WO2020135312A1 - 基于人工神经网络的动力定位推力分配装置及方法 - Google Patents

基于人工神经网络的动力定位推力分配装置及方法 Download PDF

Info

Publication number
WO2020135312A1
WO2020135312A1 PCT/CN2019/127352 CN2019127352W WO2020135312A1 WO 2020135312 A1 WO2020135312 A1 WO 2020135312A1 CN 2019127352 W CN2019127352 W CN 2019127352W WO 2020135312 A1 WO2020135312 A1 WO 2020135312A1
Authority
WO
WIPO (PCT)
Prior art keywords
thrust
neural network
artificial neural
thruster
thrust distribution
Prior art date
Application number
PCT/CN2019/127352
Other languages
English (en)
French (fr)
Inventor
王磊
李博
衣凡
汪学锋
贺华成
秦立成
Original Assignee
上海交通大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海交通大学 filed Critical 上海交通大学
Priority to US16/764,665 priority Critical patent/US11500339B2/en
Publication of WO2020135312A1 publication Critical patent/WO2020135312A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
    • G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Definitions

  • the invention relates to a thrust distribution technology for a marine structure dynamic positioning system, and in particular to a marine structure dynamic positioning thrust distribution device and method for calculating thrust loss based on an artificial neural network.
  • the dynamic positioning system is a control system that uses the propulsion of the ship or the offshore platform to keep it in the required position on the sea level as much as possible.
  • the dynamic positioning system has the advantages of easy operation, and the positioning accuracy is not affected by the water depth. It has been widely used in deep water semi-submersible drilling platforms.
  • the dynamic positioning system is generally composed of three parts: a position measurement system, a control system, and a thrust system.
  • the measurement system obtains the real-time position information of the structure through the satellite or underwater acoustic array.
  • the control system calculates the total thrust and torque required to return the marine structure to the target position based on the deviation of the actual position of the marine structure from the target position.
  • the thrust system distributes the thrust of each thruster, so that each thruster can generate the total thrust and torque required for the positioning of the marine structure, and finally achieve the dynamic positioning of the marine structure by its own thruster.
  • the thrust distribution problem is essentially a non-linear optimization problem.
  • the thrust distribution method needs to minimize the total power of all thrusters while ensuring that the combined force of all thrusters is equal to the total thrust and torque required for the positioning of the marine structure.
  • the object of the present invention is to provide a dynamic positioning thrust distribution device and method based on an artificial neural network.
  • the device and method avoid the thrust loss caused by hydrodynamic interference in the thrust distribution problem; at the same time, the forbidden angle is eliminated, the problem of the propeller corner jamming is solved, the total power of the propeller is reduced, and energy is saved.
  • the present invention is achieved by the following technical solutions.
  • a dynamic positioning thrust distribution method based on an artificial neural network including the following steps:
  • Step S1 Establish and train an artificial neural network to fit the thrust coefficients in the real-time control computer.
  • the established artificial neural network includes input layer, hidden layer and output layer.
  • ⁇ is the angle of the front thruster
  • ⁇ is the thrust coefficient of the rear thruster.
  • the method of training the artificial neural network is to use the experimental data obtained by the artificial neural network model test as the training sample ⁇ (x(1), y(1)), (x(2), y(2)),..., (x(m), y(m)) ⁇ .
  • the BackPropagation algorithm is used to optimize the input weights and offsets of the neurons of each layer of the artificial neural network, so that the output of the artificial neural network is as close as possible to the expected output to achieve the purpose of training.
  • Step S2 Add the trained artificial neural network to the thrust distribution model to obtain the following model:
  • T max , T min represent the upper and lower limits of the thrust force that each full-rotation thruster can emit
  • F x , F y , M z are the resultant forces and moments of all thrusters in the three degrees of freedom of pitch, yaw and bow respectively.
  • Step S3 Set the total force and moment of all thrusters in the three degrees of freedom of pitch, roll and bow in the real-time control computer.
  • step S4 iterative calculation of thrust distribution is performed in conjunction with the quadratic programming algorithm.
  • the mathematical model of thrust distribution is abbreviated as:
  • B k is a symmetric positive definite matrix calculated using the WHP method.
  • the speed and angle of each thruster calculated by the real-time control computer according to the thrust distribution iterative calculation are output to the speed controller and angle controller of the full-rotation thruster, thereby generating the thrust required for positioning the marine structure and realizing the positioning effect.
  • the method further includes the following steps: by repeating step S3 and step S4, continuous thrust distribution calculation can be performed.
  • a dynamic positioning thrust distribution device based on an artificial neural network, which includes a real-time control computer and a number of full-rotation thrusters distributed on the bottom of the marine structure buoy, each full-rotation thruster
