# Sixth Lichuang Electric Competition# Glass fiber reinforced plastic defect imaging inspection system based on RT-Thread and TouchGFX

* 1. Introduction to project functions

        The detection system takes on-site defect detection and health assessment of in-service FRP storage tanks, pipelines and towers in the petrochemical industry as a typical application scenario. It utilizes the characteristics of the coplanar capacitive quasi-static edge electric field in capacitive imaging technology that is sensitive to internal defects in non-metallic materials, and can effectively Detect superficial cladding damage of glass fiber reinforced materials ( GFRP ) and its composite materials, such as fiber delamination and bubbling, internal cracking and leakage, glue layer debonding and other typical forms. During the detection process, it has the characteristics of no contact with the surface to be tested and no coupling agent required.

        The system uses STM32 microcontroller as the core and has wireless network communication capabilities; the operation adopts dual control methods of touch and independent buttons, which can effectively prevent accidental touches. A complete signal processing system is designed for defect detection needs, including signal generators and acquisition cards, transimpedance amplifiers, quadrature lock-in amplifiers and supporting power systems, which greatly improves the completeness and integration of the system. Compared with traditional Capacitive imaging equipment has achieved smaller size and lighter weight. The system implements the real-time display function of defect signals based on the TouchGFX graphics framework. It adopts dual display of amplitude and phase information, which enriches the information sources and reduces the defect misjudgment rate. The lightweight cloud server is built based on the Flask framework to implement two-dimensional and three-dimensional rendering functions of defect images. The defect data is automatically uploaded to the cloud server after preliminary processing by the system. After verification by the server, it is sent to the target browser, where it is normalized on the browser side and rendered on the web page in real time. The web page can be logged in remotely to realize expert collaboration, batch and time-sharing online consultation functions. Data is automatically backed up on the server side, supports remote downloading, and ensures that defective data is reusable and traceable.

*2. Project attributes

        The main members of the team were approved by the Department of Higher Education of the Ministry of Education for the Industry-Academic Cooperation STMicroelectronics STM32 Student Innovation and Entrepreneurship Fund in 2019;

        This project is a derivative work of the above-mentioned fund , original, completely open source, and made public for the first time; it has not won any awards in other competitions and has not participated in any defense.

*3. Open source agreement

      CC-BY-NC-SA 3.0

*4. Hardware part

  4.1 Overall design of hardware system

        The hardware system is divided into a master system and a slave system based on target functions and business logic. The master system is based on the STM32F7 microcontroller, and the STM32F7 microcontroller core board and F7 core board backplane are designed around human-computer interaction requirements ; the slave system is based on the STM32G0 microcontroller. , around the sensor signal processing requirements, a signal generator, transimpedance amplifier and lock-in amplifier were designed. The main system implements human-machine interaction and slave system control functions. Its core business is to control the status of the slave system based on user input, receive data returned from the slave system, and visualize it to the front end. The WIFI module serves as a channel for uploading data to the cloud and also has a system time correction function. The slave system receives the main system instructions and is responsible for controlling the DDS to generate the target excitation signal and sampling the analog voltage signal input by the lock-in amplifier. For occasions where external synchronous trigger sampling is required, it also has a synchronous sampling function.

        The master and slave systems are independent of each other and communicate through UART . The slave system is a passive system, which only receives instructions from the master system, performs corresponding operations according to the instructions, and does not actively communicate with the master system. The system hardware structure and its internal information exchange method are shown in Figure 4-1 .

Figure 4-1 Master-slave system hardware structure diagram

  4.2 Main system hardware design

  1) Main system core board design

The core board is designed using the high-performance controller STM32F767IG launched by         STMicroelectronics ( ST ) . This microcontroller uses the ARM Cortex-M7 high-performance core. The on-chip Flash capacity is up to 1MB , the RAM capacity is up to 512kB , and the maximum operating frequency is 216MHz .

