount data is "heating" or "inserting wax." For example, the task type estimation unit 15 estimates the task type of the task performed by the task evaluation subject based on the similarity data output from the similarity calculation unit 13.

[0034] For example, the normative data stored in the normative data storage device 3 is provided with data indicating the task type so that it can be determined which task type the normative data was generated from by the norm worker. When the pre-processing unit 12 generates the normative feature data, it associates the normative feature data with data indicating the task type and outputs the associated data to the similarity calculation unit 13. The similarity calculation unit 13 generates similarity data in which, for each piece of target feature data, the target feature data, multiple similarities, the normative feature data used to calculate the similarities, and data indicating the task type are associated with each other. For example, based on the similarity data output from the similarity calculation unit 13, the task type estimation unit 15 estimates the task type associated with the largest similarity among the multiple similarities associated with the target feature data as the task type of the task corresponding to the target feature data, performed by the task evaluation subject.

[0035] Furthermore, for example, the task type estimation unit 15 may input the target feature data included in the similarity data into a machine learning model (hereinafter referred to as the "task type estimation model") that receives feature data as input and outputs information indicating the task type, thereby obtaining information indicating the task type, thereby estimating the task type of the task performed by the task evaluation subject. The task type estimation model is generated in advance and stored in a location accessible by the task type estimation unit 15. For example, the task type estimation unit 15 may directly acquire the target feature data to be input into the task type estimation model from the pre-processing unit 12. In this case, the pre-processing unit 12 associates each of the generated multiple target feature data with the target data from which the features were extracted, and outputs the resulting data to the task type estimation unit 15. Note that in FIG. 1 , the arrow from the pre-processing unit 12 to the task type estimation unit 15 is omitted.

[0036] Furthermore, for example, when estimating the task type of a task performed by a task evaluation subject, the task type estimation unit 15 may take into account the task score corresponding to the target feature amount data calculated by the score calculation unit 14. In this case, the task type estimation unit 15 may, for example, acquire task score data from the score calculation unit 14 and identify the task score corresponding to the target feature amount data from the acquired task score data. Note that the arrow from the score calculation unit 14 to the task type estimation unit 15 is omitted in FIG. For example, if the task score corresponding to the target feature amount data does not satisfy a certain value, the task type estimation unit 15 determines that the task is "unclassifiable."

[0037] The task type estimation unit 15 outputs data related to the estimated task type (hereinafter referred to as "task type estimation data") to the output control unit 17. The task type estimation data is data in which target feature amount data is associated with data indicating the task type. The task type estimation data may be data in which target feature amount data, target data corresponding to the target feature amount data, and data indicating the task type are associated with each other. Note that the task type estimation unit 15 can determine the target data corresponding to the target feature amount data from the similarity data.

[0038] The representative normative data selection unit 16 selects, from the plurality of normative data, normative data that is similar to the target data (hereinafter referred to as "representative normative data"), based on the similarity data output by the similarity calculation unit 13. The representative normative data selection unit 16 may select one or more representative normative data.

[0039] For example, the representative normative data selection unit 16 selects the largest similarity among the similarities associated with the target feature data in the similarity data, and sets the normative data associated with that similarity as the representative normative data. In this case, the similarity data is data in which, for each target feature data, the target data, multiple similarities, the normative feature data used to calculate the similarities, and the normative data are associated with each other. For example, if there are multiple largest similarities among the similarities associated with the target feature data in the similarity data, the representative normative data selection unit 16 may select any one of them and set the normative data associated with the selected similarity as the representative normative data, or may set all of the multiple normative data corresponding to the multiple similarities as the representative normative data.

[0040] Also, for example, the representative normative data selection unit 16 may select multiple (e.g., three) similarities in descending order from the similarities associated with the target feature data in the similarity data, and use the multiple (e.g., three) normative data associated with the similarities as representative normative data.

[0041] Furthermore, for example, when selecting representative normative data, the representative normative data selection unit 16 may take into consideration the task score corresponding to the target feature amount data calculated by the score calculation unit 14. In this case, the representative normative data selection unit 16 may, for example, acquire task score data from the score calculation unit 14 and identify the task score corresponding to the target feature amount data from the acquired task score data. Note that the arrow from the score calculation unit 14 to the representative normative data selection unit 16 is omitted in FIG. 1 . For example, if the task score corresponding to the target feature amount data does not satisfy a certain value, the representative normative data selection unit 16 determines that there is "no representative normative data."

[0042] The representative norm data selection unit 16 outputs data related to the selected representative norm data (hereinafter referred to as "selected norm data") to the output control unit 17. The selected norm data is data in which the target feature amount data is associated with the representative norm data selected by the representative norm data selection unit 16. The selected norm data may be data in which the target feature amount data and the target data corresponding to the target feature amount data are associated with the selected norm data. Note that the representative norm data selection unit 16 can determine the target data corresponding to the target feature amount data from the similarity data.

