技术领域technical field
本发明涉及人造肌肉控制领域，尤其涉及一种基于NARMAX模型辨识的人造神经肌肉电驱动控制方法及系统。The invention relates to the field of artificial muscle control, in particular to an artificial neuromuscular electric drive control method and system based on NARMAX model identification.
背景技术Background technique
人造肌肉是模拟生物体运动中的肌肉伸缩实现高效率的装置，可以应用于医疗、生活服务、特种环境等场景中，可以产生力量辅助人类完成工作，其主要根据智能材料通电变换效应完成工作，掌握材料的通电特性并控制通电变换是人造肌肉的关键技术壁垒。且在目前大多数现有的技术中，提出的基本只存在人造肌肉的结构、材料和制备方法，部分发明专利提出人造肌肉控制方法包括：气动、液压和离子电化学驱动方式，并未出现一种通过电驱动控制人造肌肉运动的方法。其中气动、液压驱动方式负载连接复杂，离子电化学驱动方式可能造成驱动力不足的问题。Artificial muscles are high-efficiency devices that simulate muscle expansion and contraction in the movement of living organisms. They can be used in medical treatment, life services, special environments, etc., and can generate power to assist humans in completing work. Mastering the electrification characteristics of materials and controlling the electrification transformation are the key technical barriers of artificial muscles. Moreover, in most of the existing technologies at present, only the structure, materials and preparation methods of artificial muscles are proposed. Some invention patents propose artificial muscle control methods including: pneumatic, hydraulic and ion-electrochemical drive methods, and there is no one. A method to control the movement of artificial muscles through electric actuation. Among them, the load connection of the pneumatic and hydraulic driving methods is complicated, and the ion electrochemical driving method may cause the problem of insufficient driving force.
现有技术中控制人造肌肉的方式包括液压驱动方式、电化学驱动方式。The methods for controlling artificial muscles in the prior art include hydraulic drive and electrochemical drive.
液压驱动方式中，可以以液体为介质，将液体注入不同腔体使之膨胀或收缩，配合阀门流道及液泵，通过设计腔体的结构和位置，控制人造肌肉完成多种运动形式。现有方式虽然可实现较大驱动力，但其内部腔体结构复杂、位置精确，加工、设计程序繁琐，装置采用大量刚性原件，未实现柔性设计，适应性不强。在其结构设计上，TPU软管内的布置结构复杂，固定架等结构与肌肉尺寸有关，增加了内部结构复杂度，当人造肌肉工作时，复杂的结构可能会产生工作误差。In the hydraulic drive mode, the liquid can be used as the medium, and the liquid can be injected into different cavities to make them expand or contract. With the valve channel and liquid pump, the structure and position of the cavity can be designed to control the artificial muscles to complete various forms of movement. Although the existing method can achieve a large driving force, its internal cavity structure is complex, its position is precise, its processing and design procedures are cumbersome, and the device uses a large number of rigid components, which has not realized flexible design and is not strong in adaptability. In terms of its structural design, the layout of the TPU hose is complicated, and the structures such as the fixing frame are related to the size of the muscle, which increases the complexity of the internal structure. When the artificial muscle is working, the complex structure may cause working errors.
电化学驱动方式中，可以应用电活性聚合材料通电产生形变的性质，实现人造肌肉的运动形式，以液态金属作为电极材料，柔性封装材料封装电极和电活性聚合材料，通过通电使电活性聚合材料发生形变。但这种方式在工作前，需要加热浸泡等操作，对离子交换膜去除杂质，交换膜浸泡时间久，操作过程繁琐，还需喷涂电极，对电极起到保护作用，操作过程繁琐，化学制备流程较长。In the electrochemical driving method, the property of electroactive polymer material to generate deformation when electrified can be used to realize the movement form of artificial muscles. Liquid metal is used as the electrode material, and the flexible packaging material encapsulates the electrode and the electroactive polymer material. deformed. However, before this method works, operations such as heating and soaking are required to remove impurities from the ion exchange membrane. The exchange membrane soaks for a long time, and the operation process is cumbersome. It is also necessary to spray the electrodes to protect the electrodes. The operation process is cumbersome and the chemical preparation process longer.
因此，有必要提供一种方案，解决现有的人造肌肉控制技术中存在的结构和控制程序复杂、工作误差大、操作过程繁琐的技术问题。Therefore, it is necessary to provide a solution to solve the technical problems in the existing artificial muscle control technology, such as complex structure and control program, large working error, and cumbersome operation process.
发明内容Contents of the invention
为了解决现有的人造肌肉控制技术中存在的结构和控制程序复杂、工作误差大、操作过程繁琐的技术问题，本发明提供了一种基于NARMAX模型辨识的人造神经肌肉电驱动控制方法及系统。In order to solve the technical problems of complex structure and control program, large working error and complicated operation process in the existing artificial muscle control technology, the present invention provides an artificial neuromuscular electric drive control method and system based on NARMAX model identification.
本发明提供的基于NARMAX模型辨识的人造神经肌肉电驱动控制方法，其中，人造肌肉中的收缩单元包括依次相连的多根记忆合金丝，所述方法包括步骤：The artificial neuromuscular electrical drive control method based on NARMAX model identification provided by the present invention, wherein the contraction unit in the artificial muscle includes a plurality of memory alloy wires connected in sequence, and the method includes the steps of:
构建所述收缩单元的NARMAX模型；Construct the NARMAX model of the contraction unit;
生成所述NARMAX模型的alpha多尺度小波基函数候选表达式；Generate the alpha multiscale wavelet basis function candidate expression of the NARMAX model;
根据所述收缩单元的期待伸缩曲线，从所述alpha多尺度小波基函数候选表达式中选取有效项以建立稀疏模型，计算有效项系数，其中，所述有效项的数量等于所述记忆合金丝的数量；According to the expected contraction curve of the contraction unit, effective items are selected from the candidate expression of the alpha multiscale wavelet basis function to establish a sparse model, and the effective item coefficients are calculated, wherein the number of the effective items is equal to the memory alloy wire quantity;
生成PWM信号，并根据所述有效项和所述有效项系数调节所述PWM信号；generating a PWM signal, and adjusting the PWM signal according to the significant term and the significant term coefficient;
将所述PWM信号施加于所述记忆合金丝以控制所述记忆合金丝形变，使所述收缩单元实现所述期待伸缩曲线。The PWM signal is applied to the memory alloy wire to control the deformation of the memory alloy wire, so that the shrinking unit realizes the expected contraction curve.
本发明巧妙地利用了“记忆合金丝通放电伸缩变化波形近似alpha小波基函数曲线形状”这一特点，将收缩单元的NARMAX模型使用alpha多尺度小波基函数来表达；再根据收缩单元的期待伸缩曲线，从中选取对应于记忆合金丝的有效项，估计有效项系数；将带系数的有效项作为记忆合金丝的伸缩曲线；通过PWM控制技术控制记忆合金丝实现带系数有效项的曲线形状，从而使收缩单元整体实现期待伸缩曲线。当收缩单元的期待伸缩曲线不同时，得到的有效项和有效项系数亦不同。在人造肌肉控制中，只需要选取不同有效项、调节PWM信号、将PWM信号施加于记忆金属丝即可实现人造肌肉控制，控制程序和结构简单、误差小、过程便易，解决了现有的人造肌肉控制技术中存在的结构和控制程序复杂、工作误差大、操作过程繁琐的技术问题。The present invention cleverly utilizes the characteristic of "the shape of the memory alloy wire discharge stretching change waveform approximates the shape of the alpha wavelet basis function curve", expresses the NARMAX model of the contraction unit using the alpha multi-scale wavelet basis function; Curve, from which select the effective item corresponding to the memory alloy wire, estimate the coefficient of the effective item; use the effective item with the coefficient as the expansion and contraction curve of the memory alloy wire; control the memory alloy wire through the PWM control technology to realize the curve shape of the effective item with the coefficient, thereby Make the shrink unit as a whole realize the expected stretch curve. When the expected stretching curves of the shrinkage units are different, the obtained effective term and effective term coefficient are also different. In artificial muscle control, it is only necessary to select different effective items, adjust the PWM signal, and apply the PWM signal to the memory wire to realize artificial muscle control. The control program and structure are simple, the error is small, and the process is easy, which solves the existing problems. There are technical problems in artificial muscle control technology, such as complex structure and control program, large working error, and cumbersome operation process.
