Gas compression failure and downtime is often the leading cause of unit or process downtime. Effective surveillance including proactive analytics can make a substantial improvement to plant runtime.
Complex machinery such as reciprocating compressors have a number of wear, damage and failure modes which can be difficult to detect. Often symptoms don’t appear until wear and damage are so severe that repair is needed or failure is imminent. Expensive machine condition monitoring instrumentation, hardware and software, and analytics services using huge volumes of high frequency data are often not installed or not practical even on some critical equipment and portable devices and services require many expensive and frequent service calls. An effective, low cost, continual monitoring service that makes use of available instrumentation and data to detect the onset of wear and damage is needed. Three stage, six cylinder, reciprocating compressors used to compress a relatively low volume of natural gas for liquification were experiencing frequent and varied failures through a combination of harmonic vibrations, stage imbalance, liquid carryover ring, valve, piston and rod bearing damage. Repeated efforts to balance the unit and address liquid carryover was having mixed results. The customer needed better assurance their efforts were addressing the issues.
Our first task was to process and analyze historical data to validate correlation between signals and historical damage modes determined through equipment tear-down damage reports. Our system detected the presence of liquid ingestion, as well as signals consistent with mechanical anomalies associated with ring and valve damage, stage imbalance, harmonics and eventual rod, piston and piping failure. Signals diminished when adjustments to valve lift and timing, and acoustic dampers were successful and when process alterations and scrubber modifications reduced liquid ingestion.
Sygnology Feature Extraction identified minute anomalous changes in equipment mechanical performance that are precursors to wear and damage. Feature extraction makes use of any available instrument that can serve as proxies for kinetic signals including pressure transducers, flow meters, tachometers, vibration sensors or accelerometers, etc. Our process can detect liquids entering the compressor, mechanical issues such as ring or valve wear, load and stage imbalance, mechanical binding, as well as surging pulsing and other harmonics. No new instruments or systems were required, just the time series data either online or in batch form. Periodic follow up analysis has shown periods of liquid ingestion and stage imbalance from time to time which is reported to the customer so adjustments are made to reduce further damage.
Compressor mean time to failure has more than tripled as a result of the ongoing adjustments and Sygnology Analytics not only provides confirmation that the adjustments have their intended effect, it also alerts the customer to new or repeat issues that arise.