JP6897037B2 - Workability evaluation device - Google Patents

Description

The present invention relates to a technique for evaluating workability.

At production sites such as factories, efforts are being made to evaluate the workability of product assembly work and utilize the results for process improvement such as improvement of work efficiency and safety. For example, Patent Document 1 proposes a system for evaluating assembly workability by simulation using 3D CAD data of assembled parts and a human body model. Further, in Patent Document 2, a system for evaluating whether or not there is waste in the worker's movement by taking the difference between the motion vector of the part extracted from the moving image recording the worker's movement and the reference vector obtained from the CAD data. Has been proposed.

Japanese Unexamined Patent Publication No. 2014-100773 JP-A-2009-122726

However, in the systems of Patent Documents 1 and 2, it is necessary to provide CAD data that defines the configuration and arrangement of objects involved in the work, and a program for evaluating workability is developed based on the CAD data. Must. However, in general, workability is affected not only by the composition and arrangement of objects involved in the work, but also by various factors such as the characteristics of the worker and the work environment. It is difficult to incorporate the method into the workability evaluation program in advance. Therefore, in the conventional system, there is a problem that the evaluation result and the actual workability may not match.

In the context so far, the problem has been explained by taking the assembly work as an example, but the same problem can occur in the workability evaluation in all fields.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a highly versatile technique capable of evaluating the actual workability of the work to be evaluated.

In order to achieve the above object, the present invention adopts the following configuration.

The first aspect of the present invention is a workability evaluation device that evaluates the workability of the work to be evaluated, and records information related to the operation of the worker as work data when the worker performs the work to be evaluated. A work data recording unit, a partial work classification unit that divides the entire work to be evaluated into a plurality of partial works by analyzing the movement of the worker included in the work data, and a plurality of samples of the work data. Provided is a workability evaluation device having a workability evaluation unit for obtaining an index related to workability for each partial work by evaluating the variation in the workability for each partial work.

According to this configuration, since the evaluation is performed based on the work data which is the record when the worker actually performs the work to be evaluated, the workability according to the actual work can be evaluated. Further, since it is a simple and general-purpose algorithm that evaluates the variation of work data, it can be applied to various types of workability evaluation. Furthermore, since the work to be evaluated is divided into partial work, which is a small unit of movement, there is an advantage that it is easy to associate movements between samples, find problem areas, and identify factors that hinder workability. .. Further, since the partial work classification unit classifies the partial work by analyzing the movement of the worker included in the work data, the partial work is generated in a unit suitable for evaluating the actual workability. be able to. In the present invention, the "worker" may be a person or an AI robot operated by artificial intelligence.

For example, the partial work classification unit may divide the entire evaluation target work into the plurality of partial works by analyzing the operation by deep learning based on a plurality of samples of the work data. By using machine learning by Deep Learning, it is possible to classify an unknown evaluation target work into a partial work of an appropriate unit without any special prior knowledge. In other words, the workability evaluation device can acquire a new classification ability for classifying an unknown work to be evaluated into a plurality of partial works simply by giving a plurality of samples of work data. Therefore, according to this configuration, the versatility and adaptability of the workability evaluation device can be remarkably improved.

Of course, it is also possible to classify into partial work by a method other than machine learning. For example, the partial work classification unit may detect a change point of the worker's movement based on the work data and classify the entire evaluation target work into the plurality of partial work based on the change point of the movement. Good. Even with this configuration, the workability evaluation device can classify unknown evaluation target work into a plurality of partial work simply by giving a plurality of work data samples.

The plurality of samples include a plurality of work data for each of the plurality of workers, and the workability evaluation unit calculates the average work time of each worker for each partial work, and the average work time for each partial work. It is advisable to find an index showing the variation among workers. It is considered that the difficulty level of the work can be evaluated by using an index showing the variation of the average work time among the workers.

The plurality of samples include a plurality of work data of each of the plurality of workers, and the workability evaluation unit calculates the variation in the work time of each worker for each partial work, and each worker for each partial work. It is advisable to find an index showing the variation in working hours. It is considered that the degree of freedom of the work procedure can be evaluated by using an index showing the variation in the work time of each worker.

It is preferable to have one or more sensors for detecting the movement of the operator, and the work data recording unit may record sensor data obtained from the sensors. This makes it possible to automatically collect the movements of the worker. In addition, the movement of the worker can be grasped as an objective physical quantity by the sensor data. For example, the sensor may include a motion sensor worn on the worker's body. By using the movement sensor attached to the body, it is possible to directly and accurately capture the movement of the worker's body.

The present invention can be regarded as a workability evaluation device having at least a part of the above configuration or function. The workability evaluation device may be composed of a single device or a combination of a plurality of devices. In addition, the present invention is a workability evaluation method including at least a part of the above processing, a program for causing a computer to execute such a method, or a computer-readable record in which such a program is recorded non-temporarily. It can also be regarded as a medium. Each of the above configurations and processes can be combined with each other to construct the present invention as long as there is no technical contradiction.

According to the present invention, the actual workability of the work to be evaluated can be evaluated. Further, the present invention can be universally applied to the evaluation of workability of various works.

FIG. 1 is an external view showing a hardware configuration of a workability evaluation device. FIG. 2 is a block diagram showing a functional configuration of the workability evaluation device. 3A to 3C are diagrams for explaining partial work. FIG. 4 is a diagram showing an example of assembly work. FIG. 5 is a flowchart of workability evaluation processing. 6A and 6B are diagrams showing the internal configuration of the partial work classification unit. FIG. 7 is a flowchart of the partial work classification unit. FIG. 8 is an example of a partial work list. FIG. 9 is a diagram showing an example of learning data and classification results of Deep Learning. FIG. 10 is an output example of the evaluation result.

The workability evaluation device according to the present invention is a system for evaluating the workability of any work (the work to be evaluated is referred to as "evaluation target work"). In the embodiment described below, an example in which the present invention is applied to the evaluation of workability of product assembly work in a factory is given, but the scope of application of the present invention is not limited to this example. For example, monitoring work / checking work / data entry work at production sites such as factories, operation / operation / maintenance of equipment and systems, procedures and procedures in surgery, various work in agriculture / fishery / forestry, work in office work, creators. It can be applied to workability evaluation in various fields and applications such as work of. Further, the workability evaluation result (evaluation index) generated by the apparatus according to the present invention is utilized for various purposes such as process improvement, QC (Quality Control) activity, and feedback to product design.

<Outline of workability evaluation method>
The method of evaluating workability by simulation using CAD data like the conventional system (Patent Document 1) has low versatility and practicality, and cannot consider the influence of the characteristics of the worker and the work environment. There was a problem. In view of these points, the workability evaluation method of the present embodiment causes the worker to actually perform the work to be evaluated, and analyzes the data (referred to as "work data") that records the movement of the worker at that time. By doing so, the approach of evaluating the workability of the work to be evaluated is adopted. According to this method, it is not necessary to give special definition information such as CAD data, so that it can be universally applied to all kinds of workability evaluations. In addition, since the actual work data of the worker is used, the workability is evaluated according to the actual work, taking into consideration all the influences of the design of the work object, the work procedure, the characteristics of the worker, the work environment, etc. be able to.

FIG. 1 schematically shows the workability evaluation device 100 of the present embodiment. The general flow of workability evaluation is as follows. First, the work procedure (work standard) of the work to be evaluated is set. For example, in the case of product assembly work, the order of assembling parts is defined in the work procedure. Then, the worker 110 is made to actually carry out the work according to the work procedure. At this time, the workability evaluation device 100 records the operation of the worker 110 during the work by the sensor 101 attached to the worker 110 or the object 111 used in the work. By having one or a plurality of workers perform the same work a plurality of times, it is possible to collect work data of a plurality of samples.

Here, it is assumed that the quality of workability appears in the variation in operation during work, and eventually in the variation in working time. For example, if the movements and working hours during work vary widely among workers, it is said that the workability is reduced because the work itself is difficult (because the work is not stable without skill). Can be regarded. In addition, if the movements and work times of the same worker vary widely, the degree of freedom in the work procedure is too high (because there are many choices of movements and the same work cannot be reproduced), and the workability is reduced. Can be considered to be. Based on such an assumption, the workability evaluation device 100 of the present embodiment evaluates the workability of the work to be evaluated from the variation in the work data of a plurality of samples. Details will be described later.

Further, the workability evaluation device 100 divides the entire work to be evaluated into a plurality of partial works (elemental works), and individually evaluates the workability of each partial work. This is because if the workability of each partial work is known, it becomes easy to find a problematic part that deteriorates the workability and identify a factor that hinders the workability.