进一步地，所述构建所述收缩单元的NARMAX模型的步骤中：根据输入信号和输出信号构建所述NARMAX模型，其中，所述输入信号为alpha小波基函数及其扩展项，所述输出信号为所述收缩单元的伸缩曲线。Further, in the step of constructing the NARMAX model of the contraction unit: constructing the NARMAX model according to the input signal and the output signal, wherein the input signal is an alpha wavelet basis function and its extension, and the output signal is The expansion curve of the contraction unit.
进一步地，采用正交前向回归稀疏算法从所述alpha多尺度小波基函数候选表达式中选取有效项。Further, the orthogonal forward regression sparse algorithm is used to select effective items from the candidate expressions of the alpha multiscale wavelet basis functions.
进一步地，所述生成PWM信号，并根据所述有效项和所述有效项系数调节所述PWM信号的步骤中，调节PWM信号的占空比和输出有效电平的时间。Further, in the step of generating a PWM signal and adjusting the PWM signal according to the effective item and the effective item coefficient, the duty cycle of the PWM signal and the time for outputting an active level are adjusted.
更进一步地，所述生成PWM信号，并根据所述有效项和所述有效项系数调节所述PWM信号的步骤中，通过调节计数模式，调节PWM信号输出有效电平的时间。Furthermore, in the step of generating a PWM signal and adjusting the PWM signal according to the effective item and the effective item coefficient, the time for the PWM signal to output an active level is adjusted by adjusting the counting mode.
此外，本发明还提供一种基于NARMAX模型辨识的人造神经肌肉电驱动控制系统，其中，人造肌肉中的收缩单元包括依次相连的多根记忆合金丝，所述系统包括：In addition, the present invention also provides an artificial neuromuscular electric drive control system based on NARMAX model identification, wherein the contraction unit in the artificial muscle includes a plurality of memory alloy wires connected in sequence, and the system includes:
模型构建模块，用于构建所述收缩单元的NARMAX模型；A model building block for constructing a NARMAX model of the contraction unit;
函数转换模块，用于生成所述NARMAX模型的alpha多尺度小波基函数候选表达式；Function conversion module, for generating the alpha multiscale wavelet basis function candidate expression of described NARMAX model;
有效项选取模块，用于根据所述收缩单元的期待伸缩曲线，从所述alpha多尺度小波基函数候选表达式中选取有效项以建立稀疏模型，计算有效项系数，其中，所述有效项的数量等于所述记忆合金丝的数量；An effective item selection module is used to select an effective item from the alpha multi-scale wavelet basis function candidate expressions according to the expected contraction curve of the contraction unit to establish a sparse model, and calculate an effective item coefficient, wherein the effective item The quantity is equal to the quantity of the memory alloy wire;
PWM调控模块，用于生成PWM信号，并根据所述有效项和所述有效项系数调节所述PWM信号；A PWM regulation module, configured to generate a PWM signal, and adjust the PWM signal according to the effective item and the effective item coefficient;
连接模块，用于将所述PWM信号施加于所述记忆合金丝以控制所述记忆合金丝形变，使所述收缩单元实现所述期待伸缩曲线。The connection module is used to apply the PWM signal to the memory alloy wire to control the deformation of the memory alloy wire, so that the shrinking unit can realize the desired expansion and contraction curve.
进一步地，所述模型构建模块根据输入信号和输出信号构建所述NARMAX模型，其中，所述输入信号为alpha小波基函数及其扩展项，所述输出信号为所述收缩单元的伸缩曲线。Further, the model construction module constructs the NARMAX model according to the input signal and the output signal, wherein the input signal is an alpha wavelet basis function and its extension, and the output signal is the expansion curve of the contraction unit.
进一步地，所述有效项选取模块用于采用正交前向回归稀疏算法从所述alpha多尺度小波基函数候选表达式中选取有效项。Further, the effective item selection module is used to select effective items from the alpha multi-scale wavelet basis function candidate expressions by using an orthogonal forward regression sparse algorithm.
进一步地，所述PWM调控模块用于调节PWM信号的占空比和输出有效电平的时间。Further, the PWM control module is used to adjust the duty cycle of the PWM signal and the time for outputting an effective level.
更进一步地，所述PWM调控模块用于通过调节计数模式，调节PWM信号输出有效电平的时间。Furthermore, the PWM control module is used to adjust the time for the PWM signal to output an active level by adjusting the counting mode.
采用上述技术方案，本发明基于NARMAX模型辨识的人造神经肌肉电驱动控制方法及系统，具有如下有益效果：Adopting the above-mentioned technical scheme, the artificial neuromuscular electric drive control method and system based on NARMAX model identification of the present invention have the following beneficial effects:
(1)本发明提出的人造肌肉是依据Hill肌肉力学模型构建的，其中的收缩单元是人造肌肉的关键，本发明主要对收缩单元进行创新；收缩单元的收缩变化对于人造肌肉起至关重要的作用，通过改变收缩单元自身的长度，调节弹性单元的张力变化，进而实现肌肉力变化，完成人造肌肉工作。(1) The artificial muscle proposed by the present invention is constructed according to the Hill muscle mechanics model, and the contraction unit wherein is the key of the artificial muscle, and the present invention mainly innovates the contraction unit; the contraction change of the contraction unit is crucial for the artificial muscle Function, by changing the length of the contraction unit itself, adjust the tension change of the elastic unit, and then realize the change of muscle force, and complete the artificial muscle work.
(2)本发明采用合金记忆丝作为形变材料，采用PWM技术调节电子开关控制板输出电流，因此可以不借助气、液传动，电化学材料控制人造肌肉工作，不需要过多负载，也不需采用繁琐的化学工艺制备流程，方法简单，可操作性强，可适用范围广泛；(2) The present invention uses alloy memory wire as the deformable material, and uses PWM technology to adjust the output current of the electronic switch control board, so that the artificial muscle can be controlled by electrochemical materials without excessive load and without the aid of gas and liquid transmission. The cumbersome chemical preparation process is adopted, the method is simple, the operability is strong, and the scope of application is wide;
(3)本发明采用NARMAX模型对人造肌肉伸缩曲线建立辨识模型，通过调节输入合金丝电压，根据镍钛合金记忆丝的收缩性质，实现辨识模型项，进而完成人造肌肉收缩单元的收缩工作；(3) The present invention adopts the NARMAX model to establish an identification model for the contraction curve of the artificial muscle, by adjusting the voltage of the input alloy wire, according to the contraction property of the nickel-titanium alloy memory wire, the identification model item is realized, and then the contraction work of the artificial muscle contraction unit is completed;
(4)通过有效实验证明，镍钛合金记忆丝通放电总过程的伸缩曲线近似alpha小波基函数，对合金丝通不同大小、频率的电压，使合金丝伸缩状态不同来实现多尺度alpha小波基函数，作为NARMAX模型辨识的候选项字典，采用OFR算法在字典中进行候选项选择，得到有效项作为各段合金记忆丝的伸缩曲线，此方法效率高，准确性好；(4) Through effective experiments, it is proved that the expansion and contraction curve of the nickel-titanium alloy memory wire through the discharge process approximates the alpha wavelet basis function, and the multi-scale alpha wavelet basis is realized by applying voltages of different sizes and frequencies to the alloy wire to make the alloy wire stretch and contract in different states Function, as a dictionary of candidate items for NARMAX model identification, uses the OFR algorithm to select candidate items in the dictionary, and obtains effective items as the expansion curve of each segment of alloy memory wire. This method has high efficiency and good accuracy;
(5)本发明提出的多尺度alpha小波基函数由两个特征参数确定，根据时不变非线性系统快速辨识方法建模，实现人造肌肉伸缩曲线的高拟合度，减少模型基函数数量，提高模型稀疏性，进一步提高系统辨识性能；(5) The multi-scale alpha wavelet basis function proposed by the present invention is determined by two characteristic parameters, and is modeled according to the time-invariant nonlinear system fast identification method, so as to realize the high degree of fitting of the artificial muscle contraction curve, reduce the number of model basis functions, Improve model sparsity and further improve system identification performance;
(6)本发明提出一种电驱动控制人造肌肉方法，即采用PWM技术调节触发开关控制板模块，对每段合金记忆丝的输入电压调节，实现人造肌肉的工作。(6) The present invention proposes an electric drive control artificial muscle method, that is, the PWM technology is used to adjust the trigger switch control board module, and the input voltage of each alloy memory wire is adjusted to realize the work of the artificial muscle.
