Sinopec Shanghai Petrochemical Co., Ltd. Visual Recognition Intranet Meter Reading Project (Pointer Meter)

 

一、Project background and requirements

1.1 Implementation location:

The project is located in Sinopec Shanghai Petrochemical Co., Ltd., production line: 1,500 tons of carbon fiber device spinning machine line 1-4.

1.2 Project background and requirements:

There are difficulties in data collection of mechanical instruments supporting workshop production lines. For data collection of mechanical pressure gauges, the instruments do not have specific external communication interfaces, the corresponding environment is corrosive, and the construction is difficult. It is necessary to realize real-time data collection to achieve effective monitoring of production line data indicators. According to the project background and implementation difficulty, we, as the supplier of products and supporting solutions, cooperate with the implementer (Yangtze River Institute) to customize the wireless timer camera image recognition solution to collect and record the values of the pointer pressure gauge. The camera can read the meter through image analysis, which can realize the analysis and tracing of historical data, and requires the intranet environment to realize data collection and transmission.

 

 

 

 

 

 

 

 

 

 

 

 

二、Technical parameters and implementation plan

2.1 Technical methods and routes:

In view of the pointer meters of this intelligent meter reading, as well as factors such as the detection environment, it is planned to use video photography, wireless data transmission, and visual recognition to read, store and analyze the pointer meter data; use 1 or 2 cameras to shoot a row of meters (5).

2.2 Implement hardware products and technical solutions

Hardware products, according to the requirements , need to be implemented in the intranet environment, using our company's matching timed camera (SC-GP-CAM), POE power supply version, so that the signal power supply shares the same network cable, reducing the complexity of construction wiring. Considering the corrosion situation on site, Yangtze River Institute customized the protective shell for the product.

 

Technical implementation plan

2.3 Technical indicators:

(1) The recognition accuracy rate should be ≥ 99.5% under the condition that the dial is free of dirt, water mist, broken glass, etc.; the photos must be clear, and the instrument scale and numbers can be clearly distinguished;

(2) Meter reading frequency ≤ 1 minute/time , the collection frequency can be adjusted ;

(3) The system's stable operation period is ≥ 60 days; the single fault handling time is < 24 hours, and the failure downtime rate caused by software problems is < 1% (the hardware depends on the situation);

(4) Data storage time ≥ 6 months;

(5) Protection grade IP65, working environment temperature -10°C ~ 50°C, humidity 0 ~ 95% (non-condensing);

三、On-site implementation and debugging

According to the on-site instrument layout, a total of 4 sets of cameras are provided, of which 2 sets of cameras cover 2 pointer meters and 2 sets of cameras cover 3 pointer meters . There is a corrosive environment on site to ensure that the camera is not corroded. The Yangtze River Institute follows up on the actual environment and adds a heat shield for the camera (refer to the installation example in the figure below)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

On-site installation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Docking and debugging

On-site docking and debugging, cooperate with the implementer to fine-tune the camera position, clean the camera dial (to ensure the clarity of the dial, etc.) , and eliminate the influencing factors one by one: remove the dial label, avoid water mist on the protective cover, change the installation position to avoid light source reflection, etc.

四、Implementation Results

According to the requirements of the project party, the image recognition method is used. After the data is successfully collected, the data is stored in the database , and a one-to-one correspondence is achieved between the image path and the recognition result.

 

 

 

 

 

 

 

 

 

 

 

 

Image storage path and corresponding identification value