[0043] The output control unit 17 outputs data indicating the task score to the output device 4 based on the task score data output from the score calculation unit 14. In detail, the output control unit 17 causes the display device to display the data indicating the task score.

[0044] At this time, the output control unit 17 may associate the data indicating the task score with the target data and output it to the output device 4. In other words, the output control unit 17 may cause the display device to display the data indicating the task score with the target data.

[0045] The output control unit 17 may also cause the display device to display data indicating the task score in association with data indicating the task type. Note that the output control unit 17 can associate the task score with the task type, for example, by matching the task score data output from the score calculation unit 14 with the post-task type estimation data output from the task type estimation unit 15 using the target feature amount data.

[0046] The output control unit 17 may also cause the display device to display data indicating the task score in association with the representative norm data. Note that the output control unit 17 can associate the task score with the representative norm data by, for example, matching the task score data output from the score calculation unit 14 with the selected norm data output from the representative norm data selection unit 16 using the target feature amount data.

[0047] The output control unit 17 may cause the display device to display data indicating the task score in association with the target data, data indicating the task type, and representative normative data.

[0048] FIG. 2 is a diagram illustrating an example of a screen (hereinafter referred to as the "evaluation result display screen") that the output control unit 17 displays on the display device based on the task score data output from the score calculation unit 14, the task type estimation data output from the task type estimation unit 15, and the selected normative data output from the representative normative data selection unit 16 in the first embodiment. For example, as shown in FIG. 2, the output control unit 17 displays an evaluation result display screen that displays task scores in association with task types, target data, and representative normative data. The output control unit 17 can display multiple target data in chronological order on the evaluation result display screen. In the evaluation result display screen shown in FIG. 2, the vertical axis represents surface temperature, and the horizontal axis represents time. In FIG. 2, the time-series target data is indicated by "T," and the time-series representative normative data is indicated by "N." Note that, here, it is assumed that there is one representative normative data. The evaluation result display screen shown in Figure 2 shows target data and representative normative data for the work type "heating" up to time t, and indicates that the work score for the "heating" work is "0.8". Furthermore, the evaluation result display screen shown in Figure 2 shows target data and representative normative data for the work type "brazing" after time t, and indicates that the work score for the "brazing" work is "0.2".

[0049] In other words, the evaluation result display screen shown in Figure 2 indicates that the work level of the heating work in the brazing work performed by the work evaluation subject is "0.8," which means that the heating work performed by the work evaluation subject is close to the heating work performed by a representative standard worker, and that the work evaluation subject is skilled in heating work. The evaluation result display screen also indicates that the work level of the brazing work performed by the work evaluation subject is "0.2," which means that the level of the brazing work performed by the work evaluation subject has not yet reached the level of the brazing work performed by a representative standard worker. By checking the evaluation result display screen, a manager or the like can understand that the work evaluation subject has reached the level of proficiency in heating work, but has not yet reached the level of proficiency in brazing work. Furthermore, by checking the deviation between the target data and the representative normative data on the evaluation result display screen, the manager or the like can, for example, determine which aspects of the work evaluation subject's brazing work deviate from those of the normative worker, i.e., which tasks should be prioritized in training in order for the work evaluation subject to be deemed to have reached a level of proficiency in brazing work. For example, the manager or the like can provide guidance to the work evaluation subject based on the determined results, thereby improving the work evaluation subject's skills. In this way, by displaying data indicating the work score, the output control unit 17 can provide data that helps the work evaluation subject understand the current work level and improve the work evaluation subject's skills.

[0050] FIG. 3 is a diagram illustrating another example of an evaluation result display screen displayed on a display device by the output control unit 17 based on the task score data output from the score calculation unit 14, the task type estimation data output from the task type estimation unit 15, and the selected normative data output from the representative normative data selection unit 16 in the first embodiment. The evaluation result display screen shown in FIG. 3 differs from the evaluation result display screen shown in FIG. 2 in that multiple representative normative data are included. Here, it is assumed that there are three representative normative data, and the time-series representative normative data are indicated by "N1," "N2," and "N3" on the evaluation result display screen shown in FIG. 3. For example, as shown in FIG. 3, the output control unit 17 can also display the ranges of the multiple representative normative data on the evaluation result display screen so that they can be seen. In FIG. 3, the output control unit 17 indicates the ranges of the multiple representative normative data by diagonal lines on the evaluation result display screen.

[0051] Even so-called veteran model workers have various work habits. By selecting multiple representative model workers and displaying the target data along with multiple model data on the evaluation result display screen, managers, for example, can identify tasks that require particular training among the tasks performed by the work evaluation target worker. Tasks corresponding to target data that deviate from any of the multiple model data can be considered tasks that require particular training to be closer to the tasks performed by model workers.