附图说明Description of drawings
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.
图1为人造肌肉的结构示意图。Figure 1 is a schematic diagram of the structure of the artificial muscle.
图2为记忆合金丝通电伸缩曲线图。Fig. 2 is a graph showing the expansion and contraction curve of the memory alloy wire.
图3为本发明实施例1的基于NARMAX模型辨识的人造神经肌肉电驱动控制方法的流程图。FIG. 3 is a flow chart of the artificial neuromuscular electrical drive control method based on NARMAX model identification according to Embodiment 1 of the present invention.
图4为本发明实施例2的基于NARMAX模型辨识的人造神经肌肉电驱动控制系统的结构示意图。4 is a schematic structural diagram of an artificial neuromuscular electric drive control system based on NARMAX model identification according to Embodiment 2 of the present invention.
图5为本发明基于NARMAX模型辨识的人造神经肌肉电驱动控制方法及系统的工作原理示意图。5 is a schematic diagram of the working principle of the artificial neuromuscular electrical drive control method and system based on NARMAX model identification in the present invention.
图6为本发明本发明实施例2的基于NARMAX模型辨识的人造神经肌肉电驱动控制系统中的部分结构关系示意图。FIG. 6 is a schematic diagram of partial structural relationships in the artificial neuromuscular electric drive control system based on NARMAX model identification according to Embodiment 2 of the present invention.
图7为本发明基于NARMAX模型辨识的人造神经肌肉电驱动控制方法及系统的工作阶段示意图。FIG. 7 is a schematic view of the working stages of the artificial neuromuscular electrical drive control method and system based on NARMAX model identification of the present invention.
图8为仿真算例中人造肌肉收缩单元的伸缩曲线放大图。Fig. 8 is an enlarged view of the expansion and contraction curve of the artificial muscle contraction unit in the simulation example.
图9为仿真算例中的肌肉工作曲线图。Figure 9 is the muscle work curve in the simulation example.
图10为仿真算例中的辨识结果曲线图。Figure 10 is a graph of the identification results in the simulation example.
图11为仿真算例中肌肉工作曲线和辨识结果曲线图的对比图。Fig. 11 is a comparison chart of the muscle working curve and the identification result curve in the simulation example.
图12为仿真算例中8个有效项的曲线图。Fig. 12 is a graph of 8 effective items in the simulation example.
具体实施方式Detailed ways
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述。显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本发明保护的范围。The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Apparently, the described embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.
为了解决现有的人造肌肉控制技术中存在的结构和控制程序复杂、工作误差大、操作过程繁琐的技术问题，本发明提出了一种基于NARMAX模型辨识的人造神经肌肉电驱动控制方法及系统。In order to solve the technical problems in the existing artificial muscle control technology, such as complex structure and control program, large working error, and cumbersome operation process, the present invention proposes an artificial neuromuscular electric drive control method and system based on NARMAX model identification.
本发明中的人造肌肉系统基于Hill三元肌肉结构力学模型，Hill三元肌肉结构力学模型主要包括收缩单元1和弹性单元部分。The artificial muscle system in the present invention is based on the Hill ternary muscle structure mechanical model, and the Hill ternary muscle structure mechanical model mainly includes a contraction unit 1 and an elastic unit part.
如图1所示，肌肉收缩单元1，主要由由记忆合金丝10构成，记忆合金丝10记忆合金丝10此收缩单元1主要为肌肉收缩提供张力，提供的张力大小和记忆合金丝10的数量有关；弹性单元装置分为串联弹性单元2(SE)和并联肌肉单元(PE)，其中串联弹性单元2(SE)与肌肉收缩单元1装置串联，起到的作用是肌肉的固有弹性；并联肌肉单元(PE)与肌肉收缩单元1装置和串联弹性元并联，表现松弛下的肌肉力学性质。在Hill模型中，收缩单元1、串联弹性单元2、并联弹性单元3的张力分别为FCE、FSE、FPE，肌肉总张力为FM，收缩单元1、串联弹性单元2、并联弹性单元3的拉伸长度分别为LCE、LSE、LPE，肌肉总拉伸长度为LM，满足：As shown in Figure 1, the muscle contraction unit 1 is mainly composed of memory alloy wires 10, the memory alloy wires 10 memory alloy wires 10. This contraction unit 1 mainly provides tension for muscle contraction, the tension provided and the quantity of memory alloy wires 10 Relevant; the elastic unit device is divided into a series elastic unit 2 (SE) and a parallel muscle unit (PE), wherein the series elastic unit 2 (SE) is connected in series with the muscle contraction unit 1, and the role played is the inherent elasticity of the muscle; the parallel muscle The unit (PE) is connected in parallel with the muscle contraction unit 1 device and the series elastic unit to express the mechanical properties of the muscle under relaxation. In the Hill model, the tensions of contraction unit 1, series elastic unit 2, and parallel elastic unit 3 are FCE , FSE , FPE , and the total muscle tension is FM . Contraction unit 1, series elastic unit 2, and parallel elastic unit The stretching lengths of 3 are LCE , LSE , LPE respectively, and the total stretching length of the muscle is LM , satisfying:
收缩单元1通过改变LCE，使总张力改变，实现人造肌肉工作模式。The contraction unit 1 changes the total tension by changing LCE to realize the working mode of the artificial muscle.
通过NARMAX模型建模辨识出所需的模型项作为记忆合金丝10的收缩曲线，采用PWM控制输入电压和时间调节各段记忆合金丝10的收缩。The required model item is identified as the shrinkage curve of the memory alloy wire 10 through NARMAX model modeling, and PWM is used to control the input voltage and time to adjust the shrinkage of each segment of the memory alloy wire 10 .
通过有效实验证明：一条记忆合金丝10通放电伸缩变化波形近似alpha小波基函数曲线形状，如图2所示。使用PWM(脉冲宽度调制)控制计数模式的调制方法，调节输入记忆合金丝10的输入电压和输入时间，将多段记忆合金丝10依次连接，使用不同的PWM调节电子开关控制板分别对各记忆合金丝10控制，实现人造肌肉收缩单元1的变化。Through effective experiments, it is proved that: a memory alloy wire 10-pass discharge expansion and contraction change waveform approximates the shape of the alpha wavelet basis function curve, as shown in Figure 2. Use PWM (Pulse Width Modulation) to control the modulation method of the counting mode, adjust the input voltage and input time of the input memory alloy wire 10, connect the multi-section memory alloy wire 10 in sequence, and use different PWM to adjust the electronic switch control board to control each memory alloy respectively. The wire 10 is controlled to realize the change of the artificial muscle contraction unit 1 .
本发明提供的基于NARMAX模型辨识的人造神经肌肉电驱动控制方法，其中，人造肌肉中的收缩单元1包括依次相连的多根记忆合金丝10，方法包括步骤：The artificial neuromuscular electrical drive control method based on NARMAX model identification provided by the present invention, wherein the contraction unit 1 in the artificial muscle includes a plurality of memory alloy wires 10 connected in sequence, and the method includes steps:
步骤S101：构建收缩单元1的NARMAX模型；Step S101: constructing a NARMAX model of shrinkage unit 1;
步骤S102：生成NARMAX模型的alpha多尺度小波基函数候选表达式；Step S102: Generate alpha multiscale wavelet basis function candidate expressions of the NARMAX model;
步骤S103：根据收缩单元1的期待伸缩曲线，从alpha多尺度小波基函数候选表达式中选取有效项以建立稀疏模型，计算有效项系数，其中，有效项的数量等于记忆合金丝10的数量；Step S103: According to the expected contraction curve of the contraction unit 1, select effective items from the alpha multiscale wavelet basis function candidate expressions to establish a sparse model, and calculate the effective item coefficients, wherein the number of effective items is equal to the number of memory alloy wires 10;
步骤S104：生成PWM信号，并根据有效项和有效项系数调节PWM信号；Step S104: generating a PWM signal, and adjusting the PWM signal according to the effective item and the effective item coefficient;
步骤S105：将PWM信号施加于记忆合金丝10以控制记忆合金丝10形变，使收缩单元1实现期待伸缩曲线。Step S105 : applying a PWM signal to the memory alloy wire 10 to control the deformation of the memory alloy wire 10 , so that the contraction unit 1 achieves a desired contraction curve.
