Industrial Equipment Edge Intelligence Solutions

吃了一鲸

Industrial Equipment Edge Intelligence Solutions [Copy link]

Industrial Equipment Edge Intelligence Solutions

Author: I ate a whale

1. Introduction

In recent years, IoT devices have been widely used in industrial sites. However, the parameters collected by traditional industrial equipment damage monitoring systems are relatively simple, lack intelligent analysis methods, and are difficult to adapt to complex and diverse industrial site environments. Therefore, the purpose of this solution is to use multi-source data to intelligently analyze the working status of equipment. First, the three-axis vibration acceleration (BHI160), geomagnetism (BMM150), environmental sound signal (INMP522) and temperature and humidity data (BME680) of the equipment are collected when it is working, and transmitted to the intelligent gateway through low-power Bluetooth RSL10, and the health status of industrial equipment is monitored and analyzed online using multi-sensor analysis methods. The innovation of this solution is: 1) using multi-sensor data fusion methods, extracting the features of multi-source signals to construct anomaly detection models, and distinguishing equipment damage; 2 ) deploying statistical machine learning methods to the mobile phone side , comparing the current data with the normal data distribution in the past, and checking whether the equipment is working normally, so as to adapt to the heterogeneity of various devices (equipment inherent parameters, environmental factors interference). In the future, this solution can be applied to industrial equipment status monitoring in scenarios such as rail transit equipment and machine tool working status, and can also be applied to IoT scenarios such as smart buildings and smart homes.

2. System Block Diagram

The system architecture diagram is shown above. Sensor nodes can be attached to common industrial equipment, mainly collecting vibration acceleration, environmental noise, temperature and humidity generated by the equipment when it is working. Through the anomaly detection algorithm deployed in the mobile phone, it can autonomously compare the difference between the current data distribution and the historical data distribution (assuming that the equipment has been working normally in the past). Once it is found that the data generated by the current device is significantly different from the historical data, an anomaly is reported.

3. Description of some functions

3.1 RSL10 Bluetooth 5.0

As shown in Figure 4.1, RSL10 integrates rich resources and is an ultra-low power chip that uses ARM Cortex-M3 processor and LPDSP32 DSP core, supports Bluetooth low energy technology and 2.4GHz proprietary protocol. It contains DC/DC power management module, AES128 encryption engine, DMA, 88kB data storage RAM, 384kB Flash, sampling converter, SPI master-slave bus, I2C, PWM, UART, Timers and other resources. It is powerful and has shockingly low power consumption, which makes wearable and battery-powered applications possible.

Figure 4.1 RSL10 internal resources

3.2 BHI160 sensor

As shown in Figure 4.2, the BHI160 sensor is a low-power, high-sensitivity 3-axis accelerometer and 3-axis gyroscope. It uses the I2C bus to connect to RSL10, with RSL10 as the master and BHI160 as the slave. It can also support external pin interrupt triggering, with simple physical connection and very resource-saving. The sensor can obtain gravity, linear acceleration, direction angle and angular rate, and the I2C rate is 400kHz.

Figure 4.2 BHI160 sensor external hardware circuit design

3.3 BMM150 sensor

As shown in Figure 4.3, BMM150 is a low-power, low-noise 3 -axis digital geomagnetic sensor. It uses the I2C bus to connect to the BHI160 sensor. BMM150 is the slave and BHI160 is the master. It can support external pin interrupt triggering and its application depends on the BHI160 sensor. Therefore, to use BMM150, you need to enable BHI160 and turn on the main I2C bus.

Figure 4.3 BMM150 external hardware circuit design diagram

3.4 BME680 Sensor

As shown in Figure 4.4, the BME680 sensor is a high-precision gas, pressure, humidity and temperature sensor that can obtain air quality, temperature and humidity, carbon dioxide concentration and air pressure data. It uses the I2C bus drive.

Figure 4.4 BME680 external hardware circuit design diagram

3.5 INMP522 Voice Module

INMP522 is an ultra-low noise digital microphone. The most powerful feature is that its external hardware circuit design is super simple, as shown in Figure 4.5. Through debugging, its low noise and high sensitivity characteristics have been verified, as shown in Figure 4.6.

Figure 4.5 INMP522 external hardware circuit design diagram

Figure 4.6 Voice debugging

The software part mainly includes low-level embedded software and upper-level application software.

Figure 4.7 Embedded software flow chart

First, after all sensors are initialized, a Bluetooth connection is established and data is transmitted to the upper layer application via Bluetooth. The data packet format of environmental data is: the first byte is the packet sequence, and the next 35 bytes are the direction, acceleration, angle, geomagnetism, temperature, humidity, air pressure and other data. The data packet format of sound data is: the first byte is the packet sequence, and the next 100 bytes are sound data, and each 4 bytes is a point data.

The host computer is developed using WeChat applet, which can parse and store the data sent by the sensor nodes. Since WeChat applet is a weakly typed language based on JavaScript, DataView is used in the parsing process to parse data of different byte lengths and data of different sizes; Echarts is used for curve drawing, which is very suitable for dynamic drawing, and the chart interface is relatively friendly. In addition, currently only conventional data distribution inspection methods are deployed in WeChat applet : by analyzing the multi-dimensional characteristics of multi-sensor data in the current sliding window, including the maximum value, minimum value, mean, and frequency inspection to determine whether a large offset occurs. However, in actual industrial scenarios, abnormal events will be more complicated, and we will consider integrating more methods into mobile terminals in the future, and even directly integrating some low-complexity but very effective methods into the lower-level hardware to monitor the health status of the equipment in real time.

4. Source Code

环境声信号采集源代码.zip (13.48 MB, downloads: 1)

环境多传感器数据采集源代码.zip (13.85 MB, downloads: 1)

5. Demonstration video of the work’s functions

Industrial Equipment Edge Intelligent Solutions-Industrial Equipment Edge Intelligent Solutions-EEWORLD University

6. Project Summary

At present, the data collection, transmission and analysis functions have been generally completed. The sensor node has a high degree of integration and low power consumption, and can meet the status monitoring needs of most industrial and daily equipment. There is still a lot of room for improvement in this project, such as continuing to improve the algorithm part, improving the interaction experience with users, and enabling users to customize anomaly detection algorithms according to specific actual needs. Finally, I would like to thank the organizing committee for providing this opportunity and the sharing of the participating partners. I have gained a lot~~~Keep it up~~

作品文档.docx (1.32 MB, downloads: 6)

Share Post Summary

[Industrial Equipment Edge Intelligence Solution] Part 2: Low -power 3-axis accelerometer + 3-axis gyroscope + Bluetooth debugging

[Industrial Equipment Edge Intelligence Solution] Part 1: RSL10 Development Environment Construction