Reliability assessment of key equipment for coal gasification using artificial intelligence technology

Area of Science:

• Engineering
• Reliability Engineering
• Artificial Intelligence

Background:

• Gasifier lock bucket valve systems are critical for industrial processes.
• Quantitative modeling of their dynamic failure mechanisms remains a challenge.
• Existing reliability models may not fully capture system dynamics.

Purpose of the Study:

• To develop an innovative method for quantitatively modeling dynamic failure mechanisms in gasifier lock bucket valve systems.
• To improve the accuracy of system reliability prediction and evaluation.
• To identify critical weak links within the system for targeted maintenance.

Main Methods:

• Utilizing a backpropagation (BP) neural network to optimize the prior data of a dynamic Bayesian network (DBN).
• Adapting the DBN model to a structurally adaptive BP neural network for prior parameter calibration.
• Establishing correspondence between DBN prior distribution and BP network input-output functions.
• Employing bidirectional inference analysis for dynamic system reliability evaluation.

Main Results:

• The study demonstrates that incorporating maintenance factors significantly improves system reliability (from 0.047 to 0.302 after 300 hours).
• Optimized system reliability was found to be lower than before optimization, with the gap widening over time.
• Reverse reasoning identified specific weak links: high-pressure coal powder flushing, ball seat adhesion, internal deformation, and wear.

Conclusions:

• The proposed BP-DBN optimization method provides a robust framework for dynamic reliability assessment.
• Identifying weak links through reverse reasoning enables targeted preventive maintenance strategies.
• Implementing these measures can enhance system reliability and extend operational lifespan.