DE102021210294A1 - Position-independent avoidance of cavitation on a propeller - Google Patents

Abstract

The present invention relates to a method for controlling a submarine to avoid cavitation, the submarine having a navigation device 10, a data processing device 20, a propulsion device 30 and at least one first hydrophone, with a primary control request for the propulsion device 30 being transmitted from the navigation device 10 to the data processing device 20 is sent, wherein the data processing device 20 performs an evaluation of the primary control request with regard to the effect on cavitation, wherein the data processing device 20 modifies the primary control request to secondary control requests if the data processing device 20 expects cavitation when using the primary control request when evaluating the primary control request, wherein the data processing device 20 sends the secondary control requests to the drive device 30, dadu rch characterized in that the data processing device 20 has a self-learning algorithm, the data processing device 20 being connected to the at least one first hydrophone, the self-learning algorithm being designed to evaluate the information obtained from the at least one first hydrophone for evaluation with regard to cavitation.

Description

The invention relates to a method for avoiding cavitation on a propeller when traveling at speed, not only when traveling straight ahead, in submarines not only at a constant depth.

The avoidance of cavitation with simultaneous maximum acceleration has hitherto been possible when driving at a constant depth and driving straight ahead. Acceleration is to be understood here in the physical sense and includes both positive and negative acceleration. Acceleration therefore also includes deceleration, which can be achieved in particular by turning the propeller backwards.

From the DE 10 2012 213 991 A1 a method for accelerating a propeller-driven watercraft with low noise emissions is known.

From the WO 2008/155 448 A1 a method for automatic cavitation avoidance through the use of predefined cavitation initiation curves within the engine control is known.

However, due to complex boundary conditions, it is not possible to use these methods for situations in which the submarine is not simply driving straight ahead. Even when cornering, also due to a non-neutral rudder angle, these methods are regularly no longer successful due to the different and also changing flow of the propeller. If the diving depth is changed at the same time, this applies all the more. For this reason, additional safeguards are now planned for such maneuvers in order to prevent cavitation. However, this greatly reduces the acceleration and therefore no optimal acceleration is achieved.

The object of the invention is to provide a method which enables the maximum acceleration that can be physically achieved while avoiding cavitation and thereby avoids the development of noise.

This object is achieved by a control method having the features specified in claim 1, a teaching method having the features specified in claim 9 and by a submarine having the features specified in claim 13. Advantageous developments result from the dependent claims, the following description and the drawings.

The control method according to the invention serves to control a submarine and thereby to avoid cavitation. The main issue here is cavitation, which occurs directly on the propeller. In some cases, the smallest gas bubbles are formed, which collapse again like an implosion and thereby generate noise. Thus, a vehicle that generates cavitation and that makes significant additional noise is easy to locate. Therefore, cavitation must be avoided during a corresponding operation for the safety of the submarine. The submarine has a navigation device, a data processing device, a propulsion device and at least one first hydrophone. A primary control request for the drive device is sent from the navigation device to the data processing device. This primary control request can be entered manually into the navigation device, for example. For example, the primary control requirement can be a new target speed. For example, the primary control requirement may be to come to a standstill as quickly as possible. The data processing device performs an evaluation of the primary control request in terms of the impact on cavitation. In the rapid stall example given, maximum reverse rotation of the propeller would be optimal to achieve the result, but would lead to cavitation. The data processing device modifies the primary control request to secondary control requests if the data processing device, in evaluating the primary control request, expects cavitation using the primary control request. In the example of rapid standstill mentioned, the declining acceleration of the propeller would thus have to be limited to the maximum at which cavitation can be reliably avoided. The data processing device sends the secondary control request to the driving device.

According to the invention, the data processing device has a self-learning algorithm. The data processing device is connected to a detection device for detecting cavitation. According to the invention, a detection device is any device which is suitable for the data processing device to receive a signal, as a result of which the data processing device recognizes that cavitation is occurring at the propeller. In a simple embodiment, the detection device is a data input via which a specialist tells the data processing device as soon as cavitation occurs. In a further embodiment, the detection device can also comprise the sonar system of the submarine. However, it is also possible to use at least one other sensor that is suitable for detecting cavitation, in particular on the propeller. examples other sensors can be pressure sensors, optical sensors or piezo sensors. In this case, the piezo sensors are preferably integrated directly into the propeller.

A navigation device within the meaning of the invention is to be understood broadly and includes any form of control of the submarine. For example, it can be a corresponding console into which speed and rudder position can be entered directly. Target course, target speed and target depth can also be entered and the navigation device then, based on the current course, current speed and current depth, then the necessary changes gradually. It can also be a navigation data processing device, into which a destination is entered and which then determines the control commands, including course, depth and speed, and forwards them to the appropriate components of the submarine.

