Synchronization-Free Multivariate Statistical Process Control for Online Monitoring of Batch Process Evolution

Synchronization of variable trajectories from batch process data is a delicate operation that can induce artifacts in the definition of multivariate statistical process control (MSPC) models for real-time monitoring of batch processes. The current paper introduces a new synchronization-free approach for online batch MSPC. This approach is based on the use of local MSPC models that cover a normal operating conditions (NOC) trajectory defined from principal component analysis (PCA) modeling of non-synchronized historical batches. The rationale behind is that, although non-synchronized NOC batches are used, an overall NOC trajectory with a consistent evolution pattern can be described, even if batch-to-batch natural delays and differences between process starting and end points exist. Afterwards, the local MSPC models are used to monitor the evolution of new batches and derive the related MSPC chart. During the real-time monitoring of a new batch, this strategy allows testing whether every new observation is following or not the NOC trajectory. For a NOC observation, an additional indication of the batch process progress is provided based on the identification of the local MSPC model that provides the lowest residuals. When an observation deviates from the NOC behavior, contribution plots based on the projection of the observation to the best local MSPC model identified in the last NOC observation are used to diagnose the variables related to the fault. This methodology is illustrated using two real examples of NIR-monitored batch processes: a fluidized bed drying process and a batch distillation of gasoline blends with ethanol.

Development of Profiling Methods for Contraband Firearm Volatile Odor Signatures

The early detection and location of firearm threats is critical to the success of any law enforcement operation to prevent a mass shooting event or illegal transport of weapons. Prevention tactics such as firearm detection canines have been at the front line of security tools to combat this national security threat. Firearm detection canines go through rigorous training regimens to achieve reliability in the detection of firearms as their target odor source. Currently, there is no scientific foundation as to the chemical odor signature emitted from the actual firearm device that could aid in increased and more efficient canine training and performance protocols or a better understanding of the chemistry of firearm-related odorants for better source identification. This study provides a novel method application of solid phase microextraction-gas chromatography-mass spectrometry (SPME-GC-MS) as a rapid system for the evaluation of odor profiles from firearm devices (loaded and unloaded). Samples included magazines (n = 30) and firearms (n = 15) acquired from the local law enforcement shooting range. Headspace analysis depicted five frequently occurring compounds across sample matrices including aldehydes such as nonanal, decanal, octanal and hydrocarbons tetradecane and tridecane. Statistical analysis via principal component analysis (PCA) highlighted a preliminary clustering differentiating unloaded firearms from both loaded/unloaded magazines and loaded firearm devices. These results highlight potential odor signature differences associated with different firearm components. The understanding of key odorants above a firearm will have an impact on national security efforts, thereby enhancing training regimens to better prepare canine teams for current threats in our communities.

Generalization Across Acetone Peroxide Homemade Explosives by Detection Dogs

Effective explosives detection requires dogs to generalize their response to untrained variations of targets that are related to those with which they were trained. Previous research suggests that dogs tend to be highly specific to their trained odors, and are sensitive to alterations in odor profiles. Triacetone triperoxide (TATP) is an increasingly popular homemade explosive due to the widespread accessibility of starting materials. The large variety of reagent sources and production approaches yields high variability in deployed formulations. Whether dogs trained with pure forms of TATP generalize to other variations is unknown, representing a potentially significant security gap. In the current study, we tested dogs (n = 11) previously trained to detect pure TATP with four variants: diacetone diperoxide (DADP), a homologue often created as a TATP byproduct, and three different clandestine TATP formulations designed to emulate those used by terrorists or insurgents. On average, dogs detected each untrained variant at rates equivalent to the trained TATP (ps > 0.07), with individual variability in first-trial alerts for some of the variants. Chemical analyses paralleled the canine results, showing distinct similarities and differences. For the TATP samples, the laboratory-grade was the purest sample tested and did not contain DADP or the TATP homologue that the three clandestine versions showed in their respective headspace profiles. The headspace results showed that each sample could be clearly identified as TATP, yet they showed recognizable differences due to their individual syntheses. These findings suggest that training on pure TATP may be effective for generalization to untrained variants. Further research is necessary to identify factors that influence individual variation in generalization between dogs, as well as other explosives.

Comparison of Biological Age Prediction Models Using Clinical Biomarkers Commonly Measured in Clinical Practice Settings: AI Techniques Vs. Traditional Statistical Methods

In this work, we used the health check-up data of more than 111,000 subjects for analysis, using only the data with all 35 variables entered. For the prediction of biological age, traditional statistical methods and four AI techniques (RF, XGB, SVR, and DNN), which are widely used recently, were simultaneously used to compare the predictive power. This study showed that AI models produced about 1.6 times stronger linear relationship on average than statistical models. In addition, the regression analysis on the predicted BA and CA revealed similar differences in terms of both the correlation coefficients (linear model: 0.831, polynomial model: 0.996, XGB model: 0.66, RF model: 0.927, SVR model: 0.787, DNN model: 0.998) and R2 values. Through this work, we confirmed that AI techniques such as the DNN model outperformed traditional statistical methods in predicting biological age.

Establishing Multivariate Specification Regions for Incoming Raw Materials Using Projection to Latent Structure Models: Comparison Between Direct Mapping and Model Inversion

Increasing raw material variability is challenging for many industries since it adversely impacts final product quality. Establishing multivariate specification regions for selecting incoming lot of raw materials is a key solution to mitigate this issue. Two data-driven approaches emerge from the literature for defining these specifications in the latent space of Projection to Latent Structure (PLS) models. The first is based on a direct mapping of good quality final product and associated lots of raw materials in the latent space, followed by selection of boundaries that minimize or best balance type I and II errors. The second rather defines specification regions by inverting the PLS model for each point lying on final product acceptance limits. The objective of this paper is to compare both methods to determine their advantages and drawbacks, and to assess their classification performance in presence of different levels of correlation between the quality attributes. The comparative analysis is performed using simulated raw materials and product quality data generated under multiple scenarios where product quality attributes have different degrees of collinearity. First, a simple case is proposed using one quality attribute to illustrate the methods. Then, the impact of collinearity is studied. It is shown that in most cases, correlation between the quality variable does not seem to influence classification performance except when the variables are highly correlated. A summary of the main advantages and disadvantages of both approaches is provided to guide the selection of the most appropriate approach for establishing multivariate specification regions for a given application.

Different Methods for Determining the Dimensionality of Multivariate Models

A tricky aspect in the use of all multivariate analysis methods is the choice of the number of Latent Variables to use in the model, whether in the case of exploratory methods such as Principal Components Analysis (PCA) or predictive methods such as Principal Components Regression (PCR), Partial Least Squares regression (PLS). For exploratory methods, we want to know which Latent Variables deserve to be selected for interpretation and which contain only noise. For predictive methods, we want to ensure that we include all the variability of interest for the prediction, without introducing variability that would lead to a reduction in the quality of the predictions for samples other than those used to create the multivariate model.

Improved Understanding of Industrial Process Relationships Through Conditional Path Modelling With Process PLS

Understanding how different units of an industrial production plant are operationally related is key to improving production quality and sustainability. Data science has proven indispensable in obtaining such understanding from vast amounts of historical process data. Path modelling is a valuable statistical tool to obtain such information from historical production data. Investigating how relationships within a process are affected by multiple production conditions and their interactions can however provide an even deeper understanding of the plant’s daily operation. We therefore propose conditional path modelling as an approach to obtain such improved understanding, demonstrated for a milk protein powder production plant. For this plant we studied how the relationships between different production units and steps are dependent on factors like production line, different seasons and product quality range. We show how the interaction of such factors can be quantified and interpreted in context of daily plant operation. This analysis revealed an augmented insight into the process that can be readily placed in the context of the plant’s structure and behavior. Such insights can be vital to identify and improve upon shortcomings in current plant-wide monitoring and control routines.

Novel Clinimetric Toolset to Quantify the Stability of Blood Pressure and Its Application to Evaluate Cardiovascular Function After Spinal Cord Injury

Profound dysfunction of the cardiovascular system occurs after spinal cord injury (SCI), which is a leading cause of mortality in this population. Most individuals with chronic SCI experience transient episodes of hypotensive and hypertensive blood pressure in response to daily life activities. There are currently limited tools available to evaluate the stability of blood pressure with respect to a reference range. The aim of this study was to develop a clinimetric toolset for accurately quantifying stability of the blood pressure measurements and taking into consideration the complex dynamics of blood pressure variability among individuals with SCI. The proposed toolset is based on distribution of the blood pressure data points within and outside of the clinically recommended range. This toolset consists of six outcome measures including 1) total deviation of the 90% of the blood pressure data points from the center of the target range (115 mmHg); 2) The area under the cumulative distribution curve starting from the percentage of blood pressure measurements within the range, and the percentage of values within symmetrically expanded boundary ranges, above and below the target range; 3) the slope of the cumulative distribution curve that is calculated by fitting an exponential cumulative distribution function and the natural logarithm of its rate parameter; 4) its x- and 5) y-axis intercepts; and 6) the fitting error. These outcome measures were validated using blood pressure measurements recorded during cardiovascular perturbation tests and prolonged monitoring period from individuals with chronic SCI and non-injured controls. The statistical analysis based on the effect size and intra-class correlation coefficient, demonstrated that the proposed outcome measures fulfill reliability, responsiveness and discrimination criteria. The novel methodology proposed in this study is reliable and effective for evaluating the stability of continuous blood pressure in individuals with chronic spinal cord injury.