Prognostics

What Is Prognostics?

Prognostics is the process of predicting the future health state of a machine based on real-time and historical sensor data. Prognostics algorithms enable equipment operators and manufacturers to:

  • Reduce equipment downtime by predicting critical issues in advance
  • Avoid unnecessary maintenance costs by scheduling equipment service only when necessary
  • Bring equipment back online faster by determining the root cause of failures and faults

Prognostics vs. Diagnostics

Prognostics and diagnostics are different approaches to predictive maintenance and often used in parallel. Diagnostics aims to identify the root cause of a problem that has already occurred, such as detecting a fault in a particular machine component. Prognostics aims to predict when a failure is likely to occur. This gives maintenance teams time to plan and perform maintenance before problems occur. Prognostics require a more advanced understanding of the machine’s behavior but have a greater impact on reducing unplanned downtime and maintenance costs.

Condition-based maintenance (CBM) is a maintenance strategy based on the current health state of the equipment. Learn how CBM works and how it differs from predictive maintenance.

Prognostics Algorithm Development

Prognostics algorithms are the goal of many predictive maintenance programs. To develop prognostics algorithms, you need to start with data from your machine or fleet of machines. While prognostics data needs to represent healthy and degraded states, there are several approaches to prognostics algorithm development in MATLAB® using tools like Predictive Maintenance Toolbox™.

Workflow of sensor data from machine on which algorithm is deployed.

Prognostics algorithm development workflow.

Data for Prognostics

Data for developing prognostics algorithms typically comes from machine sensors such as temperature, pressure, voltage, noise, or vibration measurements collected over time; however, since scheduled maintenance programs are often conservative, failure data for prognostics can be difficult to acquire. Sensor data can be augmented with failure data generated from physics-based models, which are tuned to match the machine parameters.

Condition Indicators

Machine data for prognostics is processed using various statistical and signal processing techniques to extract important features calle condition indicators.

Condition indicators are features that change predictably as the machine degrades. These can be any feature useful for distinguishing between healthy and faulty operations. In the case of prognostics, condition indicators are sometimes called health indicators and are used as inputs to train prognostics models.

Types of Prognostics Algorithms

To monitor the health of the equipment, apply condition indicators to train prognostics algorithms. The result is a prognostics algorithm that can predict when the next failure event is likely to occur.

Regression Models

Regression models, a type of machine learning model, are often used for basic prognostics. Regression models describe the relationship between an output variable (such as failure time) and one or more input variables (condition indicators). This can be useful when the goal is to statically predict the life of an asset, such as a battery based on its current state.

Results of a simple regression model for prognostics, predicting the cycle life of a battery.
(See example.)

However, in many prognostics applications, the goal is to accurately predict the time to failure of a particular asset based on its operating history. For that, we turn to remaining useful life (RUL) algorithms.

Remaining Useful Life Algorithms

The RUL of a machine is the expected usage time remaining before the machine requires maintenance. Specialized RUL models for prognostics in MATLAB—similarity, degradation, or survival models—depend on the amount of historical data available. They can provide continuously updated prognostics within a defined confidence interval as more data is acquired.

RUL estimation for bearing prognostics using a degradation model in MATLAB.
(See example.)

State Estimators and Identified Models

When creating an identified dynamic model that describes machine behavior, such as a state-space model or autoregressive model, you can use that model for prognostics by propagating the model forward in time and analyzing the behavior of a known condition indicator. Similarly, you can use state estimators like Kalman filters for prognostics and predict future states when failure will occur.

Physics-Based Models

Physics-based models, such as those built in Simulink and Simscape, can be used for prognostics by simulating the behavior of a machine over time. These models can be validated and calibrated using sensor data from the real machine and then used to simulate future behavior for prognostics under different operating conditions. Using these models for prognostics requires detailed knowledge of machine design and how components degrade with time.

Using models for prognostics of a faultable pump.

Physics-based model of a faultable pump for prognostics built in Simscape.
(See example.)

Deployment

Once validated, prognostics algorithms can be operationalized in an IT environment such as a server or the cloud. Alternatively, prognostics algorithms can be implemented in an embedded system directly on the equipment, enabling faster response times and significantly reducing the amount of data sent over the network.

Prognostics with MATLAB

Engineers and data scientists can quickly design and deploy prognostics algorithms in MATLAB. For tools and examples related to prognostics, explore Predictive Maintenance Toolbox.