A data processing device according to the invention is any electronic device which is suitable for executing program instructions for carrying out the method.

In a further embodiment of the invention, the detection device comprises at least one first hydrophone. The self-learning algorithm is designed to evaluate the information received from the at least one first hydrophone for evaluating the cavitation.

Particularly preferably, the detection device comprises both a data input, via which a person skilled in the art tells the data processing device as soon as cavitation occurs, and a first hydrophone. As a result, a manual first learning phase can be carried out and the acoustic information can be accessed directly in later operation.

The at least one first hydrophone is preferably part of the sonar system of the submarine. The self-learning algorithm can preferably access components of the sonar system or the complete sonar system. As a result, the number of hydrophones can preferably be significantly higher.

The self-learning algorithm is to be understood in the sense of machine learning and enables the data processing device to learn automatically from recorded ship data or to recognize the conditions under which cavitation occurs and to predict this. This takes place without the self-learning algorithm having to be explicitly programmed for this. The self-learning algorithm can continuously change or improve over time and the amount of data collected. The self-learning algorithm is, for example, methods from the group of ensemble algorithms, the group of artificial neural networks, in particular deep learning algorithms, or methods from the group of multivariate cluster analyses.

For the group of ensemble algorithms, models are combined that consist of several weaker models. Examples of self-learning algorithms in this group are bootstrapped aggregation, gradient boosting machines, gradient boosted regression trees, random trees, random forest.

The group of artificial neural networks is based on the networking of artificial neurons. The structure of the artificial neural networks can be very different. Self-learning algorithms of this group are known under the names radial basis functional network, convolution-based neural network, multilayer perceptions, convolutional neural network, deep Boltzmann machine, deep belief network.

The group of multivariate cluster analyzes includes methods for discovering similarity structures in the data. Examples of corresponding self-learning algorithms are called k-means, k-medians, expectation maximization or hierarchical clustering.

In a further embodiment of the invention, the data processing device receives the current ship data or can access them, the current ship data being selected from the group consisting of boat speed, boat acceleration, speed of the shaft, propeller thrust, torque of the propeller, rudder position, trim position of the boat, direction of the boat, drift speed, engine power, moment of inertia in the drive train, losses in the drive train, boat position, pitch angle of the propeller blades, diving depth: In particular, the changes over time and the speeds of change in the aforementioned values are also recorded. The data processing device particularly preferably receives or can access the current ship data, the current ship data being boat speed, boat acceleration, speed of the shaft, rudder position, trim of the boat, direction of the boat and engine power. These are particularly preferred for evaluating the primary control requirement in terms of impact on cavitation in Depending on boat speed, boat acceleration, shaft RPM, rudder position, boat trim, boat heading and engine power.

In a further embodiment of the invention, the data processing device receives information about changes to the outer shell. Such changes are, for example, external storage containers attached before use. These influence the flow profile around the submarine and thus also have an impact on the cavitation behavior. Therefore, for example, when assembling an external storage container, this can be entered into the data processing device. In particular, a series of maneuvers can then be carried out at the beginning of such an operation so that the self-learning algorithm can learn the difference in behavior based on the behavior without an external storage container.

Changes to the outer shell can also be smaller and more short-term. For example, it can be an open lock, for example for divers, or an outer skin flap.

In a further embodiment of the invention, the data processing device receives the current environmental data or can access them, the current environmental data being selected from the group comprising density, temperature, pressure, viscosity, concentration of dissolved gases, salt content. These current environmental data can either be real measured values, in particular from corresponding sensors. Likewise, the current environmental data can also be theoretically estimated on the basis of historical data and thus taken from a database. Likewise, some current environmental data can be estimated from real measured values. Particularly preferably, the evaluation of the primary control requirement is carried out with regard to the effect on cavitation as a function of density, temperature, viscosity, concentration of dissolved gases and/or salt content.

In a further embodiment of the invention, the control method uses the dimensionless parameters of the cavitation number and the thrust coefficient.

Here are:

p

statischer Druck

pv

Dampfdruck

ρ

Dichte des Fluids

U

Strömungsgeschwindigkeit

Cavitation number σ: $$\sigma =\frac{p-p_V}{\frac{1}{2}\cdot \rho \cdot u^2}$$

p

static pressure

pv

vapor pressure

ρ

density of the fluid

u

flow rate

T

Propellerschub

ρ

Dichte des Fluids

n

Drehzahl

D

Propellerdurchmesser

Thrust coefficient K _T : $$K_T=\frac{T}{\rho \cdot n^2\cdot D^4}$$

T

propeller thrust

ρ

density of the fluid

n

number of revolutions

D

propeller diameter

Dimensionless key figures have the advantage that their definition already allows them to be adapted to different conditions. For example, the cavitation number allows adaptability to different depths, so that the question of whether a certain behavior leads to cavitation does not have to be determined individually for each diving depth.

