JP2006171184A - Skill evaluation system and skill evaluation method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system and method for skill evaluation capable of easily and objectively evaluating the proficiency level of a worker to do the work in various works wherein the individual worker's skill is made much of. <P>SOLUTION: A device for evaluating the level of skill of the worker who needs to acquire the skill for the work comprises: a work measuring/detecting device which measures the condition of the work related to the work requiring the skill to be evaluated and detects it as quantitative behavior data and work condition data; a signal processing device which extracts and processes necessary time sequential signals out of the behavior data and the work condition data; a pattern feature extracting device which extracts a specific significant pattern from the processing result of the signal processing means; a pattern storing device 16 which stores previously established model data and information concerning its data scope/pattern; a pattern comparing device which compares the model data pattern stored in the pattern storing device with the worker's pattern extracted by the feature extracting device; a feature expressing device which selectively expresses at least one feature quantity out of the result of the comparison by the pattern comparing device; and a skill evaluation result presenting device 19 which presents an index obtained by making multi-dimensional feature amount which the feature expressing device can express into one dimension as an evaluation result. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a skill evaluation system and a skill evaluation method for objectively evaluating a skill level of an operator.

  Conventionally, there is no scientific and uniform training means to acquire the skilled skills required for special work, and the part that depends on the ability of experts is large. In general, trainers, coaches, and instructors who are specialists in the field provide individual instruction to learners. Under such circumstances, most of the conventional skill evaluation systems measure the work time of the worker, clarify the work points where the skill should be acquired by managing the time, and improve the efficiency.

For example, as described in Patent Document 1, the work required time of an operator is measured, compared with the total actual work time that is the sum of the actual work hours of the worker, and the total actual work time is the total standard work time. There is a device that detects a work process whose actual work time exceeds the standard work time in each work process, evaluates the work proficiency level of the worker, and records the evaluation result in a proficiency level database.
JP 2001-166679 A

  However, in the work learning support apparatus as described in Patent Document 1 described above, the granularity of work monitoring is coarse, and it is difficult to perform analysis that goes into the skills of each worker. In addition, in order to acquire skilled skills for special work, it is not sufficient to present analysis results based on time as disclosed in Patent Document 1 described above. Therefore, it is considered indispensable to analyze the behavior and characteristics of each worker in detail and present the characteristics in a form that the worker can understand.

  On the other hand, human work can flexibly respond to changes in work conditions such as the shape of work objects compared to automatic machines and robots, but it is significantly inferior in terms of work quality stability and work efficiency. It has been pointed out.

  On the other hand, what are the characteristics of the work performed by the worker himself during training or construction work compared to objective features found in these educational programs and the work of skilled workers, and what It has been difficult to understand objectively and quantitatively the guideline on whether it is desirable to make corrections. Recognizing your own habits, peculiarities and characteristics is an existing training system and educational program. However, it was impossible to realize.

  Traditionally, the acquisition and analysis of this information has been mainly targeted in the field of psychology, and it is very appropriate and ideal for standardization of measuring instruments and measurement environments such as a room in a research institution. However, it was usually done in a situation where the conditions and environment were completely different from the actual work site. Such idealization of measurement conditions is not appropriate for measuring equipment, which is not always suitable for the measurement equipment to reproduce the specified measurement accuracy, such as the size of the measuring device and the various measurement conditions including the temperature and humidity of the working environment itself. It can be thought that it is caused by what happens.

  In other words, in the conventional work environment, there is no application example of monitoring the work status online or supporting the work by providing information, etc., and a method for evaluating the skill of the worker based on the result of monitoring the work As an example of application, the problem of selecting a work item that changes according to work and an apparatus that does not hinder the work environment are not known, and it has been considered to be difficult.

  The present invention has been made in view of the above-described circumstances, and analyzes the work in detail and performs the work on the basis of the quantitative analysis results for various works where individual skills of workers are regarded as important. The purpose is to provide a skill evaluation system and a skill evaluation method that can easily and objectively evaluate the proficiency level of a worker and quantitatively compare the proficiency level of the worker with the proficiency level of other workers. To do.

  Another object is to make the features of multidimensional work one-dimensional and express and present it in a form that is easy for the operator to understand.

  In order to solve the above-described problem, the skill evaluation system according to the present invention is an apparatus for evaluating the skill level of an operator who needs to acquire work skills as described in claim 1, and Work measurement detection means for measuring the state of work related to work that requires a certain skill and detecting it as quantitative behavior data and work state data, and extracting necessary time series signals from the behavior data and work state data Signal processing means for processing, pattern feature extraction means for extracting a specific significant pattern from the result of signal processing performed by the signal processing means, exemplary data constructed in advance, and the data range / pattern thereof Pattern storage means for storing information, exemplary data patterns stored in the pattern storage means, and extracted by the feature extraction means A pattern comparison unit that compares the pattern of the operator, a feature expression unit that selectively expresses at least one feature amount from the result of comparison by the pattern comparison unit, and a feature expression unit that can be expressed by the feature expression unit. Skill evaluation result presenting means for presenting an index obtained by making a one-dimensional feature quantity into one dimension as an evaluation result is provided.

  The skill evaluation method according to the present invention is a method for evaluating the skill level of an operator who needs to acquire work skills as described in claim 8 in order to solve the above-described problems, and is subject to evaluation. Behavior / work state data detection and recording step to detect and record quantitative behavior data and work state data by measuring the state of work related to work that requires skills to become, and detection of this behavior / work state data A signal processing step that extracts and processes the necessary time-series signals from the behavior data and work state data detected in the recording step, and a feature extraction that extracts a specific significant pattern from the result of signal processing in this signal processing step A pattern ratio for comparing the step and the pattern of the exemplary data constructed in advance with the pattern of the worker extracted in the feature extraction step Evaluating a step and a feature expression step that selectively expresses at least one feature amount from the comparison result in the pattern comparison step, and a one-dimensional index of multidimensional feature amounts that can be expressed in the feature expression step It includes a result presentation step of performing at least one processing step of a skill evaluation result presentation step presented as a result and presenting the result.