[0052] Note that the evaluation result display screens shown in Figures 2 and 3 are merely examples. The output control unit 17 is only required to display at least the task score. For example, the output control unit 17 may display an evaluation result display screen that displays only the task score, or may display an evaluation result display screen that displays the task score and task type. The output control unit 17 may also display an evaluation result display screen that displays the task score and time-series target data, or may display an evaluation result display screen that displays the task score and representative normative data. The output control unit 17 may display an evaluation result display screen that appropriately combines the task score with the task type, target data, or representative normative data.

[0053] The operation of the work evaluation device 1 according to the first embodiment will be described. FIG. 4 is a flowchart for explaining the operation of the work evaluation device 1 according to the first embodiment. For example, when the work evaluation device 1 is turned on or when a start instruction is received from a manager or the like, the work evaluation device 1 executes the operation shown in the flowchart of FIG. 4 . For example, the manager or the like operates a touch panel display provided in the brazing device to input the start instruction. A control unit (not shown) of the work evaluation device 1 receives the start instruction input by the manager or the like and causes the data acquisition unit 11, pre-processing unit 12, similarity calculation unit 13, score calculation unit 14, work type estimation unit 15, representative normative data selection unit 16, and output control unit 17 to start their operations.

[0054] The target data acquisition unit 111 of the data acquisition unit 11 acquires, from the sensor 2, time-series target data related to the work, which has been generated when the work evaluation subject performs the work (step ST1). In detail, the target data acquisition unit 111 acquires, from the sensor 2, time-series target data related to the brazing work, which has been generated when the work evaluation subject performs the brazing work. The target data acquisition unit 111 outputs the acquired target data to the pre-processing unit 12.

[0055] The normative data acquisition unit 112 of the data acquisition unit 11 acquires, from the normative data storage device 3, a plurality of time-series normative data related to the work, which has been generated by a plurality of standard workers performing the work (step ST2). In detail, the normative data acquisition unit 112 acquires, from the normative data storage device 3, a plurality of time-series normative data related to the brazing work, which has been generated by a plurality of standard workers performing the brazing work. The normative data acquisition unit 112 outputs the acquired normative data to the pre-processing unit 12.

[0056] The pre-processing unit 12 performs pre-processing to generate object feature data related to the feature extracted from the object data acquired by the object data acquiring unit 111 in step ST1, and multiple normative feature data related to the feature extracted from each of the multiple normative data acquired by the normative data acquiring unit 112 in step ST2 (step ST3). The pre-processing unit 12 outputs the generated multiple object feature data and multiple normative feature data to the similarity calculation unit 13. The pre-processing unit 12 associates each of the generated object feature data with the object data from which the feature was extracted, and outputs the corresponding data to the similarity calculation unit 13. The pre-processing unit 12 also associates each of the generated normative feature data with the normative data from which the feature was extracted, and outputs the corresponding data to the similarity calculation unit 13.

[0057] Based on the plurality of object feature data and the plurality of normative feature data output from the pre-processing unit 12 in step ST3, the similarity calculation unit 13 compares each of the object feature data with the plurality of normative feature data and calculates the similarity of the object feature data with each normative feature data (step ST4). The similarity calculation unit 13 performs the above comparison for each of the object feature data generated by the pre-processing unit 12 and calculates the similarity of the object feature data with each normative feature data. The similarity calculation unit 13 outputs similarity data relating to the plurality of similarities calculated for each of the object feature data to the score calculation unit 14, the task type estimation unit 15, and the representative normative data selection unit 16.

[0058] The score calculation unit 14 calculates a task score indicating how well the task level of the task performed by the task evaluation target worker has reached the task levels of the tasks performed by the plurality of standard workers (step ST5) based on the similarity data output from the similarity calculation unit 13 in step ST4. The score calculation unit 14 outputs task score data relating to the calculated task score to the output control unit 17.

[0059] The task type estimation unit 15 estimates the task type of the task performed by the task evaluation subject based on the similarity data output from the similarity calculation unit 13 in step ST4 or the target feature amount data output from the pre-processing unit 12 in step ST3 (step ST6). The task type estimation unit 15 estimates the task type for each piece of target feature amount data. The task type estimation unit 15 outputs post-task type estimation data relating to the estimated task type to the output control unit 17.

[0060] The representative normative data selection unit 16 selects representative normative data from the plurality of normative data based on the similarity data output by the similarity calculation unit 13 in step ST4 (step ST7). The representative normative data selection unit 16 outputs selected normative data related to the selected representative normative data to the output control unit 17.

[0061] The output control unit 17 outputs data indicating the task score to the output device 4 based on the task score data output from the score calculation unit 14 in step ST5 (step ST8). In detail, the output control unit 17 causes the display device to display the data indicating the task score.

[0062] In the flowchart shown in FIG. 4 , the processing is performed in the order of step ST1 and step ST2, but this is merely an example. The order of the processing of step ST1 and step ST2 may be reversed, or the processing of step ST1 and step ST2 may be performed in parallel. In addition, in the flowchart shown in FIG. 4 , the processing is performed in the order of step ST5, step ST6, and step ST7, but this is merely an example. The order of the processing of step ST5, step ST6, and step ST7 may be reversed, or the processing of step ST5, step ST6, and step ST7 may be performed in parallel. However, when the task type estimation unit 15 estimates the task type of the task performed by the task evaluation subject based on the similarity data and the task score in step ST6, the task evaluation device 1 performs the processing of step ST5 before performing the processing of step ST6.

[0063] In this way, the task evaluation device 1 acquires time-series target data and multiple time-series normative data. After performing pre-processing to generate target feature data and multiple normative feature data, the task evaluation device 1 compares the target feature data with each of the multiple normative feature data and calculates the similarity of the target feature data with each of the normative feature data. The task evaluation device 1 then calculates a task score based on similarity data in which the target feature data is associated with multiple similarities. This allows the task evaluation device 1 to evaluate the quality of a task performed by a person (task evaluation subject) without requiring that the range of skill required for performing the task be previously defined in accordance with the content of the task performed by the person (task evaluation subject).

[0064] Even so-called veteran standard workers each have various work habits. When evaluating the work of a work evaluation subject (here, brazing work), it is extremely difficult to determine how to reflect the work characteristics of the standard worker in the evaluation. In other words, even if an attempt is made to evaluate the work of the work evaluation subject by comparing target data generated based on the work of the work evaluation subject with normative data generated based on the work of a standard worker, the feature values ​​of the normative data are not uniquely determined. Therefore, it is difficult to determine the degree to which the work level of the work performed by the work evaluation subject reaches the level of the work performed by a standard worker by comparing the target data with a single normative data. In contrast, the work evaluation device 1 according to the first embodiment acquires multiple normative data and compares the target data with the multiple normative data. More specifically, the work evaluation device 1 compares target feature data based on the target data with multiple normative feature data based on the multiple normative data, thereby calculating the similarity of the target feature data with each normative feature data. The work evaluation device 1 then calculates a work score based on similarity data in which the target feature data is associated with multiple similarities. This allows the task evaluation device 1 to determine the task level of the task performed by the task evaluation subject, taking into account the task characteristics of the model worker. In other words, the task evaluation device 1 can evaluate the task performed by the task evaluation subject, taking into account the task characteristics of the model worker.

[0065] The task evaluation device 1 also segments the target data into multiple segmented target data and generates target feature data for each of multiple feature amounts extracted from each of the multiple segmented target data. The task evaluation device 1 then calculates the similarity of each of the multiple generated target feature data with each of the normative feature data. This allows the task evaluation device 1 to prevent a decrease in learning efficiency due to an increase in the length of the target data. The task evaluation device 1 also prevents managers and others from having to prepare a huge amount of normative data in preparation for a decrease in the frequency of occurrence of normative data determined to be similar due to an increase in the length of the target data.

[0066] Furthermore, the task evaluation device 1 estimates the task type of the task (here, brazing) performed by the task evaluation subject based on the object feature data or similarity data. By providing the task score along with the task type, the task evaluation device 1 can provide an evaluation of the task performed by the task evaluation subject, more specifically, the task score, in a format that is easier to understand for evaluating the task performed by the task evaluation subject. For example, if the task evaluation subject is evaluated for performing a task using an incorrect procedure, the task evaluation device 1 can inform the subject that the task itself was not performed well by providing information on the task type. To give a specific example, for brazing, while the correct task procedure is heating → brazing insertion, if the task evaluation subject performs the task using a procedure such as heating → brazing insertion → heating, the task can be said to be poorly mastered.

[0067] Furthermore, the task evaluation device 1 may estimate the task type of the task performed by the task evaluation subject based on the target feature amount data or similarity data and the task score. For example, when the task evaluation device 1 estimates the task type using a task type estimation model, by also taking the task score into consideration, it becomes possible to externalize the task type estimation to some extent, which is expected to reduce the workload of estimating the task type using the task type estimation model; in other words, it is expected to improve the accuracy and efficiency of task type estimation, or reduce the amount of learning data required to generate the task type estimation model.

[0068] Furthermore, the task evaluation device 1 selects representative normative data that is similar to the target data from among the multiple normative data based on the similarity data. By providing task scores together with the representative normative data, the task evaluation device 1 can provide, for example, tasks that the task evaluation subject should practice but that have not yet reached the task level of the normative worker in an identifiable form.

[0069] Furthermore, the task evaluation device 1 may select representative normative data based on the similarity data and the task score. By selecting representative normative data in consideration of the task score, the task evaluation device 1 can improve the accuracy of the selection of representative normative data.

[0070] In the first embodiment, the activity level is represented by an activity score. However, this is merely an example. For example, the activity level may be represented by a similarity. Even in this manner, the activity evaluation device 1 can evaluate the quality of the activity of the activity evaluation subject without requiring that the range of skill required for performing the activity be previously determined in accordance with the content of the activity performed by the activity evaluation subject. The closer the activity level of the activity performed by the activity evaluation subject is estimated to be the same as the activity levels of multiple standard workers, the larger the similarity calculated by the activity evaluation device. In this case, the activity evaluation device 1 may be configured without including the score calculation unit 14. The output control unit 17 outputs data indicating the similarity to the output device 4, for example, based on the similarity data output from the similarity calculation unit 13. In this case, step ST5 may be omitted from the operation of the activity evaluation device 1 described using the flowchart of FIG. 4. Step ST5 may be omitted from the operation of the activity evaluation device 1 described using FIG. 4.

[0071] In the first embodiment described above, as shown in FIG. 1 , the task evaluation device 1 includes the task type estimation unit 15 and the representative normative data selection unit 16. However, this is merely an example. The task evaluation device 1 may be configured without the task type estimation unit 15 and the representative normative data selection unit 16. In this case, the processing of steps ST6 and ST7 in the operation of the task evaluation device 1 described using the flowchart in FIG. 4 can be omitted.

[0072] Furthermore, in the above-described first embodiment, the work evaluation device 1 may be configured without the output control unit 17. For example, the function of the output control unit 17 may be provided by a device external to the work evaluation device 1 that is connected to the work evaluation device 1 and the output device 4. For example, the work evaluation device 1 may store the work score data in an internal memory, and a device external to the work evaluation device 1 may read the work score data from the internal memory and output it to the output device 4. In this case, it is not essential that the work evaluation device 1 be connected to the output device 4.

[0073] Furthermore, in the first embodiment described above, in the task evaluation device 1, the pre-processing unit 12 divides the target data into time series, but this is merely an example. The pre-processing unit 12 may not have the function of dividing the target data into time series. The pre-processing unit 12 may extract features from the target data acquired by the target data acquisition unit 111 and generate target feature data. In this case, the pre-processing unit 12 does not need to divide the multiple normative data acquired by the normative data acquisition unit 112. The pre-processing unit 12 extracts features from each normative data acquired by the normative data acquisition unit 112 and generates normative feature data.

[0074] Furthermore, in the first embodiment described above, the output device 4 is a display device, but this is merely an example. For example, the output device 4 may be a storage device, and the output control unit 17 of the task evaluation device 1 may output task score data, task type estimation data, and selection norm data to the storage device. For example, the output control unit 17 may output data indicating the task score to the storage device based on the task score data, or may output the data indicating the task score to the storage device in association with the target data. Furthermore, the output control unit 17 may output data indicating the task score to the storage device in association with data indicating the task type based on the task score data and task type estimation data. Furthermore, the output control unit 17 may output data indicating the task score to the storage device in association with representative norm data based on the task score data and selection norm data. Furthermore, the output control unit 17 may output data indicating the task score to the storage device in association with the target data, task type data, and representative norm data based on the task score data, task type estimation data, and selection norm data.

[0075] In the first embodiment described above, it is assumed that the work evaluation device 1 is mounted on a brazing machine. However, this is merely an example. For example, the work evaluation device 1 may be mounted on a server. Furthermore, for example, some of the components constituting the work evaluation device 1, namely, the data acquisition unit 11, the pre-processing unit 12, the similarity calculation unit 13, the score calculation unit 14, the work type estimation unit 15, the representative normative data selection unit 16, and the output control unit 17, may be provided on the work device and the rest may be provided on the server, so that the work device and the server form a system.

[0076] Furthermore, in the above-described first embodiment, the work evaluated by the work evaluation device 1 is assumed to be brazing work, but this is merely an example. The work evaluated by the work evaluation device 1 may be work other than brazing that can be performed in a factory. For example, the work evaluated by the work evaluation device 1 may be soldering work, cutting work, wood screwing work, wood screwing work, or metal welding work. The work evaluated by the work evaluation device 1 may also be operating a vehicle such as an automobile, airplane, train, or ship. The work evaluation device 1 according to the first embodiment can be applied to a work evaluation device that evaluates various works that require measuring the level of skill (so-called instinct and knack work).

[0077] 5A and 5B are diagrams illustrating an example of the hardware configuration of the task evaluation device 1 according to embodiment 1. In embodiment 1, the functions of the data acquisition unit 11, the pre-processing unit 12, the similarity calculation unit 13, the score calculation unit 14, the task type estimation unit 15, the representative normative data selection unit 16, the output control unit 17, and a control unit (not shown) are realized by a processing circuit 1001. That is, the task evaluation device 1 includes a processing circuit 1001 for controlling the evaluation of tasks performed by task evaluation subjects. The processing circuit 1001 may be dedicated hardware as shown in FIG. 5A, or a processor 1004 that executes a program stored in memory as shown in FIG. 5B.

[0078] If the processing circuit 1001 is dedicated hardware, the processing circuit 1001 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.

