Service life prediction method and device for electrolytic oxygen production equipment based on deep learning

Abstract

The present invention discloses a method and device for predicting the service life of electrolytic oxygen production equipment based on deep learning. The method includes: S1, setting the detection object of each device in the electrolytic oxygen production equipment to the oxygen concentration in the oxygen production channel; S2, learning and training the oxygen concentration in the oxygen production channels of all equipment, recording the collected oxygen concentration data of all equipment in normal and abnormal conditions, using MLP neural network to perform deep learning and training on the oxygen concentration data, and constructing a model for predicting test data; S3, using the model to monitor the oxygen concentration in the oxygen production channels of all equipment in real time, and obtain the monitoring results; S4, predicting the service life of the electrolytic oxygen production equipment based on the remaining life prediction deep learning model based on the monitoring results. The present invention can effectively comprehensively consider process optimization, maintainability, operability and safety, avoid supply interruption caused by abnormality of a unit in the oxygen production process, and better predict the service life of the electrolytic oxygen production equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of electrolytic oxygen production, and more specifically, to a method and device for predicting the service life of electrolytic oxygen production equipment based on deep learning. Background Art

[0002] Electrolytic oxygen production technology is currently recognized by the world as the most suitable oxygen supply technology for medium- and long-term manned space missions. In view of the microgravity and variable pressure working environment in the spacecraft, the electrolytic oxygen production device is affected by microgravity and other conditions, and has repeatedly experienced two-phase flow control abnormalities and failed to start normally. If any unit in the oxygen production process fails, it may cause the entire oxygen production system to fail, and the quality of oxygen in space will decrease, which will greatly threaten the life support of astronauts.

[0003] The electrolytic oxygen production system adopts the basic scheme of solid polymer electrolyte (SPE) electrolysis technology and static water-gas separation technology. The working process of the electrolytic oxygen production system involves water, gas and water-gas mixture, and it is a complex system. The whole process mainly consists of five parts: oxygen supply branch, water circulation loop, hydrogen emission branch, pressure control branch, and water supply branch. At present, it is impossible to monitor the operating status of these equipment in real time, so it is impossible to predict the service life of electrolytic oxygen production equipment. For this reason, it is necessary to develop a service life prediction method and device for electrolytic oxygen production equipment based on deep learning. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for predicting the service life of electrolytic oxygen production equipment based on deep learning, so as to overcome the defects existing in the prior art.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for predicting the service life of an electrolytic oxygen production device based on deep learning comprises the following steps:

[0007] S1. Setting the detection object of each device in the electrolytic oxygen production equipment to the oxygen concentration in the oxygen production channel;

[0008] S2. Learn and train the oxygen concentration in the oxygen production channels of all devices, record the collected oxygen concentration data of all devices in normal and abnormal conditions, use MLP neural network to conduct deep learning and training on the oxygen concentration data, and construct a model for predicting test data;

[0009] S3. Use the model to monitor the oxygen concentration in the oxygen production channels of all equipment in real time and obtain the monitoring results;

[0010] S4. Predicting the service life of the electrolytic oxygen production equipment based on the remaining life prediction deep learning model according to the monitoring results.

[0011] Furthermore, the remaining life prediction deep learning model adopts a long short-term memory network LSTM model and is trained based on an MLP framework. The input of the remaining life prediction deep learning model is the remaining life of all devices, and the remaining life distribution of the devices is:

[0012] F L (w) = p{T≤1+t n |T>t n )=[F(w+t n )-F(t n )] / [1-F(t n )];

[0013] In the formula, t is time, w is the failure threshold, T represents the total use time of the equipment, t n Represents the rated usage time of the equipment.

[0014] Furthermore, the step S2 also includes: adding the model into the oxygen concentration detector, installing the oxygen concentration detector in the oxygen production channels of all the equipment, and connecting all the oxygen concentration detectors to the monitoring center through a network.

[0015] Furthermore, the monitoring results in step S3 include equipment abnormality and oxygen-generating gas leakage. When the monitoring result is equipment abnormality, the monitoring center triggers an abnormality notification to prompt the equipment abnormality. When the monitoring result is oxygen-generating gas leakage, the monitoring center notifies in real time and triggers the maintenance mode.

[0016] The present invention also provides a system according to the above-mentioned deep learning-based service life prediction method of electrolytic oxygen production equipment, comprising:

[0017] A setting module, used to set the detection object of each device in the electrolytic oxygen production equipment to the oxygen concentration in the oxygen production channel;

[0018] The model training module is used to learn and train the oxygen concentration in the oxygen production channels of all devices, record the collected oxygen concentration data of all devices in normal and abnormal conditions, use the MLP neural network to conduct deep learning and training on the oxygen concentration data, and construct a model for predicting test data;

[0019] The monitoring module is used to use the model to monitor the oxygen concentration in the oxygen production channels of all equipment in real time and obtain the monitoring results;

[0020] A life prediction module, used to predict the service life of the electrolytic oxygen production equipment according to the monitoring results;

[0021] The setting module, model training module, monitoring module and life prediction module are connected in sequence.