  According to the skill evaluation system and the skill evaluation method according to the present invention, the work is analyzed in detail for various work in which the skill of each worker is regarded as important, and the work is performed based on the quantitative analysis result. It is possible to easily and objectively evaluate an operator's proficiency level and quantitatively compare an operator's own proficiency level with other workers' proficiency levels.

  Further, since the features of the multidimensional work can be expressed and presented in a one-dimensional manner, the features of the multidimensional work can be more easily understood by an operator, an instructor, or the like. Furthermore, since the features of the multidimensional work are expressed and presented in a one-dimensional manner, even a person who has not acquired the skill can clearly judge the evaluation result of the skill of the worker.

  Embodiments of a skill evaluation system and a skill evaluation method according to the present invention will be described below with reference to the accompanying drawings.

  FIG. 1 is a schematic configuration diagram schematically showing the configuration of a skill evaluation system 10 which is an embodiment of the skill evaluation system according to the present invention.

  The skill evaluation system 10 is an apparatus for evaluating how much the operator who performs the work has acquired the required skill. That is, the skill evaluation system 10 is an apparatus that objectively evaluates the level of the skill level of the worker.

  As shown in FIG. 1, when a work to be evaluated is performed, the skill evaluation system 10 includes data relating to an environment for performing the work (hereinafter referred to as a work environment) and work itself (hereinafter referred to as work state data). And the work measurement / detection device 13 as work measurement detection means for measuring / detecting data on the behavior of the worker during work (hereinafter referred to as behavior data), and the work measurement / detection device 13 measured / detected A signal processing device 14 as signal processing means for extracting and processing a necessary time series signal from the work state data and behavior data, and a pattern for extracting a specific significant pattern from the time series signal after signal processing Pattern feature extraction device 15 as feature extraction means, and a pattern storage device for storing pre-constructed exemplary data and data range / pattern information From the comparison results of the pattern storage device 16 as the pattern storage device, the pattern comparison device 17 as the pattern comparison device for comparing the data pattern stored in the pattern storage device 16 with the pattern by the operator, and the pattern comparison device 17 A feature expression device 18 as a feature expression means for expressing a feature amount, and an index obtained by converting the multi-dimensional (plurality) feature amount that can be expressed by the feature expression device 18 into one dimension are presented as a skill evaluation result of the worker. And a skill evaluation result presentation device 19 as a skill evaluation result presentation means.

  Next, the configurations of the devices 13 to 19 included in the skill evaluation system 10 will be described more specifically. In addition, description of FIGS. 2-9 mentioned later is an example at the time of making the target operation | work evaluated by the skill evaluation system 10 into a manual welding operation | work.

  FIG. 2 is a schematic configuration diagram schematically showing the configuration of an embodiment of the work measurement / detection device 13 in the skill evaluation system 10.

  The work measurement / detection device 13 in the skill evaluation system 10 is a device for measuring and detecting quantitative work state data and behavior data, and synchronously measures, detects and records single or multiple time series signals. It has the function to do.

  In addition, the work measurement / detection device 13 uses functions for synchronously measuring, detecting, and recording a single or a plurality of time-series signals to perform work related to work required for the skill to be evaluated. A behavior / working state data detection and recording step is performed in which the state is measured and detected as quantitative behavioral data and work state data.

  By the way, the above function of the work measurement / detection device 13 is, for example, a signal synchronization that synchronizes a single or a plurality of time series signals with a sensing function of sensors such as a sensing device that measures and detects time series signals. This can be realized by providing a function and a signal recording function for recording a synchronized signal. That is, the work measurement / detection device 13 includes a sensor having a sensing function and a signal synchronization / recording unit having a signal synchronization function and a signal recording function, so that a single or a plurality of time series are provided. Signals can be measured, detected and recorded synchronously.

  More specifically, as shown in FIG. 2, the work measurement / detection device 13 includes a CCD camera 21, a gyro sensor 22, an acceleration sensor 23, a motion sensor 24, a temperature sensor 25, and a CCD camera as sensors. 21, AD converter 26 as AD conversion means for converting an analog signal obtained from each of gyro sensor 22, acceleration sensor 23, motion sensor 24, and temperature sensor 25 into a digital signal, and a plurality of digital signals in parallel as synchronization signals A multi-channel multiplexer 27 as a signal synchronization / recording means capable of recording and a welding controller 29 for manual welding work.

  The CCD camera 21 detects video information that visually captures the worker's behavior, the work environment, and the status of the work itself, that is, visual work state data and behavior data. The detected video information is output as a digital signal and input to the multi-channel multiplexer 27.

  Note that the number and specifications of the CCD camera 21 (for example, the number of pixels, focal length, operating temperature, etc.) may be appropriately selected according to the nature of the work to be evaluated.

  For example, when the CCD camera 21 is used for the purpose of measuring the status of work, the same object is measured from different directions for the purpose of calculating three-dimensional coordinates from the relative positional relationship in the camera field of view. In response to changes in workers and work positions as the work progresses, they are placed at different viewing angles, and the measurement object can be continuously captured, macroscopic worker behavior and microscopic work conditions. It is possible to use a plurality of cameras with different magnifications for the purpose of grasping the transition of the camera, and there are no restrictions or restrictions on how to use the CCD camera 21.

  The gyro sensor 22, the acceleration sensor 23, and the motion sensor 24 are devices for detecting three-dimensional information regarding the posture of the worker during work and the relative positional relationship with the work target. The detected three-dimensional information is output as a digital signal and input to the multi-channel multiplexer 27.

  The temperature sensor 25 detects temperature information of an object (for example, an object to be welded) related to the work as work state data and behavior data in order to grasp the work environment and the state of the work itself from the viewpoint of the temperature of the object. The detected temperature information is output as a digital signal and input to the multichannel multiplexer 27.