  • the propeller speed controller and the full rotation angle controller are respectively installed, and the propeller speed controller and the full rotation angle controller are connected to the real-time control computer through a data line;
  • the real-time control computer uses the above-mentioned artificial neural network-based dynamic positioning thrust distribution method to calculate the rotation speed and angle of each full-rotation thruster and output to the propeller rotation speed controller and full-rotation angle controller to generate the thrust required for marine structure positioning To achieve positioning.
  • the present invention has the following beneficial effects:
  • the dynamic positioning thrust distribution device and method based on artificial neural network provided by the present invention can accurately quantify the thrust loss on the one hand, and can expand the feasible area of the propeller rotation angle on the other hand, so that the secondary planning problem can be obtained more Optimized and reasonable results, reduce thruster power and save energy.
  • FIG. 1 is a schematic structural diagram of a dynamic positioning thrust distribution device based on an artificial neural network.
  • Figure 2 is a schematic diagram of the distribution of the semi-submersible offshore platform thrusters equipped with the present invention.
  • Figure 3 is a schematic diagram of prohibited corners.
  • FIG. 4 is a flowchart of a dynamic positioning thrust distribution method based on an artificial neural network.
  • Fig. 5 is a schematic diagram of a neural network for fitting thrust coefficients.
  • Figure 6 is a schematic diagram of the thrust distribution results.
  • This embodiment provides a dynamic positioning thrust distribution method based on an artificial neural network, which includes the following steps:
  • Step S1 Establish and train an artificial neural network to fit the thrust coefficients in the real-time control computer.
  • the established artificial neural network includes input layer, hidden layer and output layer.
  • ⁇ is the angle of the front thruster
  • ⁇ is the thrust coefficient of the rear thruster.
  • the method of training the artificial neural network is to use the experimental data obtained by the artificial neural network model test as the training sample ⁇ (x(1), y(1)), (x(2), y(2)),..., (x(m), y(m)) ⁇ .
  • the BackPropagation algorithm is used to optimize the input weights and offsets of the neurons of each layer of the artificial neural network, so that the output of the artificial neural network is as close as possible to the expected output to achieve the purpose of training.
  • Step S2 Add the trained artificial neural network to the thrust distribution model to obtain the following model:
  • T max , T min represent the upper and lower limits of the thrust force that each full-rotation thruster can emit
  • F x , F y , M z are the resultant forces and moments of all thrusters in the three degrees of freedom of pitch, yaw and bow respectively.
  • Step S3 Set the total force and moment of all thrusters in the three degrees of freedom of pitch, roll and bow in the real-time control computer.
  • step S4 iterative calculation of thrust distribution is performed in conjunction with the quadratic programming algorithm.
  • the mathematical model of thrust distribution is abbreviated as:
  • B k is a symmetric positive definite matrix calculated using the WHP method.
  • the speed and angle of each thruster obtained by the iterative calculation of the real-time control computer according to the thrust distribution are output to the speed controller and angle controller of the full-rotation thruster, thereby generating the thrust required for the positioning of the marine structure and achieving the positioning effect.
  • the method further includes the following steps: by repeating step S3 and step S4, continuous thrust distribution calculation can be performed.
  • FIG. 4 shows a flowchart of a dynamic positioning thrust distribution method based on an artificial neural network. This method can make the total force and torque of the eight full-rotation thrusters full of the total thrust and torque required for the positioning of the semi-submersible platform, and at the same time make the rotation angle of the front thruster avoid the prohibited angle area
  • First perform step one establish and train the neural network in the real-time control computer. This includes:
  • the input layer includes one neuron node
  • the first hidden layer includes 15 neuron nodes
  • the second hidden layer includes 35 neuron nodes
  • the third hidden layer includes 15 neuron nodes.
  • ⁇ is the angle of the front thruster
  • ⁇ is the thrust coefficient of the rear thruster.
  • the data of the front propeller rotation angle and the rear propeller thrust coefficient obtained from the model test are used as the training samples of the artificial neural network.
  • the training samples include 360 sets of data, ⁇ (x(1), y(1)), (x(2), y(2)), ..., (x(360), y(360)) ⁇ .
  • Each set of data is a front propeller angle and its corresponding rear propeller thrust coefficient.
  • the BackPropagation algorithm is used to optimize the input weights and offsets of neurons in each layer, so that the output of the artificial neural network is as close as possible to the expected output to achieve the purpose of training.
  • Step 1 eight times to obtain a continuous function of the rotation angle of the front propeller and the thrust coefficient of the rear propeller for the eight sets of full-rotation thrusters on the semi-submersible platform. This function can be added to the constraints of the later mathematical model of thrust distribution.
