Analysis of Medical Centrifuge Fault Diagnosis
Release time:
2012-01-06
Overview:
Fault monitoring and diagnosis of the operational status of medical centrifuge systems are achieved by collecting signals—such as vibration, balance, temperature, and pressure—from certain sensitive parts of the equipment and then extracting characteristic parameters through signal processing to identify the operating condition of the medical centrifuge. The underlying principle is that, during operation, the components of a medical centrifuge emit signals with distinct characteristics that vary depending on factors such as damage or wear of individual parts, as well as changes in voltage, current, and temperature of electronic components. Based on these characteristic parameters that specifically indicate particular faults, a fault diagnosis and prediction system utilizing artificial neural networks is employed to diagnose and forecast faults in medical centrifuges.
The process mainly comprises two stages: monitoring and diagnosis. Each of these stages consists of two sub-processes: preprocessing and feature signal extraction. Rapid and efficient extraction of features that reflect equipment fault information is crucial for fault diagnosis. The data obtained from the diagnostic object can be regarded as a time series. By segmenting and sampling this time series, the input data are mapped into points in the sample space. These data contain information about the type, severity, and location of faults. First, the input data mapped into the sample space undergo preprocessing, during which unnecessary information from the original data is removed, yielding another set of fault patterns that are then mapped back from the sample space into the data space. On the basis of the data space, invariant characteristics within the data are extracted to form an invariant fault-pattern space. After extracting the invariant characteristics of the fault patterns, the selected pattern feature vectors are subjected to quantization and compression transformations according to the diagnostic requirements and the specific nature of the problem, thereby selecting only the most useful features for fault diagnosis.
2.1 Generation and Propagation of Signals
It can be considered that the structure of the medical centrifuge system is a time-invariant linear system; therefore, the system’s vibration can be expressed as:
M【】{x}C【】{x}K【】{x}={Q}..
M【】 is the system’s mass matrix, {X} is the acceleration vector of the response, C【】 is the system’s damping matrix, {x} is the velocity vector, [KI is the system’s stiffness matrix, {x} is the displacement vector of the vibration, and Q【】 is the excitation force vector that generates vibration noise in the medical centrifuge system.
The primary sources of vibration in medical centrifuge systems include: the motor drive system, machining errors in the screen basket and rotating shaft, bearing and support misalignment, assembly imbalance, water accumulation inside the chamber caused by cracks or ruptures in test tubes, and overheating. These faults can lead to severe tilting and vibration of the centrifuge rotor during high-speed rotation. When the vibration frequency exceeds the critical threshold, it can trigger resonance throughout the entire centrifuge system, potentially resulting in serious consequences.
2.2 Signal Processing
The vibration-acoustic signals used for fault diagnosis in medical centrifuges are typically measured at a specific point near the outer casing of components such as the motor and gearbox. These signals are then analyzed to extract useful information. Fault diagnosis usually employs a mechanical system model represented by mass distribution, stiffness, and damping. The vibration-acoustic phenomena observed are associated with both moving and stationary parts within the machine, and the failure states of these parts are denoted by Fi(e), where i = 1, 2, ..., n. In the absence of faults, the initial vibrations and other frictional phenomena in the mechanical system arise from forced vibrations caused by the rotational motion and impacts of the motor and sieve basket, as well as the rotational motion of auxiliary mechanisms—vibrations that are inherently linked to mechanical motion. Such vibrations are referred to as "basic vibrations" and are denoted by c(r,t,0), where r represents the spatial location, t denotes dynamic time, and 0 signifies the machine's cumulative operating time. The intensity and frequency of these basic vibrations are determined by the type of machine. The failure of components in a medical centrifuge serves as the primary source of vibrations associated with faults, and these vibrations are denoted by Vi(t,0). Within the mechanical structure, they propagate as a function of time t and frequency f. The total response of the system—a combination of the basic vibrations and the structural acoustic responses induced by component failures—is represented by A(r,t,0) at a given spatial location r. An acceleration sensor is installed at this location, and its output is an electrical signal S(r,t,0). To obtain an analyzable model for S(r,t,0), we define the impulse transfer function from the ith failure location to the measurement point as h(ri,t,0), where ri—r(xl,y…z) is the vector connecting the measurement point to the source of vibration-acoustic generation.