From manual to AI: Paradigm shift in lead-zinc smelting process control

2025-03-11

View: 21

In the context of accelerated transformation and upgrading of 

the non-ferrous metal industry, lead and zinc smelting, as a 

typical high-energy consumption and high-pollution process 

industry, is experiencing a subversive change from “experience-driven” to

 “data intelligence”. The traditional smelting mode relies on manual 

experience to adjust process parameters, facing high energy 

consumption (comprehensive energy consumption of >1800kWh 

per ton of zinc), large fluctuations in the metal recovery rate 

(lead recovery rate of 88%-92%), and excessive emissions 

(SO₂ concentration fluctuations of ±15%) and other pain points, 

resulting in direct economic losses of more than 2.5 billion U.S. 

dollars per year. With the deep integration of artificial intelligence 

technology and smelting process, the new generation of intelligent 

control system has improved process stability to 99.7%, reduced 

energy consumption by 18%, and cut heavy metal emissions

 by 90% through the technology closed loop of real-time 

perception-autonomous decision-making-dynamic 

optimization, which has pushed lead and zinc smelting into 

a new era of “accurate and intelligent smelting”.

First, the traditional manual control of
the four major dilemmas

Lagging regulation of process parameters

Relying on manual inspection (2 hours/times) and offline 

assay (delayed 4-6 hours), the lag rate of key parameters 

(melt temperature, oxygen potential, slag type) adjustment 

reaches 78%.

Experience inheritance barrier

Operators need 5-8 years training cycle, the lead yield 

fluctuation controlled by different shifts is up to 3.5 

percentage points, the talent disconnection leads to

 the decline of process stability.

Multi-variable coupling out of control

The non-linear interaction of 20+ parameters, such as sinter 

block composition, wind-oxygen ratio, coke rate, etc., makes 

it difficult to realize global optimization manually, and

 triggers 3-5 major fluctuations in working conditions every year.

Passive response to environmental control

Flue gas SO₂ concentration relies on end-to-end management, 

the disposal delay of sudden emission peak exceeding the 

standard is more than 30 minutes, and the environmental 

protection fine accounts for 12% of the production cost.

Second, the technological breakthrough: AI 

control system of three-layer architecture innovation

1. All-dimensional perception network

Real-time monitoring of melt status: 1400+ intelligent sensors are 

deployed in sintering machine, blast furnace and electrolyzer to 

collect 40 types of parameters such as temperature (accuracy ±0.5℃), 

pressure, composition (XRF online analyzer, 0.1s/time), etc.

Multi-source data fusion engine: integrating DCS, MES, LIMS system 

data flow, building an industrial Internet of Things platform that 

processes 200,000 data points per second.

2. Intelligent decision-making core

Digital twin dynamic modeling: build a 3D smelting process 

simulation system based on Computational Fluid Dynamics 

(CFD) and Discrete Element Method (DEM), with a prediction 

accuracy of 95%.

Multi-objective optimization algorithm: Deep Reinforcement 

Learning (DRL) is used to balance the objectives of metal 

recovery rate, energy consumption, and emissions, and 

300+ sets of optimization parameter schemes are 

automatically generated every day.

3. Autonomous execution system

Intelligent valve group response time <50ms, realizing 

precise control of blast air volume (adjustment precision

 0.1m³/min) and oxygen concentration (±0.3%).

Adaptive control system can re-optimize the dosage ratio 

within 10 seconds when the composition of raw materials 

fluctuates, which guarantees the stability of the process.

Application Scenario: Intelligent Reconstruction of the 

Whole Smelting Process

1. Intelligent control of sintering process

Recognize the porosity of sintered blocks through machine 

vision (detection speed of 200 blocks/minute), and 

dynamically adjust the thickness of the material layer 

and ignition temperature.

AI model predicts the sintering end time (error <2 

minutes), which increases the sintering capacity by 

15% and reduces the powder return rate by 40%.

2. Blast furnace melting optimization

Real-time monitoring of oxygen potential distribution

 in the furnace, intelligent adjustment of oxygen-rich

 concentration (27%-33%), stabilizing the lead yield 

above 94.5%.

Melt temperature control accuracy is improved from 

±25℃ to ±5℃, and coke consumption is reduced by 12%.

3. Electrolytic refining upgrade

Based on the current efficiency digital twin model, 

dynamically optimize the pole pitch (adjustment 

precision 0.1mm) and electrolyte composition (Zn²+

 concentration control ±2g/L).

Uniformity of cathode zinc thickness increased to 

98%, DC power consumption reduced to below 2900kWh/t.

4. Environmental protection control forward

Predictive emission control system warns of SO₂ 

concentration abnormality 10 minutes in advance, 

and adjusts the operating parameters of desulfurization 

tower in a linked manner.

Arsenic, cadmium and other heavy metals emissions 

concentration fluctuations compressed to ± 5%, the 

annual emission reduction of 800 tons of hazardous wastes

IV. Benefit verification: value leap of 

intelligent manufacturing

Economic breakthrough: Reduction of zinc processing costs by 

230 yuan per ton, annual output of 300,000 tons of smelters 

with an annual profit increase of more than 60 million yuan.

Quality leap: 0# zinc grade rate increased from 88% to 99%, 

meeting GB/T 470-2022 high-end zinc ingot standard.

Energy consumption innovation: comprehensive energy 

consumption dropped to 1500kWh/t zinc, reducing 

standard coal consumption by 42,000 tons per year.

Safety upgrading: major process accidents decreased 

to 0.05/year, saving 18% on insurance costs.

Green transformation: sulfur dioxide emission concentration 

stabilized at <80mg/m³, heavy metal removal rate of 

wastewater exceeded 99.9%

V. Future picture: AI-driven new ecology 

of smelting industry

Autonomous evolution system

Construct AI model with migration learning capability, and 

complete independent development of smelting process for 

new ore types within 3 months.

Cross-process collaborative optimization

Open up the data chain of the whole process of sintering-smelting-refining 

to maximize the efficiency of global resources.

Quantum Computing Empowerment

Applying quantum neural network to process trillions of parameter 

combinations, the speed of solving complex smelting optimization 

problems is increased by 1,000 times.

Industrial meta-universe fusion

Create virtual smelting digital twins to support process engineers

 in immersive interactive optimization.

Conclusion

The paradigm shift from manual to AI not only solves the century-long 

control challenges in the lead and zinc smelting industry, but also 

redefines the competitiveness standards of the non-ferrous metals 

industry. This change has achieved a synergistic leap in quality, 

efficiency, and environmental protection by digitizing process knowledge, 

intelligentizing the decision-making process, and automating the 

execution system. With the deep penetration of 5G, edge computing, 

knowledge map and other technologies, the AI control system will 

continue to evolve and push lead and zinc smelting towards the ultimate 

goal of “zero human intervention, zero quality defects and zero pollution 

emission”, providing core support for the construction of the global green 

metal supply chain.