  The gyro sensor 22, the acceleration sensor 23, the motion sensor 24, and the temperature sensor 25 employ either a method for outputting detected information as a digital signal (digital type) or a method for outputting as an analog signal (analog type). However, when an analog system is adopted, it is necessary to AD-convert the sensor output using the AD converter 26 and then input it to the multi-channel multiplexer 27.

  The sensors 21 to 25 included in the work measurement / detection device 13 preferably use a non-contact sensing device such as the CCD camera 21 except for special conditions in a laboratory or the like. This is because when the sensors 21 to 25 are contact-type sensing devices, the sensors themselves become a hindrance to the performance of the work in actual work.

  The AD conversion device 26 converts an analog signal obtained as a monitoring signal of the state of the device used in the work into a digital signal. It should be noted that the number of analog signals input to the AD converter 26 can be handled regardless of whether it is singular or plural.

  The AD converter 26 is premised on a device that can convert an analog signal in parallel. However, the AD converter 26 may be a device that can handle only a single analog signal. However, in this case, it is necessary to prepare the same number of AD converters 26 as the number of analog signals to be converted into digital signals.

  The multi-channel multiplexer 27 has a function of recording a plurality of digital signals in parallel as synchronization signals. The digital signal recorded by the multi-channel multiplexer 27 is input to the signal processing device 14 as shown in FIG. The multi-channel multiplexer 27 may be configured integrally with at least one of the sensors 21 to 25.

  The welding controller 29 supplies power to the torch bar used for manual welding work and controls it. In addition, the control information for controlling the torch bar by the welding controller 29 is considered to be information that is useful for grasping the behavior of the worker, the work environment, and the situation of the work itself. Therefore, control information for controlling the torch bar, that is, a control signal (analog signal) output from the welding controller 29 is converted into a digital signal via the AD converter 26, and then the multi-channel. Input to the multiplexer 27.

  The work measurement / detection device 13 configured as described above realizes a function of synchronously measuring, detecting, and recording a single or a plurality of time-series signals, and can perform a behavior / work state data detection / recording step. .

  The behavior measurement / detection data detection and recording step performed by the work measurement / detection device 13 includes the monitoring of the melting phenomenon of the metal including the temperature in the welding process and the behavior of the welder's torch carrying rod and the gaze point around the weld pool. Are measured and detected by a plurality of CCD cameras 21. The gyro sensor 22, the acceleration sensor 23, and the motion sensor 24 measure and detect the relative behavior and posture change of the welder's torch bar, and measure the temperature of the molten pool (and its surroundings). Performed by the sensor 25. Further, a control signal (analog signal) is input from the welding controller 29 to the AD converter 26, and the AD converter 26 converts the input control signal from an analog signal to a digital signal.

  Subsequently, the CCD camera 21, the gyro sensor 22, the acceleration sensor 23 and the motion sensor 24, and the temperature sensor 25 transmit various information (digital signals) measured and detected to the multi-channel multiplexer 27. Further, the AD conversion device 26 transmits a control signal (digital signal) after AD conversion to the multi-channel multiplexer 27.

  Subsequently, the multi-channel multiplexer 27 that has received the digital signal records the received digital signal, that is, various information measured and detected by the sensors 21 to 25 and the control signal after AD conversion in parallel as a synchronization signal. . When the multi-channel multiplexer 27 finishes recording in parallel the various information measured and detected by the sensors 21 to 25 and the control signal after AD conversion as a synchronization signal, the behavior / working state data detection and recording step is completed.

  The work measurement / detection device 13 that has completed the behavior / work state data detection and recording step transmits the signals recorded by the multi-channel multiplexer 27 to the signal processing device 14 shown in FIG. 1 as work state data and behavior data.

  The work measurement / detection device 13 is not necessarily limited to the one shown in FIG. That is, it is not necessary to include the CCD camera 21, the gyro sensor 22, the acceleration sensor 23, the motion sensor 24, the temperature sensor 25, the AD converter 26, and the multi-channel multiplexer 27. For example, a single or a plurality of time-series signals can be received. It is not excluded to perform synchronous measurement / detection by combining measurement / detection devices that can measure and detect synchronously or by combining the CCD camera 21 and a time-series signal measurement device.

  In addition, if a plurality of CCD cameras 21 are used in combination and the corresponding arithmetic processing is performed in the feature extraction means 18 arranged at the subsequent stage of the work measurement / detection device 13, the movement of the three-dimensional motion sensor 24 is related. It is also possible to measure and detect information without contact.

  Furthermore, if a CCD camera 21 that covers the range of sensitivity up to the infrared region is prepared, the temperature distribution can be monitored. That is, the CCD camera 21 can also serve as the temperature sensor 25.

  On the other hand, regarding the synchronous measurement in the work measurement / detection device 13, various sampling cycles can be obtained depending on the performance of the measurement device, for example, regarding the mixing method of analog signals and digital signals or the setting method of the digital sampling cycle. However, for these, it is not necessary for all measurement results to always operate in synchronization. For example, if absolute time is measured separately, it is possible to realize simulated synchronous measurement by linear / nonlinear interpolation / interpolation processing between frames in the case of video and between sampling points in the case of signals.

  In manual welding work, for example, the process requires detailed analysis using a high-speed camera or the like, and high-speed sampling of signals. On the other hand, tens to 100 Hz that can be grasped by human eyes such as measurement of worker behavior. Since a case where a fast phenomenon and a slow phenomenon are mixed is assumed such that a relatively slow sampling is sufficient, the work measurement / detection device 13 is also used in a situation where such a fast phenomenon and a slow phenomenon are mixed. Are concerned about the possibility of data recording.

  However, by making the fast sampling period (fast side) a multiple of the other sampling period (slow side), basically the fast and slow side clocks are used with the clock used for fast side sampling. Since sampling can be performed synchronously, even if a fast phenomenon and a slow phenomenon are mixed, data can be recorded without any problem.

