WO2020085327A1 - Diagnosing device and diagnosing method - Google Patents

Abstract

In order to diagnose the state of an inspection target which is not exposed at an outer surface, an external appearance image of the inspection target is used to diagnose the state of the inspection target. This diagnosing device is a diagnosing device for diagnosing the state of an inspection target, and is provided with: a feature extracting unit into which sensor data for diagnosis, obtained by measuring the inspection target from the outside, are input, and which extracts at least one pre-determined feature from the sensor data for diagnosis; a score calculating unit for calculating a diagnosis score relating to the sensor data for diagnosis, on the basis of the feature input from the feature extracting unit; and a diagnosing unit which diagnoses the state of the inspection target on the basis of the diagnosis score input from the score calculating unit, and outputs a diagnosis result. The feature extracting unit includes a learned model obtained by mechanical learning using sensor data for learning including sensor data including a feature, and extracts an inferred result from the learned model, corresponding to the input sensor data for diagnosis, as a feature of the sensor data for diagnosis.

Description

Diagnostic device and diagnostic method

The present disclosure relates to a diagnostic device and a diagnostic method.

Conventionally, for example, as a method of evaluating the degree of deterioration of a steel material that causes deterioration such as corrosion, the appearance of the steel material is photographed, and the deterioration degree of the steel material is determined based on the hue and saturation of the surface of the steel material displayed in the captured image. A method for evaluation is known (for example, the deterioration diagnosis method described in Patent Document 1). In such an evaluation method, since the evaluation result can be displayed on the display unit or the like for each color, the user can recognize the deterioration at a glance.

However, the above-described evaluation method is a method of diagnosing and displaying deterioration of the outer surface of the evaluation target, and when the surface of the steel material is covered with a protective material, the captured image does not show a part that causes deterioration. In that case, the evaluation method described above cannot be used. Further, it is not realistic to remove the protective material and perform the above-described evaluation in order to evaluate the degree of deterioration of the steel material when the steel material is a large-scale facility such as a pipe or a power transmission facility in the plant. Therefore, there is a demand for a method of diagnosing the condition of the pipe without removing the protective material (that is, from the appearance of the protective material) even if the surface of the steel material is covered with the protective material.

JP, 2016-223815, A

Therefore, in order to judge the deterioration state of the pipe covered with the protective material without removing the protective material, it is conceivable to infer the deterioration of the pipe by image-processing the image data such as the appearance of the protective material. . However, since the appearance image of the pipe covered with the protective material is an image in which the pipe is not directly displayed, the diagnosis result of the deterioration state of the pipe based on the image is an inference result of the deterioration state. Therefore, when only the diagnosis result of the deterioration state of the pipe covered with the protective material is presented, the user does not know the reason for reaching the diagnosis result, and the user's satisfaction with the diagnosis result becomes low. was there.

The present disclosure has been made in view of such a problem, and an object thereof is to provide a diagnostic device and a diagnostic method that can present a diagnostic result of a diagnostic target so that the user can be satisfied. To get.

In order to solve the above problems, a diagnostic device according to an aspect of the present disclosure is a diagnostic device for diagnosing a state of an inspection target, and diagnostic sensor data obtained by externally measuring the inspection target is input. A feature extraction unit that extracts at least one predetermined feature from the diagnostic sensor data, and a diagnostic partial score associated with the feature of the diagnostic sensor data based on the feature input from the feature extraction unit. And a score calculation unit that calculates the diagnostic score of the diagnostic sensor data based on the diagnostic partial score, and based on the diagnostic score input from the score calculation unit, diagnoses and diagnoses the state of the inspection target. It is characterized by comprising a diagnostic unit for outputting a result and a display unit for displaying the diagnostic result and the diagnostic partial score.

Further, a diagnostic method according to another aspect of the present disclosure is a diagnostic method for diagnosing a state of an inspection target, the step of obtaining the sensor data for diagnosis by externally measuring the inspection target, and from the sensor data for diagnosis, A step of extracting at least one predetermined characteristic; a step of calculating a diagnostic partial score associated with the characteristic of the diagnostic sensor data based on the extracted characteristic; A step of calculating a diagnostic score of the diagnostic sensor data, a step of obtaining a diagnostic result of diagnosing a state of an inspection target based on the diagnostic score, and a step of displaying the diagnostic score and the diagnostic result on a display unit And are provided.

According to one aspect of the present disclosure, it is possible to obtain a diagnostic device and a diagnostic method capable of diagnosing a state of an inspection target without performing complicated work and disclosing the state of the inspection target to a user.

It is a block diagram showing an example of 1 composition of a diagnostic device concerning a first embodiment of this indication. FIG. 3 is a control diagram showing a configuration example of a display unit of a diagnostic device according to a first embodiment of the present disclosure. 3 is a flowchart showing an operation at the time of diagnosis of the diagnostic device according to the first embodiment of the present disclosure. FIG. 3 is a schematic diagram showing a method of calculating a diagnostic score in a diagnostic unit of the diagnostic device according to the first embodiment of the present disclosure. It is a block diagram showing an example of 1 composition at the time of learning of a diagnostic device concerning a first embodiment of this indication. 5 is a flowchart showing an operation at the time of learning of the diagnostic device according to the first embodiment of the present disclosure. It is a schematic diagram explaining a function of a feature learning part of a diagnostic device concerning a first embodiment of this indication. It is a block diagram showing an example of composition at the time of setting of a diagnostic instrument concerning a first embodiment of this indication. 5 is a flowchart showing an operation at the time of setting of the diagnostic device according to the first embodiment of the present disclosure. It is a block diagram showing an example of 1 composition of a diagnostic device concerning a second embodiment of this indication. FIG. 5 is a schematic diagram showing a configuration example of a display unit of a diagnostic device according to a second embodiment of the present disclosure. FIG. 3 is a block diagram showing an example of a hardware configuration of a diagnostic device according to the first and second embodiments of the present disclosure.

Hereinafter, the present disclosure will be described through embodiments, but the following embodiments do not limit the invention according to the claims. In addition, not all of the combinations of features described in the embodiments are essential to the solving means of the invention.

1. First Embodiment Hereinafter, a diagnostic device according to this embodiment will be described with reference to FIGS. 1 to 9. The diagnostic device according to the present embodiment has a function of extracting a feature from the input sensor data and outputting an inference result for the sensor data based on the extracted feature. The diagnostic device according to the present embodiment is, for example, a diagnostic device that diagnoses the state of the inspection target based on image data of an external image obtained by photographing the inspection target from the outside, and is particularly exposed on the outer surface. It is a diagnostic device for diagnosing the internal state of an inspection target that has not been tested. Specifically, the diagnostic device 100 according to the present embodiment diagnoses the state of the internal pipe even when the pipe is not exposed to the outer surface due to the protective material being wound on the surface of the pipe. It is a diagnostic device that can perform.
Note that the diagnostic device according to the present embodiment may use data such as acceleration, angular velocity, sound waves, magnetism, atmospheric pressure or pressure other than image data.

