| Company | FAQs

FAQs

Frequently Asked Questions About Probabilistic Technology, PredictionProbe and UNIPASS Software

What is Probabilistic Technology?

Probabilistic Technology is one of several predictive technologies aimed at allowing one to predict outcomes. The outcomes could range from how fast a plane can fly, how much a company will earn next year, how reliable a part or system is, what is the most likely failure mode for a system, etc. Sometimes these predictions can be made with statistics, where you operate the systems and gather data on the results. This approach inherently considers uncertainties in the variables that drive the outcome, but unless one is careful, there may be unknown biases or other inadequacies in the data. In any case, the statistical method can be very expensive.

Another way of predicting outcomes is by creating a mathematical or rule-based model, both of which are deterministic models. For these, you pick a value for each independent variable and calculate a value for the dependent variable, or outcome. The problem with deterministic models is that they ignore uncertainty, in that the independent variablesare usually not single-valued but are random. We might use safety factors to try to compensate for the uncertainties, but that may either lull us into thinking the safety factors are sufficient or cause us to over design the system.

The preferred way of predicting outcomes is to use Probabilistic Technology. In Probabilistic Technology, if there is data available, we use statistics to characterize the uncertainties in the independent variables. If there is not any data, we can use engineering judgment to estimate those uncertainties. These random variables then become the input to a deterministic model. Only now the outcome will not be a single-valued. Picking from a toolbox full of probabilistic methods, one of the results can be the probability density function or cumulative distribution function (pdf/cdf) curves for the desired outcome. We may also identify the most likely failure conditions. We also will have sensitivity analysis data which will tell us which random variables aremost critical. If one of the critical variables were evaluated using only engineering judgment, we might decide to create a test to gather data, and then run the analysis again. And these are only some of the benefits of Probabilistic Technology. The bottom line is that Probabilistic Technology has all the benefits of statistics and deterministic models, and yet provides much more information.

PredictionProbe’s Probabilistic Technology is a set of predictive methodologies which consider both the physical and statistical aspects of a process while systematically accounting for various types of uncertainties, including inherent uncertainties, model imperfection, lack of data, measurement error, human error, decision uncertainties, human intervention, and much more. PredictionProbe’s Probabilistic Technology allows you to: 1) Quantify, manage, and rank most important uncertainties accurately and objectively; 2) Systematically incorporate uncertainty into any process; 3) Evaluate the impact of the uncertainty on the outcome; and 4) Rank and identify the critical or most sensitive variables or uncertainties which are most important in any process or design.

When was Probabilistic Technology developed?

Probabilistic Technology began some 80 years ago with the study of the theory of structural reliability. The first phase, from about 1920 to 1960, appears in retrospect as a very slow beginning. During this period one or two dozen pioneers, scattered over many countries, worked independently on various elements of the theory of reliability applied to problems of structural strength. They questioned established thought and developed the basic concepts of random structural events, departing radically from the classical notions of structural engineering.

While considerable theoretical and practical work was done in this field, there was no comprehensive commercial software that could solve complex probabilistic problems in a practical manner. Existing software tools (e.g., FEBREL, NESSUS, PROBAN, ProFES, CALREL) had limited capabilities, were not supported, and maintained by their developers, or were not offered commercially to the public.

Finally, during the period from mid-1990 to present, as systems grew to be more complex, and the need for non-deterministic analysis became more apparent, software vendors devoted increasing attention to the development of off-the-shelf commercial codes for performing probabilistic analysis. This produced several off-the-shelf software tools with various capabilities for predictive measurement and analysis (e.g., UNIPASS, ANSYS/PDS, iSIGHT). Today, UNIPASS leads the market as the only probabilistic tool which uses all the capabilities of the laws of probability combined with deterministic and statistical methods.

What is the difference between Probabilistic Technology and Predictive Technology?