  The signal processing device 14 in the skill evaluation system 10 is a time series signal (hereinafter referred to as a skill evaluation signal) required for evaluating the skill of the work to be evaluated from the behavior data and work state data detected by the work measurement / detection device 13. A signal processing function (hereinafter referred to as a signal processing function). Further, the signal processing device 14 performs a signal processing step of extracting a necessary time-series signal from the behavior data and work state data received from the work measurement / detection device 13 and performing signal processing using the signal processing function.

  As the signal processing method performed by the signal processing device 14, various methods including a statistical signal processing method devised and proposed so far according to the signal condition, purpose, and use may be applied. it can.

  As a basic idea about the signal processing method performed by the signal processing device 14, in the case of a video signal recorded by the CCD camera 21 or the like, that is, video information obtained as two-dimensional data, a specific physical quantity is obtained by image processing. Perform processing to extract from video. At that time, according to the viewing angle and azimuth angle, regarding physical quantities such as the size of the captured object and the range of movement, the displacement and coordinates in the virtual space inside the measured image are expressed in real space. A process of converting and converting to a numerical value having a physical dimension (hereinafter referred to as measurement data conversion process) is required.

For example, the CCD camera 21 in FIG. 2 fixed at a specific viewing angle has a length corresponding to m pixels on an image in which an object having a reference actual dimension M is recorded. Assuming that the actual length L of an object of length l on the image can be expressed as shown in Equation 1 below.

  The measurement data conversion process shown as an example in Formula 1 is a very rudimentary process, but is important as a preprocessing process in signal analysis. The signal processing performed by the signal processing device 14 includes measurement data conversion processing as described above.

  FIG. 3 (both FIGS. 3A and 3B) is an explanatory diagram outlining the signal processing contents of the signal processing device 14 in the skill evaluation system 10. More specifically, FIG. 3 (A) is a schematic diagram schematically showing a peripheral image (hereinafter referred to as a molten pool peripheral image) 32 of the molten pool 31 formed in the manual welding operation, and FIG. 3 (B). These are the correlation diagrams showing the relationship between time and the position of the torch carrying rod (torch electrode tip 33).

  When the work for which the skill is to be evaluated is a manual welding work, in the signal processing step performed by the signal processing device 14, the weld pool peripheral image formed in the welding work of the welding operator as shown in FIG. The position of the torch electrode tip 33 and the position of the groove edge 34 of the base material are detected from the 32 video signals (working state data), and the temporal change of the position of the torch electrode tip with respect to the base material is converted to the actual size. Then, a temporal change in the position of the torch carrying rod (torch electrode tip 33) is obtained, and a time series signal (skill evaluation signal) relating to the position of the torch carrying rod is extracted.

  In FIG. 3B, the horizontal axis is time, and the vertical axis is displacement from the groove center line 35 (an example of the position of the torch bar). The position of the torch bar is the groove center line as time elapses. It is possible to grasp how it is displaced with respect to 35, that is, the behavior of the torch bar.

  In this way, when the skill evaluation signal is extracted based on the behavior data and the work state data received from the work measurement / detection device 13, the signal processing step is completed. The skill evaluation signal extracted from the behavior data and the work state data by the signal processing device 14 is transmitted to the pattern feature extraction device 15.

  The pattern feature extraction device 15 in the skill evaluation system 10 has a function of performing a pattern feature extraction process for extracting a specific significant pattern from a skill evaluation signal obtained as a result of signal processing by the signal processing device 14 (pattern feature extraction function). Have Further, the pattern feature extraction apparatus 15 performs a pattern feature extraction step of extracting a specific significant pattern from the received skill evaluation signal by using the pattern feature extraction function.

  When the pattern feature extraction device 15 extracts a specific significant pattern from the skill evaluation signal, if the behavior data and the work state data are measured under a relatively steady state, For example, a significant pattern is extracted by calculating basic statistics such as an average value and a variance value from a data distribution (an example of a skill evaluation signal).

  On the other hand, the calculation of basic statistics is based on the premise that behavior data and work state data are measured in a relatively steady state, so work that continues to change conditions When dealing with the unsteady state obtained from work, another processing method is required. As processing methods for unsteady states, for example, processing methods such as removal of unsteady components by frequency spectrum calculation processing in frequency space and calculation of fractal dimensions to measure the degree of unsteadiness of signals are applied. To do.

  Further, by using both the measured physical quantity and the physical model of the work process, a process of quantitatively estimating other physical quantities that are not directly measured can be performed in the pattern extraction process.

  FIG. 4 (both figures in FIGS. 4A and 4B) and FIG. 5 (both figures in FIGS. 4A and 4B) show the pattern feature extraction device 15 in the skill evaluation system 10. FIG. It is explanatory drawing outlining the content of the pattern feature extraction process.

  More specifically, FIG. 4 (A) is a correlation diagram showing the relationship between time and the voltage level of the torch bar during the manual welding operation, and FIG. 4 (B) is a time series shown in FIG. 4 (A). It is a frequency distribution diagram showing the frequency distribution obtained from data. FIG. 5 (A) is a correlation diagram (corresponding to FIG. 3 (B)) showing the time transition of the torch bar position, and FIG. 5 (B) shows a part of the waveform shown in FIG. 5 (A). It is the elements on larger scale which expanded and expressed.

  When the work whose skill is to be evaluated is a manual welding work, in the pattern feature extraction step performed by the pattern feature extraction device 15, first, in order to extract one of the significant patterns, as shown in FIG. A frequency distribution as shown in FIG. 4B is obtained from time series data (skill evaluation signal) of the voltage level (welding power supply voltage) during welding. Subsequently, the average value Va and the fluctuation range FR are calculated from the frequency distribution of the obtained voltage levels.

  On the other hand, the pattern feature extraction device 15 extracts luck per cycle of the torch bar from time-series data relating to the position of the torch bar as shown in FIG. 5A in order to extract one of the other significant patterns. The rod period T and the rod movement amplitude A are obtained, the average value and the fluctuation range for the rod movement period T and the rod movement amplitude A are obtained, and extracted as a pattern relating to the behavior of the torch rod.

