Experimental design: Achieve product value upgrade through system input optimization
The essence of experimental design is a methodology centered on "precise regulation of input variables". It breaks out of the inefficient cycle of random trial and error. By structurally combining and testing all the "resource elements" that affect the results in the entire production/R & D process, it ultimately drives up the output value of products in a targeted manner. Its core logic can be summarized as "grasp the source, find the rules, and determine the optimum": first, clarify which inputs directly affect the results; then, systematically verify the combined effects of these inputs; finally, lock in the configuration that can produce the best output.
Input variables: The "five - dimensional resource code" in the production chain
In experimental design, the term "resources" does not refer to the general "materials or human resources", but rather the five quantifiable input variables that directly determine the product outcome (commonly referred to as "personnel, machinery, materials, methods, and environment" in the industry). Each category contains key factors that influence product quality:
People: The skill boundaries of operators (such as the years of holding a certificate for welders and the programming proficiency of CNC operators), operational compliance (whether they strictly follow the SOP), and state stability (such as the fatigue level after 4 consecutive hours of work). For example, on the same production line, the defective welding rate of skilled workers may be 30% lower than that of new workers, and fatigue work will double this figure. The state fluctuations of "people" are a hidden quality hazard for many discrete manufacturing enterprises.
Machine: The performance status of the equipment (such as the calibration accuracy of the pressure sensor of an injection molding machine), maintenance frequency (such as the replacement cycle of the lathe tool), and parameter settings (such as the rotation speed of the ink roller of a printing press). For example, a pressure deviation of 10 bar in an injection molding machine will result in uneven wall thickness of plastic parts; a tool not replaced on time will cause the dimensional tolerance of metal parts to exceed the standard by 0.05 mm. The accuracy of the "machine" is the foundation for product consistency.
Materials: Differences in the properties of raw materials (such as the carbon content of steel, the melt index of plastic particles), storage conditions (such as the refrigeration temperature of pharmaceutical raw materials), and pre - treatment methods (such as the pickling time of metal sheets). For example, damp plastic particles can cause bubbles in injection - molded parts, and steel with too high carbon content can lead to brittle fracture of parts. The stability of "materials" directly determines the lower limit of product performance.
Method: The logic of the technological process (e.g., assemble first then detect vs. detect first then assemble), parameter combinations (e.g., baking temperature × time), and detection standards (e.g., manual visual inspection vs. machine vision inspection). For example, adjusting the baking temperature from 180°C to 200°C and shortening the time by 3 minutes can make the cake taste more delicate and reduce scorching. The optimization of the "method" is the key to improving efficiency.
Environment: Controllable parameters of the environment (such as the humidity in an electronics workshop, the cleanliness level in a pharmaceutical workshop, and the sterilization temperature in a food workshop). For example, if the humidity is lower than 40%, the breakage rate of textile yarns will increase by 20%, and if the cleanliness does not meet the standard, the microbial content in pharmaceuticals will exceed the limit. The stability of the "environment" is the threshold for high - end manufacturing.
These variables do not exist in isolation. For example, insufficient training of "people" will magnify the parameter deviation of "machines"; the batch differences of "materials" will offset the optimization effect of "methods". The first step in experimental design is to convert these "hidden influences" into "testable variables".
Purposeful optimization: From "single-point trial and error" to "systematic optimization search"
The "purposefulness" of experimental design lies in rejecting "isolated adjustment" and embracing "variable interaction". Traditional optimization methods often fall into the trap of "testing each time a variable is changed". For example, when the defective rate is considered high, one might first change the raw materials. If it doesn't work, then adjust the machine parameters. Finally, train the workers. This approach is not only inefficient but also prone to overlooking the synergistic effects between variables (for example, the combined effect of "worker training + machine parameter adjustment" may be more significant than individual adjustments).
The logic of experimental design is to use "structured experiments" to cover the key combinations of variables.
- Using the orthogonal experiment method, the key levels of multiple variables can be covered with the minimum number of experiments (for example, for the combination of 3 variables and 3 levels, only 9 experiments are needed to replace 27 full - factorial experiments), and the optimal solution such as 4 - hour training + 150 bar pressure + Batch A can be quickly found.
- Using the response surface method, the continuous changes of variables can be simulated (for example, the temperature ranges from 100°C to 200°C and the time ranges from 10 minutes to 30 minutes), and the "sweet spot" where the product performance (such as strength and taste) reaches the peak can be accurately located.
- By using the Taguchi method, we can focus on the anti - interference ability. For example, by optimizing the process, the product can remain qualified even when there are minor mistakes made by people or small deviations of machines, thus improving its stability.
The value of this approach lies in replacing experience with data. Instead of adjusting parameters based on "the intuition of old workers," enterprises can find "the causal relationship between input variables and output results" through experiments, shifting from "producing good products by luck" to "stably producing good products through systematic methods."
Output goal: Four specific portraits of the "more ideal product"
The ultimate goal of experimental design is to make products evolve from "qualified" to "more ideal," and the core of "more ideal" is to address the real pain points of users or the production end. Specifically, it can be broken down into four major indicators:
More stable: Reduce output variation. For example, regarding the battery life of a mobile phone, it is optimized from "an average of 48 hours with a fluctuation of ±5 hours" to "an average of 48 hours with a fluctuation of ±1 hour". Users will no longer have the annoying experience of "the battery lasts well today, but suddenly runs out tomorrow".
More compliant: Raise the performance ceiling. For example, increase the strength qualification rate of automobile parts from 95% to 99% to avoid the recall risk caused by individual defective products; increase the brightness of LED lights from 800 lumens to 1000 lumens, directly outperforming competitors.
Lower cost: Reduce input waste. For example, by optimizing the raw material ratio, the use of plastic granules can be reduced by 5% while maintaining the product strength. This can save millions of yuan in raw material costs annually. Or, by adjusting the process parameters, the scrap rate can be reduced from 5% to 2%, directly reducing the rework cost.
More efficient: Shorten the production cycle. For example, by optimizing the "human - machine" cooperation process, the cycle time of the assembly line is reduced from 60 seconds to 50 seconds, and the daily output is increased from 1,000 pieces to 1,200 pieces, with an efficiency improvement of 20%. For enterprises with tight production capacity, this is equivalent to "expanding production without adding equipment".
In short, experimental design is not a "black-box operation." Instead, it uses the cause-and-effect logic of "input-output" to transform "experience-driven" into "data-driven." It helps enterprises break out of the cycle of "firefighting optimization" and shift from "passively solving problems" to "actively designing good products." This transformation is precisely a crucial step for modern manufacturing to move from "mass production" to "refined operation."