Probabilistic Technology is a subset of Predictive Technology. In many cases, predictive technology is deterministic, meaning that given a certain process model and a set of inputs, the response or set of responses are always the same, unique values. In Probabilistic Technology, the uncertainties in the input values are explicitly modeled as some kind of a distribution of values (many choices of the distribution are possible). The distribution of input values is used to create distributions of the response values. These distributions of response values represent the uncertainties in the process given the uncertainties modeled at the input side. Each of the values in a response distribution is a possible answer with an associated probability value. Answers nearer the mean (average) value of the response distributions are more likely, or probable, than those farther away from the mean value. The deterministic response value is a member of the probabilistic response distribution and should be the limiting value as the magnitude of the input uncertainties is shrunken to zero.

What Technologies make up Predictive Technology?

Our unique and revolutionary brand of predictive technology has been formulated through the customization, innovation, and renovation of numerous key technologies, including Probabilistic Technology, Bayesian Belief Network, Classical Statistical Techniques, Deterministic Modeling Techniques, and Integration Technology. Built on the solid foundation of our unique proprietary methodologies, processes, modeling techniques, best practices, and software, our Predictive Technology is truly the advanced tech of the future here for you right now.

What is the difference between statistical approaches and Probabilistic Technology?

Statistics uses past performance data to predict future performance. For example, if one was interested in how long it takes to get to the airport from some office, he or she might collect elapsed time data from people traveling to the airport from the office to determine how long it took them to make the trip. The data could then be fitted to a statistical distribution, and one could use this to predict the probability of getting to the airport in a given time. The emphasis here is on the output time, while the input variables such as route and speed (as well as uncertainties in those variables) are ignored.

In contrast, Probabilistic Technology uses physics or rules to describe the process that yields the desired outcome and applies statistics to quantify the input (as opposed to output) variables. Thus, in our example, we might collect data on the input variables such as speed and distance, as well as data on other variables such as delays due to accidents, effect of time of day and the day of the week on traffic flow, whether the driver is aggressive or passive, etc. Using that data, we can then create a model to calculate the time to the airport. If we do not have the data available, we can use experts to estimate the missing variable’s distribution parameters.

After applying Probabilistic Technology to this problem, the output will be a much more accurate estimate of the probability of reaching the airport before a given time on a given date. In addition, this technique provides the sensitivity of the output variables to changes in the input variables or the distribution parameters for the input variables. We can find out if the output is significantly affected by lack of data or any assumptions we might have made.

The greatest benefits of Probabilistic Technology over statistics are that: (1) predictions are more accurate; (2) it is often easier to obtain data on input variables that output results; and (3) more information is available.

What is UNIPASS?

UNIPASS is a commercial, general-purpose, probabilistic software tool developed by, PredictionProbe. UNIPASS has been developed by the founders of PredictionProbe, Drs. MohammadKhalessi and Hong-Zong Lin and is continuously enhanced and updated by PredictionProbe’s Team of Predictive Technology experts. This software is a PC-based Windows 32 compliant software engine that can be utilized independently, as a stand-alone software engine, and/or integrated with deterministic software tools to perform complex probabilistic analyses. In the analyses, UNIPASS provides the basis for modeling uncertainties, computing probabilities, identifying most likely outcomes, and calculating sensitivity measures, while the deterministic software tools (if integrated) provide the computational framework for constructing complex deterministic process models. The UNIX-based process models and their associated deterministic software may be integrated with UNIPASS using PredictionProbe’s proprietary software UniFace.

Do I have to be an engineer to use UNIPASS?

No. UNIPASS is a user-friendly software tool which can be quickly learned by most anyone.In particular, UNIPASS can be used by decision makers in real time to aid in complex situations requiring timely choices to be made.

How long does it take to learn Probabilistic Technology?

PredictionProbe offers training from beginner to expert. Basic training can be done in less than one week. Please contact us for details regarding upcoming training events.

How do Monte Carlo simulations relate to Probabilistic Technology?

Monte Carlo simulation is a rudimentary form of Probabilistic Technology which was developed in the 1930’s. While Monte Carlo simulation provides the most accurate results when they work and are practical, the problem is that they often are impractical. They are time intensive, often prohibitively so, and they cannot be performed unless the model has a closed form solution. Other probabilistic methods such as First-Order Reliability Method and Second-Order Reliability Method are available which get around the limitations of Monte Carlo simulations and still provide accurate results.