  When the pattern feature extraction apparatus 15 finishes extracting all significant patterns to be extracted, the pattern feature extraction step is completed. Then, the pattern feature extraction device 15 transmits the result of the pattern feature extraction processing to the pattern storage device 16 and the pattern comparison device 17 as a pattern feature signal.

  The significant pattern extracted by the pattern feature extraction device 15 is not limited to the described pattern. That is, what is extracted from various time series data is not limited to the average value and the fluctuation range. Further, even in the case of manual welding work, the type of skill evaluation signal is not limited to the above-described voltage level time series data during welding and time series data related to the torch bar position.

  The pattern storage device 16 in the skill evaluation system 10 has a function (storage function) for recording and storing electronic information. The pattern storage device 16 can store and hold exemplary data and its data range / pattern using a storage function.

  The patterns held in the pattern storage device 16 are exemplary data and models such as time-series information measured and detected in advance by the sensors 21 to 25 in the work measurement / detection device 13. For example, data of moving image captured by the CCD camera 21 is applicable.

  The pattern storage device 16 not only functions as a device that stores the acquired data as an exemplary pattern, but also provides training for workers, confirmation of work status, confirmation of past work evaluation result history, and answers to inquiries. Is held.

  Specifically, the data measured by the work measurement / detection device 13 during the work is numerical data and numerical indices representing features relating to a series of work obtained as a result of the analysis processing by the signal processing device 14 and the pattern feature extraction device 15 in the subsequent stage. Is recorded on the computer's storage medium together with attribute values such as the structure and type of data corresponding to the (data generated by the model or data acquired for skilled workers with normative skills) It is used as a database on the storage device 16.

  It should be noted that retrieval / calling of the stored pattern is performed based on the above attribute value, and the extracted data can be processed into a format that can be expressed by the feature expression device 18 and reused at any time.

  The pattern comparison device 17 in the skill evaluation system 10 has a function (pattern comparison function) for comparing the stored data pattern with the pattern by the operator. Further, the pattern comparison device 17 performs a pattern comparison step of comparing the stored data pattern with the pattern by the operator using the pattern comparison function.

  The pattern comparison device 17 includes at least one pattern comparison processing unit, and each pattern comparison processing unit performs a pattern comparison step using a predetermined pattern comparison method.

  There are several possible pattern comparison methods employed by the pattern comparison device 17, but as a basic example, there is a method of comparing the recalled storage pattern and the work pattern of the worker on the same drawing. Can be mentioned. That is, a method of comparing time series data with the horizontal axis representing time and the vertical axis representing each feature pattern absolute value (hereinafter referred to as absolute value comparison method).

  In addition, there are several methods for comparing on the same drawing, but any method may be adopted. For example, a method of comparing with reference to difference data based on the exemplary pattern of the vertical axis while keeping the horizontal axis as time may be adopted (hereinafter referred to as a difference detection comparison method).

  On the other hand, as another pattern comparison method, there is a method of simply and effectively comparing a plurality of sets of numerical data having different dimensions, such as a pattern comparison plot using a radar chart. In addition to the pattern comparison plot using the radar chart, there is a method of expressing the construction behavior expressed by a plurality of feature values as one plot, and these methods can also be applied.

  For example, a method of normalizing the pattern feature amount of the worker by the exemplary pattern, more specifically, by dividing the pattern numerical value indicated by the worker by the numerical value indicated by the exemplary pattern, A method for expressing and comparing quantities (hereinafter referred to as a normalized comparison method) can also be applied. When the pattern feature amount of the worker is normalized by the exemplary pattern, the feature amount of the worker is greater than 1 when the feature amount is larger than the reference exemplary pattern, and 1 when the same numerical value. If it is smaller, it is represented by a value smaller than 1. Therefore, a plurality of feature quantities having different dimensions on the vertical axis can be written in parallel on the numerical axis with 1 as a reference.

  Other pattern comparison methods include statistical indicators such as average values, variance values, skewness and flatness, which are indicators of the distribution of feature values, as well as the maximum and minimum values of the range in which the feature values change, or data The variation range (maximum value-minimum value) in the change is defined as an important feature quantity, and if necessary, standardized by numerical data that serves as a reference, such as the average value, and introduces a feature quantity that expresses the stability of the parameter. A method of comparison (hereinafter referred to as a stability evaluation comparison method) can be mentioned. Furthermore, if the data is distributed around a certain value, it is also possible to employ a variance value or a standard deviation value in addition to the variation range as a parameter indicating the variation range. The parameter corresponds to a pattern and at least one feature amount included in the pattern.

  Although not shown in detail in FIG. 1, the pattern comparison device 17 performs pattern comparison using the absolute value comparison method so that the pattern comparison step can be performed by appropriately selecting the pattern comparison method described above. Normalization comparison as an absolute value comparison processing unit, a difference detection comparison processing unit as a difference detection comparison unit that performs pattern comparison using a difference detection comparison method, and a standardization unit that performs pattern comparison using a standardized comparison method A processing unit and a stability evaluation processing unit as a stability evaluation unit that performs pattern comparison using the stability evaluation comparison method are provided.

  Note that the pattern comparison device 17 does not necessarily include all of the absolute value comparison processing unit, the difference detection comparison processing unit, the standardization comparison processing unit, and the stability evaluation processing unit, and at least one of the pattern comparison processing units. As long as it has.

  When the pattern comparison device 17 performs a pattern comparison step and finishes comparing the called storage pattern with the work feature pattern by the worker, the pattern comparison step is completed. Then, the pattern comparison device 17 transmits data related to the comparison result to the feature expression device 18.

  The feature expression device 18 in the skill evaluation system 10 has a function (feature expression function) for selecting and expressing at least one feature amount from the result of comparison by the pattern comparison device 17. The feature amount desired to be expressed can be selected when the feature expression device 18 receives an input instruction from the user.