[0022] Compared with the prior art, the advantages of the present invention are: the present invention provides a method and device for predicting the service life of electrolytic oxygen production equipment based on deep learning, which realizes full monitoring of the oxygen production and supply processes through networking technology. The present invention can effectively comprehensively consider process optimization, maintainability, operability and safety, avoid supply interruption caused by abnormality of a certain unit in the oxygen production process, and better predict the service life of electrolytic oxygen production equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 This is a flow chart of the service life prediction method of electrolytic oxygen production equipment based on deep learning in the present invention.

[0025] Figure 2 It is a flow chart of the electrolytic oxygen production system of the present invention.

[0026] Figure 3 It is the life monitoring flow chart of the present invention.

[0027] Figure 4 This is a framework diagram of a service life prediction device for electrolytic oxygen production equipment based on deep learning in the present invention. DETAILED DESCRIPTION

[0028] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0029] See also Figure 1 As shown, this embodiment discloses a method for predicting the service life of an electrolytic oxygen production device based on deep learning, comprising the following steps:

[0030] Step S1, setting the detection object of each device in the electrolytic oxygen production equipment to the oxygen concentration in the oxygen production channel.

[0031] Specifically, the present invention sets the detection object as the oxygen concentration in the oxygen production channel for the 17 devices in the equipment with different original service life designs (2 years to 10 years). As shown in Table 1, it is a table of equipment products and service life of the electrolytic oxygen production system of the present invention.

[0032] Table 1

[0033]

[0034]

[0035] Step S2: learn and train the oxygen concentration in the oxygen production channels of all devices, record the collected oxygen concentration data of all devices in normal and abnormal conditions, use the MLP neural network to conduct deep learning and training on the oxygen concentration data, and construct a model for predicting test data.

[0036] Specifically, the present invention conducts learning and training on the oxygen concentration of 17 channels, records all the collected data on the oxygen concentration when the equipment is normal and abnormal, and uses an MLP neural network to conduct deep learning and training on it.

[0037] The step S2 also includes: adding the model into the oxygen concentration detector, installing the oxygen concentration detector in the oxygen production channels of all the equipment, and connecting all the oxygen concentration detectors to the monitoring center through a network.

[0038] like Figure 2 As shown in the figure, the pipeline layout is based on the system flow, and the pipeline direction is simplified overall to avoid pipelines being too long and intertwined. Oxygen concentration detectors after MLP deep learning are added to 17 channels, and Figure 3 AI networking is carried out in the form of.

[0039] Step S3: Use the model to monitor the oxygen concentration in the oxygen production channels of all devices in real time and obtain the monitoring results.

[0040] Step S4: predicting the service life of the electrolytic oxygen production equipment based on the remaining life prediction deep learning model according to the monitoring results.

[0041] Specifically, the monitoring results in step S3 include equipment abnormality and oxygen-generating gas leakage. When the monitoring result is equipment abnormality, the monitoring center triggers an abnormality notification to prompt the equipment abnormality. When the monitoring result is oxygen-generating gas leakage, the monitoring center notifies in real time and triggers the maintenance mode.

[0042] In this embodiment, Z(t) is the service life degradation of the performance index of the equipment at time t. It can be seen that Z(t) is a function of time degradation, and since the specific changes are unknown, it cannot be directly determined whether it is a linear function. It may be a linear or nonlinear function. According to the concept of service life degradation failure, when the service life degradation amount Z(t) accumulates to a certain degree W, that is, reaches the failure threshold, it can be considered that the performance of the equipment has failed, and the corresponding time at this time is the life time of the equipment.

[0043] When the oxygen concentration in the equipment shows a monotonic trend of degradation, the equipment performance failure time can be defined as:

[0044] T=int{t:Z(t)≥w}

[0045] The above formula is the service life degradation failure model of electrolytic oxygen production equipment, from which the failure law distribution can be obtained:

[0046] F(t;w)=p{Z(t)≥w}

[0047] Therefore, in general, the remaining service life of electrolytic oxygen production equipment is based on the distribution of equipment degradation failure. If the distribution of the equipment is F(t), the remaining service life distribution of the equipment at the moment is:

[0048] F L (w) = p{T≤1+t n |T>t n )=[F(w+t n )-F(t n )] / [1-F(t n )]

[0049] In the formula, t is time, w is the failure threshold, T represents the total use time of the equipment, t n Represents the rated usage time of the equipment. p is similar to f(x) and is a symbol.

[0050] It can be seen from the above formula that the formula only reflects the remaining life of the electrolytic oxygen production equipment under the condition of a single operating time, and characterizes that the failure distribution is mainly related to the current operating time of the equipment, but does not take other issues into consideration, such as the equipment is not completely the same, the equipment life cycle, the equipment maintenance, etc. Therefore, based on this principle, the present invention adopts a deep learning method, selects a long short-term memory network LSTM model, and completes the construction of a remaining life prediction deep learning model and the model training process based on the MLP framework. The effective training of the service life prediction model of the electrolytic oxygen production equipment is achieved by repeatedly manufacturing the equipment use environment under different oxygen concentrations and recording them in real time in the early stage.

[0051] See also Figure 4As shown, the present invention also provides a system for predicting the service life of electrolytic oxygen production equipment based on the above-mentioned deep learning-based method, comprising: a setting module 1, used to set the detection object of each device in the electrolytic oxygen production equipment to the oxygen concentration in the oxygen production channel; a model training module 2, used to learn and train the oxygen concentration in the oxygen production channels of all devices, record the collected oxygen concentration data of all devices in normal and abnormal conditions, use an MLP neural network to perform deep learning and training on the oxygen concentration data, and construct a model for predicting test data; a monitoring module 3, used to use a model to monitor the oxygen concentration in the oxygen production channels of all devices in real time and obtain monitoring results; a life prediction module 4, used to predict the service life of the electrolytic oxygen production equipment according to the monitoring results; wherein the setting module 1, the model training module 2, the monitoring module 3 and the life prediction module 4 are connected in sequence.

[0052] Although the embodiments of the present invention are described in conjunction with the accompanying drawings, the patent owner may make various variations or modifications within the scope of the appended claims. As long as they do not exceed the protection scope described in the claims of the present invention, they should be within the protection scope of the present invention.

Claims

1. A method for predicting the service life of electrolytic oxygen production equipment based on deep learning, characterized in that: The following steps are involved: S1. Setting the detection object of each device in the electrolytic oxygen production equipment to the oxygen concentration in the oxygen production channel; S2. Learn and train the oxygen concentration in the oxygen production channels of all devices, record the collected oxygen concentration data of all devices in normal and abnormal conditions, use MLP neural network to conduct deep learning and training on the oxygen concentration data, and construct a model for predicting test data; S3. Use the model to monitor the oxygen concentration in the oxygen production channels of all equipment in real time and obtain the monitoring results; S4. Predicting the service life of the electrolytic oxygen production equipment based on the remaining life prediction deep learning model according to the monitoring results; The remaining life prediction deep learning model adopts a long short-term memory network LSTM model and is trained based on an MLP framework. The input of the remaining life prediction deep learning model is the remaining life of all devices. The remaining life distribution of the devices is: F L (w)=p{T≤1+t n |T>t n )=[F(w+t n )-F(t n )] / [1-F(t n )]; In the formula, t is time, w is the failure threshold, T represents the total use time of the equipment, t n Represents the rated usage time of the equipment.

2. The service life prediction method of electrolytic oxygen production equipment based on deep learning according to claim 1 is characterized in that: The step S2 also includes: adding the model into the oxygen concentration detector, installing the oxygen concentration detector in the oxygen production channels of all the equipment, and connecting all the oxygen concentration detectors to the monitoring center through a network.

3. The service life prediction method of electrolytic oxygen production equipment based on deep learning according to claim 2 is characterized in that: The monitoring results in step S3 include equipment abnormality and oxygen-generating gas leakage. When the monitoring result is equipment abnormality, the monitoring center triggers an abnormality notification to prompt the equipment abnormality. When the monitoring result is oxygen-generating gas leakage, the monitoring center notifies in real time and triggers the maintenance mode.

4. A system for predicting the service life of electrolytic oxygen production equipment based on deep learning according to any one of claims 1 to 3, characterized in that: include: A setting module, used to set the detection object of each device in the electrolytic oxygen production equipment to the oxygen concentration in the oxygen production channel; The model training module is used to learn and train the oxygen concentration in the oxygen production channels of all devices, record the collected oxygen concentration data of all devices in normal and abnormal conditions, use the MLP neural network to conduct deep learning and training on the oxygen concentration data, and construct a model for predicting test data; The monitoring module is used to use the model to monitor the oxygen concentration in the oxygen production channels of all equipment in real time and obtain the monitoring results; A life prediction module, used to predict the service life of the electrolytic oxygen production equipment according to the monitoring results; The setting module, model training module, monitoring module and life prediction module are connected in sequence.

Images