Why do we need Probabilistic Technology?

Because much more accurate information is available, Probabilistic Technology allows organizations to make more informed decisions, to improve their bottom line and product performance, to design safer products with higher reliability, to ensure system availability, to minimize life cycle costs, and to quantify risk and liability.

Is Probabilistic Technology mature enough for practical applications?

Yes, because of the availability of commercial software tools, practical analytical methods, and computer power. It has been used successfully in numerous high-profile projects and with great success.

What is the best way to begin implementing Predictive Technology?

To get started, we recommend a brief introductory and planning meeting between our experts and your executives to identify which areas in your organization would best serve as a pilot for implementation of Predictive Technology.

At what stage of a project should Probabilistic Technology be implemented?

Probabilistic Technologycan provide significant benefits through the entire life cycle of process and/or product development. To maximize the benefits, probabilistic technology should be implemented as early and as broadly as possible.

Which industries have potential applications in Probabilistic Technology?

Our Probabilistic and Predictive Technologies are foundational technologies and can benefit any enterprise, including Insurance, real estate, finance, manufacturing, engineering, science, software, hardware, aerospace, weaponry, logistics, construction, and service.

What kind of projects most benefit from Predictive Technology?

Any project, whether it is to refine a business process, design a space shuttle, or determine the risk to value of a property can benefit from the use of Predictive Technology. Our CustomApplication software allows companies to fully integrate their tools with our predictive tools and processes to create a powerful System-of-Systems Software.

What kind of operating systems does the UNIPASS software require?

UNIPASS is easy to use and easy to integrate into most existing environments without a significant learning curve or additional hardware investment. UNIPASS is compatible with all MS Windows and Linux Systems.

What is Integration Technology?

Our SPISE software is the first-ever Integration Tool built entirely around predictive technology for a one of a kind work flow development and execution environment which gives the User the power to assemble a multitude of technologies into one resourceful frameworkthat enables an exponential increase in processing capability. In its user-friendly and comprehensive design, users will experience real-time execution of highly complex systems and the ability to accomplish tasks that none of the technologies could do independently.

What is Bayesian Technology?

Bayesian Technology is based on networks (also known as Bayes, belief, or decision networks) which provide a probabilistic graphical model that represent a set of variables and their conditional dependencies. Our Bayesian Technology Area focusses on the development of a proprietary approach for constructing PredictionProbe Bayesian Belief Networks (PBBN) for the practical implementation to any problem in any industry by developing capabilities that will:

  • Minimize amount of required data,
  • Improve computational efficiency for systems with a large number of random variables,
  • Improve computational efficiency for a multitude of states,
  • Allow the use of continuous random variables, with no limitations on distribution type or statistical dependency,
  • Allow computation of very small probabilities (e.g., less than 10-6), and
  • Automatic generation of states and conditional probability tables (CPTs), and more.

What are the benefits of implementing Predictive Technology?

We have leveraged our core technology to assist many of the world’s leading businesses, operating in multiple industries such as real estate, insurance, engineering, telecommunications, financing, investment markets, bioengineering, manufacturing, and many others. Using our technology, companies can quickly and accurately:

  • Minimize product development cycle time.
  • Minimize product weight.
  • Maximize product life.
  • Analyze and perform biotech research with improved speed and accuracy.
  • Diagnose medical conditions and analyze drug enhancements.
  • Optimize and streamline processes.
  • Allocate resources optimally.
  • Optimize data manipulation and knowledge management resources.
  • Optimize test plans.
  • Quantify risk, liability, reliability, and safety measures, even with limited or no data.
  • Perform Robust Design.
  • Identify root-causes objectively.
  • Identify and manage key uncertainties and critical process input variables.
  • Optimize product design and pricing.
  • Minimize warranty and repair costs.
  • Minimize inspection costs.
  • Effectively and efficiently manage data, information, and knowledge.
  • Identify data needs.
  • Predict customer behavior.
  • Enhance Six Sigma and Design for Six Sigma processes and tools.
  • Perform efficient Design for Six Sigma.
  • Realize lean thinking by effectively identifying and removing waste
  • Reduce product development, life cycle and production costs.
  • Streamline manufacturing processes.