本发明巧妙地利用了“记忆合金丝10通放电伸缩变化波形近似alpha小波基函数曲线形状”这一特点，将收缩单元1的NARMAX模型使用alpha多尺度小波基函数来表达；再根据收缩单元1的期待伸缩曲线，从中选取对应于记忆合金丝10的有效项，估计有效项系数；将带系数的有效项作为记忆合金丝10的伸缩曲线；通过PWM控制技术控制记忆合金丝10实现带系数有效项的曲线形状，从而使收缩单元1整体实现期待伸缩曲线。当收缩单元1的期待伸缩曲线不同时，得到的有效项和有效项系数亦不同。在人造肌肉控制中，只需要选取不同有效项、调节PWM信号、将PWM信号施加于记忆金属丝即可实现人造肌肉控制，控制程序和结构简单、误差小、过程便易，解决了现有的人造肌肉控制技术中存在的结构和控制程序复杂、工作误差大、操作过程繁琐的技术问题。The present invention skillfully utilizes the characteristic of "the shape of the stretching change waveform of memory alloy wire 10 through discharge is similar to the shape of the alpha wavelet basis function curve", and expresses the NARMAX model of the contraction unit 1 using the alpha multi-scale wavelet basis function; then according to the contraction unit 1 The expected expansion and contraction curve of the memory alloy wire 10 is selected therefrom, and the effective item coefficient is estimated; the effective item with the coefficient is used as the expansion and contraction curve of the memory alloy wire 10; the memory alloy wire 10 is controlled by the PWM control technology to realize the effective item with the coefficient The curve shape of the item, so that the contraction unit 1 as a whole realizes the expected stretching curve. When the expected expansion and contraction curves of the contraction unit 1 are different, the obtained effective term and effective term coefficient are also different. In artificial muscle control, it is only necessary to select different effective items, adjust the PWM signal, and apply the PWM signal to the memory wire to realize artificial muscle control. The control program and structure are simple, the error is small, and the process is easy, which solves the existing problems. There are technical problems in artificial muscle control technology, such as complex structure and control program, large working error, and cumbersome operation process.
步骤S102中，使用alpha多尺度小波基函数来表达NARMAX模型。In step S102, the alpha multiscale wavelet basis function is used to express the NARMAX model.
步骤S103中，有效项的数量S1等于收缩单元1中记忆合金丝10的数量S2，选取的多个有效项一一对应于多个记忆合金丝10。In step S103 , the number S1 of valid items is equal to the number S2 of memory alloy wires 10 in the shrinking unit 1 , and the selected valid items correspond to a plurality of memory alloy wires 10 one by one.
步骤S104和S105中，由于一条记忆合金丝10通放电伸缩变化波形近似alpha小波基函数曲线形状，因此可以通过控制记忆合金丝10通放电来控制记忆合金丝10的伸缩变化。本实施例1中，向记忆合金丝10施加调节后的PWM信号，该PWM信号能使记忆合金丝10形变至带系数有效项的曲线形状。每一记忆合金丝10实现对应的曲线形状，从而使收缩单元1整体实现期待的伸缩曲线。In steps S104 and S105, since the discharge stretching change waveform of a memory alloy wire 10 approximates the shape of the alpha wavelet basis function curve, the stretch change of the memory alloy wire 10 can be controlled by controlling the discharge of the memory alloy wire 10 . In the first embodiment, the adjusted PWM signal is applied to the memory alloy wire 10, and the PWM signal can deform the memory alloy wire 10 to a curve shape with an effective term of the coefficient. Each memory alloy wire 10 realizes a corresponding curve shape, so that the contraction unit 1 as a whole realizes an expected stretching curve.
进一步地，构建收缩单元1的NARMAX模型的步骤中：根据输入信号和输出信号构建NARMAX模型，其中，输入信号为alpha小波基函数及其扩展项，输出信号为收缩单元1的伸缩曲线。Further, in the step of constructing the NARMAX model of the contracting unit 1: constructing the NARMAX model according to the input signal and the output signal, wherein the input signal is the alpha wavelet basis function and its extension, and the output signal is the stretching curve of the contracting unit 1.
进一步地，采用正交前向回归稀疏算法从alpha多尺度小波基函数候选表达式中选取有效项。Furthermore, the orthogonal forward regression sparse algorithm is used to select effective items from the candidate expressions of alpha multiscale wavelet basis functions.
有效项的数量S1等于收缩单元1中记忆合金丝10的数量S2。实施例1中，得到alpha多尺度小波基函数候选表达式后，使用正交前向回归算法(Orthogonal ForwardRegression algorithm，OFR)从中选取S2个有效项；S2个有效项分别代表S1个记忆合金丝10的伸缩曲线函数。使用正交前向回归算法建立稀疏模型为本领域的现有技术。The number S1 of valid items is equal to the number S2 of memory alloy wires 10 in the contraction unit 1 . In embodiment 1, after obtaining alpha multiscale wavelet basis function candidate expressions, use orthogonal forward regression algorithm (Orthogonal ForwardRegression algorithm, OFR) therefrom to select S2 valid items; S2 valid items represent S1 memory alloy wires 10 respectively stretch curve function. It is the prior art in this field to use the orthogonal forward regression algorithm to establish the sparse model.
进一步地，步骤S104中，调节PWM信号的占空比和输出有效电平的时间。Further, in step S104, the duty cycle of the PWM signal and the time for outputting an active level are adjusted.
更进一步地，步骤S104中，通过调节计数模式，调节PWM信号输出有效电平的时间。Furthermore, in step S104, the time for the PWM signal to output an active level is adjusted by adjusting the counting mode.
此外，本发明还提供一种基于NARMAX模型辨识的人造神经肌肉电驱动控制系统，其中，人造肌肉中的收缩单元1包括依次相连的多根记忆合金丝10，系统包括模型构建模块41、函数转换模块42、有效项选取模块43、PWM调控模块44、连接模块45；In addition, the present invention also provides an artificial neuromuscular electric drive control system based on NARMAX model identification, wherein the contraction unit 1 in the artificial muscle includes a plurality of memory alloy wires 10 connected in sequence, and the system includes a model building module 41, a function conversion Module 42, effective item selection module 43, PWM control module 44, connection module 45;
模型构建模块41用于构建收缩单元1的NARMAX模型；The model construction module 41 is used to construct the NARMAX model of the contraction unit 1;
函数转换模块42用于生成NARMAX模型的alpha多尺度小波基函数候选表达式；The function conversion module 42 is used to generate the alpha multiscale wavelet basis function candidate expressions of the NARMAX model;
有效项选取模块43用于根据收缩单元1的期待伸缩曲线，从alpha多尺度小波基函数候选表达式中选取有效项以建立稀疏模型，计算有效项系数，其中，有效项的数量等于记忆合金丝10的数量；The effective item selection module 43 is used to select effective items from the alpha multi-scale wavelet basis function candidate expressions according to the expected expansion curve of the contraction unit 1 to establish a sparse model, and calculate the effective item coefficients, wherein the number of effective items is equal to the memory alloy wire the number of 10;
PWM调控模块44用于生成PWM信号，并根据有效项和有效项系数调节PWM信号；PWM control module 44 is used for generating PWM signal, and regulates PWM signal according to effective item and effective item coefficient;
连接模块45用于将PWM信号施加于记忆合金丝10以控制记忆合金丝10形变，使收缩单元1实现期待伸缩曲线。The connection module 45 is used to apply the PWM signal to the memory alloy wire 10 to control the deformation of the memory alloy wire 10 , so that the contraction unit 1 can achieve a desired expansion and contraction curve.
实施例2中，多根记忆合金丝10依次相连，形成收缩单元1；记忆合金丝10之间通过固定连接夹连接，即固定连接夹将通不同电压的记忆合金丝10相连接；有效项选取模块43根据收缩单元1的期待伸缩曲线，选取与记忆合金丝10数量相同的有效项；记忆合金丝10的数量根据实际情况选择，较佳地，为了达到更好的控制效果，记忆合金丝10为八条；记忆合金丝10的为镍钛合金记忆丝。本发明中记忆合金丝10的数量不限于八条。In Embodiment 2, a plurality of memory alloy wires 10 are connected in sequence to form a contraction unit 1; the memory alloy wires 10 are connected by a fixed connection clip, that is, the fixed connection clip connects memory alloy wires 10 with different voltages; effective item selection Module 43 selects the same effective item as the number of memory alloy wires 10 according to the expected expansion and contraction curve of shrinkage unit 1; the number of memory alloy wires 10 is selected according to the actual situation. Eight; the memory alloy wire 10 is a nickel-titanium alloy memory wire. The number of memory alloy wires 10 in the present invention is not limited to eight.
