Detecting Enclosed Board Channel of Data Acquisition System Using Probabilistic Neural Network with Null Matrix

Area of Science:

• Instrumentation and Measurement
• Data Science
• Control Systems Engineering

Background:

• Board channels are critical links between data acquisition systems and plant sensors.
• Flawed board channels can lead to poor-quality or erroneous data, compromising operational strategies.
• Accurate detection of board channel status is essential for maintaining system reliability.

Purpose of the Study:

• To propose a data-driven approach for detecting the status of enclosed board channels.
• To develop a method for constructing critical faulty data for training detection models.
• To validate the effectiveness of the proposed detection method using experimental data.

Main Methods:

• Utilized an error time series derived from multiple excitation signals and internal register values.
• Constructed critical faulty data using a null matrix with maximum projection, alongside healthy data, for training.
• Employed a well-trained probabilistic neural network for validating the board channel status.

Main Results:

• The proposed data-driven approach successfully detected the status of the enclosed board channel.
• The method demonstrated effectiveness in distinguishing between healthy and faulty channel data.
• Experimental validation confirmed the reliability of the probabilistic neural network in status assessment.

Conclusions:

• The developed data-driven method provides an effective means for monitoring board channel integrity.
• The technique enhances data quality by identifying and mitigating issues from faulty sensor channels.
• This approach contributes to improved reliability and accuracy in plant data acquisition systems.