Anomaly detection device for rotating machinery

Abstract

To provide a rotating machinery anomaly diagnostic device with excellent accuracy in diagnosing abnormalities in rotating machinery. [Solution] A vibration sensor for measuring the vibration of rotating machinery; an operation database for collecting and storing measurement data from the vibration sensor; a rotating machinery information database for recording at least the model, specifications, and performance information of the rotating machinery; a maintenance work history database for recording the maintenance work history of the rotating machinery; and a diagnostic processing unit for diagnosing abnormalities in the rotating machinery using the data from each of the above databases. A rotating equipment abnormality diagnosis device comprising the diagnostic processing unit which maintains a diagnostic model for each type of rotating equipment and performs the diagnosis.

Description

Technical Field

[0001] The present invention relates to an abnormal diagnosis device for rotating equipment that diagnoses abnormalities in rotating equipment such as pumps, blowers, and agitators.

Background Art

[0002] As a method for diagnosing failures of multiple rotating devices such as fans and pumps, Patent Document 1 describes a method in which, in advance, one or both of the synthetic noise and synthetic vibration during normal operation of multiple devices are measured and stored, and the stored normal values are compared with the measured values during normal operation of the multiple devices, and the abnormal state of the devices is detected from the comparison results.

[0003] Also, in the column of the prior art of Patent Document 1, when diagnosing failures of multiple and single or multiple types of devices, it is described that abnormalities of the devices are detected using the noise values or vibration values measured for each device.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In Patent Document 1, abnormalities are diagnosed by comparing the noise and vibration of rotating equipment with normal values, resulting in inferior diagnostic accuracy.

[0006] One aspect of the present invention is to provide an abnormal diagnosis device for rotating equipment with excellent diagnostic accuracy for abnormalities in rotating equipment.

Means for Solving the Problems

[0007] The present invention has the following gist.

[0008] [1] A vibration sensor for measuring the vibration of rotating machinery; an operating database for collecting and storing measurement data from the vibration sensor; a rotating machinery information database for recording at least the model, specifications, and performance information of the rotating machinery; a maintenance work history database for recording the maintenance work history of the rotating machinery; and a diagnostic processing unit for diagnosing abnormalities in the rotating machinery using the data from each of the above databases. A rotating equipment abnormality diagnosis device comprising the diagnostic processing unit which maintains a diagnostic model for each type of rotating equipment and performs the diagnosis.

[0009] [2] The rotating equipment anomaly diagnostic device [1], in which each of the aforementioned databases can be identified by a common identification code.

[0010] [3] The diagnostic processing unit has the function of extracting fault response maintenance work for rotating machinery from the maintenance work history database, identifying the time and type of fault, and training the diagnostic model by machine learning the change patterns of diagnostic data before the fault occurred. [1] or [2] An abnormality diagnostic device for rotating machinery. [Effects of the Invention]

[0011] According to one aspect of the present invention, abnormalities can be detected with high accuracy by applying a diagnostic model that is appropriate to the type and specifications of the rotating equipment.

[0012] In one aspect of the present invention, by using machine learning to train a diagnostic model with vibration data based on past failure cases, problems can be detected more accurately and earlier than with threshold judgment.

[0013] In one aspect of the present invention, it becomes possible to present past cases with similar patterns, enabling more accurate fault response and streamlining recovery operations. [Brief explanation of the drawing]

[0014] [Figure 1] This is a schematic diagram of a diagnostic device for abnormalities in rotating machinery according to an embodiment of the present invention. [Figure 2] This is a flowchart explaining the method for diagnosing abnormalities. [Modes for carrying out the invention]

[0015] Embodiments of the present invention will now be described in detail with reference to the drawings. The abnormality diagnosis device for rotating machinery according to this embodiment is for diagnosing abnormalities based on vibrations of rotating machinery.

[0016] As shown in Figure 1, in a water treatment system equipped with multiple rotating devices 1, vibration sensors 2 are installed on each rotating device 1. If the rotating device is a pump, depending on the pump model, it may be better to measure multiple locations such as the motor and pump sections, so one or more vibration sensors 2 are installed on each rotating device 1. (Pumps that do not cause harm to water treatment even if they fail and stop, such as pumps with backup units and automatic switching functions in case of failure, may be excluded.) The vibration sensor 2 may be equipped with an acceleration sensor and output acceleration data directly, or it may process acceleration data over a certain period of time (0.1 to 5 seconds) and output values ​​such as RMS value, peak acceleration, and peak frequency value obtained by frequency analysis.

[0017] Since data that has undergone statistical and analytical processing is easier to extract features related to failures from, sensors with data processing capabilities are preferable because they reduce the load on the diagnostic device 6 described later.

[0018] The data from the vibration sensor 2 is collected by the data acquisition device 3 and stored as time-series data in the operation database 4. To identify which rotating equipment 1 the vibration data belongs to, each rotating equipment 1 is assigned an identification code, and the data for each rotating equipment 1 can be searched, viewed, and retrieved using the identification code.

[0019] The operation database 4 can be either on-site or cloud-based. A cloud-based system is preferable because it allows for centralized management of multiple locations. Specifically, S.sensing (registered trademark) and Kurita-IT (registered trademark) can be applied.

[0020] The rotating equipment information database 5 stores the types, specifications, and performance information of rotating equipment such as pumps and blowers to be managed in a state related to an identification symbol for identifying the rotating equipment.

[0021] The diagnostic device 6 is composed of, for example, a computer equipped with a CPU, a communication unit, a storage unit, etc. The storage unit has, for example, a RAM, a ROM, a hard disk, etc. The function of the diagnostic processing unit 7 is realized by executing a diagnostic program stored in the storage unit.

[0022] The diagnostic device 6 extracts the data necessary for diagnosis from the rotating equipment information database 5 and the operation database 4, and diagnoses them in the diagnostic processing unit 7.

[0023] The diagnostic processing unit 7 reads out and holds a diagnostic model for each type from the diagnostic model group database 10, and performs diagnosis.

[0024] The diagnostic processing unit 7 extracts the fault response maintenance work of the rotating equipment from the maintenance work history database 8 described later, specifies the time point and type of the fault occurrence, and learns the change pattern of the diagnostic data before the fault occurrence by machine learning, and has a function of training the diagnostic model.

[0025] When an abnormality is detected as a diagnostic result, although not shown in the figure, it can be notified to the person in charge by email or output to a screen (not shown). Instead of simply classifying normal / abnormal (binary), the diagnostic result may be accumulated in the operation database 4 as numerical information between 0 (normal) and 1 (abnormal) such as the degree of abnormality.

[0026] When maintenance work is performed based on an abnormal diagnostic result, the details of the maintenance work are stored in the maintenance work history database 8.

[0027] The maintenance work history database 8 records the on-site maintenance inspection records and matters of concern such as the discovery of abnormalities together with the respective implementation / discovery dates and times. In particular, regarding the information on equipment failure discovery and equipment maintenance work, the history information is accumulated so that it can be searched by the identification symbol of the corresponding equipment.

[0028] The detailed procedure for anomaly diagnosis processing is explained based on Figure 2.

[0029] In this abnormality diagnosis process, the identification code of one rotating piece of equipment is obtained from the list of equipment to be diagnosed (which may also be a list of identification codes of equipment to be diagnosed) 9 in the rotating equipment information database 5, and a diagnosis is performed (steps (1) to (7)). Once the diagnosis of all equipment included in list 9 has been completed, the entire diagnosis process is terminated (step (8)).

[0030] To perform the diagnosis, the type information for the rotating equipment with the corresponding identification number is obtained from the rotating equipment information database 5 using the rotating equipment identification code (step (3)). Performance information and specification information may also be obtained at this time.

[0031] Using the acquired model information, a diagnostic model is retrieved from the diagnostic model database 10, which stores diagnostic models for each model (step (4)). A diagnostic model includes data items (such as acceleration RMS and frequency), the time series data period, and a parameter array obtained when machine learning is performed, for the data used in the diagnosis.

[0032] The parameter array varies depending on the machine learning method used, but for example, when diagnosing using a regression method that uses multiple vibration data, Anomaly Score = α × Acceleration RMS + β × Peak Frequency If the diagnosis is made using a model formula like this, then α and β will be retained as sequences.

[0033] In another example, when using a method that inputs acceleration RMS and peak frequency data for the past 10 days into a neural network and outputs a confidence score for abnormality / normality, the parameter array corresponds to an array of weight data within the neural network.

[0034] Furthermore, since there are various machine learning methods, it is desirable to apply the method that has learned most accurately. Therefore, the data items and period input for each method should be variable to enable diagnosis.

[0035] Based on the acquired diagnostic model and the identification code of the rotating equipment, vibration data is extracted and acquired from the operation database (step (5)).

[0036] The acquired vibration data is input into the diagnostic model to obtain the diagnostic results (step (6)).

[0037] Perform maintenance work according to the diagnostic results (step (7)).

[0038] Once steps (1) through (7) above have been performed for all rotating equipment, the diagnosis for all rotating equipment is complete (step (8)).

[0039] To enable more effective use of the present invention, maintenance work information is stored in the maintenance work history information database 8.

[0040] The above embodiments are examples of the present invention, and the present invention is not limited thereto. [Explanation of Symbols]

[0041] 1. Rotating machinery 2. Vibration Sensor 3. Data Acquisition Device 4. Driving Database 5. Rotating Machinery Information Database 6. Diagnostic equipment 7. Diagnostic Processing Unit 8. Maintenance Work History Database 9. List of items to be diagnosed 10 Diagnostic Model Database

Claims

1. A vibration sensor that measures vibrations in rotating machinery, An operating database that collects and stores measurement data from the vibration sensor, A rotary equipment information database that records at least the model, specifications, and performance information of rotating equipment, A maintenance work history database that records the maintenance work history of rotating machinery, A diagnostic processing unit that diagnoses abnormalities in rotating machinery using data from the above databases, A rotating equipment abnormality diagnosis device equipped with, The aforementioned diagnostic processing unit maintains a diagnostic model for each model and performs the diagnosis. A diagnostic device for detecting abnormalities in rotating machinery.

2. The rotating equipment abnormality diagnosis device according to claim 1, wherein each of the aforementioned databases is capable of identifying individual rotating equipment using a common identification code.

3. The abnormality diagnosis device for rotating machinery according to claim 1 or 2, wherein the diagnostic processing unit has the function of extracting fault response maintenance work for rotating machinery from the maintenance work history database, identifying the time and type of fault, and training the diagnostic model by machine learning the change patterns of diagnostic data before the fault occurred.

Images