如图1和图6所示，人造神经肌肉包括：肌肉收缩单元1、串联弹性单元2和并联肌肉单元；肌肉收缩单元1包括镍钛合金记忆丝若干、固定连接夹若干、电压调控模块(PWM调控模块44)，固定连接夹将通不同电压的镍钛合金记忆丝相连接；如图6所示，电压调控模块包括电源、PWM调节电子开关控制板若干，用于调节若干镍钛合金记忆丝的输入电压，使其达到NARMAX模型辨识后的模型项曲线；若干不同收缩程度的镍钛合金记忆丝叠加，即达到人造肌肉收缩单元1的工作状态。如图2所示，是镍钛合金通电收缩曲线。As shown in Figure 1 and Figure 6, the artificial neuromuscular includes: a muscle contraction unit 1, a series elastic unit 2 and a parallel muscle unit; the muscle contraction unit 1 includes several nickel-titanium alloy memory wires, some fixed connection clips, Regulatory module 44), the fixed connection clip will be connected with the nickel-titanium alloy memory wires of different voltages; As shown in Figure 6, the voltage regulation module includes power supply, PWM regulation electronic switch control boards, and is used to regulate some nickel-titanium alloy memory wires The input voltage makes it reach the model term curve after NARMAX model identification; a number of nickel-titanium alloy memory wires with different contraction degrees are superimposed to reach the working state of the artificial muscle contraction unit 1. As shown in Figure 2, it is the contraction curve of nickel-titanium alloy.
进一步地，模型构建模块41根据输入信号和输出信号构建NARMAX模型，其中，输入信号为alpha小波基函数及其扩展项，输出信号为收缩单元1的伸缩曲线。Further, the model construction module 41 constructs the NARMAX model according to the input signal and the output signal, wherein the input signal is the alpha wavelet basis function and its extended term, and the output signal is the stretching curve of the contraction unit 1 .
进一步地，有效项选取模块43用于采用正交前向回归稀疏算法从alpha多尺度小波基函数候选表达式中选取有效项。Further, the effective item selection module 43 is used to select effective items from the alpha multiscale wavelet basis function candidate expressions by using an orthogonal forward regression sparse algorithm.
进一步地，PWM调控模块44用于调节PWM信号的占空比和输出有效电平的时间。Further, the PWM regulation module 44 is used to adjust the duty ratio of the PWM signal and the time for outputting an effective level.
更进一步地，PWM调控模块44用于通过调节计数模式，调节PWM信号输出有效电平的时间。Furthermore, the PWM regulation module 44 is used to adjust the time for the PWM signal to output an active level by adjusting the counting mode.
下面结合具体原理对本发明进行说明。The present invention will be described below in conjunction with specific principles.
结合图7所示，本发明可分为两个阶段，包括时不变非线性伸缩曲线快速辨识阶段(NARMAX模型辨识阶段)和PWM控制调制阶段(电驱动控制阶段)。As shown in FIG. 7 , the present invention can be divided into two stages, including a time-invariant nonlinear stretch curve rapid identification stage (NARMAX model identification stage) and a PWM control modulation stage (electric drive control stage).
(一)时不变非线性伸缩曲线快速辨识阶段(1) Fast identification stage of time-invariant nonlinear stretching curve
此阶段主要分为两个具体步骤，将输入信号和系统输出信号，构建时不变非线性系统模型；采用OFR算法对alpha多尺度小波基函数进行有效项选择，建立稀疏模型，将多段合金记忆丝通放电伸缩变化曲线作为有效项。This stage is mainly divided into two specific steps. The input signal and the system output signal are used to construct the time-invariant nonlinear system model; the OFR algorithm is used to select the effective items of the alpha multi-scale wavelet basis function, and the sparse model is established, and the multi-segment alloy memory The expansion and contraction curve of wire discharge is taken as an effective item.
alpha小波基函数的表达式如下：The expression of the alpha wavelet basis function is as follows:
其中，in,
(a)构建时不变非线性系统模型(a) Constructing a time-invariant nonlinear system model
时不变非线性系统的NARMAX模型表达式如下：The NARMAX model expression of the time-invariant nonlinear system is as follows:
其中，y(t)表示辨识系统输出，u(t)表示系统输入，e(t)表示系统噪声，f(·)表示非线性函数，ny、nu、ne表示延迟阶次。Among them, y(t) represents the identification system output, u(t) represents the system input, e(t) represents the system noise, f(·) represents the nonlinear function,ny ,nu , nee represent the delay order.
对于时不变非线性模型可通过如下线性回归函数表示：For the time-invariant nonlinear model, it can be expressed by the following linear regression function:
其中，为系统输入、输出回归项的非线性组合向量，θi为相应的回归参数，M为候选项的个数，N为离散数据点的个数。in, is the nonlinear combination vector of system input and output regression items, θi is the corresponding regression parameter, M is the number of candidate items, and N is the number of discrete data points.
式(2)可以表示为：Formula (2) can be expressed as:
其中in
(b)采用经典OFR算法对模型进行有效项的选择：首先将矩阵正交分解表示为：(b) Use the classic OFR algorithm to select effective items for the model: firstly, the The matrix orthogonal decomposition is expressed as:
其中A为上三角阵，W为正交阵：Where A is an upper triangular matrix and W is an orthogonal matrix:
且W满足：And W satisfies:
D中的元素di满足：The elements di in D satisfy:
正交矩阵元素中的w1,w2,…,wM向量构成的空间与向量构成的空间完全一致，由此式(3)可表示为：The space formed by w1 ,w2 ,…,wM vectors in the orthogonal matrix elements and The space formed by the vectors is exactly the same, so formula (3) can be expressed as:
式中模型参数向量κ转化为辅助参数向量g＝[g1,g2,…,gM]T满足如下关系：In the formula, the model parameter vector κ is transformed into an auxiliary parameter vector g=[g1 ,g2 ,…,gM ]T satisfies the following relationship:
g＝D-1WTY-D-1WTε (12)g＝D-1 WT YD-1 WT ε (12)
g的估计量为The estimator of g is
通过模型参数向量κ与辅助参数向量g的关系式Aκ＝g得出：Through the relationship Aκ=g between the model parameter vector κ and the auxiliary parameter vector g, it can be obtained:
通过误差减小率准则(Error Reduction Ratio,ERR)依次挑选模型有效项，ERR表达式如下：The effective items of the model are selected sequentially through the error reduction rate criterion (Error Reduction Ratio, ERR), and the expression of ERR is as follows:
其中，Y为系统输出序列，wi为正交化的候选项序列，M为候选项数，为辅助参数的估计值。Among them, Y is the system output sequence, wi is the orthogonalized candidate sequence, M is the number of candidates, is the estimated value of the auxiliary parameter.
在有效项的选取过程中，每一步都通过比较该步骤中正交化后候选项序列对应的ERRi值确定该步的有效项，而后每一步候选项都与前一步的所有有效项进行Schmidt正交化后，计算ERR值，确定该步的有效性，依次进行，具体步骤如下：In the selection process of effective items, each step determines the effective items of this step by comparing the ERRi value corresponding to the sequence of candidates after orthogonalization in this step, and then the candidates of each step are compared with all valid items of the previous step for Schmidt After orthogonalization, calculate the ERR value, determine the validity of this step, and proceed in sequence. The specific steps are as follows:
Step1：确定第一个有效项，对候选项计算各项辅助参数和ERR值：Step1: Determine the first valid item, for the candidate Calculate various auxiliary parameters and ERR values:
Step k：确定第k个有效项，对其余候选项依次与前k-1个有效项依次进行Schmidt正交化，计算各项辅助参数和ERR值：Step k: Determine the kth effective item, perform Schmidt orthogonalization on the remaining candidates and the first k-1 effective items in turn, and calculate various auxiliary parameters and ERR values:
选取有效项作为人造肌肉的时不变非线性伸缩曲线辨识模型的模型结构。The effective item is selected as the model structure of the time-invariant nonlinear stretch curve identification model of artificial muscle.
(二)PWM控制调制阶段(2) PWM control modulation stage
此阶段目的通过调节PWM的计数模式，控制合金记忆丝两端的电压和输入时间，使各段记忆丝产生不同的收缩变化，以此作为辨识模型的输入。生成辨识模型后的alpha多尺度小波基函数的有效项，主要分为两个具体步骤，通过PWM技术调节占空比，以此实现各段合金记忆丝分别通入不同电压，使合金记忆丝通放电形成不同伸缩变换尺度下的基函数曲线；通过调节计数模式，调节PWM输出有效电平的时间，使合金记忆丝伸缩程度形成不同平移变换尺度下的基函数曲线。The purpose of this stage is to control the voltage and input time at both ends of the alloy memory wire by adjusting the counting mode of the PWM, so that each segment of the memory wire produces different shrinkage changes, which are used as the input of the identification model. The effective items of the alpha multiscale wavelet basis function after the identification model is generated are mainly divided into two specific steps, and the duty cycle is adjusted through PWM technology, so as to realize that each segment of the alloy memory wire is connected to different voltages, so that the alloy memory wire can pass through The discharge forms the basis function curves under different stretching transformation scales; by adjusting the counting mode, the time for PWM to output the effective level is adjusted, so that the stretching degree of the alloy memory wire forms the basis function curves under different translation transformation scales.
下面基于仿真算例和具体数据验证本发明提出的基于NARMAX模型辨识的电驱动控制的人造神经肌肉，若人造肌肉收缩单元1的收缩曲线是正弦变化的，如图8所示，为人造肌肉收缩单元1的工作收缩曲线，选用图2的alpha小波基函数及其扩展项作为模型候选项。The following is based on simulation examples and specific data to verify the artificial neuromuscular based on NARMAX model identification proposed by the present invention. If the contraction curve of the artificial muscle contraction unit 1 changes sinusoidally, as shown in FIG. 8, it is artificial muscle contraction For the working contraction curve of unit 1, the alpha wavelet basis function and its extensions in Figure 2 are selected as model candidates.
在仿真算例中，人造肌肉工作曲线的样本序列长度N＝1024，模型输入为alpha小波基函数及其扩展项u(t)、u(t-n)，输入项的序列长度M＝1024，所以产生了2047个扩展项u(t),…,u(t-2047)。在其仿真示例中，用8段合金记忆丝连接完成人造肌肉工作，只需在模型输入项中挑选8项即可。根据图8所示的人造肌肉收缩单元1的工作收缩曲线，并利用模型输入项产生的辨识结果如图9至图12所示，模型参数辨识结果列示在表1中。In the simulation example, the sample sequence length of the artificial muscle working curve is N=1024, the model input is the alpha wavelet basis function and its extended items u(t), u(t-n), and the sequence length of the input item is M=1024, so the generated 2047 expansion items u(t),...,u(t-2047) are obtained. In its simulation example, 8 sections of alloy memory wire are used to complete the artificial muscle work, and only 8 items need to be selected in the model input items. According to the working contraction curve of the artificial muscle contraction unit 1 shown in FIG. 8 , the identification results generated by using the model input items are shown in FIGS. 9 to 12 , and the identification results of the model parameters are listed in Table 1.
表1仿真算例模型参数辨识结果Table 1 Simulation results of model parameter identification
图12显示了仿真算例中8个有效项的曲线图，包括第一曲线61、第二曲线62、第三曲线63、第四曲线64、第五曲线65、第六曲线66、第七曲线67和第八曲线68。Fig. 12 shows the graphs of 8 effective items in the simulation example, including the first curve 61, the second curve 62, the third curve 63, the fourth curve 64, the fifth curve 65, the sixth curve 66, and the seventh curve 67 and the eighth curve 68 .
仿真算例的均方误差为0.1002，ERR总和为99.9803％，利用本发明提出的NARMAX模型辨识方法得到的模型输入项和系数，将其作为每段合金记忆丝的形变曲线；通过调节PWM的占空比和计数模式，实现各段合金记忆丝的收缩，完成人造肌肉的工作模式。The mean square error of the simulation calculation example is 0.1002, and the ERR sum is 99.9803%. The model input items and coefficients obtained by the NARMAX model identification method proposed by the present invention are used as the deformation curve of each section of alloy memory wire; Empty ratio and counting mode realize the contraction of each section of alloy memory wire and complete the working mode of artificial muscle.
人造肌肉控制中，通过控制肌肉结构模型中的收缩单元1变化，实现人造肌肉工作模式。在控制收缩单元1的过程中，将收缩单元1作为一个人造肌肉子系统，该子系统由镍钛合金记忆丝、电源、PWM调节电子开关控制板构成，将合金记忆丝多段相连接作为一个整体收缩单元1；通过PWM产生脉冲控制信号，控制收缩单元1中的分段合金丝收缩，通过调节每段合金丝产生相变收缩的时间实现合金丝组长度控制。基于Hill三元肌肉结构力学模型，产生需要的肌肉力。In the artificial muscle control, the working mode of the artificial muscle is realized by controlling the change of the contraction unit 1 in the muscle structure model. In the process of controlling the contraction unit 1, the contraction unit 1 is used as an artificial muscle subsystem, which is composed of a nickel-titanium alloy memory wire, a power supply, and a PWM adjustment electronic switch control board, and the alloy memory wire is connected in multiple sections as a whole Shrinking unit 1: Generate a pulse control signal through PWM to control the shrinkage of segmented alloy wires in the shrinking unit 1, and realize the length control of the alloy wire group by adjusting the time for each segment of alloy wire to produce phase transition shrinkage. Based on the Hill ternary muscle structure mechanics model, the required muscle force is generated.
本发明基于NARMAX模型辨识方法，通过对整体收缩单元1的收缩曲线进行非线性自回归滑动平均(NARMAX)模型建模，采用经典正交前向回归稀疏(OFR)算法对展开的时不变参数模型进行有效项选择，通过调节PWM电子开关控制板的占空比和计数模式，调节各段输入镍钛合金记忆丝的通电电压和时间，使其产生任意长度的形变量和收缩变化，实现收缩单元1的变化。The present invention is based on the NARMAX model identification method, by performing nonlinear autoregressive moving average (NARMAX) model modeling on the shrinkage curve of the overall shrinkage unit 1, and adopting the classic orthogonal forward regression sparse (OFR) algorithm to expand the time-invariant parameters The model selects effective items. By adjusting the duty cycle and counting mode of the PWM electronic switch control board, the energized voltage and time of each segment input to the nickel-titanium alloy memory wire are adjusted to produce deformation and shrinkage changes of any length to achieve shrinkage. Unit 1 Variations.
通过有效实验证明：一条合金记忆丝通放电伸缩变化波形近似alpha小波基函数曲线，可以将其作为模型项选择，通过脉冲宽度调制(PWM)技术，控制PWM的计数模式，来调节多段合金记忆丝通放电时间和电压大小，使合金记忆丝实现与OFR算法选中的有效项相同的伸缩状态。Through effective experiments, it is proved that: a stretching change waveform of an alloy memory wire through discharge approximates the alpha wavelet basis function curve, which can be selected as a model item, and the multi-segment alloy memory wire can be adjusted by controlling the PWM counting mode through pulse width modulation (PWM) technology Through the discharge time and voltage, the alloy memory wire can achieve the same expansion and contraction state as the effective item selected by the OFR algorithm.
本发明中的人造肌肉系统是依据Hill肌肉力学模型构建的，其中的收缩单元1是人造肌肉的关键，本发明主要对收缩单元1的收缩变化进行创新；收缩单元1的收缩变化对于人造肌肉起至关重要的作用，通过改变收缩单元1自身的长度，调节各弹性单元的张力变化，进而实现肌肉力变化，完成人造肌肉工作。The artificial muscle system in the present invention is constructed according to the Hill muscle mechanics model, wherein the contraction unit 1 is the key of the artificial muscle, and the present invention mainly innovates the contraction change of the contraction unit 1; the contraction change of the contraction unit 1 plays an important role in the artificial muscle The crucial role is to adjust the tension of each elastic unit by changing the length of the contraction unit 1 itself, thereby realizing the change of muscle force and completing the work of the artificial muscle.
人造肌肉系统采用采用合金记忆丝作为形变材料，如20cm长、0.3mm直径的镍钛合金丝，通电完成收缩，断电延展回复。经有效实验证明：合金记忆丝通放电总过程的伸缩曲线近似alpha小波基函数曲线，且通不同大小、频率的电压对应小波基函数尺度不同，将多段合金记忆丝利用PWM技术通过触发开关控制板模块输入脉冲信号，调节至合金记忆丝最适形变电压，作为驱动人造肌肉运动的基础。The artificial muscle system uses alloy memory wire as the deformable material, such as a nickel-titanium alloy wire with a length of 20cm and a diameter of 0.3mm. Effective experiments have proved that the expansion and contraction curve of the alloy memory wire through the discharge process is similar to the alpha wavelet basis function curve, and the voltages of different sizes and frequencies correspond to different wavelet basis function scales. The multi-segment alloy memory wire uses PWM technology to pass through the trigger switch control panel The module inputs the pulse signal and adjusts to the optimum deformation voltage of the alloy memory wire, which serves as the basis for driving the artificial muscle movement.
人造肌肉系统包括：肌肉收缩单元1装置、弹性单元装置；肌肉收缩单元1装置包括镍钛合金记忆丝若干、固定连接夹若干、电压调控系统，固定连接夹将通不同电压的镍钛合金记忆丝相连接；电压调控模块包括电源、PWM调节电子开关控制板若干，用于调节若干镍钛合金记忆丝的输入电压，使其达到NARMAX模型辨识后的模型项曲线；若干不同收缩程度的镍钛合金记忆丝叠加，即达到人造肌肉收缩单元1的工作状态，实现人造肌肉总张力变化。The artificial muscle system includes: a muscle contraction unit 1 device and an elastic unit device; the muscle contraction unit 1 device includes a number of nickel-titanium alloy memory wires, a number of fixed connection clips, and a voltage regulation system. The fixed connection clips will pass through nickel-titanium alloy memory wires of different voltages. The voltage regulation module includes a power supply and a number of PWM regulating electronic switch control boards, which are used to adjust the input voltage of several nickel-titanium alloy memory wires to make it reach the model term curve after NARMAX model identification; several nickel-titanium alloys with different shrinkage degrees The memory filaments are superimposed, that is, the working state of the artificial muscle contraction unit 1 is reached, and the total tension of the artificial muscle can be changed.
人造肌肉总张力的变化的通过总体伸缩曲线表示，通过调节收缩单元1的收缩变化曲线，实现对串、并联弹性单元3张力调节；对收缩单元1调节采用n根合金记忆丝，将收缩单元1的伸缩曲线通过NARMAX模型构建辨识系统，采用正交前向回归算法，对不同尺度的alpha小波基函数挑选出有效项，通过控制输入电流和时相延时构建有效项的形变曲线。The change of the total tension of the artificial muscle is represented by the overall expansion curve. By adjusting the contraction change curve of the contraction unit 1, the tension adjustment of the series and parallel elastic units 3 is realized; the adjustment of the contraction unit 1 uses n alloy memory wires, and the contraction unit 1 The expansion and contraction curve of the NARMAX model is used to build the identification system, and the orthogonal forward regression algorithm is used to select effective items for the alpha wavelet basis functions of different scales, and the deformation curve of the effective items is constructed by controlling the input current and time delay.
本发明的有益效果在于：The beneficial effects of the present invention are:
(1)本发明提出的人造肌肉是依据Hill肌肉力学模型构建的，其中的收缩单元1是人造肌肉的关键，本发明主要对收缩单元1进行创新；收缩单元1的收缩变化对于人造肌肉起至关重要的作用，通过改变收缩单元1自身的长度，调节弹性单元的张力变化，进而实现肌肉力变化，完成人造肌肉工作；(1) The artificial muscle proposed by the present invention is constructed according to the Hill muscle mechanics model, wherein the contraction unit 1 is the key of the artificial muscle, and the present invention mainly innovates the contraction unit 1; the contraction variation of the contraction unit 1 plays an important role for the artificial muscle It plays an important role, by changing the length of the contraction unit 1 itself, adjusting the tension change of the elastic unit, and then realizing the change of muscle force, and completing the artificial muscle work;
(2)本发明采用合金记忆丝作为形变材料，采用PWM技术调节电子开关控制板输出电流，因此可以不借助气、液传动，电化学材料控制人造肌肉工作，不需要过多负载，也不需采用繁琐的化学工艺制备流程，方法简单，可操作性强，可适用范围广泛；(2) The present invention uses alloy memory wire as the deformable material, and uses PWM technology to adjust the output current of the electronic switch control board, so that the artificial muscle can be controlled by electrochemical materials without excessive load and without the aid of gas and liquid transmission. The cumbersome chemical preparation process is adopted, the method is simple, the operability is strong, and the scope of application is wide;
(3)本发明采用NARMAX模型对人造肌肉伸缩曲线建立辨识模型，通过调节输入合金丝电压，根据镍钛合金记忆丝的收缩性质，实现辨识模型项，进而完成人造肌肉收缩单元1的收缩工作；(3) The present invention adopts the NARMAX model to establish an identification model for the contraction curve of the artificial muscle, by adjusting the voltage of the input alloy wire, according to the contraction properties of the nickel-titanium alloy memory wire, the identification model item is realized, and then the contraction work of the artificial muscle contraction unit 1 is completed;
(4)通过有效实验证明，镍钛合金记忆丝通放电总过程的伸缩曲线近似alpha小波基函数，对合金丝通不同大小、频率的电压，使合金丝伸缩状态不同来实现多尺度alpha小波基函数，作为NARMAX模型辨识的候选项字典Ω，采用OFR算法在字典中进行候选项选择，得到有效项作为各段合金记忆丝的伸缩曲线，此方法效率高，准确性好；(4) Through effective experiments, it is proved that the expansion and contraction curve of the nickel-titanium alloy memory wire through the discharge process approximates the alpha wavelet basis function, and the multi-scale alpha wavelet basis is realized by applying voltages of different sizes and frequencies to the alloy wire to make the alloy wire stretch and contract in different states Function, as the candidate item dictionary Ω for NARMAX model identification, the OFR algorithm is used to select the candidate items in the dictionary, and the effective items are obtained as the expansion curve of each segment of the alloy memory wire. This method has high efficiency and good accuracy;
(5)本发明提出的多尺度alpha小波基函数由两个特征参数确定，根据时不变非线性系统快速辨识方法建模，实现人造肌肉伸缩曲线的高拟合度，减少模型基函数数量，提高模型稀疏性，进一步提高系统辨识性能；(5) The multi-scale alpha wavelet basis function proposed by the present invention is determined by two characteristic parameters, and is modeled according to the time-invariant nonlinear system fast identification method, so as to realize the high degree of fitting of the artificial muscle contraction curve, reduce the number of model basis functions, Improve model sparsity and further improve system identification performance;
(6)本发明提出一种电驱动控制人造肌肉方法，即采用PWM技术调节触发开关控制板模块，对每段合金记忆丝的输入电压调节，实现人造肌肉的工作。(6) The present invention proposes an electric drive control artificial muscle method, that is, the PWM technology is used to adjust the trigger switch control board module, and the input voltage of each alloy memory wire is adjusted to realize the work of the artificial muscle.
