Machinery Lubrication magazine published by Noria Corporation
Issue link: https://www.e-digitaleditions.com/i/1356397
38 | March - April 2021 | www . machinerylubrication.com 38 | March - April 2021 | www . machinerylubrication.com CONDITION MONITORING stoichiometric AN method which overcomes the limitations inherent to the chemometric approaches currently in use. is was prompted by the need for both onsite CM monitoring as well as concurrent determinative AN analyses in a remote mining site in Papua New Guinea. With potentiometric titration not being an option and direct-read AN systems considered inadequately deterministic, the automated stoichiometric method was re-examined and reconfigured. e FTIR unit selected for this work was a compact Agilent 5500t with an open architecture TumblIR® accessory (Figure 1), requiring only a few drops of sample and simply wiped between analyses. e method reconfigured was once used for edible oil anal- ysis where high analytical throughputs were not paramount, but accuracy was. is used a paired split-sample approach which involves some additional sample preparation, but its benefits are substantive, including: • Independent of oil type • Avoidance of chemometrics • Facilitating acid type differentiation (strong vs. weak) • Robust analytical rate of ~20 samples/hr. Calibration e method employs pure oleic acid as a primary calibrant added at various levels to a neutral hydrocarbon oil for calibration stan- dards. Each standard is split and a reagent-free solvent-diluent is added to one part (Ion) and a reagent-diluent containing NaPhenolate added to the other (In). ese are scanned as a sequential pair, the first a background scan followed by the sample scan producing a differential absorbance spectrum, -Log (In/Ion) = ΔAbsn. is process gives rise to two measurable signals (changes in NaPhenolate and oleic acid absorptions) which facilitate AN determination as well as acid differ- entiation (see Figure 2) in real samples. Generic Implementation With access to basic spectral data collected by an FTIR, there is little to differentiate instru- ments. e main bottleneck is data processing, calibration, prediction and reporting. ese limitations have been greatly improved by SpectraGryph®, a free, generic, post-spectral data processing software package. It is powerful and easy to use in this application and when combined with Excel, facilitates calibration, prediction and reporting of results for on-site AN analysis. Implementation Basically any FTIR equipped with a standard flow cell loaded by aspiration or an open archi- tecture accessory can be used. Preparing only two 2.5 ml samples (sample-reagent-diluent and sample-blank-diluent, prepared with 1:4 oil:di- luent ratio) was more than sufficient using a 100 µm "wipeable" TumblIR® accessory integral to the Agilent 5500 used and handled as per the protocol presented in Figure 3. For both calibration and sample analysis, the 2nd derivative differential spectra were used and measured for the NaPhenolate absorptions at 1589 cm-1 (Figure 4a) and related by linear regression to the mg Oleic Acid/ml (Figure 4b) which represents total acidity (AcidityTOTAL). Similarly, the COOH signal available (~1710 cm-1), representative of weak acids (Acidity- COOH) present in unknown samples was calibrated. Both acidity values are converted to mg KOH/g (AN). In AN terms, the calibra- tion precision (Figure 4b) is < ±0.10 over a range of 0-4 AN. With these two AN measures, the strong acids can be determined by difference: AN STRONG = AN TOTAL – AN COOH is additional acid differentiation capability may be useful in gauging the relative corro- siveness of oils, especially if combined with moisture information. Performance e methods' performance are illustrated in Figure 5, where an oxidized mineral oil containing both weak and strong acids has been serially diluted with acid-free, ester-based oil. e ANTOTAL and ANCOOH values for blends were determined and plotted as a function of dilution. Both measures track linearly even though the spectral signature of the oil varies continuously by mixing the two dissimilar oils in differing proportions. Even so, the method succeeds in tracking AN because each sample analyzed serves as its own refer- ence. e oil matrix changes are accounted for and ratioed out accordingly, leaving only the spectral changes associated with the acid- base reaction to be measured. All competing chemometric-based FTIR methods would fail this test as the spectral changes induced cannot be modeled or anticipated. As such, it addresses their key limitation: the need to know the oil type and to have modeled it in advance in order to have any hope of estimating AN. Even in the best of circumstances, the predicted AN is an approximation without even considering common confounding issues. Conclusion Conventional titrimetric AN and BN methods are slow, expensive and environ- Figure 1. Tumblr® open architecture accessory which is well suited to rapid manual lubricant analysis. SPLIT SAMPLE Figure 2. Facsimile paired split-sample calibration spectra illustrating how two distinct measurable signals are generated in the different spectra while the oil matrix component is simultaneously ratioed out. The arrows indicate the signal direction as acidity increases with that of oleic acid facilitating measurement of weak (COOH) acids and NaPhenolate the total acidity.