[Structure of diagnostic device]
As shown in FIG. 1, the diagnostic device 100 is a pipe inferred based on image data of a pipe covered with a protective material (hereinafter referred to as a pipe with a protective material) photographed by a camera 200 and an external image of the periphery thereof. It has a function of displaying the state of as the inference result on the display unit 10. Hereinafter, the external image of the pipe with the protective material and its surroundings is referred to as "diagnostic image", and the image data thereof is referred to as "diagnostic image data". The diagnostic device 100 diagnoses the deterioration state of the surface of the protective material extracted from the diagnostic image data and the surface of the pedestal on which the pipe is placed, and as a result of inferring the deterioration state of the pipe, the corrosion level of the pipe or the necessity of inspecting the pipe or The inspection necessity degree is displayed on the display unit 10.

In addition, the diagnostic device 100 uses the image data of the appearance of the pipe with the protective material or the image data of the deteriorated portion around the pipe related to the deterioration of the pipe with the protective material captured by the camera 200 to detect the deteriorated portion as a machine. Has the ability to learn. Hereinafter, the image of the appearance of the pipe with the protective material or the altered portion generated in the peripheral portion thereof is referred to as a learning image, and the image data thereof is referred to as learning image data.
Further, the diagnostic device 100 has a function of setting a criterion for determining a pipe state. The diagnostic criteria are image data of an external appearance image of the pipe with a protective material and its surroundings, the degree of deterioration of which has been determined by the camera 200 in advance, and the degree of deterioration of the pipe taken by the external appearance image (for example, of the pipe). Thickness reduction amount) and. Hereinafter, an image of the appearance of the pipe with a protective material and its surroundings, the degree of deterioration of which is known in advance, is referred to as a setting image, and the image data thereof is referred to as setting image data.
The camera 200 and the diagnostic device 100 will be described below.

(camera)
The camera 200 is a known imaging device, and generates image data (learning image data, setting image data, or diagnostic image data) by photographing the protective material-attached pipe to be inspected and its periphery. The camera 200 outputs image data to the diagnostic device 100 via an external memory that is attachable to and detachable from the camera 200 and the diagnostic device 100, a wired network or a wireless network to which the camera 200 is connected, and the like. The image may be either a still image or a moving image. In the case of a moving image, the still image is cut out to diagnose the internal state of the inspection target.

(Learning image data)
The learning image data captured by the camera 200 is image data of an image of a deteriorated portion which is estimated to be generated on the surface of the protective material or its peripheral portion due to deterioration of the pipe. In many cases, the pipe with the protective material is arranged on a pedestal in the plant. Therefore, the deterioration associated with the deterioration of the pipe with the protective material appears on, for example, the surface of the protective material or the surface of the pedestal arranged below the pipe with the protective material. The image data for learning is generated by capturing an image of a deteriorated portion that appears in a pipe with a protective material or a pedestal and that includes an element effective for determining the deterioration of the pipe. Examples of effective elements include the paint color (discoloration) of the surface of the protective material and the pedestal, peeling of the protective material, deformation of the surface of the protective material, and rust adhered to the pedestal.

The setting image data captured by the camera 200 is the image data of the appearance image of the pipe with the protective material and its surroundings, the degree of deterioration of which is known in advance. The setting image data is image data that includes, for example, the appearance of the pipe with the protective material and its surroundings that was previously captured. For the setting image data, by actually measuring the pipes corresponding to the setting image data, data such as data relating to the thickness of the pipes (for example, the amount of decrease in the thickness of the pipes), which is an index of the internal state, becomes clear. There is.

The diagnostic image data 202 photographed by the camera 200 is data of an image of not only the appearance of the pipe with the protective material but also an area including the pedestal and the like around the pipe with the protective material. The diagnostic image data 202 is obtained by photographing a region in which the appearance may change due to deterioration of the pipe inside the inspection target.
By generating, it is possible to more appropriately diagnose the internal state of the inspection target.

(Diagnostic device)
The diagnostic device 100 includes a feature extraction unit 20, a score calculation unit 30, a diagnostic unit 40, an image synthesis unit 50, and a display unit 10. Hereinafter, each part of the diagnostic device 100 will be described.

(Feature extraction unit)
The feature extraction unit 20 has a data analysis unit 22, a feature inference unit 24, and a feature learning unit 26.
The feature extraction unit 20 receives the learning image data, the setting image data, and the diagnostic image data 202, and extracts at least one predetermined feature from these image data.

The data analysis unit 22 analyzes the learning image data, the setting image data, and the diagnostic image data 202. The data analysis unit 22 calculates and outputs an element used for score calculation from these image data. The data analysis unit 22 outputs the analysis result to the feature inference unit 24 as a feature amount. Further, the data analysis unit 22 outputs the analysis result to the score calculation unit 30 as a feature. As a result, the feature extraction unit 20 can extract, as a feature, the information about the appearance of the protective material-attached pipe or the alteration that occurs in the pedestal or the like around the pipe, in association with the deterioration of the pipe, based on the image data.

The feature inference unit 24 has a learned model 242 obtained in advance by machine learning using the image data for learning in the feature learning unit 26. The feature extraction unit 20 extracts the input image data and the inference result of the learned model for the feature of the image data as the feature of the diagnostic image data.

(Data analysis section)
The data analysis unit 22 analyzes the input image data (learning image data, setting image data, and diagnostic image data). The data analysis unit 22 features an index of alteration appearing in the image based on the colors, lines, and shapes in the images (learning images, setting images, and diagnostic images) corresponding to the image data, and combinations thereof. calculate.
As an indicator of the deteriorated part such as the color of the paint on the surface of the protective material (discoloration), the peeling of the protective material, and the deformation of the surface of the protective material, for example, the area of the discolored surface of the protective material or the frame of the protective material shown in the image Examples include the amount of peeling of the material, the number and area of deformation of the protective material surface.

In the data analysis unit 22, the characteristic points of the image data may be extracted as the characteristic amount from the input image data. As one of the algorithm for extracting the feature points, for example, ORB (Oriented FAST and Rotated BRIEF), SIFT (Scale-invariant feature transform), SURF (Speed-Upped Robust Feature), Harris corner, KAZE, etc. are used. Such an algorithm is stored in a storage unit (not shown) in the feature extraction unit 20 as a feature point extraction program, and is expanded in, for example, a RAM (Random Access Memory) (not shown) during the extraction process of the feature amount and the like.
When performing the feature point extraction process, the data analysis unit 22 performs, for example, a size change process for enlarging or reducing the input image data to a predetermined size, or a smoothing process for reducing noise in the input image data. You may perform pretreatment, such as.

The data analysis unit 22 outputs the analysis result of the image data to the feature inference unit 24 or the score calculation unit 30. For example, the data analysis unit 22 performs edge detection from the image data, and the presence or absence of an altered portion of the protective material surface detected as an area surrounded by the edge (for example, the presence or absence of coating deterioration, discoloration, rust) and the area thereof. It is output to the score calculation unit 30 by using the area and the like as features. In addition, the data analysis unit 22 obtains a feature inferred based on one or a plurality of feature quantities extracted by the image analysis (for example, the presence or absence of deformation or peeling of the surface of the protective material) and the result of the image analysis. Is also output to the feature inference unit 24 as a feature amount.
The data analysis unit 22 also outputs the extracted features to the image synthesis unit 50.

(Feature inference section)
The feature inference unit 24 has a learned model 242 that extracts an inference result for the feature amount input from the data analysis unit 22 as a feature of the diagnostic image data 202. Further, as the input to the learned model 242, the diagnostic image data 202 may be input instead of the analysis result of the data analysis unit 22. The feature inference unit 24 outputs the inference result of the learned model 242 to the score calculation unit 30 as a feature. The feature inference unit 24 also outputs the extracted features to the image synthesis unit 50.

(Feature learning section)
The feature learning unit 26 generates the learned model 242. The feature learning unit 26 generates a learned model 242 by machine learning based on the input learning image data, initial model and learning parameters. The method of machine learning will be described later.

(Score calculator)
The score calculation unit 30 calculates the diagnostic score of the diagnostic image data based on the features input from the feature extraction unit 20. The “diagnostic score” is a score indicating the state of the pipe calculated based on certain diagnostic image data. The "diagnostic score" is a comprehensive diagnostic standard of diagnostic image data calculated based on a score (diagnostic partial score) indicating a diagnostic standard for each feature such as surface unevenness and discoloration of the protective material surface. It is the indicated score (total score). The total score is obtained by weighting and adding each diagnostic partial score.

The score calculation unit 30 may calculate the diagnostic score based on the characteristics input from the characteristic extraction unit 20, data related to the surrounding environment of the pipe, data related to the pipe contents, and the like. . As the data on the surrounding environment of the pipe, a place where the pipe to be diagnosed is installed, weather of the place, temperature, atmospheric pressure, humidity and the like can be used. As the data related to the pipe contents, the type of contents, temperature, pressure, flow velocity, etc. can be used.

The diagnostic partial score is calculated based on the features input from the feature extraction unit 20 and a feature-specific score (coefficient) preset for each feature. Further, the score calculation unit 30 may calculate a plurality of diagnostic partial scores for each of the plurality of diagnostic image data.

The "diagnosis partial score" and the "diagnosis score" may be obtained from the inference result of the learned model (not shown) when the feature is input. As the learning data of the learned model, data including the above-mentioned characteristics and data indicating the presence or absence of deterioration of the pipe or the degree of deterioration is used.
Further, as the inference result of the learned model when the feature is input, the degree of deterioration of the pipe may be directly obtained instead of the “diagnostic score”. The diagnostic score will be described in detail later.

The total score is used, for example, as a diagnostic criterion for diagnosing the condition of the pipe. In this case, the diagnostic partial score is used for calculating the total score and is a score shown to a user such as an inspector as a basis for the diagnosis.
The score calculation unit 30 outputs the diagnostic partial score and the total score to the display unit 10 and the diagnostic unit 40.

(Diagnostic department)
The diagnosis unit 40 diagnoses the internal state of the protective material-equipped pipe based on the diagnosis score input from the score calculation unit 30, and outputs the diagnosis result. The diagnosis unit 40 diagnoses the internal state of the pipe with a protective material based on the above-mentioned total score, for example. The diagnostic unit 40 determines the diagnostic rank of each diagnostic image data by comparing the predetermined total score with at least one threshold value. The diagnosis unit 40 outputs the determined diagnosis rank to the display unit 10.

The diagnostic unit 40 also includes a diagnostic standard setting unit 42 that sets a threshold for determining the diagnostic standard in the diagnostic unit 40. The diagnostic standard setting unit 42 stores the setting image data and the data (for example, the amount of decrease in the thickness of the pipe) that is an index of the internal state obtained by measuring the inside of the pipe with the protective material corresponding to the setting image data. Display on the display. A user who is a threshold value setting person inputs a threshold value setting instruction to the diagnostic reference setting unit 42 based on the displayed data. When setting the above-mentioned threshold, the diagnostic standard setting unit 42 sets a standard value, which is set by the user with respect to data (amount of decrease in thickness of the pipe) that is an index of the internal state of the pipe and determines whether or not to inspect the pipe. It may be configured to set. An upper margin and a lower margin can be further set in the above-mentioned reference value.

(Image synthesis section)
The image synthesizing unit 50 synthesizes the mark corresponding to the feature extracted by the feature extracting unit 20 or the feature inferring unit 24 with the image corresponding to the input diagnostic image data to create an analyzed image. The characteristic mark is, for example, a line indicating the edge of a pipe or a gantry, a circle or a mark (see FIG. 2) indicating a deformation or discolored portion of the surface of the protective material, and even if it is indicated by a different color for each characteristic. good. The analyzed image data combined by the image combining unit 50 is output to the display unit 10.

(Display)
The display unit 10 displays the diagnosis result output from the diagnosis unit 40.
As shown in FIG. 1, the display unit 10 has a diagnosis result display area 110 that shows the diagnosis result of the diagnosis unit 40. In addition, the display unit 10 displays a diagnostic partial score display area 120 that displays a diagnostic partial score and the like as a basis for the diagnosis, and a post-analysis image 134 and a pre-analysis diagnostic image 132 that are combined by the image combining unit 50. Diagnostic image display area 130.

FIG. 2 shows a specific example of the display unit 10. As shown in FIG. 2, in the diagnostic result display area 110 of the display unit 10, the total score calculated based on the diagnostic image data and the diagnostic result (rank) determined based on the total score are displayed. In FIG. 11, the rank “R-1” corresponding to the total score “XXX” is shown. In the diagnostic device 100, for example, ranks "R-1" to "R-4" are set. "R-1" is a rank estimated to have the best pipe condition (low possibility of pipe corrosion), and "R-4" has the worst pipe condition (low possibility of pipe corrosion). High) is the estimated rank. For example, “R-1” and “R-2” are ranks that indicate that follow-up observation is required, and “R-3” and “R-4” are quantitative inspections that actually measure deterioration of pipes (for example, reduction of pipe thickness). It can be a rank indicating that ultrasonic examination for measuring the amount is essential.

Note that the diagnosis result may indicate "quantitative inspection required" or "observation required" instead of the predetermined rank. Further, the user may perform a sampling inspection such as a quantitative inspection or a visual inspection on a part of the piping of the diagnostic image data which has been diagnosed as “necessary observation”. The diagnosis result indicated by the diagnosis device 100 does not merely indicate to the user whether or not the inspection is necessary, but is an index indicating a pipe having a high possibility of deterioration.

In the diagnostic partial score display area 120, the diagnostic partial score and the name of the feature corresponding to the diagnostic partial score are displayed in association with each other. The diagnostic partial score and the name of the feature displayed in the diagnostic partial score display area 120 may be one or plural. In the diagnostic partial score display area 120, when a plurality of diagnostic partial scores and feature names corresponding to the diagnostic partial scores are displayed, they are displayed as graphs such as a radar chart, a bar graph, a line graph, and a scatter diagram. It may be. FIG. 2 illustrates a case where a plurality of diagnostic partial scores and a plurality of feature names are associated with each other and displayed as a radar chart in the diagnostic partial score display area. This makes it easy to visually recognize the diagnostic partial score of each feature.
By displaying a plurality of diagnostic partial scores and feature names together with the total score and rank on the display unit 10, the diagnostic device 100 can show the basis of the diagnosis to the user.

In the diagnostic image display area 130, the post-analysis image 134 input from the image composition unit 50 is displayed. Further, the diagnostic image 132 may be displayed together with the post-analysis image 134 in the diagnostic image display area 130. The diagnostic image 132 is, for example, a diagnostic image corresponding to the diagnostic image data 202.
By displaying the post-analysis image 134 in the diagnostic image display area 130, the diagnostic device 100 can indicate to the user which part of the diagnostic image 132 has which characteristics. In addition, the diagnostic image 100 is displayed in the diagnostic image display area 130 together with the post-analysis image 134, so that the diagnostic device 100 indicates to the user the state of the pipe with protective material whose features are extracted by the feature extraction unit 20. You can

Here, in the diagnostic image display area 130 of FIG. 2, a diagnostic image 132 including the pipe P and the gantry S in which the features F1 to F4 appear is shown. In the diagnostic image display area 130, a post-analysis image 134 of the diagnostic image 132 is shown. In the post-analysis image 134, for example, lines L1 and L2 indicating the edge of the pipe P and a line L3 indicating the edge of the gantry S are combined with the diagnostic image 132. Further, in the post-analysis image 134, lines O1 and O2 showing the outer shapes of the features F1 and F2 analyzed by image analysis, and the features obtained as the inference result of the learned model 242 are obtained with respect to the diagnostic image 132, for example. Marks M3 and M4 shown are combined. Accordingly, the user can see at a glance the features obtained by the image analysis and the features obtained as the inference result of the learned model 242.

[Operation of each part of the diagnostic device]
Hereinafter, the operation of each unit of the diagnostic device 100 will be described.
In the following, the operation for diagnosing the piping state using the diagnostic image data 202, the operation for learning the feature using the learning image data, and the threshold for determining the diagnostic standard using the setting image data. The operation at the time of setting will be described respectively.

(Operation during diagnosis)
First, with reference to FIG. 1, FIG. 3 and FIG. 4, an operation at the time of diagnosing a pipe state using the diagnostic image data 202 will be described. FIG. 3 is a flow chart for explaining the operation at the time of diagnosing the piping condition using the diagnostic image data 202, and FIG. 4 is a schematic diagram showing an example of a method for calculating the diagnostic partial score and the total score.

First, at the start of diagnosis, the learned model 262 generated by the feature learning unit 26 prior to diagnosis is copied to the feature inference unit 24 to generate a learned model 242. The learned model 262 may be used as the learned model 242.
Subsequently, the diagnostic image data 202 obtained by photographing the area including the appearance of the pipe with the protective material and the surrounding pedestal and the like is input to the feature extraction unit 20 of the diagnostic device 100 (step S11).
Next, the data analysis unit 22 of the feature extraction unit 20 performs image analysis on the input diagnostic image data 202 (step S12). The data analysis unit 22 outputs the image analysis result to at least one of the feature inference unit 24 and the score calculation unit 30 and the image synthesis unit 50.

In step S12, when the obtained analysis result is the area of the discolored portion of the protective material surface, the number of surface deformations, or the like, the data analysis unit 22 outputs the analysis result to the score calculation unit 30 as a feature.
Further, in step S12, the data analysis unit 22 outputs the obtained analysis result to the feature inference unit 24 as a feature amount. When it is necessary for the feature inference unit 24 to obtain the inference result of the learned model 242, the feature inference unit 24 displays the analysis results such as the area of the discolored portion of the protective material surface and the number of surface deformations together with the feature amount. You may enter it in 24.

Next, the feature inference unit 24 of the feature extraction unit 20 obtains a feature as an inference result of the learned model 242 for the input feature amount (step S13). When the feature amount is input, the feature inference unit 24 outputs the inference result (feature) inferring the feature of the diagnostic image data 202 having the feature amount to the score calculation unit 30.

Subsequently, the score calculation unit 30 calculates a diagnostic score based on the features input from the data analysis unit 22 and the feature inference unit 24 (step S14). In step S14, first, a diagnostic partial score for each of a plurality of features is calculated based on the input features, and each diagnostic partial score is added to calculate a total score of diagnostic image data.

As shown in an example in FIG. 4, the diagnostic partial score includes feature-specific scores xa to xh predetermined for each feature (feature A to feature H shown in FIG. 4) and corresponding feature amounts (ya to yh). The feature-specific score is a score that is set for each feature so as to have the same tendency as the diagnosis result when a person determines deterioration of the pipe on the diagnostic image of the pipe with protective material. Further, the “feature amount” is a quantitative index of the feature existing in the diagnostic image data 202, and indicates, for example, the area or the number. For example, when the feature is rust, the rust area or the like is shown.

In FIG. 4, the diagnostic partial score A for the features 1 to 8 is indicated by xa · ya. Similarly, the diagnostic partial score B to the diagnostic partial score H are also represented by xb · yb, xc · yc, ... Xh · yh, respectively. The total score can be obtained by adding or weighting the diagnostic partial score A to the diagnostic partial score H. The total score may be a score adjusted to have a predetermined number of digits by multiplying a coefficient after adding the diagnostic partial score A to the diagnostic partial score H. The total score is output to the diagnosis unit 40 as a score used for diagnosis of diagnostic image data.

Subsequently, the diagnosis unit 40 compares the input total score with a preset threshold value, and outputs a diagnosis result of diagnosing the internal state of the pipe with the protective material (step S15). The diagnostic unit 40 outputs the diagnostic rank for the diagnostic image data as the diagnostic result.

Finally, the diagnosis result input from the diagnosis unit 40 is displayed on the display unit 10 (step S16). In step S16, the diagnostic rank for the diagnostic image data and the total score of the diagnostic image data are displayed as the diagnostic result in the diagnostic result display area 110 of the display unit 10. Further, in the diagnostic partial score display area 120 of the display unit 10, each diagnostic partial score and the name of the feature corresponding to the diagnostic partial score are displayed in association with each other. In the diagnostic image display area 130, the post-analysis image 134 input from the image composition unit 50 is displayed. A diagnostic image 132 before analysis may be displayed in the diagnostic image display area 130.

(Operation during learning)
Next, with reference to FIG. 5 and FIG. 6, an operation at the time of learning a feature using the learning image data 204 will be described. At the time of learning, a learned model 242 obtained by machine learning using the learning image data 204 is generated. 5 and 6 are a block diagram and a flowchart for explaining the operation at the time of learning a feature using the learning image data 204. As shown in FIG. 5, when learning a feature using the learning image data 204, the learned model 242 is generated using the feature extraction unit 20 of the diagnostic device 100.

First, a plurality of learning image data 204 obtained by photographing the appearance of the protective material-attached pipe or the altered portion generated on the pedestal or the like around it are input to the feature extraction unit 20 of the diagnostic device 100 (step S21). The learning image data 204 is image data obtained by photographing a portion of the diagnostic image data 202 in which a characteristic, which is a viewpoint to be noticed at the time of diagnosis, is generated as an alteration. It is preferably entered.

Next, the data analysis unit 22 of the feature extraction unit 20 performs image analysis using the ORB algorithm or the like on the input learning image data 204, and extracts the feature amount of the learning image data 204 (step S22). ). The data analysis unit 22 outputs the feature amount (for example, the magnitude and direction of the brightness change) extracted by the image analysis to the feature learning unit 26.

Subsequently, the feature learning unit 26 performs machine learning using the feature amount input from the data analysis unit 22 to generate a learned model 262 (step S23). As a machine learning algorithm, a known one can be used according to the initial model 302 to be used. At this time, as shown in FIG. 5, the initial model 302 and the learning parameter 304 are input to the data analysis unit 22. The initial model 302 may be stored in advance in a storage unit (not shown) of the feature learning unit 26.

The initial model 302 input to the feature learning unit 26 may be any model as long as the learning image data 204 can be used as an input. As the initial model 302 input to the feature learning unit 26, a model such as a support vector machine, a neural network, or a random forest can be used. When the feature learning unit 26 receives the image data 204a, the feature learning unit 26 learns the initial model 302 so as to output the output value indicating the feature (feature type and feature position) 204b included in the image data 204a. A case where a neural network is used as an initial model will be described as an example.

As shown in FIG. 7, the neural network as the initial model 302 to be learned includes an input layer 302a, an intermediate layer (hidden layer) 302b, and an output layer 302c in order from the input side. The intermediate layer 302b is not limited to one layer and may include two or more intermediate layers 302b.
Each layer has one or more neurons. Each of the input layer 302a, the intermediate layer (hidden layer) 302b, and the output layer 302c includes one or more neurons. For example, the number of neurons in the input layer 302a can be set according to the number of pixels in the input image data 204a. The number of neurons in the intermediate layer 302b can be set appropriately according to the embodiment. Further, the output layer 302c can be set according to the number of types of features to be analyzed.

The neurons in adjacent layers are connected appropriately, and weights (connection weights) are set for each connection. In the example of FIG. 7, each neuron is connected to all the neurons in the adjacent layers, but the connection of the neurons is not limited to such an example, and is set appropriately according to the embodiment. You may

The learning parameter 304 is a setting value set to obtain the learned model 262. The feature learning unit 26 scans and sets the parameters with which the learned model 262 can provide the optimum solution. For example, as shown in FIG. 7, by using the learning image data 204 that includes at least one of the features A to H and the position of the feature in the image is known, the learning image The learning parameter 304 is adjusted so that the output value indicating the feature 204b included in the data 204 is output. To adjust the learning parameters 304, a grid search that tries all combinations of parameters that are considered to be appropriate, a random search that randomly tries combinations of parameters, or the like is used, and the combinations of these parameters are input as the learning parameters 304. To be done.

Finally, the learned model 262 generated in step S23 is stored in a storage unit (not shown) of the feature learning unit 26 (or the feature extraction unit 20) (step S24). As described above, the learned model 262 is generated by machine learning using the learning image data 204.

(Operation when setting)
Next, with reference to FIG. 8 and FIG. 9, an operation at the time of setting a threshold value for determining a diagnostic standard using the setting image data 206 will be described. 8 and 9 are a block diagram and a flow chart for explaining the operation at the time of setting the threshold value using the setting image data 206. As shown in FIG. 8, when setting the threshold value using the setting image data 206, the threshold value is set using the feature extraction unit 20, the score calculation unit 30, the diagnosis unit 40, and the display unit 10 of the diagnostic device 100.

First, the setting image data 206 obtained by photographing the area including the appearance of the pipe with the protective material and the surrounding pedestal and the like is input to the feature extraction unit 20 of the diagnostic device 100 (step S31). The setting image data 206 is, for example, image data including the appearance of the pipe with the protective material and its surroundings that was previously photographed. By actually measuring the pipe corresponding to the setting image data, the pipe thickness reduction amount can be calculated in advance. The data that is an index of the internal state such as is clear.

Next, the data analysis unit 22 of the feature extraction unit 20 performs image analysis using the ORB algorithm or the like on the input setting image data 206 (step S32). As an image analysis algorithm, for example, ORB is used. The data analysis unit 22 outputs the result obtained by the image analysis to at least one of the feature inference unit 24 and the score calculation unit 30. In step S32, similar to step S12, the feature is output from the data analysis unit 22 to the feature inference unit 24 or the score calculation unit 30 according to the content of the obtained analysis result.

Next, the feature inference unit 24 of the feature extraction unit obtains a feature as an inference result of the learned model 242 for the input analysis result (step S33). The feature inference unit 24 outputs the obtained inference result (feature) to the score calculation unit 30.
Then, the score calculation unit 30 calculates a setting score based on the features input from the data analysis unit 22 and the feature inference unit 24 (step S34). The setting score is a score similar to the diagnostic score. In step S34, similar to step S14, a partial diagnostic score is calculated for each of a plurality of features based on the input features, and the partial diagnostic scores are added to calculate a total score of the setting image data.

Subsequently, the diagnostic criterion setting unit 42 of the diagnostic unit 40 sets a threshold value based on the total score input from the score calculation unit 30 (step S35). At this time, the diagnostic unit 40 is input with data that is an index of the internal state, such as the pipe thickness reduction amount corresponding to each setting image data 206 input to the feature extraction unit 20. Data that serves as an index of the internal state is also input from the diagnosis unit 40 to the display unit 10.

As shown in FIG. 6, in step S35, for example, the total score and the pipe thickness reduction amount corresponding to each setting image data 206 are displayed on the display unit 10. It is preferable that the total score and the pipe thickness reduction amount corresponding to each setting image data 206 are displayed in ascending or descending order with the total score as a reference. In the display unit 10, the total score is shown by a solid line, and the corresponding pipe thickness reduction amount is shown by a scatter diagram.

The user inputs a threshold setting instruction to the diagnostic reference setting unit 42 via an input unit (not shown) based on the total score and the pipe thickness reduction amount corresponding to each setting image data 206 displayed on the display unit 10. To do. For the threshold setting instruction, for example, a numerical value of the threshold may be input, or the slider displayed on the display unit 10 may be moved with a mouse or the like to determine the threshold. The user uses, as the threshold value, for example, a threshold value TH1 of rank "R-1" and a rank "R-2", a threshold value TH2 of rank "R-2" and a rank "R-3", and a rank "R-3". ] And the rank “R-4” can be input as the total scores as the threshold TH3. Here, for example, “R-1” and “R-2” are ranks indicating the follow-up observations, and “R-3” and “R-4” are actual measurement of deterioration of the pipe such as the amount of decrease in the thickness of the pipe. It can be set as a rank indicating that a quantitative test (for example, ultrasonic test) is required.

In addition, the threshold value may be set so that the diagnosis result leans toward the safer side according to the contents of the pipe. For example, when the content passing through the pipe is a material having the property of increasing the deterioration of the pipe, by setting the above-mentioned threshold value TH2 to a low value, the percentage of pipes judged to require quantitative inspection increases, and the diagnostic result is Lean toward the safer side.
Further, in the diagnostic standard setting unit 42, the threshold value may be automatically set based on a condition previously input to the diagnostic standard setting unit 42, instead of inputting a threshold setting instruction.

As described above, the diagnostic standard setting unit 42 sets the threshold value used at the time of diagnosis. One or more threshold values are set.

(Diagnostic program)
A diagnostic program executed by the diagnostic device 1 according to this embodiment will be described. The diagnostic device 1 diagnoses according to a program that causes a computer to execute the following operations (a) to (f). The following programs are recorded on a recording medium such as a hard disk drive or a memory, or a DVD disk or an optical disk such as Blu-ray (registered trademark). The following programs may be distributed via the Internet. Further, the following program may be recorded in the cloud server and each step may be executed via the Internet.

(A) Externally measuring an inspection target to obtain diagnostic sensor data (b) Extracting at least one predetermined feature from the diagnostic sensor data (c) Based on the extracted feature Calculating a diagnostic partial score associated with a characteristic of the diagnostic sensor data (d) calculating a diagnostic score of the diagnostic sensor data based on the diagnostic partial score (e) a diagnostic score Based on the above, obtaining a diagnosis result of diagnosing the state of the inspection target (f) displaying the diagnosis score and the diagnosis result on the display unit

<Effects of First Embodiment>
The diagnostic device 100 according to the first embodiment has the following effects.
(1) The diagnostic device 100 calculates a diagnostic partial score associated with the feature of the diagnostic image data 202, and displays the diagnostic partial score on the display unit 10 as the basis for the diagnosis. Therefore, the diagnosis basis can be shown to the user at a glance.
(2) In the diagnostic device 100, the inspection target is based on the characteristics (alteration such as discoloration or deformation of the surface of the protective material covering the pipe) that occurs outside the inspection target in relation to the state inside the inspection target (corrosion of the pipe, etc.). It is possible to diagnose the internal condition.

(3) The diagnostic device 100 has a learned model 242 obtained by machine learning using the learning image data 204, and the learned model 242 shows the inference result for the learning image data 204 as a diagnostic image. It is extracted as a feature of the data 202. Therefore, the features of the learning image data 204 can be accurately extracted.
(4) The diagnostic device 100 displays the plurality of diagnostic partial scores A to H and the names of the features (features A to H) corresponding to the diagnostic partial scores A to H as a graph such as a radar chart. Therefore, it is possible to show at a glance to the user which features are more prominent.

(5) The diagnostic device 100 displays the post-analysis image and the pre-analysis diagnostic image 132 in which the analysis results of the diagnostic image data 202 are combined on the display unit 10. Therefore, the user can visually recognize the features existing in the diagnostic image data.
(6) The diagnostic device 100 can selectively use both the feature extraction by image processing and the feature extraction (using the learned model 242) by machine learning. Therefore, more accurate and faster diagnosis can be performed.

(7) The diagnostic device 100 has a diagnostic standard setting unit 42 capable of setting a threshold value for determining a diagnostic standard at the time of diagnosis in the diagnostic unit 40. Therefore, it is possible to set an appropriate threshold value based on restrictions (for example, cost, period, safety, and man-hours) at the time of diagnosis and subsequent quantitative tests.
(8) The diagnostic device 100 further uses the diagnostic image data 202 and the actual measurement value of the deterioration obtained by the quantitative test as learning data and diagnostic data when a quantitative inspection of the piping state is performed after the diagnosis. Enables accurate diagnosis.

(9) In many cases, the pipe with the protective material is arranged on the pedestal in the plant, and corrosion is often observed at the place where the pipe with the protective material is in contact with the pedestal. In order to carry out a quantitative inspection using a measuring instrument at such a place, it is necessary to install scaffolds, lift pipes, and the like, which is very costly. Conventionally, the degree of deterioration of the pipe has been judged by an inspector based on an empirical relationship with the appearance state of the pipe with a protective material. For this reason, since the quantitative inspection is also performed on the pipe that originally does not require the quantitative inspection, unnecessary inspection cost is incurred.
In the diagnostic device 100, a learned model is generated by using effective elements for corrosion estimation extracted from the deteriorated portion appearing around the pipe with the protective material as teacher data, and the inference result of the learned model is used at the time of diagnosis. As a result, it is possible to more appropriately determine the state of the pipe and reduce unnecessary quantitative inspections, so that it is possible to reduce the cost required for quantitative inspections.

2. Second Embodiment Hereinafter, a diagnostic device 300 according to a second embodiment of the present disclosure will be described with reference to FIGS. 10 and 11.
The diagnostic device 300 according to the second embodiment is related to the first embodiment in that the display unit 310 has a multiple target evaluation result display area 350 for displaying diagnostic results for multiple diagnostic image data 202A to 202E. Different from the diagnostic device 100.
The display on the display unit 310 will be described below. Further, the feature extracting unit 20, the score calculating unit 30, the diagnosing unit 40, and the image synthesizing unit 50 are similar to the feature extracting unit 20, the score calculating unit 30, the diagnosing unit 40, and the image synthesizing unit 50 of the diagnostic device 100, and thus will be described. Is omitted.

(Display)
In the display unit 310 of FIG. 10, each display area is schematically shown. The display unit 310 of the diagnostic device 300 has a single target evaluation result display area 340 that displays a diagnostic result, a diagnostic partial score, and the like regarding one diagnostic image data 202A. In addition, the display unit 310 displays a plurality of diagnostic image data (for example, five diagnostic image data 202A to 202E) diagnostic results and diagnostic partial scores side by side for each diagnostic image data 202A to 202E. It has an evaluation result display area 350.

FIG. 11 shows a specific example of the display unit 310. The display unit 310 displays the total score and the diagnostic partial score in the multi-object evaluation result display area 350, for example, in a graph such as a bar graph, a scatter diagram, or a line graph. In the display unit 310 shown in FIG. 11, the total score is shown by a line graph, and the diagnostic partial scores A to D for the features A to D are shown as four bar graphs. Further, in the display unit 310, the diagnostic partial score E to the diagnostic partial score H for the characteristic E to the characteristic H are shown in a scatter diagram by “□”, “◯”, “Δ”, “x” and the like. In the multi-object evaluation result display area 350, the diagnostic partial scores of the diagnostic image data 202A to 202E are displayed at positions close to each diagnostic image data. In addition, the diagnostic partial score groups of the diagnostic image data 202A to 202E are displayed side by side.

Further, the display unit 310, similar to the display unit 10 of the diagnostic device 100, displays the diagnostic result display area 110 for displaying the diagnostic result of one diagnostic image data 202A and the diagnostic partial score of one diagnostic image data 202A. It has a diagnostic partial score display area 120 to be displayed. Further, the display unit 310 has a diagnostic result display area 110, a diagnostic partial score display area 120, and a diagnostic image display area 130, similar to the display unit 10 of the diagnostic device 100. In the diagnostic result display area 110, diagnostic images (diagnostic image 132 before analysis and image 134 after analysis) of one diagnostic image data 202A are displayed. The diagnostic image 132 is, for example, a diagnostic image corresponding to the diagnostic image data 202A.

Further, the display unit 310 has a rank display area 360 in which a table 362 showing the score range for each diagnosis result (rank) and a graph 364 showing the ratio of the diagnostic image data diagnosed for each rank are displayed. There is. The display unit 310 has a diagnostic image data selection area 370 in which a diagnostic result or the like is displayed in the single target evaluation result display area 340 and which enables selection of one image data of the diagnostic image data 202A to 202E. have. The user selects a display indicating any one of the plurality of diagnostic image data 202A to 202E displayed in the diagnostic image data selection area 370, and thus the single target evaluation result display area 340 is selected. The diagnostic result of any one of the diagnostic image data 202A to 202E is displayed.

<Effects of Second Embodiment>
The diagnostic device 300 according to the second embodiment has the following effects in addition to the effects (1) to (9) of the diagnostic device 100 according to the first embodiment.
(10) The diagnostic device 100 has a single target evaluation result display area 340 for displaying a detailed diagnostic result, a diagnostic partial score, and the like on one diagnostic image data 202A on the display unit 10. Further, the diagnostic device 100 has a multiple target evaluation result display area 350 for displaying the diagnostic results and the diagnostic partial scores regarding the diagnostic image data 202A to 202E. The diagnostic device 100 also displays these diagnostic results in graphs such as a radar chart, a bar graph, a scatter diagram, or a line graph. Therefore, the diagnostic device 100 can show the user a detailed diagnostic result for each diagnostic image data and a comparison result between a plurality of diagnostic image data at a glance.

3. Hardware Configuration An example of the hardware configuration of the diagnostic device 1 according to the first embodiment and the diagnostic device 2 according to the second embodiment will be described in detail below.
As shown in FIG. 12, the diagnostic device 1 includes a storage device 1001, a computing device 1002 including a processor, and a communication interface (I / F) 1003 for transmitting / receiving information to / from an external device via a communication path (network) 1110. It has hardware resources. The external device is, for example, an imaging device (not shown) or a computer 1120 that stores an image. The diagnostic device 1 executes information processing based on a predetermined program and data stored in the storage device 1001, thereby diagnosing the state of the inspection target from at least the input image of the inspection target, and the diagnostic score and the diagnostic result. And are displayed on the display unit.

The arithmetic unit 1002 includes a CPU (Central Processing Unit) as a hardware processor, 1002A, a RAM (Random Access Memory) 1002B, and a ROM (Read-Only Memory) 1002C. The arithmetic device 1002 executes specific information processing using predetermined data and a predetermined program stored in the storage device 1001. The hardware processor is not limited to the CPU, and various processors such as GPU (Graphics Processing Unit), ASIC (Application Specific Integrated Circuit), and FPGA (Field-Programmable Gate Array) can be used.

The storage device 1001 is an auxiliary storage device including a memory such as a hard disk drive and a solid state drive. The storage device 1001 stores a predetermined program and predetermined data necessary for specific information processing by the arithmetic device 1002. In the present disclosure, in the storage device 1001, the arithmetic device 1002 executes diagnostic sensor data acquisition processing, feature extraction processing, diagnostic partial score calculation processing, diagnostic score calculation processing, diagnostic result acquisition processing, diagnostic result display processing, and the like. A program 1008A, a score 1008B such as a diagnostic partial score and a diagnostic score (total score), a feature name 1008C, an image 1008D such as the diagnostic image 132 and the post-analysis image 134, and a model 1008E such as a learned model 242. I remember. Each of the program 1008A, the score 1008B, the feature name 1008C, the image 1008D, and the model 1008E has a predetermined area (for example, a program storage area, a score storage area, a feature name storage area, an image storage area, and a model storage area) in the storage device 1001. Etc.) may be stored separately. Further, the program 1008A, the score 1008B, the feature name 1008C, the image 1008D, and the model 1008E are stored in the separately configured program storage unit, score storage unit, feature name storage unit, image storage unit, and model storage unit, respectively. It may be configured to.

Note that the predetermined data and the predetermined program may not always be stored in the storage device 1001. For example, when the arithmetic device 1002 executes specific information processing, a part or all of the information may be acquired from another device via the communication path 1110. Further, the predetermined data and the predetermined program may be read from the storage medium 112 via the drive 111 described later when the arithmetic device 1002 executes the specific information processing.

The communication I / F 1003 is, for example, a wired LAN module, a wireless LAN module, or the like, and is an interface for performing wired or wireless communication with another device via the communication path 1110. In the present disclosure, the communication I / F 1003 is configured to receive predetermined data and the like from the computer 1120 and other devices (not shown) described below via the communication path 1110.

The diagnostic device 1 may further include an input device 113 and an output device 114 such as a display device. A mouse, a keyboard, a touch panel, or the like can be used as the input device 113. A display, a speaker, or the like can be used as the output device 114. The diagnostic device 1 may further include a drive 111 such as a CD drive or a DVD drive for reading data and programs stored in the storage medium 112.
The diagnostic device 2 also has a similar configuration.

Although the specific configurations of the present disclosure have been described above with the respective embodiments, the scope of the present disclosure is not limited to the exemplary embodiments illustrated and described, and the present disclosure is intended. It also includes all embodiments that produce an equivalent effect. Moreover, the scope of the present disclosure is not limited to the combination of inventive features defined by the claims, but can be defined by any desired combination of specific features of each disclosed feature.

10, 310 Display unit 20 Feature extraction unit 22 Data analysis unit 24 Feature inference unit 26 Feature learning unit 30 Score calculation unit 40 Diagnostic unit 42 Diagnostic reference setting unit 50 Image synthesis unit 100, 300 Diagnostic device 110 Diagnostic result display area 120 Diagnostic use Partial score display area 130 Diagnostic image display area 132 Diagnostic image 134 Analyzed image 200 Cameras 202, 202A to 202E Diagnostic image data 204 Learning image data 206 Setting image data 242 Learned model 302 Initial model 304 Learning parameter 340 Single target evaluation result display area 350 Multiple target evaluation result display area 360 Rank display area 370 Diagnostic image data selection area

Claims (20)

A diagnostic device for diagnosing the condition of an inspection target,
A feature extraction unit that receives diagnostic sensor data obtained by measuring the inspection target from the outside and extracts at least one predetermined feature from the diagnostic sensor data,
Based on the feature input from the feature extraction unit, a diagnostic partial score associated with the feature of the diagnostic sensor data is calculated, and diagnosis of the diagnostic sensor data is performed based on the diagnostic partial score. A score calculation unit for calculating a score for
Based on the diagnostic score input from the score calculation unit, a diagnostic unit that diagnoses the state of the inspection target and outputs a diagnostic result,
A display unit that displays the diagnosis result and the diagnostic partial score,
A diagnostic device including. The score calculation unit calculates the diagnostic partial score based on a feature-specific score associated with the feature, and calculates a total score as the diagnostic score.
The diagnostic device according to claim 1, wherein the diagnostic unit diagnoses a state of the inspection target based on the total score. The diagnostic device according to claim 2, wherein the display unit has a diagnostic partial score display area that displays at least one of the diagnostic partial scores and the name of the feature corresponding to the diagnostic partial score in association with each other. The score calculation unit, for the diagnostic sensor data, calculates a plurality of the diagnostic partial score,
The diagnostic device according to claim 3, wherein the display unit displays a plurality of the diagnostic partial scores and the names of the features corresponding to the diagnostic partial scores as a graph in the diagnostic partial score display area. The diagnostic device according to claim 4, wherein the display unit displays a radar chart by associating a plurality of the diagnostic partial scores with a plurality of the feature names in the diagnostic partial score display area. The diagnostic device according to any one of claims 2 to 5, wherein the display unit has a single target evaluation result display area that displays the diagnostic result and the diagnostic partial score regarding one piece of the diagnostic sensor data. The diagnostic device according to claim 6, wherein the display unit has a plurality of target evaluation result display areas that display the diagnostic results and the diagnostic partial scores relating to the plurality of diagnostic sensor data side by side for each diagnostic sensor data. . The diagnostic device according to claim 7, wherein the display unit displays the total score and the diagnostic partial score in a bar graph, a scatter diagram, or a line graph in the multi-target evaluation result display area. It further includes a feature learning unit that generates a learned model that outputs the feature included in the diagnostic sensor data input to the feature extraction unit by machine learning based on learning sensor data including sensor data having the feature. The diagnostic device according to any one of claims 1 to 8. The diagnostic device according to claim 9, wherein the feature extraction unit includes a feature inference unit that outputs, as the feature, an inference result of the learned model based on the diagnostic sensor data input to the feature extraction unit. The diagnostic device according to claim 10, wherein the feature extraction unit includes a data analysis unit that analyzes the diagnostic sensor data and the learning sensor data. The diagnostic device according to claim 11, wherein the data analysis unit outputs the analysis result to the feature inference unit as a feature amount or outputs the result to the score calculation unit as the feature. The diagnostic device according to claim 12, wherein the feature inference unit extracts an inference result of the learned model for the feature amount input from the data analysis unit as the feature of the diagnostic sensor data. The diagnostic device according to any one of claims 11 to 13, wherein the diagnostic sensor data and the learning sensor data are diagnostic image data and learning image data that are image data. The diagnostic device according to claim 14, wherein the data analysis unit performs analysis based on a color, a line, a shape in an image corresponding to the image data, and a combination thereof. An image combining unit that creates a post-analysis image by combining a mark indicating the feature extracted by the feature extracting unit with respect to an image corresponding to the input diagnostic image data,
16. The diagnosis according to claim 14, wherein the display unit has a diagnostic image display region for displaying an image corresponding to the diagnostic image data and the analyzed image input from the image combining unit. apparatus. The diagnostic device according to any one of claims 1 to 16, wherein the diagnostic unit includes a diagnostic standard setting unit that sets a threshold value for determining a diagnostic standard in the diagnostic unit. The diagnostic standard setting unit is set based on setting sensor data including sensor data having the characteristics and data serving as an index of a state obtained by measuring the inspection target corresponding to the sensor data. The diagnostic device according to claim 17, wherein the threshold value is set. The inspection target is a pipe covered with a protective material,
The data that is an index of the state is data regarding the thickness of the pipe,
The diagnostic device according to any one of claims 1 to 18, which comprises: A diagnostic method for diagnosing a condition of an inspection target,
A step of measuring the inspection object from the outside to obtain diagnostic sensor data;
Extracting at least one predetermined characteristic from the diagnostic sensor data;
Calculating a diagnostic partial score associated with the feature of the diagnostic sensor data, based on the extracted feature;
Calculating a diagnostic score of the diagnostic sensor data based on the diagnostic partial score;
A step of obtaining a diagnosis result of diagnosing the state of the inspection object based on the diagnostic score;
Displaying the diagnostic score and the diagnostic result on a display unit;
A diagnostic method comprising.

Images