Neural Network Processor for Liquid Chromatography Pump Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Liquid chromatography pump units face challenges in accurately maintaining flow rates, leading to variations in pressure and retention time, making it difficult to diagnose and address minor defects promptly, which can render the process unreliable.
Innovation Solution
A method involving a trained neural network processor connected to sensors and a pump controller, capable of detecting various pump faults and providing diagnostic outputs to facilitate timely maintenance, using a combination of Multilayer Perceptrons, Convolutional Neural Networks, and Long Short-Term Memory layers to analyze sensor signals and control signals, allowing for predictive maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a neural network processor is implemented for pump diagnosis, then diagnostic accuracy and reliability are improved, but device complexity increases
Solution Approach 1:
A trained neural network processor is introduced as an intermediary diagnostic device that receives sensor signals from the pump system and automatically identifies faults. The neural network acts as a mediator between raw sensor data and diagnostic conclusions, processing multiple sensor inputs (pressure, temperature, vibration, flow rate) simultaneously to provide accurate fault detection without requiring complex manual analysis systems
Solution Approach 2:
The patent replaces traditional mechanical diagnostic methods (manual inspection, physical testing) with an intelligent neural network-based diagnostic system. The neural network processor substitutes complex mechanical diagnostic procedures with automated signal processing and pattern recognition, thereby improving diagnostic reliability while the automation actually reduces operational complexity despite adding computational components
2Measurement precision
If multiple sensors are deployed for comprehensive monitoring, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent merges multiple sensor types (pressure sensors, temperature sensors, vibration sensors, flow rate sensors) into a unified diagnostic system fed by the neural network processor. Instead of treating each sensor as a separate monitoring system, they are combined into an integrated multi-sensor framework where the neural network simultaneously processes all sensor inputs to detect faults, thereby improving measurement precision without proportionally increasing system complexity
Solution Approach 2:
The neural network processor serves as a universal diagnostic platform that handles multiple sensor types and multiple fault detection functions simultaneously. A single neural network system performs diverse diagnostic tasks (detecting seal leaks, valve failures, pump cavitation, bearing defects) using multiple sensors, making the sensor system multi-functional and reducing overall complexity compared to having separate dedicated systems for each function
3Productivity
If the pump operates continuously without interruption, then productivity is improved, but undetected faults may worsen leading to reliability issues
Solution Approach 1:
The neural network diagnostic system performs preliminary fault detection by continuously analyzing sensor signals and identifying early signs of pump faults before they develop into serious problems. The system proactively detects incipient faults (seal wear, valve degradation, bearing deterioration) and alerts operators to take preventive action, thereby maintaining both high productivity through continuous operation and high reliability through early fault intervention
Solution Approach 2:
The patent implements a continuous feedback loop where sensor signals are constantly fed to the neural network processor, which provides real-time diagnostic feedback about pump health status. This feedback mechanism enables continuous monitoring during operation, allowing the system to maintain productivity while simultaneously detecting and reporting faults that could compromise reliability, enabling timely maintenance decisions
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure provides a method of training a neural network processor (7) for providing diagnostic information of a controlled liquid chromatography pump unit comprising an execution of a sequence of steps wherein the neural network processor (7) is trained with input signals (PS, PC, OS) obtained from a simulated version (S1, S3) of the controlled liquid chromatography pump unit and associated sensors, while modifying the simulated version of the liquid chromatography pump unit 10 to a pump fault simulation signal (S3). Dependent on a value of the pump fault simulation signal (PFC) the simulated version (S1, S3) of the liquid chromatography pump unit (1) simulates operation of the liquid chromatography pump unit (1) free from faults or the operation thereof with one or more pump faults. The trained neural network processor (7) obtained therewith can be integrated with a controlled liquid chromatography pump unit (1) to provide for auto-diagnostic capabilities or can be used in a separate diagnostic unit for diagnosing one or more controlled liquid chromatography pump unit (1) not having auto-diagnostic capabilities.