IBM Maximo Predict 8.5.0
IBM® Maximo® Predict is an application in IBM Maximo Application Suite that you can use to improve your assets' and locations' reliability and, for assets, predict downtime, degradation, and failures.
What is Maximo Predict?
By using Maximo Predict, you can use artificial intelligence (AI) and your performance data, maintenance records, inspection reports, and environmental data to predict downtime, degradation, and failures and to track imminent failures and maintenance schedules. Maximo Predict includes all of the features that are available in Maximo Health and the features that are described in the following sections.
For more information about how to deploy the application, see Deploying and activating Maximo Predict.
For more information about how to get started with the application, see Getting started.
Predict your assets' futures
You can create groups of assets and then work with your data scientist to generate different types of predictions for those assets, such as current failure probability or failure dates. Your data scientist can use the group ID and the default notebooks to build and train instances of a predictive model. After a model is trained and deployed, for each asset, a Predictions section is populated with predictions and related data for the asset. For example, asset records might contain the current probability of different failure modes, the number of days until a specific failure mode might occur, or if detected anomalies signal a probable failure. You can also detect operating context-specific anomalies.
- WS notebooks. Use these notebooks for testing machine learning pipelines. These notebooks do not connect to a data lake or Maximo system.
- Data Quality Learn (DQLearn) notebooks. Use these notebooks for determining data quality for a data set.
- PMI notebooks. These notebooks use Maximo Predict pipelines. After the WS and DQLearn notebooks are used to test and determine data quality, use the PMI notebooks to build models for Maximo Predict.
Additionally, your data scientist can configure custom notebooks that are either extensions of default notebooks or completely custom, but the models must be deployed in Watson™ Machine Learning.
The following table describes the specific default notebooks:| Notebook | Use case |
|---|---|
| AutoInspect_TS_MultiVariate.ipynb | Using a DQLearn pipeline for multivariate data set. |
| AutoInspect_TS_UniVariate.ipynb | Using a DQLearn pipeline for univariate data set. |
| AutoInspection_TS_Label.ipynb | Using a DQLearn pipeline for time series characteristics. |
| Auto_Imputation.ipynb | Automatic imputation of missing values in a data set that has a single index (asset ID). |
| Auto_Imputation_MultiIndex.ipynb | Automatic imputation of missing values in a data set that has multiple columns in the index, such as asset ID and timestamp or site. |
| Custom0-RUL-LTSM-WML.ipynb | Training a remaining useful life (RUL) model and deploying the model on Watson Machine Learning. |
| Custom1-RUL-LoadData.ipynb | Loading data for an RUL model. |
| Custom2-RUL-Scoring.ipynb | Invoking an RUL model that is deployed on Watson Machine Learning from Maximo Monitor to score new data. |
| Custom3-RUL-in-Predict-UI.ipynb | Displaying results from an RUL model in Maximo Predict. |
| DQLearn_AD.ipynb | Using a DQLearn a pipeline for anomaly detection. |
| DQLearn_FPA.ipynb | Using a DQLearn pipeline for failure probability analysis. |
| DQLearn_Label.ipynb | Using a DQLearn pipeline for failure labels. |
| DQLearn_MissingPatternAnalysis.ipynb | Using a DQLearn pipeline for missing value analysis. |
| DQLearn_Transactional_RegularUI.ipynb | Using a DQLearn pipeline for to any data set that is not a time series data set and generating graphical output. |
| DQLearn_Timeseries_Pipeline.ipynb | Using a DQLearn pipeline for univariate time series data with notebook-level interactive plots. |
| DQLearn_Timeseries_Pipeline_MultiColumn.ipynb | Using a DQLearn pipeline for multivariate time series data by using a pipeline. |
| DQLearn_Transactional_MinimalUI.ipynb | Using a DQLearn pipeline for independent and identically distributed (IID) data with minimal notebook-level plots. |
| DQLearn_Transactional_Pipeline.ipynb | Using a DQLearn pipeline for IID data by using a pipeline |
| DQLearn_TimeSeries_RegularUI.ipynb | Using a DQLearn pipeline for IID data by using a pipeline with minimal notebook-level plots. |
| DataLoader-Vehicle-Multiclass.ipynb | Loading vehicle data that includes multiple failure classes. |
| Explainable_FailureProbability_Dataloader.ipynb | Loading data for explainable failure probability analysis model. |
| Explainable_FailureProbability_Training.ipynb | Training and deploying a failure probability analysis model that includes explainability. |
| Explainable_FailureProbability_Scoring.ipynb | Scoring the explainable failure probability analysis model. |
| Explainble_FailureProbability_Results.ipynb | Retrieving and displaying explanations for failure probability analysis model. |
| Explainable_PredictedFailureDate_Dataloader.ipynb | Loading data for explainable predicted failure date model or time to event model. |
| Explainable_PredictedFailureDate_Training.ipynb | Training and deploying a predicted failure date model or time to event model that includes explainability. |
| Explainable_PredictedFailureDate_Scoring.ipynb | Scoring the explainable predicted failure date model or time to event model. |
| Explainable_PredictedFailureDate_Results.ipynb | Retrieving and displaying explanations for explainable predicted failure date model or time to event model. |
| FastStart2021Loader Multi.ipynb | Loading data to train different models. |
| FastStart2021Loader-New.ipynb | Loading data to train different models. |
| FastStartDataLoader_Unsupervised_AD.ipynb | Loading data to train unsupervised anomaly detection models. |
| PMI - Anomaly Detection-UnSupervised.ipynb | Detecting anomalies by using unsupervised machine learning techniques. |
| PMI - Anomaly Detection-SemiSupervised.ipynb | Detecting anomalies by using semi-supervised machine learning techniques. This notebook requires labeled data. |
| PMI - End of Life Curve.ipynb | Determine the end of life curve or asset degradation curve for a group of assets. |
| PMI - Failure Probability-Binary Classification.ipynb | Completing binary-type failure classification. |
| PMI - Failure Probability-MultiClassification.ipynb | Completing multiple-type failure classification. |
| PMI - Predicted Failure Date-Smart Regression.ipynb | Predicting failure dates by using a smart regression pipeline. |
| PMI - Predicted Failure Date.ipynb | Predicting failure date by using survive analysis techniques. |
| PMI use corrective maintenance.ipynb | Completing failure pattern analysis by using corrective maintenance records instead of failure records. |
| PMI_Use_Aggregated_Data_in_Monitor | Using aggregated data in Maximo Monitor to predict failure dates. |
| SmartClassificationExample.ipynb | Using a smart classification pipeline. |
| SmartRegressionExample.ipynb | Using a smart regression pipeline. |
| WS - Anomaly Detection - SemiSupervised.ipynb | Detecting anomalies by using semi-supervised machine learning techniques. This notebook requires labeled data. |
| WS - Anomaly Detection - UnSupervised.ipynb | Detecting anomalies by using unsupervised machine learning techniques. |
| WS - End of Life Curve.ipynb | Determine the end of life curve or asset degradation curve for a group of assets. |
| WS - Failure Probability - BinaryClassification.ipynb | Completing binary-type failure classification. |
| WS - Failure Probability - MultiClassification.ipynb | Completing multiple-type failure classification. |
| WS - Predicted Failure Date-Survive Analysis.ipynb | Predicting failure dates or time to an event by using survival analysis techniques. |
| WS - Root Cause Analysis.ipynb | Determining factors that caused failure and generating a failure contribution breakdown. |
| PMI - Model Development Tutorial.ipynb | Get started developing models by using the default notebooks. |
| PMI - Custom Model Development | Get started developing models by using custom notebooks. |
After Maximo Predict is deployed, more documentation for the models is available in the notebooks.
Track imminent failures and maintenance schedules
Use work queues to track assets that have a high probability of failure or assets that will fail before the next scheduled generation of a preventive maintenance work order.