Batch process reliability

N.J. Scenna. Some aspects of fault diagnosis in batch processes. Reliability Engineering and... 

The process for drug B has been shown to operate within narrow limits and yield finished dosage forms that are therapeutically equivalent, as measured by standard product release criteria. There is no reason to believe subsequent batches will perform differently as long as all conditions remain static. Despite this generally favorable prognosis, additional work is necessary to provide the assurance of process reliability expected today. 

There remains the unanswered requirement to demonstrate blend uniformity of active ingredients B2. This issue might be addressed by testing the blends of a series of batches until sufficient data are accumulated to consider the process reliable. Hwang et al. have provided some insight into establishing an in-process blend test [15]. The vali-... 

In this paper we would like to describe a new design, based on gas chromatographic analysis of the monomer mixture, for production of constant composition copolymers and its application to emulsion copolymerization. This design was already shortly described and applied to solution copolymerization (3) of methylmethacrylate and vinylidene chloride. Since then, the apparatus was made more simple, more reliable and more accurate. It is actually monitored by an analogic computering system which keeps the ratio of the monomers constant by controlling the addition of one of them. The process based on it can be called corrected batch process because the initial value of this ratio is kept up to the end. 

Unsteady-state or dynamic simulation accounts for process transients, from an initial state to a final state. Dynamic models for complex chemical processes typically consist of large systems of ordinary differential equations and algebraic equations. Therefore, dynamic process simulation is computationally intensive. Dynamic simulators typically contain three units (i) thermodynamic and physical properties packages, (ii) unit operation models, (hi) numerical solvers. Dynamic simulation is used for batch process design and development, control strategy development, control system check-out, the optimization of plant operations, process reliability/availability/safety studies, process improvement, process start-up and shutdown. There are countless dynamic process simulators available on the market. One of them has the commercial name Hysis [2.3]. 

G. Maria and D.W.T. Rippin, Modified Integral Procedure (MIP) as a Reliable Short-Cut Method in Mechanistic Based ODE Kinetic Model Estimation Non-Isothermal and (Semi-)Batch Process Cases, Comp. Chem. Eng. 19 (1995) S709-S714. 

An optimal control strategy for batch processes using particle swam optimisation (PSO) and stacked neural networks is presented in this paper. Stacked neural networks are used to improve model generalisation capability, as well as provide model prediction confidence bounds. In order to improve the reliability of the calculated optimal control policy, an additional term is introduced in the optimisation objective function to penalise wide model prediction confidence bounds. PSO can cope with multiple local minima and could generally find the global minimum. Application to a simulated fed-batch process demonstrates that the proposed technique is very effective. 

Keywords Batch processes. Neural networks. Particle swam optimisation. Reliability. 

It must be stressed that statistical sampling is not infallible and in particular can never reliably detect rogue pockets of excessive variability, since its confidence relies on faults being randomly distributed throughout the batch under examination. This is not always the case, since in any batch process isolated rogue defects can occur which may escape the defect detection system operated by supplier or customer. 

Independent of the mode of operation, the most critical point during the course of the process corresponds to the time at which the maximum driving temperature difference between jacket and reaction mixture temperature occurs. Due to this fact the safety assessment focuses its efforts on the most reliable prediction of this point and its stability. In all three cases this critical point of a batch process is mathematically characterized by the condition dT/dt = 0 for the heat balance. 

The applicability of radioanalytical chemistry methods is tested initially with radioactive tracers and realistic mock samples. A tracer that can be measured reliably and conveniently, such as a radionuclide that emits gamma rays, is preferred. If initial tracer tests are successful, tests are repeated with the media that will be analyzed. These tests must demonstrate that the radionuclide of interest is recovered consistently with good yield and that no interfering radionuclides or solids remain. The extent of reproducibility is determined by analyzing acmal samples in replicate for chemical and radionuclide yield. Replicate samples are identical samples from the same batch, processed and counted separately to assess the variability or uncertainty in the analysis. The recommended test statistic for a duplicate analysis is computed using Eq. (10.27) ... 

For tactical production planning roughly assigning production quantities to production sites suffices for deriving reliable mid-term plans. On the operational level, however, a more detailed view on the involved production processes is needed. In particular, the planning of batch processes is a challenging task as it involves... 

It is also necessary for the user to take into account that fine chemicals for laboratory use are usually produced by batch processes rather than by continuous methods of manufacture. Therefore, because specifications set minimum standards of purity, the analysis of one or two batches of a given reagent from a supplier will not give reliable general information concerning the impurity profile of that product. 

The in-line method seems to be the most straightforward in that the sensing device is embedded within the process such as is thecase with a thermocouple. The problem with in-line analyzers is that except for the most ru ed and reliable measurement devices, there can be problems when the analyzer fails. With an in-line analyzer the process typically must be shut down to calibrate, fix or change the analyzer. This may not be a major problem for batch processes whose cycle times are fairly short, but is a major problem for continuous processes, which are designed to operate for years without a shutdown. 

It also follows that as the pH of a waste stream varies the soluble metal content will also vary. Some processes, especially batch processes, are prone to variation in the chemical content of their waste streams as a direct consequence of intermittent variations in pH. As a result, all metal removal methods must be coupled to effective and reliable pH measurement of the waste stream. 

---