以上仅为本发明的较佳实施例，并不用以限制本发明，凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。The above are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention Inside.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910764266.6ACN110471284B (en) | 2019-08-19 | 2019-08-19 | Artificial neuromuscular electric drive control method and system based on NARMAX model identification |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910764266.6ACN110471284B (en) | 2019-08-19 | 2019-08-19 | Artificial neuromuscular electric drive control method and system based on NARMAX model identification |
| Publication Number | Publication Date |
|---|---|
| CN110471284Atrue CN110471284A (en) | 2019-11-19 |
| CN110471284B CN110471284B (en) | 2024-08-20 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910764266.6AActiveCN110471284B (en) | 2019-08-19 | 2019-08-19 | Artificial neuromuscular electric drive control method and system based on NARMAX model identification |
| Country | Link |
|---|---|
| CN (1) | CN110471284B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111897212A (en)* | 2020-06-09 | 2020-11-06 | 吉林大学 | Multi-model joint modeling method for magnetron shape memory alloy actuators |
| CN115348882A (en)* | 2019-12-02 | 2022-11-15 | 通用医疗公司 | Systems and methods for patient positioning during radiation therapy |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5092901A (en)* | 1990-06-06 | 1992-03-03 | The Royal Institution For The Advancement Of Learning (Mcgill University) | Shape memory alloy fibers having rapid twitch response |
| US6243624B1 (en)* | 1999-03-19 | 2001-06-05 | Northwestern University | Non-Linear muscle-like compliant controller |
| WO2011098056A1 (en)* | 2010-02-09 | 2011-08-18 | Friedrich-Schiller-Universität Jena | Device for simulating the motion behavior of a natural muscle |
| CN107967395A (en)* | 2017-12-11 | 2018-04-27 | 北京航空航天大学 | A fast identification method for time-varying nonlinear systems based on beta wavelet basis function expansion |
| CN109598175A (en)* | 2017-09-30 | 2019-04-09 | 北京航空航天大学 | It is a kind of based on before multi-wavelet bases function and transothogonal to the Time-Frequency Analysis Method of recurrence |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5092901A (en)* | 1990-06-06 | 1992-03-03 | The Royal Institution For The Advancement Of Learning (Mcgill University) | Shape memory alloy fibers having rapid twitch response |
| US6243624B1 (en)* | 1999-03-19 | 2001-06-05 | Northwestern University | Non-Linear muscle-like compliant controller |
| WO2011098056A1 (en)* | 2010-02-09 | 2011-08-18 | Friedrich-Schiller-Universität Jena | Device for simulating the motion behavior of a natural muscle |
| CN109598175A (en)* | 2017-09-30 | 2019-04-09 | 北京航空航天大学 | It is a kind of based on before multi-wavelet bases function and transothogonal to the Time-Frequency Analysis Method of recurrence |
| CN107967395A (en)* | 2017-12-11 | 2018-04-27 | 北京航空航天大学 | A fast identification method for time-varying nonlinear systems based on beta wavelet basis function expansion |
| Title |
|---|
| S. A. BILLINGS 等: "The wavelet-NARMAX representation: A hybrid model structure combining polynomial models with multiresolution wavelet decompositions", 《INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE》* |
| S. A. BILLINGS 等: "The wavelet-NARMAX representation: A hybrid model structure combining polynomial models with multiresolution wavelet decompositions", 《INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE》, vol. 36, no. 3, 20 February 2005 (2005-02-20), pages 137 - 152* |
| VAHID HASSANI 等: "A survey on hysteresis modeling, identification and control", 《MECHANICAL SYSTEMS AND SIGNAL PROCESSING》* |
| VAHID HASSANI 等: "A survey on hysteresis modeling, identification and control", 《MECHANICAL SYSTEMS AND SIGNAL PROCESSING》, no. 49, 13 May 2014 (2014-05-13), pages 209 - 233* |
| 孙宝元: "《现代执行器技术》", 30 April 2003, 吉林大学出版社, pages: 257 - 245* |
| 应申舜 等: "基于形状记忆合金弹簧阵列的人工肌肉设计与研究", 《中国机械工程》* |
| 应申舜 等: "基于形状记忆合金弹簧阵列的人工肌肉设计与研究", 《中国机械工程》, vol. 19, no. 15, 15 August 2008 (2008-08-15), pages 1782 - 1786* |
| 张琛 等: "《微执行器》", 31 December 2005, 上海交通大学出版社, pages: 211 - 239* |
| 杨宜良: "《新型驱动器及其应用》", 28 February 1998, 机械工业出版社, pages: 137 - 144* |
| 杨彬: "形状记忆合金的行为特性测试与模型辨识", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》* |
| 杨彬: "形状记忆合金的行为特性测试与模型辨识", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》, no. 02, 15 February 2017 (2017-02-15), pages 1 - 74* |
| 杨鹏: "磁控形状记忆合金执行器非线性建模及滑模控制方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》* |
| 杨鹏: "磁控形状记忆合金执行器非线性建模及滑模控制方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》, no. 09, 15 September 2016 (2016-09-15), pages 1 - 32* |
| 殷跃红: "《骨骼肌力产生机理、仿生及应用》", 31 August 2017, 国防工业出版社, pages: 243* |
| 毛晨曦 等: "《形状记忆合金监测与控制智能系统》", 30 June 2007, 东北林业大学出版社, pages: 17* |
| 王扬威 等: "形状记忆合金丝驱动的柔性机械臂建模与实验", 《浙江大学学报（工学版）》* |
| 王扬威 等: "形状记忆合金丝驱动的柔性机械臂建模与实验", 《浙江大学学报（工学版）》, vol. 52, no. 4, 30 April 2018 (2018-04-30), pages 628 - 634* |
| 谢长珍: "a尺度正交的多尺度函数的构造", 《工程数学学报》* |
| 谢长珍: "a尺度正交的多尺度函数的构造", 《工程数学学报》, vol. 20, no. 2, 31 May 2003 (2003-05-31), pages 113 - 117* |
| 邱子辉: "基于形状记忆合金的复合材料驱动研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》* |
| 邱子辉: "基于形状记忆合金的复合材料驱动研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》, no. 07, 15 July 2012 (2012-07-15), pages 1 - 71* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115348882A (en)* | 2019-12-02 | 2022-11-15 | 通用医疗公司 | Systems and methods for patient positioning during radiation therapy |
| US12329987B2 (en) | 2019-12-02 | 2025-06-17 | The General Hospital Corporation | Systems and methods for patient positioning during radiotherapy |
| CN111897212A (en)* | 2020-06-09 | 2020-11-06 | 吉林大学 | Multi-model joint modeling method for magnetron shape memory alloy actuators |
| CN111897212B (en)* | 2020-06-09 | 2022-05-31 | 吉林大学 | Multi-model combined modeling method of magnetic control shape memory alloy actuator |
| Publication number | Publication date |
|---|---|
| CN110471284B (en) | 2024-08-20 |
| Publication | Publication Date | Title |
|---|---|---|
| JP6609072B2 (en) | EAP actuator and driving method | |
| CN110471284A (en) | Artificial neural's muscle electric drive control method and system based on NARMAX Model Distinguish | |
| CN109617405B (en) | DC/DC converter modeling method based on harmonic state space | |
| US11322677B2 (en) | Actuator device based on an electroactive polymer | |
| CN117318480A (en) | Modulation strategy design method and system of DC-DC converter based on reinforcement learning | |
| CN114970335A (en) | Asymmetric rate-dependent hysteresis modeling method based on play operator | |
| CN102096747A (en) | Method and device for simulating power electronic system | |
| CN116760289B (en) | A discrete sliding mode control algorithm for DCDC buck converter based on genetic algorithm | |
| Illenberger et al. | The integrated self priming circuit: an autonomous electrostatic energy harvester with voltage boosting | |
| CN116700424A (en) | Dynamic thermoelectric maximum power tracking method based on improved flying mouse searching algorithm | |
| CN105871210A (en) | Finite time passive control method for Buck converter | |
| CN113541170B (en) | Grid-connected inversion control method and system for emergency power supply of fuel cell | |
| CN103226655B (en) | Be suitable for the improvement EMTP algorithm of PWM converter averaging model | |
| JP7173868B2 (en) | EAP actuator and driving method | |
| CN111208863B (en) | Method and device for fast tracking control of maximum power point of photovoltaic cells | |
| Zhang et al. | Design of a Multi-connection Pneumatic Artificial Muscle | |
| CN111446855A (en) | Boost DC-DC converter with multiple basic units | |
| CN116565960A (en) | Method and system for inverter grid-connected control based on full artificial neural network | |
| CN109728743A (en) | A new method for electromechanical response estimation of electroactive polymers | |
| CN107272815A (en) | A kind of unimodal MPPT methods of photovoltaic based on SQP Trust Region Algorithms | |
| Morava | On formal groups and geometric quantization | |
| Hamdi et al. | Design of bio-inspired muscle sarcomere structure using a hybrid hydrogel-IPMC actuator | |
| CN118143955B (en) | Foot track planning method for foot robot based on pneumatic artificial muscle driving | |
| Zhang et al. | Characterisation, modelling and hysteresis compensation in a tubular dielectric elasomer actuator | |
| CN115967301B (en) | Ultralow hysteresis switch type piezoelectric ceramic independent driving method integrating feedforward correction |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |