In this project, I will explore different Failure Analysis techniques in various industrial applications. The project is ongoing, so more datasets and more analysis will be added to the project. I will keep the project organized by using dedicated folders for each dataset.

I follow CRISP-DM methodology in my analysis following these steps:

1. Business Understanding: Defining business objectives Assessing the current situation Determining data mining goals Producing a project plan

2. Data Understanding:   Gathering initial data Describing data Exploring data Verifying data quality

3. Data Preparation: Selecting data Cleaning data Constructing data (feature engineering) Integrating data Formatting data

4. Modeling: Selecting modeling techniques Designing tests Building the model Assessing the model

5. Evaluation: Evaluating results Reviewing the process Determining the next steps

6. Deployment: Planning deployment Monitoring and maintenance Reviewing the project Finalizing the project

Bearing Fault Detection 01

DE - drive end accelerometer data

FE - fan end accelerometer data

IR - Inner Race

OR - Outer Race

NB - Normal Baseline

Fault Diameters: 0.007" and 0.021"

-> OR-21 = 0.021" fault on outer diameter

Drive End (DE) Accelerometer: Near the motor drive end. Fan End (FE) Accelerometer: Near the opposite end (fan side). The presence of both DE and FE accelerometer data helps in analyzing how bearing faults propagate along the shaft and how their signatures differ at different positions.

Each accelerometer captures different aspects of fault propagation:

(A) Drive End (DE) Accelerometer Directly affected by shaft forces and misalignment. Captures stronger vibration signals from inner race, ball, and drive-end outer race faults. Fault signals do not get dampened as much since they are closer to the source.

(B) Fan End (FE) Accelerometer Located further from the main load and drive forces. Captures weaker but more global vibration signals. Outer race faults on the fan-end bearing may be easier to detect here. Some shaft-related issues (misalignment, unbalance) may be more pronounced here.

Benefits of Having Both Sensors

1️⃣ Identifying Fault Location If a fault is stronger in DE data, it is likely at the drive-end bearing. If a fault is stronger in FE data, it is likely at the fan-end bearing.

2️⃣ Understanding Fault Propagation Vibration signals travel through the shaft differently. Having both datasets allows engineers to study how faults spread from one end of the system to another.

3️⃣ Differentiating Between Local & Global Vibrations DE detects more localized effects from faults near the motor load. FE detects overall system vibrations, including shaft misalignment or resonance.

4️⃣ Reducing Misdiagnosis If only one sensor were used, some faults could be missed or misinterpreted. For example, a shaft misalignment issue may be subtle on the DE but stronger on the FE.