        Considering that the designed TouchGFX interface has rich elements and has high requirements on system Flash and memory resources, a 32MB large-capacity DRAM memory W9825G6KH is mounted as an external RAM through an external memory controller ( FMC ) ; at the same time, a 32MB large-capacity Flash is mounted as an external RAM. External Flash . The display uses a 7 -inch LCD screen driven by LTDC in RGB565 mode and connected using a 0.5MM-40P clamshell FPC connector. The core board system structure is shown in Figure 4-2 .

Figure 4-2 Core board system structure diagram

        微控制器工作电压3.3V，显示屏输入电压5V。为简化供电设计，使用BTB连接器从底板获取5V供电，经BTB连接器向显示屏供电。微控制器工作所需的3.3V电压经ST低压差线性稳压器LD1117S33TR从5V电压中获得。LD1117S33TR最大输出电流800mA，在输出电流为800mA时，压降仅为1.2V，纹波抑制达到75dB（120Hz），可部分抑制底板DC-DC电源的高频噪声。如图4-3所示。

图4-3  核心板PCB实物及原理图

  2）主系统底板设计

        底板设计有WIFI模块ESP-12S，两个串口通信接口，一个独立按键接口。板上电源经由BTB连接器对核心板供电，核心板IO经由BTB连接器与底板各模块相连。底板简化结构如图4-4所示。

图4-4 底板简化结构图

        考虑WIFI模块ESP-12S的上电电流较大（电源供电流能力应大于500mA）。为减小核心板3.3V供电压力，减小LDO发热，在底板上单独设置一颗ST LD1117S33TR低压差线性稳压器为底板各器件供电。同时考虑底板二次开发时可能有使用5V电源的需求，单独使用一颗ST L7805CDT低压差线性稳压器产生5V电压。由于核心板上显示屏接口兼顾显示屏供电，而显示屏采用5V供电，功耗在1.5 2W。此功耗较大，已不适用于单颗LDO供电方案。使用降压型DC-DC功率变换器，从供电电池处取得7.4-12.V电压，经BUCK变换为5V。该电压即为核心板5V电压唯一来源。如图4-5所示。

图4-5 底板PCB实物及原理图

  3）主系统电源设计

       主系统采用单电源供电，电池输入电压范围7.4-13.2V。配电方式综合考虑了核心板和底板的供电需求，底板从电池取电，核心板从底板取电，触摸屏从核心板取电。电源结构简图如图4-6所示。

图4-6 主系统电源结构简图

        底板配有三套不同的供电系统，板上低压差线性稳压器LDO1与LDO2分别为底板器件提供3.3V和5V电压。而DC-DC变换器通过BTB连接器为核心板供电，输送功率最大可达10W。核心板从BTB连接器取得5V供电，经FPC连接器给触摸屏供电。核心板上STM32F7系统经低压差线性稳压器LDO取得3.3V电能供给。

  4.3 从系统硬件设计

        从系统接收并处理主系统下达的指令，同时采集探头数据并回传给主系统。从系统主要由STM32G0最小系统、信号发生器、跨阻放大器和锁相放大器构成。最小系统使用串口接收主系统指令，通过SPI控制DDS芯片AD9833产生电压信号，并经片内ADC对锁相放大器输出的直流电压进行采样。系统结构简图如图4-7所示。

图4-7 从系统结构简图

  1）STM32G0最小系统与信号发生器设计

        最小系统采用意法半导体公司（ST）推出的新系列微控制器STM32G071CB，其内部集成多通道12位ADC，满足对两路电压信号采样的需求。信号发生器由DDS芯片AD9833和电压放大电路组成。由于DDS芯片输出电压范围在38mV-650mV，电压幅值过低，不满足直接驱动探头的需求，因此需要使用运算放大器对其进行适当放大。考虑激励信号频率10kHz，输出电压幅值不应小于1V，此处将意法半导体公司（ST）通用运算放大器TSV912配置为同相运算放大器，放大倍数最高可达6倍。系统通过MCU的SPI接口控制DDS芯片产生目标信号，同时经采样电路采集锁相器输出的电压信号。由于被采样电压信号为交流信号，而MCU的ADC为单电源转换器，需要通过外部电路将交流信号转换为可采样的直流信号。使用外部同步功能时，同步电路将外部同步信号转化为适配板上数字电路的脉冲信号并送微控制器IO触发同步采样，信号转化功能通过光电耦合器实现。

图4-8 信号发生器与采样模块

  2）跨阻放大器与正交锁相放大器设计

      跨阻放大器（TIA）作为信号调理电路的一部分，直接与外部探头相连，可以将探头输出的微弱电流信号转换为电压信号，并将其放大至目标摆幅，输入下一级锁相电路。跨阻放大器的两个基本特征是：1、输入偏置电流极小；2、输入阻抗极大。此处使用意法半导体公司（ST）通用运算放大器TSV912，该放大器的输入偏置电流为1pA，带宽为8MHz。跨阻放大器PCB实物如图4-9。

图4-9 跨阻放大器

        正交锁相器的两路输入分别与跨阻放大器输出信号和探头激励信号相接，利用上述方法对这两路信号做乘法运算，再经低通滤波后输出探头信号的实部和虚部，输送至信号采集模块进行数字化。由于跨阻放大模块采用单电源供电，为充分利用输出电压范围，对输出信号施加了2.5V的直流偏置电压。然而锁相器采用双电源供电，输入相敏检波器的电压不能包含直流分量，需在前端设计阻容耦合器。正交锁相器PCB实物如图4-10。

图4-10 正交锁相放大器

*5、软件部分

  5.1 主系统软件设计

        主系统使用RT-Thread实时操作系统与ST TouchGFX图形框架。主机部分使用MCU型号为STM32F767IGT6，使用外部器件包括：RGB接口的7寸1024*600分辨率显示器，型号为FT5426的触摸芯片，型号为ESP8266的Wifi模块，型号为W9825G6KH的SDRAM，型号为W25Q256的QSPI Flash。使用RT-Thread创建七个线程以及构建虚拟文件系统，进程如图5-1所示。主系统软件总体结构如图5-2所示。

图5-1 RT-Thread进程

图5-2 主系统软件总体结构

        自定义了数据传输模式，所有的数据传输都以 结尾，主系统向探头发送[C]0001[E]表示请求探头基准电压，向探头发送[C]0002[E]表示请求探头背景电压，向探头发送[C]0003[E]表示请求接收探头数据（包括幅值与相位信息），向探头发送[C]0004[E]表示请求接收探头数据（只有幅值信息）。探头向主系统发送[D1]XXX[E]表示探头基准电压值，向主系统发送[D2]XXX[E]表示探头背景电压值，向主系统发送[D3]XXX[E]表示探头采集到的幅值，向主系统发送[D4]XXX[E]表示探头采集到的相位信息。

        屏幕的触摸芯片为FT5426，IIC接口。为了增强程序的鲁棒性与可移植性，触摸程序分为软件IIC驱动、FT5426驱动、Touch触摸接口三部分。当有触摸动作时，FT5426会产生脉冲沿，MCU接收到外部中断触发后进行一次采样。由于FT5426触摸的灵敏性，触摸一次会出现四次或以上的中断信号，在每次采样后比较结果，如果不同则同步到Touch触摸接口。Touch触摸接口预留了触摸数据的全局变量，供TouchGFX使用。

        RT-Thread中对物联网开发做了相应的支持，比如AT组件，集成了AT指令；SAL组件， AT Socket 接口的抽象，实现标准 BSD Socket API；netdev 组件，用于抽象和管理 AT 设备生成的网卡设备相关信息，提供 ping、ifconfig、netstat 等网络命令；AT Device 软件包，针对不同设备的 AT Socket 移植和示例文件，以软件包的形式给出。移植NTP同步网络时间，需要开启本地RTC，选择RTC时钟为外部时钟，开启RTC后，在开机时连接wifi，将同步的NTP网络时间更新到RTC上，后面的系统时间由RTC负责。

        由于单片机性能的特性，TouchGFX进程的优先级不能过高，否则会影响底层驱动的运行，优先级也不能过低，否则会使刷新变慢，导致帧数下降出现卡顿现象。设置TouchGFX进程优先级为15，栈大小为20k，完全满足TouchGFX运行条件。在TouchGFX Designer中创建六个界面。分别是屏保界面、桌面界面、标定界面、检测一界面，检测二界面、系统信息界面。

        屏保界面：当处于其他界面，三分钟没有操作时，就会返回到屏保界面。桌面界面：桌面上放置了两个应用程序，一个是检测程序，另一个是系统信息程序。标定界面：标定界面显示了基准电压值与背景电压值的大小，每秒刷新三次。有四个按钮，分别为获取基准电压，获取背景电压，进入检测一界面，进入检测二界面。当点击获取基准电压按钮时，会向探头发送相应命令，探头会发送基准电压，并且同步到显示上，背景电压的同步也是如此。如图5-3所示。

图5-3 桌面与标定界面

        检测一界面：从标定界面进入，此界面可绘制探头传输的容值与相位信息，容值的绘图范围为探头背景电压上下20%波动。此界面包括四个按钮，分别是：开始、暂停、停止与复位。绘图时分为两个状态，运行态与暂停态。初始状态为暂停态，当点击开始按钮后，进入运行态，在handleTickEvent中绘制图形，每秒绘制6个点，无论容值与相位信息是否发生改变都会绘图。点击暂停按钮，进入暂停态，暂停绘图，点击开始按钮可恢复绘图。点击停止按钮，进入暂停态，并将绘图清空。点击复位按钮，绘图清空并进入运行态，进行绘图。检测二界面：从标定界面进入，此界面绘制探头容值，与检测界面一不同的是，此界面绘图时容值的范围时随着容值而变化的，即当有更大的容值，就同步绘图的最高值设置，同理最小值设置也是如此。

图5-4 检测界面

        系统信息界面：此界面用来展示TouchGFX设计的时钟表盘，在每次进入handleTickEvent时，都会同步本地RTC时间到时钟表盘，即每秒钟刷新表盘。

  5.2 从系统软件设计

        从系统主要包含指令接收与处理、DDS控制、信号采样与外部同步等功能。其中指令接收与处理程序是从系统与外部环境交互的唯一接口。DDS控制程序驱动AD9833芯片产生目标频率的正弦信号。信号采样程序负责目标信号的量化与滤波。外部同步功能通过外部中断捕获同步信号，经由信号采样程序实现同步采样。信号发生器输出信号在主函数while(){}循环前设置，信号频率与波形在程序中一次性设定，系统上电后不可更改。指令接收与处理程序从主系统接收指令，对指令内容进行判断，激活对应的数据发送形式，并调用信号采集程序获取采样信号。在使用外部信号同步采样时，同步机制通知指令接收与处理程序采样，实现同步功能，如图5-3所示。

图5-3 从系统软件结构简化框图

The signal generator is driven directly         using the open source driver chili-DDS developed by the team. chiliDDS is a set of drivers implemented in C language and adapted to different DDS . The driver does not depend on specific underlying hardware, and only needs to implement a few very simple functions to run when transplanting. Here, the STM32G071 peripheral SPI2 is used as the AD9833 driver interface, and the host-only sending mode is used. Taking AD9833 as an example, copy the driver files chilis9833.c and chilis9833.h to the software project; complete the two macros FSYNC_HIGH and FSYNC_LOW in chilis9833.c , and implement the chilis9833_SPI_Init() and chilis9833_SPI_Transmit_2Bytes() functions to complete the transplant.

       Before transmitting object surface defect data to the server, the client needs to establish a communication connection with the server. Since the client sends transient data packets, in order to save client and server resources, they use HTTP POST to transmit data. The client first sends the "Start" command to the server through the routing port opened by the server, such as " http://ip_address:5555/data_up" , informing the server that data is about to be sent to it. After the server receives the command packet, it changes to Listening state, collecting defective data packets sent by the client. After receiving the data, the server sends the "Finished" command to the client to end this communication. Subsequently, the server will perform preliminary processing on the received data packets, including format conversion and normalization, to prepare for data visualization.

*6. System testing and evaluation

        The system supports two working modes: manual scanning and external synchronous scanning. In manual scanning mode, the operator holds the probe to quickly detect defects on the target material. In the external synchronous scanning mode, the S- shaped surface is scanned through the gantry. After the scanning is completed, the data is transmitted to the cloud server, and the two-dimensional and three-dimensional cloud images of the defects can be directly viewed on the web page.

  6.1 Manual scan test

        Select the FRP specimen shown in Figure 6-1 for manual scanning testing. The specimen is 20mm thick, and three round holes with a diameter of 10mm are prefabricated on it . Among them, hole A is a through hole, hole B and hole C are blind holes, hole B has a buried depth of 1mm , and hole C has a buried depth of 2mm . Use a handheld probe to perform line scanning on the above three defects, and the detection results are shown in Figure 6-2 .

Figure 6-1 FRP sample

Figure 6-2 FRP sample test results

        The upper part of the detection interface is the amplitude waveform, and the lower part is the phase waveform. At defect A , the amplitude change is the largest, about 60mV ; at defect B , the amplitude change is about 30mV ; at defect C , the amplitude change is about 15mV . At the three defect positions, the phase is disturbed, and the range of the disturbance is positively correlated with the amplitude change. Tests show that this system with the probe shown has good detection capabilities for volumetric defects.

  6.2 Automatic scanning test

       Select the FRP sample shown in Figure 6-3 . The sample is 2mm thick, and an equilateral triangular through hole with a side length of 30mm is prefabricated on it . The probe shown in Figure 3.3.13 is held on the bench to perform surface scanning on the through hole, and 25 coordinates are divided in the X and Y directions respectively , with a single step length of 2mm , and a total of 625 original point data are obtained. The original point data is processed by the program and interpolated into an imaging array with 50 coordinates in the X and Y directions respectively, and a total of 2500 points. The imaging array is formatted and uploaded to the server, and then displayed on the web page after rendering. Its two-dimensional and three-dimensional defect imaging are shown in Figure 6-4 .

Figure 6-3 FRP triangular defect specimen

Figure 6-4 Two-dimensional and three-dimensional imaging effects of defects

        The left side is a two-dimensional defect cloud image with clear defect boundaries, which more accurately reflects the true shape and damage area of ​​the defect. The right side shows the three-dimensional imaging effect. There is an obvious height drop at the defect, which is displayed as a dark blue depression. The vertical drop in the depression is large and the bottom surface is flat, which better reflects the depth information of the defect.

*7. Work summary

        This work is a portable FRP defect detection system. The work is based on the safety issues of FRP storage tanks, pipes, and towers in industrial environments. The system adopts coplanar capacitance non-destructive testing technology and utilizes the characteristic of coplanar capacitance quasi-static fringe electric field in capacitance imaging technology that is sensitive to internal defects of non-metallic materials. It can effectively detect superficial cladding of fiber-reinforced plastics ( FRP ) and its composite materials. Damage, during the inspection implementation process, has the characteristics of no contact with the surface to be tested and no need for couplant coupling.

       The instrument has two main innovations. One is the two-dimensional and three-dimensional defect inversion function based on the cloud server. The defect data is automatically uploaded to the cloud server after preliminary processing by the system. After verification by the server, it is sent to the target browser, where it is normalized on the browser side and rendered on the web page in real time. The web page can be logged in remotely to realize expert collaboration, batch and time-sharing online consultation functions. Data is automatically backed up on the server side, supports remote downloading, and ensures that defective data is reusable and traceable. The second is the real-time display function of feedback signal waveform based on TouchGFX , which uses dual display of amplitude and phase to provide more available information and achieve higher accuracy.

        The system uses STMicroelectronics' STM32F7 high-performance microcontroller as the core, has network connection capabilities, is equipped with a 7- inch touch screen, and is operated through the TouchGFX graphical interface. The hardware system is divided into a master system and a slave system based on target functions and business logic. The master system is based on the STM32F7 microcontroller, and the STM32F7 microcontroller core board and F7 core board backplane are designed around human-computer interaction requirements ; the slave system is based on the STM32G0 microcontroller. , around the sensor signal processing requirements, a signal generator, transimpedance amplifier and lock-in amplifier were designed. Experimental results show that this work can better detect superficial defects in fiberglass composite materials and has broad application prospects.

*8. BOM list (summary of main components)

*9. Contest LOGO verification