Predictive and Data-Driven Optimal Process Maintenance

Predictive and Data-Driven Optimal Process Maintenance

The SDG Solution

Talking about how the government could have helped Uncle Sam to save billions, Thomas Bertram Lance, the Director of the Office of Management and Budget in Jimmy Carter's 1977 administration, stated: “If it ain't broke, don't fix it". He explains: "That's the trouble with government: Fixing things that aren't broken and not fixing things that are broken".

Hopefully times have changed and many success stories show the importance of acting beforehand to keep thing working smoothly and to make proper decisions.

Managers in the industrial industry have recently developed a wider vision which take into account the value of an effective predictive approach as the best strategy to maximize productivity by minimizing downtime and high repair costs. By relying on cutting-edge technologies, accurate and real-time data machines provide meaningful insights that empower technical staff to monitor the processes, predict equipment failure as well as implement corrective measures.


4 important reasons why Predictive Maintenance makes sense

  1. Lets you save time and fixing costs
  2. Predicts major faults and maintenance problems
  3. Allows optimal maintenance operations
  4. Supports optimal management of assets

The architecture of every predictive maintenance solution assume the presence of a Learning System, that have the function of learning from available training data, highlighting inefficiencies in the past operations, detecting collective/high risk patterns and pointing out markers of preventive alarms.

Once the first stage is complete, a Predictive System starts the analysis of real-time data, predicting patterns driving to lower efficiency, down-time or faults, providing preventive alerts as well as giving advices on the best actions for the predicted risk scenario.

When the information gathering phase is done, the Performance Measurement generate reports of the plants/processes and of the system behaviour, allowing interactive analysis (visual and numeric) of data.


SDG Predictive Maintenance Solution can be defined as a Multi-Engine Dynamic Learning System which uses different engines according to data ad objectives, among these:

  • Multiscale Detection of Anomaly in Time Sequence Data
    Use Data-adaptive Singular Spectrum Analysis, Robust Density Spectral Analysis and Permutation Distribution Clustering / Multiscale Permutation Entropy /Multiscale PCA+Persistent Local Homology
  • Multiscale Prediction of Failure Events
    Use Predictive Models (SVM, DynamicTreed Gaussian Processes, Deep Boltzmann Machines, Logistic Model Trees)
  • Modeling and Prediction of Survival Times of components
    Use Extended Cox Models, Frailty Parametric Models, Survival Random Forests
  • Causal Analysis to discover relationships and influences in a complex system
    Use Bayesian Networks and Link Analysis

The benefits coming from the adoption of a Predictive Maintenance approach are clear, first and foremost the wide offer of high quality results you can trust and will lead your decision-making. Want to make your business more effective and farsighted?

Rely on proven strategies for monitoring your critical process equipment and systems, scheduling maintenance and plan your growth.


"If it ain't broke... Focus on predictive maintenance to keep things working” - SDG Group.