An on-line real-time data analysis method for quality and efficiency enhancement by intelligent manufacturing

Tang Honghua

Internet + continuous fermentation, cloud computing, large data, IoT, mobile Internet and other new technologies, rapid development and application, many enterprises are in the use of these new technologies to accelerate the transformation of the upgrade.

Many manufacturing enterprises have collected a large number of processes, equipment and quality data through DCS, PLC or MES. For example, a 5000 ton/day cement enterprise is about 10,000 processes per second, equipment data, 1 days about 3.456G, in addition to hundreds of quality data.

This article will introduce how to use the production process, equipment and quality inspection of large data, timely detection of quality anomalies, reduce secondary waste, improve product quality, increase the economic efficiency of enterprises to help enterprises transformation and upgrading. Large data analysis strategy based on control chart large data analysis strategy based on control chart
(1) Select characteristic values, such as reaction temperature, velocity, electrode eccentricity, etc. (2) to determine sampling frequency, packet, analysis time and execution period, it is necessary to design according to the characteristics of characteristic value:
A the minute level, for example, special electrode production 60-80 per minute, measuring the eccentricity every 30 times, can be 5min minutes to 1 groups, each group has 10 data, each analysis can be observed 2 hours (24 sets of data), the implementation cycle of 5min; b) Hourly level: For example, chemical enterprise reactor temperature monitoring, Can be 10min once, 1 hours 6 times, for a group, monitoring 1 days (24 groups of data), the execution period is 1 hours; c) Team level: for example, a chemical enterprise electrolytic cell electrolyte concentration, every 2 hours, can be by team, 4 times per class, monitoring 1 weeks (21 sets of data), the implementation cycle of one Class 8 hours d) Day class : For example, a chemical enterprise equipment vibration, monitoring 1 times every 4 hours, in days for the unit group, 4 data per group, monitoring 1 months (30 groups of data), the implementation cycle for one day.
(3) Obtaining data from a real-time database or relational database according to established rules. (4) According to the control chart statistic method calculates the average value, the standard deviation (may use the Open Source Tool R language to be possible to calculate the correlation statistic values and the drawing). (5) Drawing the control chart and displaying it to the electronic Kanban; (6) Exception decision rule and Exception handling flow (detailed later)
(7) The system defines the clock in the background automatically executes, if discovers the unusual, automatically triggers.
Key points: 1, eigenvalue selection 2, eigenvalue statistical parameters of the design 3, anomaly determination rules
Second, the control chart principle (slightly), interested readers of their own Baidu. III. guidelines for exception types of Control charts 1:1 pips fall outside area A (the idea gets out of control)
Guideline 2:9 Consecutive points fall on the same side of the centerline Guideline 3: Continuous 6-point increment or decrement guideline 4: Consecutive 14 dots are always up and down. Rule 5:3 points in a continuous 2 point fall on the same side of the centerline B. Guideline 6:5 points in a continuous 4 point fall on the same side of the Central line, Area C. Guideline 7:15 consecutive points in the Central line with both sides of the C zone guideline 8:8 points in a row on both sides of the center line and No 1 o'clock in the C zone Note: For details, please refer to the National standard gb/t4091-2001 of the People's Republic of The system according to set rules, will be pushed to the relevant positions, such as "chlorine press displacement average 6 consecutive hours, please dispose as soon as possible", and pushed to the lead. Conclusion using the process of production process, equipment status, quality inspection data, through the abnormal analysis of control charts, timely detection of quality anomalies, reduce secondary waste, improve product quality, increase enterprise economic benefits, to help enterprises transformation and upgrading.

application of artificial intelligence in industrial enterprises--the prediction of the end point of copper smelting

Application and exploration of artificial intelligence in industrial enterprises

Prediction of the end point of----copper smelting

Tang Honghua (Ufida Network Technology Co., Ltd.)

2017.10

Artificial intelligence technology more and more into the social production and life, AI in the prediction, classification, clustering and so have a good adaptability. There are many prediction scenes in industrial enterprises, the author tries to use the neural network method to predict the smelting end of copper smelting enterprises, and it is quite good to apply the results preliminarily. Special share to everyone.

1, the problem