[0079] When the processing circuit is a processor 1004, the functions of the data acquisition unit 11, pre-processing unit 12, similarity calculation unit 13, score calculation unit 14, task type estimation unit 15, representative norm data selection unit 16, output control unit 17, and a control unit (not shown) are realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in memory 1005. The processor 1004 reads and executes the program stored in memory 1005 to execute the functions of the data acquisition unit 11, pre-processing unit 12, similarity calculation unit 13, score calculation unit 14, task type estimation unit 15, representative norm data selection unit 16, output control unit 17, and a control unit (not shown). In other words, the task evaluation device 1 includes memory 1005 for storing a program that, when executed by the processor 1004, results in the execution of steps ST1 to ST8 of FIG. 4 described above. In addition, the program stored in memory 1005 can also be said to cause the computer to execute the processing procedures or methods of the data acquisition unit 11, the pre-processing unit 12, the similarity calculation unit 13, the score calculation unit 14, the task type estimation unit 15, the representative normative data selection unit 16, the output control unit 17, and a control unit not shown. Here, the memory 1005 may be, for example, a non-volatile or volatile semiconductor memory such as a RAM, a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), or an EEPROM (Electrically Erasable Programmable Read-Only Memory), a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, or a DVD (Digital Versatile Disc).

[0080] Note that the functions of the data acquisition unit 11, pre-processing unit 12, similarity calculation unit 13, score calculation unit 14, task type estimation unit 15, representative normative data selection unit 16, output control unit 17, and a control unit (not shown) may be partially implemented by dedicated hardware and partially implemented by software or firmware. For example, the data acquisition unit 11 may be implemented by a processing circuit 1001 as dedicated hardware, while the pre-processing unit 12, similarity calculation unit 13, score calculation unit 14, task type estimation unit 15, representative normative data selection unit 16, output control unit 17, and a control unit (not shown) may be implemented by a processor 1004 reading and executing programs stored in memory 1005. The task evaluation device 1 also includes an input interface device 1002 and an output interface device 1003 that communicate with devices such as the sensor 2, the normative data storage device 3, and the output device 4 via wired or wireless communication.

[0081] As described above, according to the first embodiment, the task evaluation device 1 is configured to include a target data acquisition unit 111 that acquires time-series target data related to the task generated by a task subject performing the task, a normative data acquisition unit 112 that acquires multiple time-series normative data related to the task generated by multiple norm workers performing the task, a pre-processing unit 12 that generates target feature data related to features extracted from the target data acquired by the target data acquisition unit 111 and multiple normative feature data related to features extracted from each of the multiple normative data acquired by the normative data acquisition unit 112, and a similarity calculation unit 13 that compares each of the target feature data with the multiple normative feature data and calculates the similarity of the target feature data with each normative feature data. Therefore, the task evaluation device 1 can evaluate the quality of the task of a person (task evaluation subject) without requiring that a range of skill for performing the task be previously defined in correspondence with the content of the task of the person (task evaluation subject).

[0082] Furthermore, according to the first embodiment, in the task evaluation device 1, the pre-processing unit 12 classifies the target data into a plurality of post-classification target data, generates target feature data for each of a plurality of feature amounts extracted from each of the plurality of post-classification target data, and the similarity calculation unit 13 calculates the similarity of the target feature data with each of the normative feature data for each of the plurality of target feature data generated by the pre-processing unit 12. This allows the task evaluation device 1 to prevent a decrease in learning efficiency due to an increase in the length of the target data. Furthermore, the task evaluation device 1 prevents a manager or the like from having to prepare a huge amount of normative data in preparation for a decrease in the frequency of occurrence of normative data determined to be similar due to an increase in the length of the target data.

[0083] Furthermore, according to the first embodiment, the task evaluation device 1 can be configured to include a score calculation unit 14 that calculates a task score indicating how closely the task level (task level) performed by the task evaluation subject compares with the task levels (task levels) performed by a plurality of standard workers, based on similarity data in which the target feature data is associated with a plurality of similarities calculated by the similarity calculation unit 13. Therefore, the task evaluation device 1 can evaluate the quality of the task of a person (task evaluation subject) without requiring that a range of skill required for performing the task be previously defined in correspondence with the content of the task performed by the person (task evaluation subject) to be evaluated.

[0084] Furthermore, according to the first embodiment, the task evaluation device 1 can be configured to include a task type estimation unit 15 that estimates the task type of the task performed by the task evaluation subject based on the object feature data or similarity data. Therefore, the task evaluation device 1 can provide the task score along with the task type, and can provide the evaluation of the task performed by the task evaluation subject, more specifically, the task score, in a format that is easier to understand for evaluating the task performed by the task evaluation subject. For example, if the task evaluation subject is evaluated for performing a task using an incorrect procedure, the task evaluation device 1 can inform the subject that the task itself was not performed well by providing information on the task type. To give a specific example, for brazing, while the correct task procedure is heating → brazing insertion, if the task evaluation subject performs the task using a procedure such as heating → brazing → heating, the task can be considered to be an insufficiently mastered task.