By the way, in order to realize the workability evaluation for each partial work, it is necessary to divide the work data of the work to be evaluated into the data for each partial work. This process may be performed manually by the evaluator, but it is not realistic because it is difficult to extract the corresponding partial work from a plurality of samples and a huge amount of man-hours are required. However, it is not possible to simply divide the work data because there are variations in operation and work time for each worker and each time the work is performed. Therefore, the workability evaluation device 100 of the present embodiment analyzes the movement of the worker based on the work data and automates the classification into the partial work. Details will be described later.

Although a person is illustrated as the worker 110 in FIG. 1, an AI robot operated by artificial intelligence may be used as the worker 110. For example, even in the process in which an AI robot learns the optimum operation of a work by machine learning such as deep learning, the operation during the work and the work time vary, and the workability of the work is evaluated based on this variation. It is possible.

<Configuration of workability evaluation device>
The configuration of the workability evaluation device will be described with reference to FIGS. 1 and 2. FIG. 1 is an external view showing a hardware configuration of a workability evaluation device, and FIG. 2 is a block diagram showing a functional configuration of the workability evaluation device.

(Hardware configuration)
The workability evaluation device 100 includes one or more sensors 101 and an information processing device 102 as main hardware.

The information processing device 102 is, for example, a hardware resource such as a hardware processor (CPU), a memory, a storage device (storage of a hard disk, a semiconductor disk, etc.), an input device (keyboard, mouse, touch panel, etc.), a display device, a communication device, and the like. It can be configured by a general-purpose computer having. The function of the workability evaluation device 100 shown in FIG. 2 is realized by loading the program stored in the storage device into the memory and executing it by the processor. The information processing device 102 may be configured by one computer or may be configured by distributed computing by a plurality of computers. Further, a part of the functions of the information processing device 102 may be realized by a cloud server. Further, in order to speed up the processing, it is possible to realize a part or all of the functions of the information processing apparatus 102 by using dedicated hardware (for example, GPU, FPGA, ASIC, etc.).

The sensor 101 is a device for recording the operation of the worker 110 as data. Any type and type of sensor may be used as long as the operation of the worker 110 can be directly or indirectly detected or estimated. For example, the sensor 101 includes a sensor that senses the worker 110, a sensor that senses an object (hereinafter referred to as a “work target”) 111 that the worker 110 handles during work, and an interposition between the worker 110 and the work target 111. Includes sensors that sense objects. Hereinafter, an example of the sensor 101 will be given, but the workability evaluation device 100 does not need to be provided with all the sensors 101, and the necessary sensors 101 may be provided according to the device configuration, the type and content of work, the application, and the like.

As the sensor 101 that senses the worker 110, a motion sensor that senses the movements of the head, the line of sight, the hands, the feet, the torso, and the like can be used. For example, by using an acceleration sensor or an angular acceleration sensor attached to the left and right wrists of the worker 110, it is possible to detect the movements of the right hand and the left hand in the assembly work. Further, by attaching an acceleration sensor or an angular acceleration sensor to the fingertip, the movement of the fingertip can be detected. It is also possible to sense the movement of the worker 110 and the relationship with surrounding objects by analyzing the moving image taken by the image sensor (camera) instead of the sensor attached to the worker 110. Alternatively, the movement of the worker 110 can be sensed by detecting the marker attached to the worker 110 with a magnetic sensor or an infrared sensor.

The sensor 101 that senses the worker 110 includes a face image sensor that senses facial expressions, eye movements, and how the face moves, a heartbeat sensor that senses heartbeat, a pulse sensor that senses pulse, and hands, feet, neck, and so on. Myoelectric sensor that senses muscle movement with electrodes attached to each part of the body such as the torso, blood pressure sensor that senses blood pressure, brain wave sensor that senses brain waves in each area of the brain, blood flow sensor that senses blood flow in each part, An image sensor or the like that senses the direction of the line of sight of the worker 110 or the point of gaze can be used.

The form of the sensor 101 does not matter. For example, a sensor provided in a smartphone may be used, or a sensor provided in a wearable device such as a smart watch or a smart glass may be used.