In a further aspect, the invention relates to a learning method for the self-learning algorithm for the control method according to the invention. The learning method has at least two learning steps. In a first training step, various maneuvers are performed with the submarine, with a person manually notifying the data processing device when cavitation can be detected via the at least one first hydrophone. As a result, the self-learning algorithm can learn in a first step which noises are to be evaluated as cavitation. It is particularly preferable for these maneuvers to be carried out in the first training step in the classic manner at a constant depth and when driving straight ahead. In a second learning step, various maneuvers are carried out and the self-learning algorithm automatically determines, based on the knowledge gained in the first learning step, from the data of the at least one first hydrophone whether or not cavitation is present. As a result, the self-learning algorithm assigns cavitating properties or non-cavitating properties to the driving conditions. Since this then no longer requires manual training by an experienced person, in the second training step the maneuvers driven can easily be extended to include cornering or a simultaneous change in depth.

In a further embodiment of the invention, the self-learning algorithm proposes maneuvers that should be driven in the second learning step or in a later learning step. This allows the self-learning algorithm to achieve optimal forecasting accuracy.

In a further embodiment of the invention, the second training step includes running a predetermined group of maneuvers. By specifying maneuvers, the most likely occurrences can be specified in a targeted manner. These maneuvers may include, for example, maximum deceleration from high speed together with simultaneous depth change and/or maximum port or starboard steering to include, for example, possible evasive maneuvers. Other maneuvers can be, for example, those that could occur when divers are released and resumed.

In a further embodiment of the invention, the second learning step includes the traversing of a predetermined period of time. By specifying a period of time, a corresponding value does not have to be placed in advance on the selection of possible maneuvers, but a time, in particular during the testing of the submarine, is used to record as many and as varied a maneuver as possible and thus to make the most comprehensive possible prediction receive.

In another embodiment of the invention, the self-learning algorithm reports a confidence score based on the experience gained through the second learning step. As a result, without specifying a specific number of maneuvers and/or a specific minimum time, training can be carried out until the confidence value has reached a specified minimum value. The confidence value is in particular the probability with which the self-learning algorithm correctly predicts the absence of cavitation in a maneuver that has not yet been carried out. For example, the trust value can be determined from the reliability (correctness) of the last predictions made, since this allows conclusions to be drawn about the correctness of future decisions as well.

In a further embodiment of the invention, the training method has a third training step, the third training step being carried out during normal operation of the submarine. The system thus learns continuously. For this purpose, the self-learning algorithm can preferably be switched to two different operating modes. In a first operating mode, the self-learning algorithm reliably avoids cavitation. This first mode of operation is the standard mode. In a second operating mode, the self-learning algorithm should also go beyond the area that is definitely not cavitating in order to determine other areas that are definitely not cavitating. In this case, cavitation is definitely accepted during operation in the second operating mode. The second operating mode can therefore be used, for example and in particular, during a transfer trip or during targeted training.

In a further aspect, the invention relates to a control method according to the invention, the self-learning algorithm being trained using the teaching method according to the invention.

In a further aspect, the invention relates to a submarine with a data processing device for carrying out the control method according to the invention and for carrying out the teaching method according to the invention.

• 1 Kavitationszahl in Abhängigkeit vom Schubbeiwert
• 2 Schematisches Unterseeboot

The method according to the invention is explained in more detail below with reference to an exemplary embodiment illustrated in the drawings.

• 1 Cavitation number as a function of the shear coefficient
• 2 Schematic submarine

In 1 shows the relationship between the cavitation number σ and the shear coefficient K _T . In the case of straight cruising on an even keel, there is a limit indicated by the dashed line. Below the dashed line cavitation occurs, above no cavitation. By defining the cavitation number and the thrust coefficient, theoretically any operating points of the propeller can be set in the diagram. If you now add the other parameters such as cornering, trim of the submarine, non-neutral rudder position and the like, the relationship becomes extremely complex, since both the operating point and the cavitation deployment limit change. It is therefore comparatively complex when all parameters are taken into account. According to the invention, a self-learning algorithm is therefore used in order to find a structure in this data space and also to be able to make reliable predictions for operating states that have not yet been experienced. At any point in time, a submarine has an operating point in this diagram, which is defined by the coordinates of the cavitation number σ and the thrust coefficient K _T . The cavitation number σ depends on the flow velocity, for example. If this is changed at the location of the propeller, for example by changing the rudder position (with otherwise unchanged parameters, the operating point will shift within the diagram. At the same time, however, the position of the dashed line will also change that separates the range of non-cavitating operating conditions above the dashed line from the cavitating operating conditions below the dashed line. Especially when the current conditions change, this leads to very complex changes, until the submarine has subsequently reached a new stationary stable state.

2 shows a very schematic representation of a submarine. The submarine has a propulsion device 30 which is connected to a propeller 40 via a shaft. The control commands are input via a navigation device 10 . A data processing device 20 is arranged between the navigation device 10 and the drive device 30 . The data processing device 20 has the self-learning algorithm, which modifies the control request of the navigation device 10 if this would lead to cavitation. The control request is then sent from the data processing device 20 to the drive device and ultimately results in a changed movement of the propeller 40. So that the self-learning algorithm can also learn which operating states lead to cavitation, the data processing device is connected via the sonar control 60 with two sonars 50 shown here tied together.

Reference List

10

navigation device

20

data processing device

30

drive device

40

propeller

50

sonar

60

sonar control

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• DE 102012213991 A1 [0003]
• WO 2008155448 A1 [0004]

Claims (13)

Control method of a submarine to avoid cavitation, wherein the submarine has a navigation device (10), a data processing device (20), a propulsion device (30) and at least one first hydrophone, wherein a primary control request for the propulsion device (30 ) is sent to the data processing device (20), the data processing device (20) performing an evaluation of the primary control request with regard to the effect on cavitation, the data processing device (20) modifying the primary control request to secondary control requests when the data processing device (20) at the evaluation of the primary control request expects cavitation when using the primary control request, wherein the data processing device (20) sends the secondary control requests to the drive device (30), thereby characterized t that the data processing device (20) has a self-learning algorithm, wherein the data processing device (20) is connected to a detection device for detecting cavitation. control procedure claim 1 , characterized in that the detection device comprises at least one first hydrophone, the self-learning algorithm for evaluating the information obtained from the at least one first hydrophone being designed for evaluation with regard to cavitation. Control method according to one of the preceding claims, characterized in that the data processing device (20) receives the current ship data or can access them selected from the group consisting of boat speed, boat acceleration, speed of the shaft, propeller thrust, torque of the propeller, rudder position, trim position of the boat , direction of the boat, drift speed, engine power, moment of inertia in the drive train, losses in the drive train. control procedure claim 3 , characterized in that the data processing device (20) receives at least the current ship data or can access them from the group consisting of boat speed, boat acceleration, speed of the shaft, rudder position, trim position of the boat, direction of the boat, engine power. control procedure claim 4 , characterized in that the evaluation of the primary control request with regard to the effect on cavitation is carried out as a function of boat speed, boat acceleration, speed of the shaft, rudder position, trim position of the boat, direction of the boat and engine power. Control method according to one of the preceding claims, characterized in that the data processing device (20) receives the current environmental data or can access them selected from the group comprising density, temperature, viscosity, concentration of dissolved gases, salt content. control procedure claim 6 , characterized in that the assessment of the primary control requirement with regard to the effect on cavitation is carried out as a function of density, temperature, viscosity, concentration of dissolved gases and/or salinity. Control method according to one of the preceding claims, characterized in that the control method uses the dimensionless parameters of the cavitation number and the thrust coefficient. Learning method of the self-learning algorithm for the control method according to one of the preceding claims, wherein the learning method has at least two learning steps, wherein in a first learning step different maneuvers are performed with the submarine and a person manually notifies the data processing device (20) when about the at least one first hydrophone cavitation can be detected, with different maneuvers being driven in a second training step and the self-learning algorithm automatically determines cavitation from the data of the at least one first hydrophone on the basis of the knowledge gained in the first training step, and with the self-learning algorithm assigning cavitating properties or non-cavitating properties to the driving states assigns properties. learning procedure claim 9 , characterized in that the second learning step comprises driving through a predetermined group of maneuvers. learning procedure claim 9 , characterized in that the second learning step includes the driving down a predetermined period of time. Learning method according to one of the claims 9 until 11 , characterized in that the training method comprises a third training step, the third training step being carried out during normal operation of the submarine. Submarine with a data processing device (20) for carrying out the control method according to one of Claims 1 until 8th and to carry out the learning process according to one of the claims 9 until 12 .

Images