Vibration analysis online is useful until you try to trust it without supervision
I spent years running condition monitoring programs before most of this content existed on the internet. The good resources are out there. The bad ones make you worse at your job because they teach you patterns that don't match reality. Here is what actually works and what you should watch out for. The best starting point for Free Online Vibration Analysis Training is not a polished course site. It is whatever is available from the equipment manufacturers themselves, combined with open-source signal processing tutorials. Most bearing and pump manufacturers publish application notes that are denser than a textbook chapter but infinitely more accurate than a YouTube video edited for engagement. Start with the basics of FFT. Every single vibration problem comes back to someone misunderstanding the transform or not knowing what aliasing looks like on a spectrum. I can tell you exactly how many technicians I have worked with who misidentified a 1x RPM runout as a misalignment because their sampling rate was set too low and they never checked the Nyquist condition. It happens constantly.
Here is the practical order I recommend. Learn time waveform first, then envelope demodulation, then you can move into spectral analysis. Most free courses flip this. They dump you into frequency domain immediately and you end up thinking vibration is just bars on a graph. It is not. The time domain tells you things the spectrum hides. Impact events, rubbing, mechanical looseness, structural resonance all show clearly in the waveform before they ever appear anywhere near their harmonic content in the FFT. I remember running a diagnosis on a 3000 RPM motor-driven fan that was failing every eight months. The spectrum looked clean across the board. Envelope bands were flat. We had replaced bearings three times. The techs on the previous calls had been doing exactly what the online videos showed them to do, looking at the overall RMS and the basic 1x and 2x readings. Nothing showed up as abnormal. What finally caught it was the time waveform on the horizontal bearing housing. A repeating spike every revolution, buried under the normal running signature. It was a loose coupling guard hitting the shaft once per rotation. The guard touch was so mild it showed zero envelope activity and nearly nothing in the traditional spectrum. You would have walked away from that machine saying everything was fine after months of data collection. The fix was a quarter-turn of a setscrew and some shims. Twenty minutes of work after six months of failed bearing replacements.
That is why I say time waveform matters before spectrum. Free courses rarely drill this hard enough. Look for free resources that use real field data, not simulated spectra generated by software. There is a noticeable difference. Simulated data is clean, well-conditioned, and teaches you nothing about the noise and complexity you will actually see at a plant. Companies like SKF, Bosch Rexroth, and Emerson publish condition monitoring guides that include raw data files and photos from actual installations. These are gold. Search for their application note libraries rather than buying into a branded course platform. Another thing that almost nobody covers adequately in free material is sensor selection and mounting. You can have the best analysis software in the world but if you are measuring through a magnetic base on a painted surface with a flaking coating, your data is garbage. Paint thickness and surface roughness change the frequency response of your measurement chain. I have seen technicians miss a developing gear tooth fracture by over 40 percent because they compared a freshly ground test surface reading taken with a stud-mounted sensor against a baseline taken through a magnet on mill scale. The mounting method changed the sensitivity curve enough to make the fault invisible.
Get the Full Details
If you want a structured free path, try this sequence. First, work through any basic acoustics or mechanical vibration fundamentals available from university extension pages or open courseware. You need to understand harmonic series, damping, resonance, and phase before you touch a spectrum analyzer. Second, find the manual for a vibration meter you can actually afford to buy used and read every troubleshooting section. Third, use free software like MATLAB or Python with SciPy to generate your own test signals and run them through windowing functions. Seeing how a Hanning window suppresses sidelobes versus a Rectangular window in real time builds more intuition than any lecture. Phase analysis is where most free training falls apart. It gets mentioned in a paragraph and never actually practiced. Understanding phase is what separates a technician who can distinguish misalignment from soft foot from one who just guesses at both. If a free course does not give you exercises with two-channel phase measurements, skip it. There are videos showing shaft trajectory and orbit plots from real machinery. Watch those. They are rare and valuable. One counter-intuitive point that will save you trouble. Overall vibration level, measured as RMS velocity or acceleration, is a terrible standalone diagnostic tool. It tells you how much vibration exists but almost nothing about why. A machine can be in terrible condition with a low overall reading if the energy is concentrated in a narrow frequency band that the broadband meter smooths over. Conversely, a healthy machine with loose structural components can read high on overall but have no fault requiring action. Always go deeper than the overall number. Always.
Similarly, many beginners fixate on bearing fault frequencies without checking if those frequencies are even present in the system. Bearing defect calculations depend on exact geometry, rotational speed, and contact angle. If your bearing model does not match the actual installed bearing, your calculated BPFO or BSF values are fiction. I once chased a phantom bearing defect for three weeks because the part number on the nameplate did not match the bearing actually inside the housing. A different manufacturer had substituted during a repair and the geometry was slightly off. The calculated fault frequencies were wrong, the alerts triggered constantly, and the vibration was completely normal for that assembly. The fix was recalculating using the correct bearing dimensions from the actual parts bin. For hands-on practice without access to industrial machinery, a small bench setup goes a long way. A variable speed motor, a few different pulleys, a single bearing mounted in a housing, and a cheap USB accelerometer from AliExpress or similar will let you explore how unbalance, misalignment, and bearing defects appear across speeds. Spend a weekend collecting data at different RPMs and you will internalize more than a week of watching lecture videos. The data is yours. You make mistakes with it. You learn from those mistakes directly. The biggest limitation of free online training is the absence of feedback. You can watch someone demonstrate a diagnosis and nod along confidently. Then you get to a real machine and cannot replicate the process. The gap between passive watching and active doing is enormous. The workaround is simple. Record your own data on equipment around you, even household appliances. A washing machine, a desk fan, a car engine at idle. Practice taking measurements, generating spectra, and writing a one-paragraph report. This builds the muscle memory that video watchers never develop.
Another hard truth most free material ignores. Data collection protocol matters more than data analysis. Two technicians measuring the same pump at different points will produce incomparable datasets. If you are building a trend, the measurement points must stay identical. Mark them with paint. Use a fixture. Write down the exact procedure and stick to it. I have seen trending programs destroyed because someone moved a measurement point three centimeters and the new location picked up a structural resonance that masked the actual degradation signature. The trend looked flat for two years while the bearing was grinding itself to dust. Free training is enough to get you competent if you approach it with discipline and a healthy skepticism toward anything presented as a shortcut. It will not make you an expert. No free course can replace supervised field experience. But it will give you the foundation to stop making the obvious mistakes and start asking the right questions when you do encounter a problem in the field.