The work target 111 is a part to be assembled in the case of assembly work, and the device in the case of operation / operation of the device. Since the state of the work target 111 can also affect the operation of the worker 110, using the information obtained by sensing the state of the work target 111 as indirect information may be useful for accurate evaluation of the movement of the worker 110. There is. For example, a sensor that detects the spatial position of the work target 111, a sensor that detects the posture of the work target 111, a sensor that detects the state of the work target 111 (changes in states such as acceleration, temperature, color, and shape), and work. An example is an example of a sensor that detects the environmental state of the target 111 (physical quantity and information related to the environment such as ambient temperature and humidity).

The inclusions between the worker 110 and the work target 111 are tools and devices used for assembly in the case of assembly work. Since the state of such inclusions can also affect the movement of the worker 110, it is possible to use the information obtained by sensing the state of the inclusions as indirect information, which is useful for accurate evaluation of the movement of the worker 110. There is sex. For example, a sensor that detects the spatial position of the inclusion, a sensor that detects the posture of the inclusion, a sensor that detects the state of the inclusion (change in the state such as temperature, color, shape, etc.), and the environmental state of the inclusion ( Examples include sensors that detect physical quantities and information related to the environment such as ambient temperature and humidity. Further, a sensor that detects a force acting on the inclusions by the operation of the operator 110, a physical quantity related to the inclusions (acceleration, vibration, cutting sound, etc.) and the like may be used. For example, in the case of driving an automobile, a sensor attached to the driving operation unit may detect steering operation, accelerator operation, brake operation, various switch operations, and the like.

When the worker 110 is an AI robot, an encoder used for controlling the motor of the AI robot, a camera included in the AI robot, and the like can also be used as the sensor 101.

(Functional configuration)
As shown in FIG. 2, the workability evaluation device 100 includes a sensor information input unit 201, a work target information input unit 202, a work data recording unit 203, a partial work classification unit 204, a workability evaluation unit 205, and a result output unit 206. Have.

The sensor information input unit 201 is a function of acquiring sensor data from the sensor 101. The work target information input unit 202 is a function of acquiring work target information (for example, evaluation target work type, work procedure, etc.). The work data recording unit 203 is a function of generating an operation record (work data) from the start to the end of the evaluation target work based on the information acquired by the sensor information input unit 201 and recording it in the storage device. For example, by having a plurality of workers perform the same work a plurality of times, a plurality of work data having different workers and proficiency levels are recorded.

The partial work classification unit 204 is a function of dividing and classifying the entire evaluation target work into a plurality of partial works based on the work data. The workability evaluation unit 205 is a function of obtaining an index related to workability for each partial work by evaluating the variation of the work data for each partial work. The result output unit 206 is a function of outputting the evaluation result of the workability evaluation unit 205.

<Definition of work>
In this embodiment, "work" is defined as "an act performed by a worker using a part or all of his / her body to accomplish a set purpose (task)". Therefore, the movement of the worker always occurs in the process of carrying out the work. In addition, a "work procedure" is set in advance for the work, but as long as the work procedure is followed, the specific operation is arbitrary. Therefore, when recording work data, there may be some differences in the operations performed in the process of performing the work, depending on the worker performing the work or the degree of proficiency in the work.

One "work" can be divided into a plurality of "partial work". As shown in FIG. 3A, a plurality of partial works may be set to be continuous in time, or as shown in FIG. 3B, a plurality of partial works may be set discretely. The number of partial works and the setting method are arbitrary, but as shown in FIG. 3C, in order to facilitate the correspondence of partial works between different work data, the partial work is performed in units of a certain amount of operation. It is preferable to set it.

<Workability evaluation process>
Next, the operation of the workability evaluation device 100 will be described by taking the assembly work in the factory as an example.

(Example of assembly work)
FIG. 4 is an example of product assembly work. It is assumed that three parts 402a, 402b, and 402c having different lengths are attached to the base part 401 having three holes. The parts 402a in the parts box 403a, the parts 402b in the parts box 403b, and the parts 402c in the parts box 403c are attached to the holes 401a, the holes 401b, and the holes 401c in the base part 401, respectively. As a work procedure
It is assumed that only the assembly order of (1) inserting the part 402a into the hole 401a (2) inserting the part 402b into the hole 401b (3) inserting the part 403c into the hole 401c is given. Since the base component 401 is provided with frames 401d and 401e in the front and rear, it is necessary to avoid the frames 401d and 401e when attaching the component 402a, the component 402b, and the component 402c.

(Operation of the device)
The flow of the workability evaluation process by the workability evaluation device 100 will be described with reference to the flowchart of FIG.

(1) Recording of work data The work to be evaluated is performed by the worker a plurality of times, and each work data is recorded by the work data recording unit 203 (step S500). In the example of FIG. 4, the work of sequentially attaching the three parts 402a, 402b, and 402c to the base part 401 is recorded. That is, for each of the three parts 402a, 402b, 402c, the parts are picked up from the parts box and moved toward the holes of the base part 401 while avoiding the frame and other objects, and the parts and holes Information on a series of operations is recorded, in which the tip of the part is inserted into the hole in a spatial orientation, the fitting is completed, and the hand is released.

The start and end of the work (start and end of data recording) may be instructed by the operator by pressing a button or the like, or may be automatically determined based on the signal of the sensor 101. The work may be carried out by one worker a plurality of times, or by a plurality of workers once or a plurality of times. In order to make the best use of the know-how and skills due to the difference in movements between people and the proficiency, it is desirable to have a plurality of workers carry out the work multiple times. As a result, a plurality of work data are acquired for the same work. Individual work data is also called a "work sample" or simply a "sample".

The work data includes at least the sensor data acquired from the sensor information input unit 201. Further, if necessary, the work data may include work target information acquired from the work target information input unit 202.

In the present embodiment, when the multidimensional vector records the position and orientation of the right wrist of the worker 110 every unit time (for example, every 0.1 seconds) by the motion sensor 101 attached to the right wrist of the worker 110. Series data is recorded as work data. For example, 30 samples can be obtained by having three workers perform the work 10 times each and record the movements.

(2) Classification of partial work Next, the partial work classification unit 204 analyzes the movement of the worker based on the work data of a plurality of samples, thereby classifying the entire movement of the work to be evaluated into a plurality of partial work, and at the same time. The work data of each sample is divided into data for each partial work (step S501).

Classification of partial work can be done based on changes in movement. Although the movement and work time of the worker are slightly different for each work sample, the change in movement is almost the same when viewed in small partial work units because there are restrictions due to the work procedure and the design of the assembly target. For example, focusing on the change point of the motion vector representing the movement of the hand of the worker 110 (position, direction, velocity, angular acceleration, acceleration, angular acceleration, etc.), the acceleration / deceleration / stop / change of direction of the hand can be known. It is possible to capture changes in movement (switching of partial movements). Alternatively, a classification technique based on Deep Learning may be used. For example, a recurrent neural network (RNN), which is a deep learning technique, can be used to train time-series data. When time-series data of each of a plurality of work samples is given as learning data and learning is performed by Deep Learning, a classifier that classifies the entire work into a predetermined number of partial works is automatically generated. In this embodiment, it is assumed that the partial work is automatically classified by data analysis, but a part of the classification work may be performed by a person.

When classifying the partial work, the information detected from the work target may be used. In addition, information detected from inclusions such as tools, actuators, and other machines can be added for classification. For example, it is possible to detect the vibration, reaction force, position, operating status, etc. of the tool and use it as classification information. Alternatively, the sensor information of the adjacent robot can be used as the classification information. Classification may be performed according to the attributes of the worker and the attributes of the work target. The partial work does not necessarily have to match the unit of work as seen by a person. It is only necessary that the corresponding partial work can be accurately extracted from each of the plurality of work samples. Appropriate classification can be assessed using the similarity between samples belonging to the same classification. If properly classified, the similarity between samples will be high.

FIG. 6A shows a configuration example in which movements are classified based on changes in movements as an example of the internal configuration of the partial work classification unit 204. The partial work classification unit 204 includes a motion vector extraction unit 601, a change point detection unit 602, a similarity point detection unit 603, a classification generation unit 604, and a classification result output unit 605.

The processing flow of the partial work classification unit 204 will be described with reference to the flowchart of FIG. 7. First, the motion vector extraction unit 601 reads one sample (work data) (step S700). The motion vector extraction unit 601 reads out time-series data of a multidimensional vector that records the position and orientation of the worker's right wrist from the work data, and by time-differentiating it, the motion vector has velocity and angular velocity as elements. Generate series data (step S701). The change point detection unit 602 detects the change point of the worker's movement by analyzing the time-series data of the motion vector (step S702). For example, it is possible to detect when the movement starts from a stopped state, when the movement is stopped, when the speed is suddenly increased / decreased, when the moving direction is changed, and the like as the change points of the movement. The change point detection unit 602 creates a change point list that describes the detected change points (step S703). By performing the processing of steps S700 to S703 for all the samples (step S704), a list of change points for each of the 30 samples is obtained.

Next, the similarity detection unit 603 selects one of the plurality of samples as the representative sample, refers to the change point list, and sets each of the change points in the representative sample from among the change points in the other samples. Similarities with similar movement changes are detected (step S705). This is a process of searching for a corresponding point of operation among a plurality of samples. The representative sample may be selected in any way, and in the present embodiment, the sample having the smallest number of change points among the plurality of samples is selected as the representative sample. In some cases, no similarities are found for one change point, and in other cases, multiple similarities are found. Stretch matching and genetic algorithms can be used to detect similarities. The similarity detection unit 603 creates a similarity list describing similarities corresponding to each change point in the representative sample (step S706).

Next, the classification generation unit 604 divides the entire operation of the work in the representative sample into a plurality of partial work based on the change point list of the representative sample (step S707). If the representative sample contains few change points (for example, several), the unit separated by the change points may be a partial work. When the number of changing points is large (for example, several tens or more), several singular points may be extracted from the changing points included in the changing point list, and the unit separated by the singular points may be regarded as a partial work. .. In this case, it is preferable to preferentially select a point where the change is particularly large, a point where the moving direction changes, a point where similar points exist in other samples, and the like as singular points. The user may perform a part of the classification work of the partial work. For example, the partial work classified by the classification generation unit 604 can be presented to the display device, and the user can modify (integrate or subdivide the partial work).

Next, the classification generation unit 604 divides the time-series data of the motion vector of the representative sample into the data for each partial work, and also for each partial work for the time-series data of the motion vector of another sample based on the similarity list. (Step S708). Finally, the classification result output unit 605 creates and outputs a partial work list (step S709).

FIG. 8 shows an example of a partial work list. This partial work list is a list in which information for identifying a worker (worker ID, etc.), work time for each partial work, time-series data of a motion vector for each partial work, and the like are described for each sample.

Up to this point, an example of the configuration and processing in the case of classifying the motion based on the change in the motion has been described, but it is also possible to classify the motion by using the technique of Deep Learning. In that case, the partial work classification unit 204 creates a work data input unit 610 for inputting work data, a deep neural network 611 for classifying the input work data, and a partial work list based on the classification result, as shown in FIG. 7B. It can be configured by the classification result output unit 612 that is created and output. Also in the case of this configuration, the classification result by the deep neural network 611 may be presented so that the user can modify it.

When the technique of Deep Learning is used, for example, it is possible to learn about classifying the operation from the start point to the end point of the work to be evaluated into N stages (for example, 100 stages). At this time, the work data input unit 610 may input the work data of a plurality of samples obtained by having many workers perform the work to be evaluated a plurality of times as learning data. In each work data, the data (position of the worker's hand, etc.) is the same at the start point and the end point of the work, but the data may be slightly different for each work data in the middle of the work (between the start point and the end point). By learning to classify them, it is possible to obtain a deep neural network 611 capable of classifying parts having similar data. In this case, each part classified into N stages is connected to the adjacent part. Then, in the actions of all the workers, similar parts are classified into the same classification.

FIG. 9 shows an example of learning data and classification results of Deep Learning. Each curve drawn on the graph shows the work data. The vertical axis of the graph is the value of the data, and the horizontal axis is the time. In the present embodiment, since the horizontal acceleration of the worker's hand and the position information of the hand or part are used as the work data, the data value is a multidimensional vector, but in FIG. 9, the data is shown for convenience. The value of is shown in one dimension. Further, the total work time (time length from the start point to the end point) may differ for each work data, but in FIG. 9, the start point and the end point are the same (the total work time length of each work data is the same). (To be) standardized. In Deep Learning, classification is performed based not only on the acceleration information of the hand but also on the position information of the hand or the part. Therefore, movements in which the hands are close to each other are classified into the same part. Below the graph, an example of the classification result by Deep Learning is shown. In this example, the entire work data is classified into six stages, and each part corresponds to a partial work.

(3) Evaluation of workability The workability evaluation unit 205 selects partial work one by one from the partial work list (step S502) and evaluates the workability of each partial work. In this embodiment, as an index showing the workability of partial work,
(Indicator 1) Variation in average working time among workers, and
(Indicator 2) Variation in working time for each worker,
Is calculated (steps S503, S504).

The index 1 is an index showing the difficulty level of the partial work (for example, low assembling property), and the larger the variation, the higher the difficulty level of the partial work and the lower the workability. Further, the index 2 is an index showing the degree of freedom of the work procedure, and the larger the variation, the higher the degree of freedom (that is, ambiguity) of the work procedure and the lower the workability. For "variation", for example, variance, standard deviation, range (difference between maximum value and minimum value) and the like can be used.

For example, when the average work time of the workers A, B, and C regarding the partial work W is x _WA , x _WB , x _WC , and the variance is y _WA , y _WB , y _WC , the variation of the index 1 is large. means that x _{WA, x WB,} the variance of _{x WC} is large, the representative of the _{y WA,} y _WB, each value or _y WA of _{y WC, y WB, y WC} and variations of the index 2 is greater It means that the value (for example, average value, intermediate value, maximum value, etc.) is large.

After evaluating the workability of all the partial works included in the partial work list (YES in step S505), the process proceeds to step S506.

(4) Output of evaluation result In step S506, the result output unit 206 outputs the evaluation result of the workability evaluation unit 205 to the display device or the external device.

FIG. 10 shows an output example of the evaluation result. In this example, the evaluation results of three items of "evaluation", "work difficulty", and "procedure freedom" are output for each partial work of the evaluation target work W001 and W002. "Work difficulty" indicates the magnitude of the variation of the above-mentioned index 1 in three stages of low / medium / high, and "procedure freedom" indicates the magnitude of the variation of the above-mentioned index 2 in low / medium. It is shown in three stages of medium / high. In addition, "evaluation" is a comprehensive evaluation of the result of work difficulty and the result of procedure freedom, and the workability of partial work is shown in four stages of ◎ / ○ / △ / ×. By outputting such evaluation results, problem areas (partial work with low workability) and factors that reduce workability in the work to be evaluated (whether the work is too difficult or the work procedure is ambiguous) Can be easily identified, which can be useful for reviewing product design and improving work procedures.

In addition, the output of the evaluation result may include information indicating the minimum value of the working time of each partial work and the operation (moving route, moving speed, etc.) at that time. By outputting such information, it becomes easy to grasp the degree of improvement in workability and set improvement targets. In addition, the information recorded when each partial work is performed may be included in the output of the evaluation result. This makes it easier to estimate the factors that reduce workability and to study how to improve them.

According to the workability evaluation device 100 of the present embodiment described above, since the evaluation is performed based on the work data which is the record when the worker actually performs the work to be evaluated, the workability is in line with the actual work. Can be evaluated. Further, since it is a simple and general-purpose algorithm that evaluates the variation of work data, it can be applied to various types of workability evaluation. Furthermore, since the work to be evaluated is divided into partial work, which is a small unit of movement, there is an advantage that it is easy to associate movements between samples, find problem areas, and identify factors that hinder workability. .. Moreover, the workability evaluation device 100 of the present embodiment can automatically classify into partial work by motion analysis based on work data. Therefore, the versatility and adaptability of the workability evaluation device 100 can be remarkably improved.

It should be noted that the configuration of the above-described embodiment is merely a specific example of the present invention, and is not intended to limit the scope of the present invention. The present invention can adopt various concrete configurations as long as it does not deviate from the technical idea.

For example, the technical idea disclosed in the present specification can also be specified as the following invention.
(Appendix 1)
The present invention is a workability evaluation device for evaluating the workability of the work to be evaluated.
It has a hardware processor, memory operably linked to the hardware processor, and storage.
The hardware processor arranges a program stored in the storage in the memory, and functions as a computer.
A work data recording unit that records information related to the operation of the worker as work data in the storage when the worker performs the work to be evaluated.
A partial work classification unit that divides the entire work to be evaluated into a plurality of partial works by analyzing the movements of workers included in the work data.
A workability evaluation unit that obtains an index related to workability for each partial work by evaluating the variation of the work data in a plurality of samples for each partial work.
Have.
(Appendix 2)
The present invention is a workability evaluation method for evaluating the workability of the work to be evaluated.
A step of recording information related to the operation of the worker as work data in the storage when the worker performs the work to be evaluated by at least one hardware processor.
A step of classifying the entire work to be evaluated into a plurality of partial works by analyzing the behavior of the worker included in the work data by at least one hardware processor.
A step of obtaining an index related to workability for each partial work by evaluating the variation of the work data in a plurality of samples for each partial work by at least one hardware processor.
Have.

100: Workability evaluation device, 101: Sensor, 102: Information processing device, 110: Worker, 111: Object (work target)
201: Sensor information input unit, 202: Work target information input unit, 203: Work data recording unit, 204: Partial work classification unit, 205: Workability evaluation unit, 206: Result output unit 401: Base parts, 401a, 401b, 401c: Hole, 401d, 401e: Frame, 402a, 402b, 402c: Parts, 403a, 403b, 403c: Parts box 601: Vector extraction unit, 602: Change point detection unit, 603: Similarity detection unit, 604: Classification generation Unit, 605: Classification result output unit 610: Work data input unit, 611: Deep neural network, 612: Classification result output unit

Claims (9)

A workability evaluation device that evaluates the workability of the work to be evaluated.
A work data recording unit that records information related to the movement of the worker as work data when a worker including at least one of a human or an AI robot operated by artificial intelligence performs the work to be evaluated.
A partial work classification unit that divides the entire work to be evaluated into a plurality of partial works by analyzing the movements of workers included in the work data.
A workability evaluation unit that obtains an index related to workability for each partial work by evaluating the variation of the work data in a plurality of samples for each partial work.
Have,
The plurality of samples include a plurality of work data of each of the plurality of workers.
The workability evaluating unit calculates an average work time of each worker in each partial work obtains an index representing the variation among workers of the average working time for each part work, variations represented in the index A workability evaluation device characterized in that the larger the value, the higher the work difficulty level of the partial work, and the evaluation result of the work difficulty level for each partial work is output. Before SL workability evaluating unit calculates a variation of the working time of each worker in each partial work obtains an index for each partial work represents the variation of the working time of each worker, variation represented in the index The workability evaluation device according to claim 1, wherein the larger the value is, the higher the degree of freedom of the work procedure of the partial work is evaluated, and the evaluation result of the degree of freedom of the work procedure for each partial work is output. Claim 1 or 2 is characterized in that the partial work classification unit divides the entire work to be evaluated into the plurality of partial works by analyzing the operation by deep learning based on a plurality of samples of the work data. The workability evaluation device described in 1. The partial work classification unit is characterized in that it detects a change point of the worker's movement based on the work data and divides the entire evaluation target work into the plurality of partial work based on the change point of the movement. The workability evaluation device according to claim 1 or 2. It has one or more sensors that detect the movement of the worker
The workability evaluation device according to any one of claims 1 to 4, wherein the work data recording unit records sensor data obtained from the sensor. The workability evaluation device according to claim 5, wherein the sensor includes a motion sensor attached to the body of an operator. It is a workability evaluation method that evaluates the workability of the work to be evaluated by a computer.
A step of recording information related to the operation of the worker as work data when a worker including at least one of a human or an AI robot operated by artificial intelligence performs the work to be evaluated.
A step of classifying the entire work to be evaluated into a plurality of partial works by analyzing the movement of the worker included in the work data.
A workability evaluation step for obtaining an index related to workability for each partial work by evaluating the variation of the work data in a plurality of samples for each partial work.
An output step that outputs the evaluation result for each partial work,
Have,
The plurality of samples include a plurality of work data of each of the plurality of workers.
Wherein in the workability evaluating step calculates an average work time of each worker in each partial work obtains an index representing the variation among workers of the average working time for each part work, variations represented in the index The larger the value, the higher the difficulty of the partial work.
In the output step, a workability evaluation method is characterized in that an evaluation result of a work difficulty level for each partial work is output. Prior Symbol workability evaluating step calculates the variation of the working time of each worker in each partial work obtains an index for each partial work represents the variation of the working time of each worker, variation represented in the index The larger the value, the higher the degree of freedom in the work procedure of the relevant partial work.
The workability evaluation method according to claim 7, wherein in the output step, an evaluation result of the degree of freedom of the work procedure for each partial work is output. A program for causing a computer to execute each step of the workability evaluation method according to claim 7 or 8.

Images