Method for detecting cavitation in cryogenic pumps

Abstract

The invention relates to a method for detecting cavitation in a cryogenic pump drawing cryogenic liquid from a reservoir, by acquiring operating data including at least temperature and pressure measurements to define a thermodynamic state of the liquid in the reservoir, calculating a predicted thermodynamic state of the liquid at the pump inlet using thermodynamic equations and an artificial intelligence model trained on the operating data and physical parameters of the piping, pump, and reservoir, comparing the predicted thermodynamic state to a liquid saturation curve, generating an alert signal when the predicted thermodynamic state exceeds a predefined threshold value, and generating a message using a communication device. Abstract figure: Fig. 1

Description

Title of the invention: Method for detecting cavitation in cryogenic pumps

[0001] The present invention relates to the field of industrial equipment monitoring, and more particularly to a method for detecting cavitation in cryogenic pumps, especially those used in high-pressure liquid oxygen conditioning centers for medical or industrial applications. The invention also relates to the cavitation detection system implementing the method of the invention.

[0002] Cryogenic pumps, used for example to pressurize liquid oxygen, are subject to cavitation. Cavitation corresponds to the vaporization, often partial, of the cryogenic liquid, typically at the inlet or in the pressurization chamber of the pump, due to a drop in pressure and / or an increase in temperature that brings the fluid close to its saturation conditions. When saturation conditions are reached, vapor bubbles form. The cavitation phenomenon reduces the volumetric efficiency of the pump and can cause significant damage due to pressure shocks induced by the bursting of vapor bubbles.

[0003] Prior art solutions for identifying cavitation are known. These solutions focus primarily on detecting cavitation once it occurs, for example, by measuring the temperature upstream of the pump, sometimes using thermal imaging cameras. One such known solution is disclosed in US patent 8292501B1.

[0004] Prior art solutions do not allow for anticipating this phenomenon. Furthermore, detection generally triggers a pump shutdown to prevent damage, which interrupts the industrial process. This interruption leads to operational delays and necessitates time-consuming cooling and restart procedures.

[0005] The present invention aims to remedy these drawbacks by proposing a method which makes it possible to prevent the pumps from stopping and a monitoring system implementing this method.

[0006] The invention relates to a method for detecting cavitation in a cryogenic pump that draws a cryogenic liquid from a reservoir at a given flow rate and is connected to said reservoir by means of a pipeline, said pump and said reservoir each being equipped with sensors configured to perform at least one temperature measurement, one pressure measurement and one suction flow rate measurement, the method being carried out by means of a processing unit and comprising the steps:

[0007] a. acquisition of operating data including at least the measurement of temperature and pressure in the tank allowing the definition of a thermodynamic state of the liquid in the tank;

[0008] b. calculation of a predicted thermodynamic state of the liquid at the pump inlet by means of thermodynamic equations modeling the variation of temperature and pressure through the pipeline as a function of the measurement of the suction flow rate and by means of an artificial intelligence model;

[0009] c. compare said predicted thermodynamic state to a stored saturation curve (Cs) of said cryogenic liquid;

[0010] d. generate an alert signal when said predicted thermodynamic state exceeds a predefined threshold value;

[0011] e. generate a message by means of a communication device.

[0012] The invention further relates to a monitoring system configured to implement the method of the invention and comprising sensors for measuring said operating data, and a processing unit configured to execute said hybrid predictive model and generate said alert signal.

[0013] The invention will now be better understood with reference to the following detailed description, given by way of illustration but not limitation, with reference to the attached figure where:

[0014] [Fig-1] is a schematic representation of a liquid saturation curve cryogenic depending on temperature and pressure.

[0015] The cavitation detection method of the invention relates to a cryogenic storage system comprising a cryogenic fluid storage tank, for example, for liquid oxygen. The system also includes a pump. This pump is connected to the cryogenic tank by means of a pipeline. For example, the pump is connected to the tank via its inlet through the pipeline. The pump draws the liquid from the tank at a given suction flow rate. A set of sensors in the system allows the acquisition of operating data. For example, sensors measure the temperature and pressure within the tank, and more particularly in the region at the bottom of the tank, as well as the temperature and pressure in the pipeline. Sensors also measure the liquid suction flow rate, as well as, for example, environmental conditions.Environmental conditions include, in particular, the temperature outside the installation and the amount of sunlight.

[0016] The process of the invention is carried out by means of a processing unit implementing a computer program (or programs), also called algorithms, which executes the steps described below.

[0017] Initially, an acquisition of operating data is carried out, including at least the measurement of temperature and pressure in the tank. These measurements allow the definition of a thermodynamic state of the liquid in the tank, for example, at the flow rate of the aspirated liquid. In one embodiment, the thermodynamic state of the liquid in the tank is determined from temperature and pressure measurements at the bottom of a storage column, or, for example, from measurements of the pressure of the gas head and the liquid height in the tank.

[0018] Then we proceed to acquire environmental conditions, for example outside temperature and sunshine.

[0019] These environmental conditions data and the defined thermodynamic state are provided as input to a hybrid predictive model which will calculate a so-called predicted thermodynamic state of the liquid at the pump inlet.

[0020] The hybrid predictive model is stored within the processing unit.

[0021] This hybrid predictive model includes, in particular, thermodynamic equations and an artificial intelligence model. The thermodynamic equations model the variation of temperature and pressure across the pipeline as a function of the measured suction flow rate and environmental conditions.

[0022] The hybrid predictive model also includes an artificial intelligence model trained to refine the thermodynamic state of the fluid at the pump inlet based on operating data and the physical parameters of the pump and the reservoir. For example, the artificial intelligence model takes into account the cross-section and length of the pipeline, but also, preferably, the storage time and the size of the reservoir.

[0023] Typically, the hybrid predictive model makes it possible to model at least a temperature variation and a pressure loss of the liquid through the pipeline under measured environmental conditions.

[0024] The user thus obtains a predicted thermodynamic state of the liquid at the pump inlet. The artificial intelligence model is, for example, a Physics-Informed Neural Network (PINN). This model uses both real-time sensor data, thermodynamic equations, and physical equations to predict the state of a pipe, tank, or pump. This AI model allows for the calibration of unknown or difficult-to-measure parameters specific to each installation: pipe geometry, external insulation, etc. Thus, the definition of the predicted thermodynamic state is more precise.

[0025] Depending on the case, the method of the invention calculates continuously or intermittently the thermodynamic state at the inlet of the pump.

[0026] Next, the predicted thermodynamic state will be compared to a saturation curve Cs of the cryogenic liquid. This saturation curve Cs is stored in the processing unit, for example by means of flash memory or similar. Such The curve is represented in [Fig. 1]. [Fig. 1] is therefore a Pressure-Temperature diagram illustrating the saturation curve Cs of a liquid. The method allows the determination of the thermodynamic state of the liquid in the reservoir and the predicted thermodynamic state of the liquid at the pump inlet. With reference to [Fig. 1], each thermodynamic state can, for example, be in phase I or II. For example, phase I corresponds to a state in which the liquid is in a subcooled liquid state. Phase II corresponds to a state in which the liquid has transformed into superheated vapor.

[0027] Then, an alert signal is triggered when the predicted thermodynamic state exceeds a predefined threshold value. This predefined threshold value is, for example, represented by a defined area near the saturation curve Cs. This defined area is shown hatched in [Fig. 1]. This hatched area can represent a safety margin before cavitation occurs. Thus, if the predicted thermodynamic state is within this hatched area, an alert is generated. In other words, when the safety margin is exceeded, the alert is generated. This alert can be an audible and / or visual signal and / or a vibration. Thanks to this hybrid approach combining physical modeling and artificial intelligence to predict the thermodynamic state of the fluid at the pump inlet, the risk of cavitation is anticipated.

[0028] In one embodiment, the processing unit can then generate a message using a communication device. The communication device is, for example, a computer screen or a mobile phone communicating with the computer program. This message might, for example, prompt the user to reduce the pump's suction flow rate, increase the pump inlet pressure, or decrease the pump inlet temperature. Thus, by generating alerts and / or messages, the method of the invention reduces the risk of damage to the system, and in particular to the pump. The user can maintain a sufficient safety margin to prevent any damage. The availability and efficiency of the system are improved.

[0029] In one embodiment, the process of the invention is applied to a cryogenic pump for the pressurization of liquid oxygen in a high-pressure oxygen conditioning center.

[0030] The invention also relates to a monitoring system configured to implement the method of the invention. This monitoring system includes sensors for measuring operating data, and a processing unit configured to execute the hybrid predictive model and generate the alert signal. The processing unit includes a (micro)processor configured to execute the computer program(s).

Claims

1.

2. Demands A method for detecting cavitation in a cryogenic pump that draws a cryogenic liquid from a reservoir at a given flow rate and is connected to said reservoir by means of a pipeline, said pump and said reservoir each being equipped with sensors configured to perform at least one temperature measurement, one pressure measurement, and one suction flow rate measurement, the method being carried out by means of a processing unit and comprising the following steps: - a. acquisition of operating data including at least the measurement of temperature and pressure in the tank allowing the definition of a thermodynamic state of the liquid in the tank; - b. calculation of a predicted thermodynamic state of the liquid at the pump inlet by means of thermodynamic equations modeling the variation of temperature and pressure through the pipeline as a function of the measurement of the suction flow rate and by means of an artificial intelligence model; - c. compare said predicted thermodynamic state to a stored saturation curve (Cs) of said cryogenic liquid; - d. generate an alert signal when said predicted thermodynamic state exceeds a predefined threshold value in response to the comparison when the comparison shows a discrepancy. A monitoring system configured to implement the method according to claim 1 and comprising sensors for measuring said operating data, and a processing unit configured to execute said hybrid predictive model and generate said alert signal.

Images