[0085] Furthermore, according to the first embodiment, the task evaluation device 1 includes a task type estimation unit 15 that estimates the task type of the task performed by the task evaluation subject based on the target feature amount data or similarity data in which the target feature amount data is associated with multiple similarities calculated by the similarity calculation unit 13, and the task type estimation unit 15 can be configured to estimate the task type of the task performed by the task evaluation subject based on the target feature amount data or similarity data and the task score calculated by the score calculation unit 14. Therefore, the task evaluation device 1 can improve the accuracy and efficiency of task type estimation, or reduce the amount of learning data used to generate a task type estimation model.

[0086] Furthermore, according to the first embodiment, the task evaluation device 1 can be configured to include a representative normative data selection unit 16 that selects representative normative data from among the multiple normative data that is deemed to be similar to the target data, based on similarity data in which the target feature data is associated with the multiple similarities calculated by the similarity calculation unit 13. Therefore, the task evaluation device 1 can provide task scores together with the representative normative data, and can provide the data in a form that enables identification of tasks that the task evaluation subject should practice and that have not yet reached the task level of the normative worker, for example.

[0087] Furthermore, according to the first embodiment, the task evaluation device 1 includes a representative normative data selection unit 16 that selects, from the plurality of normative data, representative normative data that is deemed to be similar to the target data, based on similarity data in which the target feature data is associated with the plurality of similarities calculated by the similarity calculation unit 13. The representative normative data selection unit 16 can be configured to select, from the plurality of normative data, representative normative data that is deemed to be similar to the target data, based on the similarity data and the task score calculated by the score calculation unit 14. Therefore, the task evaluation device 1 can improve the accuracy of the selection of representative normative data by selecting representative normative data in consideration of the task score.

[0088] Furthermore, according to the first embodiment, the task evaluation device 1 can be configured to include an output control unit 17 that outputs data indicating the task score calculated by the score calculation unit 14 to the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0089] Furthermore, according to the first embodiment, the task evaluation device 1 can be configured to include an output control unit 17 that associates data indicating the task score calculated by the score calculation unit 14 with data indicating the task type estimated by the task type estimation unit 15 and outputs the data to the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0090] Furthermore, according to the first embodiment, the task evaluation device 1 can be configured to include an output control unit 17 that associates data indicating the task score calculated by the score calculation unit 14 with the representative normative data selected by the representative normative data selection unit 16 and outputs the data to the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0091] Furthermore, according to the first embodiment, the output control unit 17 in the task evaluation device 1 can be configured to associate data indicating the task score with the target data and output the data to the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0092] Furthermore, according to the first embodiment, the output control unit 17 in the task evaluation device 1 can be configured to display data indicating the task score on the display device, which is the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0093] Furthermore, according to the first embodiment, in the task evaluation device 1, the output control unit 17 can be configured to associate data indicating the task score with data indicating the task type estimated by the task type estimation unit 15 and display the data on the display device, which is the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0094] Furthermore, according to the first embodiment, in the task evaluation device 1, the output control unit 17 can be configured to associate data indicating the task score with the representative normative data selected by the representative normative data selection unit 16 and display the data on the display device, which is the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0095] Furthermore, according to the first embodiment, the output control unit 17 in the task evaluation device 1 can be configured to associate data indicating the task score with the target data and display the data on the display device, which is the output device 4. Therefore, the task evaluation device 1 can provide data that helps to understand the current task level of the task evaluation subject and to help improve the skills of the task evaluation subject.

[0096] Any of the components of the embodiments may be modified or omitted.

[0097] The work evaluation device according to the present disclosure can evaluate the quality of a person's work without requiring that the range of skill required to perform the work be previously divided in accordance with the content of the work of the person to be evaluated.

[0098] 1 Work evaluation device, 2 Sensor, 3 Normative data storage device, 4 Output device, 11 Data acquisition unit, 111 Target data acquisition unit, 112 Normative data acquisition unit, 12 Pre-processing unit, 13 Similarity calculation unit, 14 Score calculation unit, 15 Work type estimation unit, 16 Representative normative data selection unit, 17 Output control unit, 1001 Processing circuit, 1002 Input interface device, 1003 Output interface device, 1004 Processor, 1005 Memory.

Claims

1. a target data acquisition unit that acquires time-series target data related to a task, the time-series target data being generated when a task evaluation subject performs the task; a normative data acquiring unit that acquires a plurality of time-series normative data related to the work, the normative data being generated by a plurality of norm workers performing the work; a pre-processing unit that generates target feature data relating to features extracted from the target data acquired by the target data acquisition unit, and a plurality of normative feature data relating to the features extracted from each of the plurality of normative data acquired by the normative data acquisition unit; a similarity calculation unit that compares the target feature data with a plurality of the reference feature data and calculates a similarity between the target feature data and each of the reference feature data; A work evaluation device comprising:

2. The pre-processing unit divides the target data into a plurality of divided target data, and generates the target feature amount data for each of the plurality of feature amounts extracted from each of the plurality of divided target data, The similarity calculation unit calculates the similarity between the target feature data and each of the norm feature data for each of the plurality of target feature data generated by the pre-processing unit.

2. The task evaluation device according to claim 1,

3. and a score calculation unit that calculates a task score indicating how much a level of the task performed by the task evaluation subject has reached the task levels of the tasks performed by the plurality of reference workers, based on similarity data in which the target feature data is associated with the plurality of similarities calculated by the similarity calculation unit.

2. The task evaluation device according to claim 1,

4. The score calculation unit calculates the task score for each of the plurality of object feature amount data generated by the pre-processing unit based on the object feature amount data and the similarity data.

4. The task evaluation device according to claim 3.

5. The score calculation unit calculates the task score from a maximum value of the similarity for each of the normative feature data.

4. The task evaluation device according to claim 3.

6. The score calculation unit calculates the task score from an average value of the similarity for each of the normative feature data.

4. The task evaluation device according to claim 3.

7. The score calculation unit calculates the task score by weighting the similarity for each of the normative feature data.

4. The task evaluation device according to claim 3.

8. a task type estimation unit that estimates a task type of the task performed by the task evaluation subject based on the target feature amount data or similarity data in which the target feature amount data is associated with the multiple similarities calculated by the similarity calculation unit. The task evaluation device according to claim 1 , further comprising:

9. a task type estimation unit that estimates a task type of the task performed by the task evaluation subject based on the target feature amount data or based on the similarity data in which the target feature amount data is associated with a plurality of the similarities calculated by the similarity calculation unit, The task type estimation unit is The task type of the task performed by the task evaluation subject is estimated based on the object feature amount data or the similarity data and the task score calculated by the score calculation unit.

4. The task evaluation device according to claim 3.

10. a representative normative data selection unit that selects representative normative data that is similar to the target data from among the plurality of normative data based on similarity data in which the target feature data and the plurality of similarities calculated by the similarity calculation unit are associated with each other. The task evaluation device according to claim 1 , further comprising:

11. a representative normative data selection unit that selects representative normative data that is similar to the target data from among the plurality of normative data based on the similarity data in which the target feature data is associated with the plurality of similarities calculated by the similarity calculation unit, The representative normative data selection unit The representative normative data that is similar to the target data is selected from among the plurality of normative data based on the similarity data and the task score calculated by the score calculation unit.

4. The task evaluation device according to claim 3.

12. an output control unit that outputs data indicating the task score calculated by the score calculation unit to an output device The task evaluation device according to claim 3 , further comprising:

13. a task type estimation unit that estimates a task type of the task performed by the task evaluation subject based on the target feature amount data or based on the similarity data in which the target feature amount data is associated with a plurality of the similarities calculated by the similarity calculation unit, an output control unit that outputs, to an output device, data indicating the task score calculated by the score calculation unit in association with data indicating the task type estimated by the task type estimation unit. The task evaluation device according to claim 3 , further comprising:

14. a representative normative data selection unit that selects representative normative data that is similar to the target data from among the plurality of normative data based on the similarity data in which the target feature data is associated with the plurality of similarities calculated by the similarity calculation unit, an output control unit that outputs data indicating the task score calculated by the score calculation unit to an output device in association with the representative normative data selected by the representative normative data selection unit; The task evaluation device according to claim 3 , further comprising:

15. The output control unit outputs data indicating the task score to the output device in association with the target data.

15. The task evaluation device according to claim 12, wherein the task evaluation device is a task evaluation device that performs a task evaluation process.

16. the output device is a display device; The output control unit causes the display device to display data indicating the task score.

13. The task evaluation device according to claim 12.

17. the output device is a display device; The output control unit causes the display device to display data indicating the task score in association with data indicating the task type estimated by the task type estimation unit.

14. The task evaluation device according to claim 13.

18. the output device is a display device; The output control unit causes the display device to display data indicating the task score in association with the representative normative data selected by the representative normative data selection unit.

15. The task evaluation device according to claim 14.

19. The output control unit causes the display device to display data indicating the task score in association with the target data.

19. The task evaluation device according to claim 16,

20. a step of a target data acquisition unit acquiring time-series target data related to a task, the time-series target data being generated when a task evaluation subject performs the task; a reference data acquiring unit acquiring a plurality of time-series reference data related to the work, the reference data being generated by a plurality of reference workers performing the work; a pre-processing unit generating object feature data relating to features extracted from the object data acquired by the object data acquiring unit, and a plurality of normative feature data relating to the features extracted from each of the plurality of normative data acquired by the normative data acquiring unit; a similarity calculation unit comparing the target feature data with each of the plurality of normative feature data and calculating a similarity between the target feature data and each of the normative feature data; A work evaluation method comprising: