JP2021131310A - Piping deterioration diagnosis device and piping deterioration diagnosis system - Google Patents

Abstract

To provide a piping deterioration analysis device and a piping deterioration analysis system capable of more easily analyzing deterioration of piping.SOLUTION: A piping deterioration diagnosis device 100 includes: a piping deterioration model creation unit 40 for creating a learning model of a relation between deterioration diagnosis determination for a piping sample inspected by piping extraction in the past and various parameters regarding piping of the piping sample; and a piping deterioration estimation unit 50 for estimating the state of deterioration of the piping which is a deterioration diagnosis target based on the learning model and the various parameters regarding the piping which is the deterioration diagnosis target. The piping deterioration diagnosis device creates a learning model indicating the relation based on information on the deterioration diagnosis determination of the piping sample inspected by piping extraction in the past and data of the various parameters regarding the piping of the piping sample inspected by piping extraction in the past, and can estimate the state of deterioration of the piping which is the deterioration diagnosis target based on the learning model and data of the various parameters regarding the piping which is the deterioration diagnosis target.SELECTED DRAWING: Figure 1

Description

The present invention relates to a pipe deterioration diagnosis device and a pipe deterioration diagnosis system for diagnosing a deterioration state of a pipe.

Conventionally, as a method of evaluating the deterioration status of piping such as corrosion and wall thinning, after grasping the characteristics of equipment by preliminary investigation, the parts to be inspected are selected from each part of the piping, and X is selected for each selected part. A technique is known in which an appropriate inspection method is selected from line inspection, endoscopy, ultrasonic inspection, water quality analysis, macroscopic observation, etc., and a predetermined inspection is performed (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2014-202603

However, in the case of the technique of Patent Document 1, it is necessary to stop a part of the equipment to be inspected so that water does not flow to the piping under inspection, and the work is performed for each inspection method. There was a problem that the inspection would be relatively large because it had to be prepared.

An object of the present invention is to provide a pipe deterioration diagnosis device and a pipe deterioration diagnosis system capable of diagnosing deterioration of pipes more easily.

In order to achieve the above object, the present invention is a pipe deterioration diagnostic apparatus.
A piping deterioration model creation unit that creates a learning model of the relationship between the deterioration diagnosis judgment of a piping sample that has been extubated in the past and various parameters related to piping of the piping sample using a machine learning method.
A pipe deterioration estimation unit that estimates the deterioration state of the pipe subject to deterioration diagnosis based on the learning model and various parameters related to the pipe subject to deterioration diagnosis.
To have.

In the case of a pipe deterioration diagnosis device having such a configuration, the relationship between them is based on the information on the deterioration diagnosis judgment of the pipe sample that has been extubated and inspected in the past and the data of various parameters related to the piping of the pipe sample that has been extubated and inspected in the past. It is possible to create a learning model showing the above, and estimate the deterioration state of the piping subject to deterioration diagnosis based on the learning model and the data of various parameters related to the piping subject to deterioration diagnosis.
In other words, the deterioration state of the pipe is estimated and the renewal / replacement time of the pipe is examined without performing X-ray inspection, endoscopy, ultrasonic inspection, etc. for the pipe subject to deterioration diagnosis. be able to.
With such a pipe deterioration diagnosis device, deterioration of the pipe can be diagnosed more easily.

Also, preferably
The various parameters related to the piping include at least one of a parameter relating to the use of the piping, a parameter relating to the joint core of the piping, a parameter relating to the water source of the piping, and a parameter relating to the aging of the piping.

By using the learning model created based on the data of such parameters, it is possible to more preferably diagnose the deterioration of the pipe.
Of course, if it is a learning model created based on data that includes all four parameters related to the use of the pipe, the parameters related to the joint core of the pipe, the parameters related to the water source of the pipe, and the parameters related to the aging of the pipe. , It is possible to diagnose the deterioration of the piping even better.

Also, preferably
The pipe deterioration model creation unit relates the deterioration diagnosis judgment of a predetermined number of pipe samples randomly extracted from all the pipe samples that have been extubated in the past and various parameters related to the pipes of the extracted pipe samples. A plurality of preliminary models were created by performing machine learning of sex a plurality of times, and the learning model was created by optimizing the plurality of preliminary models by averaging them.

By doing so, even if there is a bias due to an accidental pattern in the pipe sample that has been extubated in the past, it is possible to create a learning model that suppresses the influence of the bias.

Further, in order to achieve the above object, the present invention has a piping deterioration diagnosis system.
An information storage means for storing information on deterioration diagnosis judgment of piping samples that have been extubated and inspected in the past,
Parameter storage means for storing data of various parameters related to piping of piping samples that have been extubated and inspected in the past,
A piping deterioration model creating means for creating a learning model of the relationship between the deterioration diagnosis determination information and data of various parameters related to the piping by using a machine learning method.
Based on the learning model and various parameters related to the pipe subject to deterioration diagnosis, a pipe deterioration estimation means for estimating the deterioration state of the pipe subject to deterioration diagnosis, and a pipe deterioration estimation means.
To have.

In the case of a pipe deterioration diagnosis system having such a configuration, the relationship between the information on the deterioration diagnosis judgment of the pipe sample that has been extubated and inspected in the past and the data of various parameters related to the piping of the pipe sample that has been extubated and inspected in the past. It is possible to create a learning model showing the above, and estimate the deterioration state of the piping subject to deterioration diagnosis based on the learning model and the data of various parameters related to the piping subject to deterioration diagnosis. Then, the deterioration of the pipe can be diagnosed more easily.

According to the present invention, deterioration of piping can be diagnosed more easily.

It is a block diagram which shows the functional structure of the pipe deterioration diagnosis apparatus of this embodiment. It is a flowchart which shows the operation example of a pipe deterioration diagnosis apparatus. It is explanatory drawing about the machine learning which creates the learning model which shows the relationship between various parameters about piping of a piping sample, and deterioration diagnosis judgment of a piping sample. It is explanatory drawing about the data of various parameters about piping. It is explanatory drawing of the mode which estimates and outputs the deterioration state of the pipe which is the deterioration diagnosis target based on a learning model. It is explanatory drawing about the modification of the machine learning which creates a learning model.

Hereinafter, embodiments of the pipe deterioration diagnosis device and the pipe deterioration diagnosis system according to the present invention will be described in detail with reference to the drawings. However, although the embodiments described below are provided with various technically preferable limitations for carrying out the present invention, the scope of the present invention is not limited to the following embodiments and illustrated examples.
In this embodiment, among the pipes, the water supply / drainage pipe through which city water, sewage, etc. flows is targeted for diagnosis.

FIG. 1 is a block diagram showing a specific example of the functional configuration of the pipe deterioration diagnosis device 100 of the present embodiment. The pipe deterioration diagnosis device 100 includes a CPU (Central Processing Unit), a memory, an auxiliary storage device, and the like connected by a bus, and executes a program.
As shown in FIG. 1, for example, the pipe deterioration diagnosis device 100 includes a deterioration diagnosis determination information storage unit 10, a pipe parameter storage unit 20, an information input unit 30, a pipe deterioration model creation unit 40, and a pipe deterioration estimation unit by executing a program. It functions as a device including a unit 50 and a display unit 60.

Even if all or part of each function of the pipe deterioration diagnosis device 100 is realized by using hardware such as ASIC (Application Specific Integrated Circuit), PLD (Programmable Logic Device), and FPGA (Field Programmable Gate Array). good. The program may be recorded on a computer-readable recording medium. The computer-readable recording medium is, for example, a flexible disk, a magneto-optical disk, a portable medium such as a ROM or a CD-ROM, or a storage device such as a hard disk built in a computer system. The program may be transmitted over a telecommunication line.

The deterioration diagnosis determination information storage unit 10 and the piping parameter storage unit 20 are configured by using a storage device such as a magnetic hard disk device or a semiconductor storage device.
The deterioration diagnosis judgment information storage unit 10 stores the deterioration diagnosis judgment information of the pipe sample that has been extubated and inspected in the past, and the piping parameter storage unit 20 stores various parameter data related to the piping of the pipe sample that has been extubated and inspected in the past. Remember.

The information input unit 30 has a pointing device such as a keyboard and a touch panel, and receives an input operation from the outside. The input information and data are stored in the deterioration diagnosis determination information storage unit 10 and the piping parameter storage unit 20, respectively.
Further, the information input unit 30 receives input of data of various parameters related to the piping to be diagnosed for deterioration.
The information input unit 30 is also used as an operation unit, for example, when various programs are manually started or terminated, or when the display content is switched by the display unit 60.

The pipe deterioration model creation unit 40 has stored in the deterioration diagnosis judgment information storage unit 10 the information on the deterioration diagnosis judgment of the pipe sample that has been extubated and inspected in the past, and has been stored in the pipe parameter storage unit 20 in the past. Machine learning processing is performed using the data of various parameters related to the piping of the pipe sample that has been extubated and inspected as teacher data.
A machine learning technique for creating or updating a learning model by supervised learning is applied to the pipe deterioration model creating unit 40. For example, a machine learning algorithm such as the k-nearest neighbor method (kNN) is applied to create a learning model.
The pipe deterioration model creating unit 40 learns the relationship between the information of the deterioration diagnosis determination of the pipe sample and the data of various parameters related to the pipe of the pipe sample by performing the machine learning process based on the teacher data. Specifically, the pipe deterioration model creating unit 40 creates or updates a learning model showing the relationship between the information of the deterioration diagnosis determination of the pipe sample and the data of various parameters related to the pipe of the pipe sample.
Then, the pipe deterioration model creation unit 40 outputs the created or updated learning model to the pipe deterioration estimation unit 50.

The pipe deterioration estimation unit 50 estimates the deterioration state of the pipe to be diagnosed based on the learning model created or updated by the pipe deterioration model creation unit 40 and the data of various parameters related to the pipe to be diagnosed.
The evaluation result / diagnosis result of the deterioration state of the pipe subject to deterioration diagnosis, which is estimated by the pipe deterioration estimation unit 50, is output to the display unit 60.

The display unit 60 has a display device such as a liquid crystal display or an organic EL display, and displays the evaluation result / diagnosis result of the deterioration state of the pipe to be diagnosed by the pipe deterioration estimation unit 50.

Next, an operation example of the pipe deterioration diagnosis device 100 of the present embodiment will be described with reference to the flowchart of FIG.

First, the information on the deterioration diagnosis determination of the pipe sample that has been extubated in the past and the data of various parameters related to the piping of the pipe sample that has been extubated in the past are accumulated (step S1).
For example, as shown in FIG. 3, the parameters related to the piping of the piping sample in the present embodiment include "use", "position", "aging", "pipe type", "caliber", "connection method", "part", "joint core", and "joint core". Eleven parameters of "water source", "location", and "heat insulating material" were set.
Further, as shown in FIG. 3, five items of "straight pipe portion", "connection portion", "joint portion", "blockage rate", and "comprehensive" are set as the items for determining the deterioration diagnosis of the piping sample in the present embodiment. ..
The "straight pipe part" is a deterioration diagnostic standard determined from the wall thinning rate of the base material, and the "connection part" and "joint part" are the loss ratio of the base material at the connection point between the straight pipe and the joint (deficiency of the screw part). Deterioration diagnostic criteria judged from (ratio)," blockage rate "is a deterioration diagnostic standard judged from the void ratio (void ratio inside the pipe (blockage ratio inside the pipe)" excluding the substance part adhering to the inside of the pipe in the cross section of the pipe. , "Comprehensive" refers to the worst judgment among the results judged by "straight pipe part", "connection part", "joint part", and "occlusion rate", respectively.
Then, the data of various parameters related to the piping of the piping sample, which is input via the information input unit 30, is stored and accumulated in the deterioration diagnosis determination information storage unit 10.
Further, the information of the deterioration diagnosis determination of the piping sample input via the information input unit 30 is stored and accumulated in the piping parameter storage unit 20.
Then, data and information of hundreds, thousands, and tens of thousands of samples are accumulated.
The parameters and items shown here are merely examples, and other parameters and items may be set, and the illustrated parameters and items may not necessarily be set. However, the various parameters related to piping include at least one of the parameters related to the "use" of the piping, the parameters related to the "joint core" of the piping, the parameters related to the "water source" of the piping, and the parameters related to the "aging" of the piping. Make sure it is included. Since it was confirmed that the accuracy of deterioration judgment decreases when the learning model is created excluding the four elements of "application", "joint core", "water source", and "aging", these four elements are necessary for deterioration judgment. I decided that there was.
For example, when the parameters of these four elements are not included, the correct answer rate decreases from 63% to 50% in the "straight pipe part", and the correct answer rate decreases from 71% to 46% in the "connection part". , The correct answer rate decreased from 74% to 53% in the "joint part", the correct answer rate decreased from 69% to 55% in the "blockage rate", and the correct answer rate decreased from 70% to 57% in the "overall". .. The "correct answer rate" means the rate of the number of samples in which the judgment result of the evaluation and the comprehensive evaluation judgment which is the correct answer judgment match.
In addition, the sample No. shown in FIG. 1-No. The data and information input to each parameter and each item in No. 3 are for explaining the present embodiment and are different from the actual data.

Regarding the input of data of various parameters related to piping, for example, as shown in FIG. 4, the input of data of "use" is predetermined as water supply; 0, sewage; 1, kitchen drainage; 2, drainage; 3. It is done by inputting the given numerical value.
Further, as shown in FIG. 4, for example, as shown in FIG. 4, the data of "position" is input by inputting predetermined numerical values such as upstream; 0, midstream; 1, downstream; 2, unknown; 3. ..
In addition, for the "aged" data, the value of the number of years of use of the pipe is entered.
Regarding the input of "pipe type" data, for example, as shown in FIG. 4, VP (PVC pipe); 0, VLP (PVC lining steel pipe); 1, DVLP (hard PVC lining steel pipe); 2, CIP ( Cast iron pipe); 3, DCIP (Ductile cast iron pipe); 4, SGP (Carbon steel pipe); 5, MD (Flexible joint); 6, SUS304 (Stainless steel pipe); 7, HIVP (Impact resistant PVC pipe); 8 and a predetermined numerical value are input.
In addition, for the "caliber" data, the value of the caliber of the pipe is input.
Regarding the input of the data of the "connection method", for example, as shown in FIG. 4, the input of a predetermined numerical value is as follows: screw; 0, welding; 1, mechanical or insertion; 2, adhesion; 3. I tried to do it.
Regarding the input of "part" data, for example, as shown in FIG. 4, the horizontal pulling portion; 0, the vertical pipe portion; 1, the joint portion; 2, the valve connection portion; 3, the joint portion + the valve connection. Part; 4 and a predetermined numerical value are input.
Further, as shown in FIG. 4, for example, the data of the "joint core" is input by inputting a predetermined numerical value such as none; 0, pipe end; 1, whole surface; 2.
Regarding the input of "water source" data, for example, as shown in FIG. 4, city water; 0, well water (after treatment); 1, city water + well water; 2, well water (before treatment); 3 , Sewage; 4, kitchen drainage; 5, drainage; 6, and so on.
Regarding the input of "location" data, for example, as shown in FIG. 4, indoor; 0, outdoor; 1, culvert; 2, soil; 3, unknown; 999, and so on. I tried to do it.
Further, as shown in FIG. 4, for example, as shown in FIG. 4, the data of the "heat insulating material" is input by inputting a predetermined numerical value such as Yes; 0, No; 1, Unknown; 999.
In machine learning, it is necessary to set some number in the input data, so even if it is unknown, the number must be input. In the opportunity learning to extract features from the difference in numbers and learn, by setting the unknown to "999" etc., it is possible to learn as an outlier (singular point) because it is two digits different from other input data, and learning. It is possible to reduce the impact on the results.

Further, regarding the input of the information of the deterioration diagnosis judgment of the piping sample, the five-grade evaluation of A (worst) to E (best) is input as the information.
For the "straight pipe part", for example, the diagnosis of the wall thinning rate, for the "connection part", for example, the diagnosis of the corroded state of the threaded part, and for the "joint part", for example, the diagnosis of the wall thinning rate, the "occlusion rate". For example, the diagnosis of the state of obstruction in the pipe was made, and the diagnosis of "comprehensive" was made for the comprehensive evaluation, and each was evaluated on a 4-point scale.
The "straight pipe part" is a straight pipe part, the "joint part" is an L-shaped pipe part connecting the "straight pipe part", and the "connection part" is a "straight pipe part" or "joint part". It is the connection part of the end part.

Next, the pipe deterioration model creation unit 40 determines the deterioration diagnosis judgment of the pipe sample stored in the pipe parameter storage unit 20 and the data of various parameters related to the pipe of the pipe sample stored in the deterioration diagnosis judgment information storage unit 10. Machine learning is performed on the relationship with information (see FIG. 3), and a learning model showing the relationship is created (step S2).

Then, after creating a learning model (learning model for diagnosing deterioration of piping), input of data of various parameters related to the piping to be diagnosed for deterioration is accepted (step S3).
The various parameters related to the piping subject to deterioration diagnosis are the same as the various parameters related to the piping accumulated for creating the learning model, and the data of the various parameters related to the piping subject to deterioration diagnosis are obtained in advance. Has been acquired.

Then, when data of various parameters related to the pipe to be diagnosed for deterioration is input by the user via the information input unit 30, for example, as shown in FIG. 5, the pipe deterioration estimation unit 50 receives the input various parameters. Based on the data and the learning model, the deterioration state of the pipe to be diagnosed for deterioration is estimated, and the deterioration estimation result for the pipe is displayed on the display unit 60 and output (step S4).
Here (in FIG. 5), the piping No. of the deterioration diagnosis target. As data of various parameters for 1, "use;0""position;1""aged;25""pipetype;1""caliber;32""connectionmethod;0""part;2""jointcore;0" As a result of inputting "water source; 0", "location;0", and "heat insulating material; 0", the pipe No. of the deterioration diagnosis target. It shows the mode in which the evaluations of "connection portion; C", "joint portion; C", "blockage rate; D", and "comprehensive;C" are output as the deterioration estimation result in 1.
In the deterioration estimation result, "evaluation D" is evaluated as "currently usable", "evaluation C" is evaluated as "regular diagnosis is recommended", and "evaluation B" is evaluated as "3 to 5 years later". Is the evaluation of "Piping renewal time", and "Evaluation A" is the evaluation of "Piping renewal time (the pipe should be renewed immediately)".

As described above, according to the pipe deterioration diagnosis device 100 of the present embodiment, the information on the deterioration diagnosis determination of the pipe sample that has been extubated in the past and the data of various parameters related to the piping of the pipe sample that has been extubated in the past can be obtained. Based on this, a learning model showing the relationship between them can be created, and the deterioration state of the piping subject to deterioration diagnosis can be estimated based on the learning model and data of various parameters related to the piping subject to deterioration diagnosis.
That is, it is possible to estimate the deterioration state of the pipe and examine the renewal time of the pipe without performing X-ray inspection, endoscopy, ultrasonic inspection, etc. for the pipe subject to deterioration diagnosis. .. As a result, the cost for maintenance of piping can be suppressed.
As described above, the pipe deterioration diagnosis device 100 can more easily diagnose the deterioration of the pipe.

For example, as shown in Table I below, the correct answer rate of the deterioration evaluation model by the pipe deterioration diagnosis device 100 of the present embodiment was 70% accuracy in total. Even in the case of an incorrect answer, the judgment to the dangerous side (which makes a good judgment on the bad one) was 7% accurate.
This result can be said to be superior in terms of accuracy as compared with the current visual inspection.

In Table I, "number of samples" means the number of piping samples for which the evaluation was performed.
“Match” means the number of samples in which the judgment result of the evaluation and the comprehensive evaluation judgment which is the correct answer judgment match.
"Disagreement (safety)" means the number of samples judged to the safe side (good ones are badly judged) even if the answer is incorrect.
"Disagreement (danger)" means the number of samples that are judged to the dangerous side (bad ones are judged well) even if the answer is incorrect.
The "correct answer rate" means the rate of the number of samples in which the judgment result of the evaluation and the comprehensive evaluation judgment which is the correct answer judgment match.
The "wrong answer rate (safety)" refers to the rate at which a judgment is made to the safe side (good is judged bad) even if the answer is wrong.
The "wrong answer rate (danger)" refers to the rate at which a dangerous answer is judged (a bad one is judged well) even if the answer is wrong.

The present invention is not limited to the above embodiment.
The pipe deterioration model creation unit 40 of the pipe deterioration diagnosis device 100 includes information on deterioration diagnosis determination of a predetermined number of pipe samples randomly extracted from all the pipe samples that have been extubated and inspected in the past, and the extracted pipe samples. A plurality of preliminary models may be created by performing machine learning on the relationship with the data of various parameters related to the piping of the above, and the plurality of preliminary models may be averaged to create a learning model.

For example, as shown in FIG. 6, two-thirds of the piping samples are randomly extracted from all the piping samples that have been extubated in the past, and the information on the deterioration diagnosis judgment of the piping sample and various parameters related to the piping of the piping sample are displayed. Based on the data, create a preliminary model showing the relationship between randomly extracted piping samples. In order to prevent overfitting, the extracted data set is divided into five, and each of them is used as teacher data, and preliminary machine learning is performed five times to create five preliminary models. The preliminary machine learning is not limited to 5 times, and may be appropriately changed depending on the number of data used for learning. Overfitting refers to learning in which unknown piping sample data is not taken into consideration and excessively matches the possessed piping sample data.
Then, the pipe deterioration model creation unit 40 creates a learning model optimized by the five preliminary models created by random extraction of the five patterns.
The optimization process can be performed by various methods such as averaging of preliminary models, parameter tuning of machine learning algorithms, and extraction of input parameter combinations having the highest correct answer rate.
Further, the sample data used for learning may be preprocessed to be converted into a number from 0 to 1 by standardization or normalization.

By creating a learning model by machine learning of such a method, even if there is a bias due to an accidental pattern in the piping sample that has been extubated in the past, it is possible to create a learning model that suppresses the influence of that bias. can.

The application of the present invention is not limited to the above-described embodiment, and can be appropriately changed without departing from the spirit of the present invention, and evaluation indexes and parameters may be added or changed.

10 Deterioration diagnosis judgment information storage unit 20 Piping parameter storage unit 30 Information input unit 40 Piping deterioration model creation unit 50 Piping deterioration estimation unit 60 Display unit 100 Piping deterioration diagnosis device

Claims (4)

A piping deterioration model creation unit that creates a learning model of the relationship between the deterioration diagnosis judgment of a piping sample that has been extubated in the past and various parameters related to piping of the piping sample using a machine learning method.
A pipe deterioration estimation unit that estimates the deterioration state of the pipe subject to deterioration diagnosis based on the learning model and various parameters related to the pipe subject to deterioration diagnosis.
A piping deterioration diagnostic device characterized by having. The various parameters related to the piping include at least one of a parameter relating to the use of the piping, a parameter relating to the joint core of the piping, a parameter relating to the water source of the piping, and a parameter relating to the aging of the piping. The piping deterioration diagnosis device according to claim 1. The pipe deterioration model creation unit relates the deterioration diagnosis judgment of a predetermined number of pipe samples randomly extracted from all the pipe samples that have been extubated in the past and various parameters related to the pipes of the extracted pipe samples. The pipe deterioration diagnosis according to claim 1 or 2, wherein a plurality of preliminary models are created by performing machine learning of sex a plurality of times, and the plurality of preliminary models are optimized to create the learning model. Device. An information storage means for storing information on deterioration diagnosis judgment of piping samples that have been extubated and inspected in the past,
Parameter storage means for storing data of various parameters related to piping of piping samples that have been extubated and inspected in the past,
A piping deterioration model creating means for creating a learning model of the relationship between the deterioration diagnosis determination information and data of various parameters related to the piping by using a machine learning method.
Based on the learning model and various parameters related to the pipe subject to deterioration diagnosis, a pipe deterioration estimation means for estimating the deterioration state of the pipe subject to deterioration diagnosis, and a pipe deterioration estimation means.
Piping deterioration diagnosis system characterized by having.

Images