  In addition, the feature expression device 18 uses at least one feature expression method based on the result of comparison by the pattern comparison device 17 using the feature expression function, so that at least one feature quantity can be selectively selected. The feature expression step to express is performed.

  Several feature expression methods adopted by the feature expression device 18 are conceivable. Basically, the comparison result is expressed in accordance with the pattern comparison method adopted by the pattern comparison device 17. As an example of the feature expression method, if the model data pattern used as the reference and the pattern of the worker are expressed by a single index, the horizontal axis represents the time. In addition, there is a method in which the state of change of the corresponding physical quantity is written together with the exemplary data pattern and the worker's pattern with the corresponding physical quantity as the vertical axis.

  FIG. 6 is an explanatory diagram for explaining an example in which the feature expression device 18 in the skill evaluation system 10 includes both a work pattern of an expert who is an exemplary data pattern and a pattern of the worker for a single feature amount (index). FIG.

  In FIG. 6, the horizontal axis represents time, and the vertical axis represents a single feature value, the execution direction (presenting the direction as an angle), and tracking and comparing the construction directions accompanying the work progress of skilled workers and workers, respectively. Yes. According to FIG. 6, since the worker always shows a lower value than the skilled worker, it is possible to easily compare and grasp the work method including the work posture and the like.

  Further, as another feature expression method, there is a method in which the horizontal axis indicates the distinction between the worker and the exemplary data, and the median value or average value of each pattern is expressed as a dot or bar chart on the vertical axis. If a non-stationarity that changes slowly is also observed for the above average value calculation method, the average value or median value is obtained for the data range on a specific time axis, and the data range is moved sequentially. An average value or the like can be calculated, and the state of parameter transition can be represented by a moving average plot.

  Furthermore, as another feature expression method, when a non-stationary signal itself is interpreted as the expression of one work pattern and has a certain reproducibility, a multidimensional space is obtained as a combination of a plurality of acquired data pattern transitions. A method of tracking the state transition trajectory above can be considered.

  More specifically, it is assumed that a plurality of different time-series data are obtained when the behavior of the skilled worker and the behavior of the worker to be compared and the process of change in the physical state in the work are measured. For example, from among multiple feature parameters of a multi-dimensional feature, select a parameter that is particularly sensitive to the skill where the difference is noticeable or evaluated, and set a one-dimensional or three-dimensional space. Expressed by a trajectory in space with time.

  Employing such a method of tracking the state transition trajectory in a multidimensional space is effective as a method of expressing a skill feature of a type that is difficult to express with a statistical feature amount.

  FIG. 7 is an explanatory diagram for explaining an example in which the feature expression device 18 in the skill evaluation system 10 includes a work pattern of an expert who is an exemplary data pattern and a pattern of the worker for a plurality of feature amounts (indexes). It is.

  The plot shown in FIG. 7 is a radar chart comparison plot of multidimensional feature values. The radar chart comparison plot of multi-dimensional features can express multiple features as polygonal patterns, so it is usually difficult to compare work processes performed by humans with a single physical quantity alone It is effective when considered.

  According to FIG. 7, a plurality of feature amounts, that is, seven feature amounts from skill parameter 1 to skill parameter 7 are digitized, and which feature amount is significantly different between the expert and the worker? Can be easily identified and grasped.

  On the other hand, the feature expression device 18 uses a plurality of CCD cameras 21 as shown in FIG. 8 in cases where quantitative expression of features is difficult, such as work where analysis and evaluation of the process is not sufficiently performed. A plurality (four in FIG. 8) of moving images 40a to 40d recorded simultaneously can be simultaneously reproduced and displayed on the display unit 41 of the feature expression device 18. Even if the operator or the instructor who instructs the worker only looks at the moving images 40a to 40d displayed on the display unit 41, a certain effect can be exhibited. In particular, work visualization and work skills that were difficult to record by remote video recording that did not hinder work by recording and playing back details in detail using a small measuring device. Recognition can be shared.

  When at least one of the above-described feature expression methods is adopted and at least one or more feature amounts are selectively expressed, the feature expression step is completed.

  The skill evaluation result presentation device 19 in the skill evaluation system 10 includes a presentation content selection unit as a presentation content selection unit, a specific feature amount automatic selection presentation processing unit and a specific feature amount manual selection presentation processing unit as a presentation feature amount automatic selection unit. And a skill score calculation unit as skill score calculation means.

  The presentation content selection unit of the skill evaluation result presentation device 19 selects a feature amount presentation method (feature amount representation method selection function) and a feature amount to be presented when presenting the skill evaluation result. It has a function (feature quantity selection function).

  As a result of comparing the worker's pattern with the exemplary data pattern, the specific feature amount automatic selection / presentation processing unit automatically performs a process of preferentially presenting a feature amount in which the difference between both patterns is particularly significant. It has a function (specific feature automatic priority presentation function). For example, the specific feature amount automatic selection / presentation processing unit can automatically select and present a feature amount having the most significant difference between the two patterns. In addition, the number of feature quantities presented by the specific feature quantity automatic selection and presentation processing unit is not limited to one, and by updating the setting of the number of selections, it is possible to present a plurality of top ranks for the most prominent feature quantities.

  The specific feature amount manual selection presentation processing unit performs a process of selecting and presenting an arbitrary feature amount of both patterns as a result of comparing the worker's pattern with the exemplary data pattern (specific feature amount manual (Selective presentation function). Therefore, when a feature quantity selection request is input from the user, the specific feature quantity manual selection presentation processing unit can select and present the feature quantity for which the selection request has been made.

  The skill score calculation unit has a function (multi-dimensional feature quantity one-dimensionalization function) that presents as an evaluation result an index obtained by one-dimensionalizing a multi-dimensional feature quantity that can be expressed on the display unit 41 by the feature expression device 18. Have.