  • Step 2 Add the trained artificial neural network to the thrust distribution model to obtain the following model:
  • c i is a constant related to the fluid density ⁇ and the propeller diameter D of each full-rotation thruster.
  • T max , T min represent The upper and lower limits of the thrust that each full-rotation thruster can deliver.
  • the total force and total torque of all thrusters in the three degrees of freedom are set in step 3 below.
  • Steps 1 and 2 above are the preparatory steps for the thrust distribution calculation, and are completed before the thrust distribution calculation is performed. Steps 1 and 2 need only be completed once, without repetition, and memory is used repeatedly for thrust distribution calculations.
  • Step 4 Combine the quadratic programming algorithm to perform iterative calculation of thrust distribution.
  • the angle values ⁇ i of the eight sets of thrusters are taken out and brought into the artificial neural network of each thruster respectively to obtain the thrust loss coefficients of the eight sets of thrusters.
  • the optimal solution for thrust distribution x [162.66,153.64,136.74,116.73,103.07,107.24,142.96,155.25,
  • the angle of thrust distribution and the magnitude of thrust are shown in Figure 6.
  • the circle in the figure represents the thruster, the line segment starting from the center of the circle represents the thrust force, the direction of the line segment is the thrust direction, and the size of the line segment is the thrust force.
  • the speed and angle of each thruster calculated by the real-time control computer according to the thrust distribution method are output to the speed controller and angle controller of the full-rotation thruster, thereby generating the thrust required for the positioning of the marine structure and achieving the positioning effect.
  • This embodiment provides a dynamic positioning thrust distribution device based on an artificial neural network, which includes a real-time control computer and a number of full-rotation thrusters distributed at the bottom of the buoy of the marine structure, each full-rotation thruster is equipped with a propeller speed
  • the controller and the full rotation angle controller, the propeller speed controller and the full rotation angle controller are connected to the real-time control computer through a data line;
  • the real-time control computer uses the artificial neural network-based dynamic positioning thrust distribution method provided in Embodiment 1 to calculate the rotation speed and angle of each full-rotation propeller and output it to the propeller rotation speed controller and the full-rotation angle controller to generate the ocean.
  • Fig. 1 is a schematic structural diagram of a dynamic positioning thrust distribution device based on an artificial neural network.
  • a schematic diagram of a dynamic positioning thrust distribution device based on an artificial neural network includes a real-time control computer 1 and several distributed structures in marine structures A full rotation propeller 4 at the bottom of the pontoon, each full rotation propeller 4 is respectively installed with a propeller speed controller 2 and a full rotation angle controller 3, and the propeller speed controller 2 and the full rotation angle controller 3 pass through the data line Connect with real-time control computer.
  • FIG 2 is a schematic diagram of the distribution of the semi-submersible offshore platform thrusters of the present invention, which is equipped with a total of eight sets of full-rotation thrusters.
  • the angle of rotation of the front propeller should be avoided by setting the prohibited angle to avoid the area where the thrust of the rear propeller is lost during the meeting.
  • the real-time control computer uses the artificial neural network-based dynamic positioning thrust distribution method provided in Example 1 to calculate the rotation speed and angle of each full-rotation thruster and output it to the propeller speed controller and full-rotation angle controller, thereby generating the marine structure
  • the thrust required for positioning solves the above problems and achieves positioning.
  • the above two embodiments of the present invention provide a thrust distribution device and method for dynamic positioning of marine structures based on artificial neural networks to calculate thrust loss.
  • the thrust distribution method is a quadratic planning problem in the optimization problem by considering the front propeller rotation angle Using artificial neural network, the thrust coefficient of the rear thruster can be calculated in the constraint conditions of the thrust distribution problem. Then, based on the sequential quadratic programming algorithm, the optimization problem of minimizing the thruster power is solved, and the thrust distribution scheme on the full-rotation thruster is obtained.
  • the thrust distribution device adopts the thrust distribution method, calculates the rotation speed and angle of each full-rotation propeller, and outputs it to the propeller rotation speed controller and the full-rotation angle controller to generate the thrust required for the positioning of the marine structure and achieve the positioning.
  • the present invention can accurately quantify the thrust loss, on the other hand, it can expand the feasible area of the propeller rotation angle, so that the quadratic planning problem can be more optimized and Reasonable results reduce propeller power and save energy.

Landscapes

  • Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Automation & Control Theory (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Feedback Control In General (AREA)
  • Other Liquid Machine Or Engine Such As Wave Power Use (AREA)

Abstract

一种基于人工神经网络的动力定位推力分配方法及装置，该方法为最优化问题中的二次规划问题，通过考虑前推进器转角，利用人工神经网络，可以计算推力分配问题约束条件中的后推进器推力系数；随后基于序列二次规划算法求解使推进器功率最小化的最优化问题，获得全回转推进器上的推力分配方案。该方法及装置通过引入推力系数的概念，一方面可以精确地量化推力损失，另一方面可以扩大推进器回转角的可行区域，从而使二次规划问题能获得更为优化与合理的结果，降低推进器功率，节约能源。

Description

基于人工神经网络的动力定位推力分配装置及方法 技术领域
本发明涉及一种海洋结构物动力定位系统推力分配技术，具体地，涉及一种基于人工神经网络计算推力损失的海洋结构物动力定位推力分配装置及方法。
背景技术
随着人类对海洋资源的开发规模的逐步扩大，在恶劣海况中作业的海洋结构物定位系统的发展，已成为海洋工程高技术装备研发中的重要课题之一。动力定位系统是一种利用船舶或海洋平台自身推进器，使其尽可能地保持在海平面上要求的位置的控制系统。动力定位系统具有操作简便，定位精度不受水深影响等有点，已经广泛应用于深水半潜钻井平台。动力定位系统一般由位置测量系统，控制系统，推力系统三部分构成。测量系统通过卫星或水声列阵获得结构物的实时位置信息。控制系统依据海洋结构物的实际位置与目标位置的偏差计算出使海洋结构物回复到目标位置所需的总推力与扭矩。推力系统对各推进器进行推力分配，进而使各推进器产生海洋结构物定位所需的总推力和扭矩，最终实现海洋结构物通过自身推进器的动力定位。
推力分配问题本质上是一个非线性最优化问题，推力分配方法需要在保证所有推进器的合力等于海洋结构物定位所需的总推力与扭矩的同时，使所有推进器的总功率最小化。
在推力分配问题中，必须考虑的一个关键因素是全回转推进器之间的水动力干扰问题。对于半潜钻井平台，串联排列的两个全回转推进器，前推进器尾流会影响后推进器，造成后推进器产生的实际推力小于其理论推力，这种现象称之为推力损失。推力损失会造成半潜平台严重的能源浪费。在现有技术中，为避免推力损失，前推进器的回转角会设置禁止区域，前推进器的转角不能停留在进入禁止角区域，防止其尾流影响后推进器。但是，由于的禁止角存在，推进器在旋转时无法穿过禁止角，从而出现推进器转角卡顿在禁止角边缘的问题。无法得到 推力分配的最优解，推进器总功率无法实现最小化，造成能源浪费。
发明内容
针对现有技术中存在的上述不足，本发明的目的是提供一种基于人工神经网络的动力定位推力分配装置及方法。该装置及方法避免了在推力分配问题中，由于水动力干扰造成的推力损失；同时又取消了禁止角，解决了推进器转角卡顿的问题，使推进器总功率减小，节约能源。
本发明是通过以下技术方案实现的。
根据本发明的一个方面，提供过了一种基于人工神经网络的动力定位推力分配方法，包括以下步骤：
步骤S1，在实时控制计算机中建立并训练人工神经网络拟合推力系数。
其中：
建立的人工神经网络，包括输入层、隐藏层和输出层。输入向量为x＝[α]，α为前推进器转角，输出向量为y＝[η]，η为后推进器推力系数。
训练人工神经网络的方法为，将人工神经网络模型试验得到的试验数据作为训练样本{(x(1)，y(1))，(x(2)，y(2))，...，(x(m)，y(m))}。使用Back Propagation算法最优化人工神经网络各层神经元的输入权值以及偏置，使得人工神经网络的输出尽可能地接近期望输出，以达到训练的目的。
步骤S2，将训练完成的人工神经网络加入推力分配模型，得到如下模型：
Figure PCTCN2019127352-appb-000001
Figure PCTCN2019127352-appb-000002
Figure PCTCN2019127352-appb-000003
Figure PCTCN2019127352-appb-000004
T min≤T i≤T max
其中，c i为每个全回转推进器与流体密度ρ和螺旋桨直径D有关的常数，T max，T min表示每个全回转推进器所能发出推力大小的上限和下限，F x，F y，M z分 别为设定的纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩。
步骤S3，在实时控制计算机中设定纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩。
步骤S4，结合二次规划算法进行推力分配迭代计算。推力分配数学模型简记为：
Figure PCTCN2019127352-appb-000005
其中，x＝[T 1，T 2，...，T 8，α 1，α 2，...，α 8]，α i，为第i个推进器的转角，i＝1，2，...，8，给定初值
Figure PCTCN2019127352-appb-000006
开始迭代计算，对于迭代过程中的某一中间值
Figure PCTCN2019127352-appb-000007
将每个推进器的转角变量带入训练好的人工神经网络，得到人工神经网络输出的第i个推进器的推力系数
Figure PCTCN2019127352-appb-000008
并带入推力分配模型。将推力分配模型线性化，得到线性约束的子问题：
Figure PCTCN2019127352-appb-000009
其中，B k为使用WHP方法计算的对称正定矩阵。求解线性约束的二次规划最优化问题，得到目标函数在x k处得下降方向d k与步长α k，更新x k+1＝x k+α kd k。若新的解小于容许误差，则得到推力分配的最优解，否则继续迭代。
将实时控制计算机根据推力分配迭代计算得到的各推进器转速与角度输出至全回转推进器转速控制器和角度控制器，从而产生海洋结构定位所需的推力，实现定位效果。
优选地，还包括如下步骤：重复步骤S3和步骤S4，即可进行连续的推力分配计算。
根据本发明的另一个方面，提供了一种基于人工神经网络的动力定位推力分配装置，包括一台实时控制计算机和若干分布在海洋结构物浮筒底部的全回转推进器，各全回转推进器上分别安装有螺旋桨转速控制器和全回转角度控制器，所述的螺旋桨转速控制器和全回转角度控制器通过数据线与实时控制计算机相连；
所述实时控制计算机采用上述基于人工神经网络的动力定位推力分配方法，计算每一个全回转推进器的转速与角度并输出至螺旋桨转速控制器和全回转角 度控制器，产生海洋结构定位所需推力，实现定位。
优选地，若干全回转推进器分布在海洋结构物底部的两个浮筒的前部和后部。与现有技术相比，本发明具有如下有益效果：
本发明所提供的基于人工神经网络的动力定位推力分配装置及方法，一方面可以精确地量化推力损失，另一方面可以扩大推进器回转角的可行区域，从而使二次规划问题能获得更为优化与合理的结果，降低推进器功率，节约能源。
附图说明
通过阅读参照以下附图对非限制性实施例所作的详细描述，本发明的其它特征、目的和优点将会变得更明显：
图1为基于人工神经网络的动力定位推力分配装置结构示意图。
图2为装备了本发明的半潜式海洋平台推进器分布示意图。
图3为禁止角示意图。
图4为基于人工神经网络的动力定位推力分配方法流程图。
图5为用于拟合推力系数的神经网络示意图。
图6为推力分配结果示意图。
具体实施方案
下面对本发明的实施例作详细说明：本实施例在以本发明技术方案为前提下进行实施，给出了详细的实施方式和具体的操作过程。应当指出的是，对本领域的普通技术人员来说，在不脱离本发明构思的前提下，还可以做出若干变形和改进，这些都属于本发明的保护范围。
实施例1
本实施例提供过了一种基于人工神经网络的动力定位推力分配方法，包括以下步骤：
步骤S1，在实时控制计算机中建立并训练人工神经网络拟合推力系数。
其中：
建立的人工神经网络，包括输入层、隐藏层和输出层。输入向量为x＝[α]，α为前推进器转角，输出向量为y＝[η]，η为后推进器推力系数。
训练人工神经网络的方法为，将人工神经网络模型试验得到的试验数据作为训练样本{(x(1)，y(1))，(x(2)，y(2))，...，(x(m)，y(m))}。使用Back Propagation算法最优化人工神经网络各层神经元的输入权值以及偏置，使得人工神经网络的输出尽可能地接近期望输出，以达到训练的目的。
步骤S2，将训练完成的人工神经网络加入推力分配模型，得到如下模型：
Figure PCTCN2019127352-appb-000010
Figure PCTCN2019127352-appb-000011
Figure PCTCN2019127352-appb-000012
Figure PCTCN2019127352-appb-000013
T min≤T i≤T max
其中，c i为每个全回转推进器与流体密度ρ和螺旋桨直径D有关的常数，T max，T min表示每个全回转推进器所能发出推力大小的上限和下限，F x，F y，M z分别为设定的纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩。
步骤S3，在实时控制计算机中设定纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩。
步骤S4，结合二次规划算法进行推力分配迭代计算。推力分配数学模型简记为：
Figure PCTCN2019127352-appb-000014
其中，x＝[T 1，T 2，...，T 8，α 1，α 2，...，α 8]，α i，为第i个推进器的转角，i＝1，2，...，8，给定初值
Figure PCTCN2019127352-appb-000015
开始迭代计算，对于迭代过程中的某一中间值
Figure PCTCN2019127352-appb-000016
将每个推进器的转角变量带入训练好的人工神经网络，得到人工神经网络输出的第i个推进器的推力系数
Figure PCTCN2019127352-appb-000017
并带入推力分配模型。将推力分配模型线性化，得到线性约束的子问题：
Figure PCTCN2019127352-appb-000018
其中，B k为使用WHP方法计算的对称正定矩阵。求解线性约束的二次规划最优化问题，得到目标函数在x k处得下降方向d k与步长α k，更新x k+1＝x k+α kd k。若新的解小于容许误差，则得到推力分配的最优解，否则继续迭代。
将实时控制计算机根据推力分配迭代计算得到的各推进器转速与角度输出至全回转推进器转速控制器和角度控制器，从而产生海洋结构定位所需的推力，实现定位效果。
优选地，还包括如下步骤：重复步骤S3和步骤S4，即可进行连续的推力分配计算。
为了使本实施例实现的技术手段、创作特征、达成目的与功效易于明白了解，下面结合附图对本实施例进一步详细阐述。
针对实施例1中所指出的动力定位系统推力损失与推进器转角卡顿问题，图4展示了一种基于人工神经网络的动力定位推力分配方法流程图。该方法可以使八个全回转推进器的合力与合力矩满半潜平台定位所需的总推力与转矩，同时使前推进器的回转角度避开禁止角区域
首先执行步骤一，在实时控制计算机中建立并训练神经网路。具体包括：
1)建立人工神经网络，其结构如图5所示，包括输入层、三个隐藏层和输出层。其中，输入层包括一个神经元节点，第一隐藏层包括15个神经元节点，第二隐藏层包括35个神经元节点，第三隐藏层包括15个神经元节点。输入向量为x＝[α]，α为前推进器转角，输出向量为y＝[η]，η为后推进器推力系数。
2)训练人工神经网络。将模型试验得到的前桨回转角与后桨推力系数数据作为人工神经网络的训练样本，训练样本共包括360组数据，{(x(1)，y(1))，(x(2)，y(2))，...，(x(360)，y(360))}。每组数据为一个前桨角度以及其对应的后桨推力系数。使用Back Propagation算法最优化各层神经元的输入权值以及偏置，使得人工神经网络的输出尽可能地接近期望输出，以达到训练的目的。
重复进行八次步骤一，得到半潜平台八套全回转推进器关于前桨回转角度 与后桨推力系数的连续函数。该函数可以加入到后文推力分配数学模型的约束条件中。
步骤二，将训练完成的人工神经网络加入推力分配模型，得到如下模型：
Figure PCTCN2019127352-appb-000019
Figure PCTCN2019127352-appb-000020
Figure PCTCN2019127352-appb-000021
Figure PCTCN2019127352-appb-000022
T min≤T i≤T max
其中，c i为每个全回转推进器与流体密度ρ和螺旋桨直径D有关的常数，示例性的，c i＝1，i＝1，2，...，8，T max，T min表示每个全回转推进器所能发出推力大小的上限和下限，示例性的T max＝800，T min＝50，F x，F y，M z分别为设定的纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩，在后文步骤三中设定具体数值。
以上步骤一和步骤二为推力分配计算的准备步骤，在推力分配计算进行之前完成。步骤一和步骤二只需完成一次即可，无需重复，记忆反复用于推力分配计算。
步骤3，在实时控制计算机中设定纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩。示例性的，设定F x＝1000，F y＝-344，M z＝-10000。
步骤4，结合二次规划算法进行推力分配迭代计算。首先给定初值，x 0＝[50，50，50，50，50，50，50，50，0，0，0，0，0，0，0，0]。取出八套推进器的转角值α i分别带入每个推进器的人工神经网络，得到八套推进器的推力损失系数。示例性的，对于α 1＝α 2＝…＝α 8＝0，通过人工神经网络输出的推力系数η 1＝η 2＝…＝η 8＝1。将推力系数带入推力分配模型，将推力分配模型线性化，得到线性约束的子问题：
Figure PCTCN2019127352-appb-000023
其中，B k为使用WHP方法计算的对称正定矩阵(Schittkowski K.The nonlinear programming method of Wilson,Han,and Powell with an augmented Lagrangian type line search function[J].Numerische Mathematik,1982,38(1):115-127.)。求解线性约束的二次规划最优化问题，使用有效集方法求解(Shanno D F.Numerical Optimization by Jorge Nocedal；Stephen J.Wright[J].Siam Review,2001,43(3):572-574.)，得到目标函数在x k处得下降方向d k与步长α k，更新x k+1＝x k+α kd k。若新的解小于容许误差，则得到推力分配的最优解，否则将x k+1中的推进器角度带入人工神经网络，得到推力损失值η i(i＝1，2，...，8)，带入线性约束的子问题，继续迭代，直至得到最优解。对于设定的F x＝1000，F y＝-344，M z＝-10000，得到推力分配最优解x＝[162.66，153.64，136.74，116.73，103.07，107.24，142.96，155.25，
-23.2，-9.0，-41.3，-32.3，-9.0，-10.4，-8.9，-17.0]。推力分配的角度和推力大小如图6所示。图中圆圈表示推进器，以圆心为起点的线段代表推力，线段方向为推力方向，线段大小为推力大小。
将实时控制计算机根据推力分配方法计算得到的各推进器转速与角度输出至全回转推进器转速控制器和角度控制器，从而产生海洋结构定位所需的推力，实现定位效果。
实施例2
本实施例提供了一种基于人工神经网络的动力定位推力分配装置，包括一台实时控制计算机和若干分布在海洋结构物浮筒底部的全回转推进器，各全回转推进器上分别安装有螺旋桨转速控制器和全回转角度控制器，所述的螺旋桨转速控制器和全回转角度控制器通过数据线与实时控制计算机相连；
所述实时控制计算机采用实施例1所提供的基于人工神经网络的动力定位推力分配方法，计算每一个全回转推进器的转速与角度并输出至螺旋桨转速控制器和全回转角度控制器，产生海洋结构定位所需推力，实现定位。
优选地，若干全回转推进器分布在海洋结构物底部的两个浮筒的前部和后部。
为了使本实施例实现的技术手段、创作特征、达成目的与功效易于明白了解，下面结合附图对本实施例进一步详细阐述。
图1为基于人工神经网络的动力定位推力分配装置结构示意图，如图所示，一种基于人工神经网络的动力定位推力分配装置示意图，包括一台实时控制计算机1和若干台分布在海洋结构物浮筒底部的全回转推进器4，各全回转推进器4上分别安装有螺旋桨转速控制器2和全回转角度控制器3，所述的螺旋桨转速控制器2和全回转角度控制器3通过数据线与实时控制计算机相连。
图2装备了本发明的半潜式海洋平台推进器分布示意图，它一共装备八套全回转推进器。距离较近的两个推进器为一组。若一组推进器中的前推进器(以下简称前桨)角度达到某些位置时，如图3所示，其尾流对后推进器(以下简称后桨)产生较大的水动力干扰，从而造成后桨推力损失。在实际工程应用中，通过设置禁止角使前桨回转角应避开会造成后桨推力损失的区域。但是禁止角的存在，推进器在旋转时无法穿过禁止角，从而推进器的角度停留在禁止角边缘，形成推进器转角卡顿的现象。无法得到推力分配的最优解，推进器总功率无法实现最小化，造成能源浪费。实时控制计算机采用实施例1所提供的基于人工神经网络的动力定位推力分配方法，计算每一个全回转推进器的转速与角度并输出至螺旋桨转速控制器和全回转角度控制器，进而产生海洋结构定位所需推力，解决了以上问题，实现了定位。
本发明上述两个实施例提供的一种基于人工神经网络计算推力损失的海洋结构物动力定位推力分配装置及方法，推力分配方法为最优化问题中的二次规划问题，通过考虑前推进器转角，利用人工神经网络，可以计算推力分配问题约束条件中的后推进器推力系数。随后基于序列二次规划算法求解使推进器功率最小化的最优化问题，获得全回转推进器上的推力分配方案。推力分配装置采用推力分配方法，计算每一个全回转推进器的转速与角度并输出至螺旋桨转速控制器和全回转角度控制器，产生海洋结构定位所需推力，实现定位。本发明上述两个实施例通过引入推力系数的概念，本发明一方面可以精确地量化推力损失，另一方面可以扩大推进器回转角的可行区域，从而使二次规划问题能获得更为优化与合 理的结果，降低推进器功率，节约能源。
以上显示和描述了本发明的基本原理、主要特征和本发明的优点。本领域的技术人员应该了解，本发明不受上述例子的限制，上述实例和说明书中描述的只是说明本发明的原理，在不脱离本发明精神和范围的前提下本发明还会有各种变化和改进，这些变化和改进都落入要求保护的本发明范围内。

Claims (6)

  1. 一种基于人工神经网络的动力定位推力分配方法，其特征在于，包括以下步骤：
    步骤S1，在实时控制计算机中建立并训练人工神经网络拟合推力系数；
    步骤S2，将训练完成的人工神经网络加入推力分配模型，得到如下模型：
    Figure PCTCN2019127352-appb-100001
    Figure PCTCN2019127352-appb-100002
    Figure PCTCN2019127352-appb-100003
    Figure PCTCN2019127352-appb-100004
    T min≤T i≤T max
    其中，c i为每个全回转推进器与流体密度ρ和螺旋桨直径D有关的常数，T max，T min分别表示每个全回转推进器所能发出推力大小的上限和下限，F x，F y，M z分别为设定的纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩；
    步骤S3，在实时控制计算机中设定纵荡、横荡和艏摇三个自由度方向上所有推进器合力与合力矩；
    步骤S4，结合二次规划算法进行推力分配迭代计算；推力分配数学模型简记为：
    Figure PCTCN2019127352-appb-100005
    其中，x＝[T 1，T 2，...，T 8，α 1，α 2，...，α 8]，α i，为第i个推进器的转角，i＝1，2，...，8， 给定初值
    Figure PCTCN2019127352-appb-100006
    开始迭代计算，对于迭代过程中的某一中间值
    Figure PCTCN2019127352-appb-100007
    将每个推进器的转角变量带入训练好的人工神经网络，得到人工神经网络输出的第i个推进器的推力系数
    Figure PCTCN2019127352-appb-100008
    并带入推力分配模型；将推力分配模型线性化，得到线性约束的子问题：
    Figure PCTCN2019127352-appb-100009
    其中，B k为使用WHP方法(Wilson-Han-Powell方法)计算的对称正定矩阵，该方法为推力分配算法中使用的经典算法；使用有效集方法求解线性约束的二次规划最优化问题，该方法为推力分配算法中使用的经典算法，得到目标函数在x k处得下降方向d k与步长α k，更新x k+1＝x k+α kd k；若新的解小于容许误差，则得到推力分配的最优解，否则继续迭代；
    将实时控制计算机根据推力分配迭代计算得到的各推进器转速与角度输出至全回转推进器转速控制器和角度控制器，从而产生海洋结构定位所需的推力，实现定位效果。
  2. 根据权利要求1所述的基于人工神经网络的动力定位推力分配方法，其特征在于，所述步骤S1中，建立的人工神经网络，包括输入层、隐藏层和输出层；输入向量为x＝[α]，α为前推进器转角，输出向量为y＝[η]，η为后推进器推力系数。
  3. 根据权利要求1所述的基于人工神经网络的动力定位推力分配方法，其特征在于，所述步骤S1中，训练人工神经网络的方法为，将人工神经网络模型试验得到的试验数据作为训练样本{(x(1)，y(1))，(x(2)，y(2))，...，(x(m)，y(m))}；使用人工神经网络经典算法Back Propagation算法最优化人工神经网络各层神经元的输入权值以及偏置，使得人工神经网络的输出尽可能地接近期望输出，以达 到训练的目的。
  4. 根据权利要求1至3中任一项所述的基于人工神经网络的动力定位推力分配方法，其特征在于，还包括如下步骤：
    重复步骤3和步骤4，进行连续的推力分配计算。
  5. 一种基于人工神经网络的动力定位推力分配装置，其特征在于，包括实时控制计算机(1)和若干分布在海洋结构物底部的全回转推进器(4)，其中，每一个全回转推进器(4)上分别安装有螺旋桨转速控制器(2)和全回转角度控制器(3)，所述螺旋桨转速控制器(2)和全回转角度控制器(3)分别通过数据线与实时控制计算机(1)连接；
    所述实时控制计算机采用权利要求1至4中任意一项所述的基于人工神经网络的动力定位推力分配方法，计算每一个全回转推进器(4)的转速与角度并输出至螺旋桨转速控制器(2)和全回转角度控制器(3)，产生海洋结构定位所需推力，实现定位。
  6. 根据权利要求5所述的基于人工神经网络的动力定位推力分配装置，其特征在于，若干全回转推进器(4)分布在海洋结构物底部的两个浮筒的前部和/或后部。
PCT/CN2019/127352 2018-12-26 2019-12-23 基于人工神经网络的动力定位推力分配装置及方法 WO2020135312A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/764,665 US11500339B2 (en) 2018-12-26 2019-12-23 Dynamic positioning and thrust distribution device and method based on artificial neural network

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811597787.9 2018-12-26
CN201811597787.9A CN109683479B (zh) 2018-12-26 2018-12-26 基于人工神经网络的动力定位推力分配装置及方法

Publications (1)

Publication Number Publication Date
WO2020135312A1 true WO2020135312A1 (zh) 2020-07-02

Family

ID=66189297

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/127352 WO2020135312A1 (zh) 2018-12-26 2019-12-23 基于人工神经网络的动力定位推力分配装置及方法

Country Status (3)

Country Link
US (1) US11500339B2 (zh)
CN (1) CN109683479B (zh)
WO (1) WO2020135312A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109709970A (zh) * 2018-12-25 2019-05-03 哈尔滨工程大学 一种水下机器人六自由度推力分配优化方法
CN112947254A (zh) * 2021-03-29 2021-06-11 武汉船用电力推进装置研究所（中国船舶重工集团公司第七一二研究所） 一种工程船的多推进器联合操控方法
CN113115355A (zh) * 2021-04-29 2021-07-13 电子科技大学 一种d2d系统中基于深度强化学习的功率分配方法
US20210223747A1 (en) * 2018-12-26 2021-07-22 Shanghai Jiao Tong University Dynamic positioning and thrust distribution device and method based on artificial neural network
CN115407664A (zh) * 2022-11-01 2022-11-29 北京航天自动控制研究所 一种基于神经网络训练的非程序制导方法

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114194347B (zh) * 2022-01-05 2022-12-27 广东海洋大学 半潜式海洋平台的动力定位方法、装置、设备及介质
CN117150826B (zh) * 2023-10-31 2024-01-23 中铁十四局集团有限公司 适用于同步推拼系统的油缸推力精确分配方法及系统
CN117341953A (zh) * 2023-12-04 2024-01-05 中国船舶集团有限公司第七〇七研究所 基于船舶喷水推进器的推力分配方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002173086A (ja) * 2000-12-08 2002-06-18 Mitsubishi Heavy Ind Ltd 洋上プラットフォームの制御方法
CN102508431A (zh) * 2011-11-04 2012-06-20 江苏科技大学 一种海洋钻井平台动力定位系统推力分配方法
CN109683479A (zh) * 2018-12-26 2019-04-26 上海交通大学 基于人工神经网络的动力定位推力分配装置及方法

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6317658B1 (en) * 1997-10-15 2001-11-13 The Boeing Company Neurocomputing control distribution system
US7257470B2 (en) * 2004-10-29 2007-08-14 Rosemount Aerospace Inc. Fault isolation method and apparatus in artificial intelligence based air data systems
CN101963784A (zh) * 2010-09-07 2011-02-02 江苏科技大学 基于支持向量机和广义预测控制算法的船舶动力定位方法
CN102455708A (zh) * 2010-10-28 2012-05-16 上海振华重工(集团)股份有限公司 船舶动力定位的船舶模型在线辨识和控制系统和方法
CN104050369B (zh) * 2014-06-17 2016-09-14 上海交通大学 基于综合动力定位能力的推力敏感性分析方法
CN104112045B (zh) * 2014-07-07 2017-04-26 上海交通大学 基于动力定位能力综合标准的推力器局部最优配置方法
CN106200384A (zh) * 2016-08-10 2016-12-07 浙江大学 一种应用于冗余驱动系统的双直线电机协同控制方法
US10464670B2 (en) * 2017-02-22 2019-11-05 Gopro, Inc. Variable condition motor controller
CN107992648B (zh) * 2017-11-15 2021-04-20 南京航空航天大学 一种自适应rbf神经网络进行航空发动机推力估计算法
CN108333925A (zh) * 2018-01-18 2018-07-27 上海交通大学 考虑输入和终端约束的船舶推力分配自适应优化方法
MX2020007636A (es) * 2018-01-18 2020-10-12 Safe Marine Transfer Llc Unidad submarina de control electrico inteligente.
US11561690B2 (en) * 2018-04-22 2023-01-24 Jmp Statistical Discovery Llc Interactive graphical user interface for customizable combinatorial test construction
CN112379591B (zh) * 2020-10-22 2022-05-03 智慧航海（青岛）科技有限公司 一种考虑推进器性能的推力分配优化方法
CN112506060B (zh) * 2020-12-15 2021-11-02 南通大学 一种混群优化算法的船舶推力分配方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002173086A (ja) * 2000-12-08 2002-06-18 Mitsubishi Heavy Ind Ltd 洋上プラットフォームの制御方法
CN102508431A (zh) * 2011-11-04 2012-06-20 江苏科技大学 一种海洋钻井平台动力定位系统推力分配方法
CN109683479A (zh) * 2018-12-26 2019-04-26 上海交通大学 基于人工神经网络的动力定位推力分配装置及方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
XIAODI ZHANG: "Research on Multi-step Optimization Algorithm of Ship Thrust Distribution", MASTER’S DISSERTATION, 31 July 2016 (2016-07-31), CN, pages 1 - 98, XP009521701 *
YAN ZHU: "Research on Thrust Distribution of Dynamic Positioning System", MASTER DEGREE THESIS OF JIANGSU UNIVERSITY OF SCIENCE AND TECHNOLOGY, 31 March 2016 (2016-03-31), pages 1 - 98, XP009521647 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109709970A (zh) * 2018-12-25 2019-05-03 哈尔滨工程大学 一种水下机器人六自由度推力分配优化方法
CN109709970B (zh) * 2018-12-25 2022-01-14 哈尔滨工程大学 一种水下机器人六自由度推力分配优化方法
US20210223747A1 (en) * 2018-12-26 2021-07-22 Shanghai Jiao Tong University Dynamic positioning and thrust distribution device and method based on artificial neural network
US11500339B2 (en) * 2018-12-26 2022-11-15 Shanghai Jiao Tong University Dynamic positioning and thrust distribution device and method based on artificial neural network
CN112947254A (zh) * 2021-03-29 2021-06-11 武汉船用电力推进装置研究所（中国船舶重工集团公司第七一二研究所） 一种工程船的多推进器联合操控方法
CN113115355A (zh) * 2021-04-29 2021-07-13 电子科技大学 一种d2d系统中基于深度强化学习的功率分配方法
CN113115355B (zh) * 2021-04-29 2022-04-22 电子科技大学 一种d2d系统中基于深度强化学习的功率分配方法
CN115407664A (zh) * 2022-11-01 2022-11-29 北京航天自动控制研究所 一种基于神经网络训练的非程序制导方法

Also Published As

Publication number Publication date
CN109683479B (zh) 2021-07-06
US20210223747A1 (en) 2021-07-22
CN109683479A (zh) 2019-04-26
US11500339B2 (en) 2022-11-15

Similar Documents

Publication Publication Date Title
WO2020135312A1 (zh) 基于人工神经网络的动力定位推力分配装置及方法
Carrica et al. Near-surface self propulsion of a generic submarine in calm water and waves
Belibassakis et al. Hydrodynamic performance of flapping wings for augmenting ship propulsion in waves
Bertram Practical ship hydrodynamics
CN106773722A (zh) 一种基于人工鱼群算法的船舶动力定位系统推力分配方法
Suastika et al. Effects of the application of a stern foil on ship resistance: A case study of an Orela crew boat
Delen et al. Direct CFD simulations of standard maneuvering tests for DARPA Suboff
Liu Mathematical modeling of inland vessel maneuverability considering rudder hydrodynamics
Dai et al. Experimental and numerical investigation on maneuvering performance of small waterplane area twin hull
CN117521556B (zh) 一种内孤立波影响下潜艇姿态预测方法
Fu et al. Research on thrust allocation of dynamic positioning ship with cycloidal propeller
Greve et al. Design of the Propulsion System for the Autonomous XLUUV MUM
CN116541951A (zh) 一种基于改进的鱼鹭算法的船舶推力分配方法
Chen et al. Numerical simulation of near-surface turn maneuver of a generic submarine
de Arcos et al. A numerical study on the hydrodynamics of a floating tidal rotor under the combined effects of currents and waves
CN107221964A (zh) 一种动力定位海洋平台多发电机组调度方法
Wang et al. CFD investigations of ship propulsion performance at different trim angles
Elkafas Assessment of station keeping capability for dynamically positioned offshore supply vessel
Le Pivert et al. Routing optimisation for towing a floating offshore wind turbine under weather constraints
CN118228514B (zh) 一种浮式海洋平台推进器推力优化分配方法
Lee et al. Hydrodynamic design of SWATH for offshore wind turbine transportation and installation
Piekło et al. Dynamic positioning capability assessment based on optimal thrust allocation
Liu et al. The Numerical Investigation on Hydrodynamic Performance of Twisted Rudder during Self-propulsion
Hao et al. Multiobjective optimization of low-speed sail-assisted VLCC performance
Salem et al. Dynamic positioning system capability analysis for offshore support vessel

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19905136

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 04.11.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19905136

Country of ref document: EP

Kind code of ref document: A1