These are just a small sampling of the unlimited applications and vertical potential harnessed by PredictionProbe’s Predictive Technology all which can seamlessly integrate into your company’s Information Technology, financial systems, and other information or knowledge management systems. From the world’s largest technology-driven companies to multi-state franchises to Wall Street, PredictionProbe’s software products focus on maximizing profits and removing costs from processes and business models, while adding to the bottom line a minimum 5% savings, and in many cases, much more.

How are statistically based predictions different than probabilistic predictions?

Statically based predictions use variability models of fixed-process outcomes or responses, whereas probabilistic predictions use uncertainties model of the inputs to potentially dynamic processes. The accuracy of statistically based predictions depends upon the number and the quality of the response data points used to construct them. In contrast, probabilistic predictions model the uncertainties of inputs within an existing process. This uncertainty modeling can be done completely independent of having any data points and this uncertainty modeling, as well as the underlying physics modeling, can be modified by the user over time to improve accuracy. In addition, probabilistic predictions from various physics or uncertainty models for the same process, of different fidelities and using assumptions, can be easily integrated into more comprehensive predictive models.

Is a Design of Experiment (DOE) analysis as useful as a probabilistic analysis?

Design of Experiments (DOE) is one of a set of three powerful statistically based techniques that are quite complimentary to probabilistic technology, the other two being Analysis of Variance (ANOVA) and Response Surface Modeling (RSM). In engineering (for example), DOE uses ANOVA and RSM together in an iterative way to suggest new test points within a process domain that should be acquired to improve the physics modeling of a given process at low (or perhaps minimum) cost. The ANOVA technique assesses the variability or uncertainty of a proposed test point for a chosen and specific RSM modeling structure. When paired with probabilistic technology, as has been done in numerous PPI applications, these technologies enable the rapid and low-cost improvement (reduction of uncertainty) for physics models in any given domain.

Is Probabilistic Technology used in Artificial Intelligence?

Artificial Intelligence (AI) is a way to ultimately model the complexity of human decision making, in which the brain takes stimulus (inputs) from the senses and the surroundings to suggest or automatically enable a response. The response might vary from “fight or flight” to simply performing a routine function or choosing a way to express a thought or an emotion. In humans, this decision-making process also adapts through a dynamic learning process, to improve the “quality” of the response over time, perhaps for greater safety, speed, or efficiency. Ultimately, these are also the goals of combining probabilistic and Bayesian technologies into a single package such as the PPI framework, SPISE. Rudimentary AI, like a statistically based prediction, models only the outcomes or responses of a process. There is no fundamental reason why AI could not consider the uncertainty of inputs to a process, as well as the likelihood of a given response from the process, as numerous PPI applications have done successfully.

How can I be confident in my model with limited data?

When using a statistically-based prediction, the accuracy of the prediction is numerically bound to the number of data points that are used, as well as the quality of the data points used, though the later aspect (quality) is rarely considered. The accuracy of statistically based predictions improves only extremely slowly as the number of data points used increases. The quality aspect of the data means that they all have the same inherent assumptions and fidelity. None of these issues come into consideration with probabilistic technology (PT). The user of PT creates explicit uncertainty models of the inputs to a process. These models are then used to create a distribution of outputs or responses from a given process model. By varying the assumptions used to create the input uncertainty models, different output distributions are created for the same responses and the robustness of the uncertainty models can be judged by the effect the uncertainty choices have upon the outputs. When small changes in the uncertainty models make large changes in the response distributions, this is an indication that uncertainties are great and that more data can be obtained to improve the process model. In contrast, if the output response distributions show little variation when the input uncertainty models are changed, the user gains can be confident that they have adequately captured the important features of the input uncertainties.

What kind of technology support does PredictionProbe offer its customers?

PredictionProbe is a customer-centered organization which prides itself on providing 5-Star customer service and support. Our trained technical support team are skilled in communicating the many aspects of our technology and operating functions of our software. Our support includes software maintenance, software updates, and telephonic, email and/or bulletin board customer support.