  The evaluation result presentation device 19 configured in this way preferentially and automatically selects a feature quantity in which the difference between the two patterns is particularly significant as a result of comparing the worker's pattern with the exemplary data pattern. Can be presented. When a selection request for a feature quantity desired by the user such as an operator or an instructor who instructs the worker is received, the skill evaluation result is displayed on the display unit 43 of the evaluation result presentation device 19 according to the content of the received selection request. be able to. Furthermore, an index obtained by making the multidimensional feature quantity that can be expressed on the display unit 41 by the feature expression device 18 one-dimensional can be calculated as a skill score.

  On the other hand, the skill evaluation result presentation device 19 uses the multi-dimensional feature quantity one-dimensionalization function to present an index obtained by making the multi-dimensional feature quantity that can be expressed by the feature expression device 18 one-dimensional as an evaluation result. An evaluation result presentation step is performed.

  FIG. 9 is an explanatory diagram illustrating an example of the skill evaluation result presented on the display unit 43 of the skill evaluation result presentation device 19.

  The skill evaluation result displayed on the display unit 43 of the skill evaluation result presentation device 19 uses, for example, a time series data plot 46 of work feature values and a pattern feature obtained from the time series data plot 46 of work feature values as indices. The work feature quantity radar charts 47 and 48 and the numerical values obtained by making the multi-dimensional (plurality) pattern features obtained from the work feature quantity time series data plot 46 into one dimension, that is, the work skill score 49 are included.

  In the example of the skill evaluation result shown in FIG. 9, the radar chart 47 displaying the absolute value of the work feature value and the radar chart 48 displaying the variation (relative value) of the work feature value are arranged in the center of the screen, Are arranged with time series data plots 46 of a plurality of work feature amounts. A work skill score 49 is arranged at the upper center of the screen. Of course, the type, number, and arrangement of the charts and plots to be presented are arbitrary and are not limited to those shown in FIG.

  Next, a method for the skill evaluation result presentation device 19 to score the skill of the worker, that is, a method for expressing a multidimensional feature value as a one-dimensional index will be described.

  The calculation of the one-dimensional skill index (work skill score S) is performed by linear superposition of a plurality of feature parameters. At the time of calculating the skill score S, there are Q items of data relating to the model worker from the time-series data extracted in advance, that is, the feature data of the model data, and the total N depending on the difference of the model worker and the number of trials. Assume that a set of work data has been obtained.

Based on the above assumptions, the skill score S is calculated by first selecting K items, which are recognized as effective for skill evaluation, from among Q item feature values (feature parameters), and selecting the selected feature parameter X _ij ( .., K; j = 1, 2,..., N), an average value μ _i and a fluctuation range σ _i are obtained. The average value μ _i and the fluctuation range σ _i can be obtained by the following formula 2.

Note that the fluctuation range σ _i may be obtained by an equation for obtaining a normal statistical standard deviation, that is, the following Equation 3.

When the average value μ _i and the fluctuation range σ _i are obtained, the feature parameter X _ij is then normalized by a statistical method using the average value μ _i and the fluctuation range σ _i . Here, the normalized feature parameter x _ij normalized feature parameters X _ij can be obtained by Equation 4 below.

Subsequently, a correlation matrix R and an inverse matrix R ⁻¹ (= A) of the correlation matrix R are obtained. The correlation matrix R and the inverse matrix A of the correlation matrix R can be obtained by the following Equation 5.

Next, for the data of the worker who is the subject of skill evaluation, select the K feature parameters that are recognized as effective for skill evaluation from the Q item feature values (feature parameters) as in the case of the model data, and select them. The characteristic parameter Y _i (i = 1, 2,..., K) is normalized with the average value μ _i and the fluctuation range σ _i , and the one-dimensional skill index, that is, the work skill score is calculated using the following Equation 6. S is obtained.

  The skill evaluation result presenting step is completed when the skill evaluation result presenting device 19 calculates the work skill score S and presents it as an evaluation result on the display unit 43 of the skill evaluation result presenting device 19 using the above-described mathematical expressions 2 to 6. To do. For example, as shown in FIG. 9, the calculated work skill score S is presented together with at least one of the time-series data and the extracted index.

  As described above, the skill evaluation result presentation device 19 has the specific feature amount automatic priority presentation function and the specific feature amount automatic selection presentation function, so that information adaptively according to the skill level, habit, and learning preference of the worker. Can be processed and provided. That is, it is possible to make a presentation that has an effect of promoting the work improvement of the worker.

  In addition, the processing results can be presented to the worker after the work is completed by a method of expressing the signal level / feature level between the single feature values, so that the objective is based on the data obtained by measuring the work in detail and detail. Skill evaluation results can be provided.

  In the embodiment, the skill evaluation system 10 includes the feature expression device 18 and the skill evaluation result presentation device 19 as separate and independent devices, but may be configured using the same display device. That is, the display unit 41 and the display unit 43 may be the same. One device (for example, the pattern feature extraction device 15) may include another device (for example, the pattern comparison device 17).

  According to the skill evaluation system and the skill evaluation method according to the present invention, the result of skill evaluation is obtained by processing a series of physical data measured in the work process. The acquisition process can be made objective.

  Further, in the skill evaluation system according to the present invention, the skill evaluation result presentation device 19 has a specific feature quantity automatic priority presentation function and a specific feature quantity automatic selection presentation function, so according to the skill level of the worker, the skill, and the learning preference. Therefore, it is considered that information can be processed and provided adaptively, skills can be acquired more efficiently than before, and skill evaluation results can be easily compared with each other.

  Furthermore, the mutual comparison of the skill evaluation results is facilitated, so that even if the administrator is an unskilled administrator, the skill acquisition process is clarified, and the effect of introducing the apparatus can be easily grasped.

  As a result of obtaining such effects, in the field where the skill of the worker influences the quality of the product, for example, in the field of manufacturing handicraft products, the period required for stabilizing the quality, improving the efficiency of work training, and acquiring the skill It can contribute to shortening.

1 is a schematic configuration diagram of a skill evaluation system according to the present invention. 1 is a schematic configuration diagram schematically showing a configuration example of a work measurement / detection device in a skill evaluation system according to the present invention. FIG. 3 is an explanatory diagram outlining the signal processing content of the signal processing device in the skill evaluation system according to the present invention, and FIG. 3 (A) is a schematic diagram schematically showing a peripheral image of a molten pool formed in a manual welding operation. FIG. 3B is a correlation diagram showing the relationship between time and the position of the torch bar. FIG. 4A is an explanatory diagram outlining the contents of pattern feature extraction processing of the pattern feature extraction device in the skill evaluation system according to the present invention, and FIG. 4A is the relationship between time and the voltage level of the torch bar during manual welding work. FIG. 4B is a frequency distribution diagram showing the frequency distribution obtained from the time-series data shown in FIG. It is explanatory drawing which outlined the content of the pattern feature extraction process of the pattern feature extraction apparatus in the skill evaluation system which concerns on this invention, and FIG. 5 (A) is a correlation diagram (FIG. 3 (B) showing the time transition of a torch rod position. 5B) is a partially enlarged view showing a part of the waveform shown in FIG. 5A in an enlarged manner. Explanatory drawing explaining an example which the feature expression apparatus in the skill evaluation system which concerns on this invention described the exemplary data pattern and the said worker's pattern together about the single feature-value. Explanatory drawing explaining an example which the feature expression apparatus in the skill evaluation system which concerns on this invention wrote together the exemplary data pattern and the said worker's pattern about the some feature-value. Explanatory drawing showing an example which reproduces and displays the some moving image recorded with the CCD camera on the display part of the feature expression apparatus in the skill evaluation system which concerns on this invention. Explanatory drawing explaining an example of the skill evaluation result which the skill evaluation result presentation apparatus in the skill evaluation system which concerns on this invention shows on the display part.

Explanation of symbols

10 Skill Evaluation System 13 Work Measurement / Detection Device (Work Measurement Detection Unit)
14 Signal processing device (signal processing means)
15 Pattern feature extraction device (Pattern feature extraction means)
16 pattern storage device (pattern storage means)
17 Pattern comparison device (pattern comparison means)
18 Characteristic expression device (characteristic expression means)
19 Skill evaluation result presentation device (skill evaluation result presentation means)
21 CCD Camera 22 Gyro Sensor 23 Acceleration Sensor 24 Motion Sensor 25 Temperature Sensor 26 AD Converter 27 Multi-Channel Multiplexer 29 Welding Controller 31 Weld Pool 32 Weld Pool Peripheral Image 33 Torch Electrode Tip 34 Base Material Groove Edge 35 Groove Center lines 40a, 40b, 40c, 40d Moving image 41 Display unit 43 of characteristic expression device Display unit 46 of skill evaluation result presentation device 46 Time series data plot 47 Radar chart (absolute value display)
48 Radar chart (relative value display)
49 Work skill score

Claims (8)

In a device that evaluates the skill level of a worker who needs to acquire work skills,
Work measurement detection means for measuring the work state related to the work that requires the skill to be evaluated and detecting it as quantitative behavior data and work state data;
A signal processing means for extracting and processing a necessary time series signal from the behavior data and work state data;
Pattern feature extraction means for extracting a specific significant pattern from the result of the signal processing by the signal processing means;
Pattern storage means for storing pre-constructed exemplary data and information on the data range / pattern;
Pattern comparison means for comparing the pattern of exemplary data stored by the pattern storage means with the pattern of the worker extracted by the feature extraction means;
Feature expression means for selectively expressing at least one feature quantity from the result of comparison by the pattern comparison means;
A skill evaluation system comprising skill evaluation result presentation means for presenting, as an evaluation result, an index obtained by one-dimensionalizing a multidimensional feature quantity that can be expressed by the feature expression means.   2. The skill according to claim 1, wherein the pattern comparison unit includes a difference detection comparison unit that detects and compares the difference between the worker pattern extracted by the feature extraction unit and an exemplary data pattern. Evaluation system. The skill evaluation system according to claim 1, wherein the pattern comparison unit includes a normalization unit that statistically normalizes features of the work performed by the worker with reference to an exemplary data pattern. 2. The pattern comparison unit includes a stability evaluation unit that evaluates the stability of at least one of the worker pattern extracted by the feature extraction unit and a feature amount included in the pattern. Skill evaluation system. The skill evaluation result presentation means includes a presentation content selection means for selecting a part of feature quantities from among a plurality of feature quantities and displaying an original signal subjected to feature extraction together with a signal processing result. Item 1. The skill evaluation system according to Item 1. The skill evaluation result presentation means automatically selects a feature amount indicating a result far from the most exemplary data or pattern based on the feature of the work by the worker, and automatically presents the feature amount to be presented to the worker. The skill evaluation system according to claim 1, further comprising selection means. The skill evaluation result presenting means includes skill score calculating means for calculating, as a skill score, an index obtained by converting a multidimensional feature quantity that can be expressed by the feature expressing means into one dimension. Skill evaluation system. It is a method to evaluate the skill level of workers who need to acquire work skills,
A behavior / work state data detection recording step for measuring and recording quantitative behavior data and work state data by measuring a work state related to a work that requires the skill to be evaluated;
A signal processing step for extracting and processing a necessary time series signal from the behavior data and the work state data detected in this behavior / work state data detection recording step;
A feature extraction step of extracting a specific significant pattern from the result of signal processing in this signal processing step;
A pattern comparison step for comparing a pattern of exemplary data constructed in advance with the pattern of the worker extracted in the feature extraction step;
A feature expression step that selectively expresses at least one feature quantity from the comparison result in the pattern comparison step and an index obtained by making the multidimensional feature quantity that can be expressed in this feature expression step into one dimension are presented as evaluation results. A skill evaluation method comprising: a result presentation step of performing at least one processing step of the skill evaluation result presentation step to be performed and presenting the result.

Images